Desenvolvimento Económico Português No Espaço Europeu 2002

Conferência do Banco de Portugal
24 - 25 Maio 2002
Desenvolvimento Económico
Português no Espaço Europeu:
Determinantes e Políticas
Portuguese Economic
Development in the European Context:
Determinants and Policies
comunicações
proceedings
Banco de Portugal
Av. Almirante Reis, 71
1150-012 Lisboa
Economic Research Department
Printed by
Mira Dupla, Artes Gráficas Lda.
Printing number: 750
Duty copy no. 208266/04
ISBN 972-9479-83-6
Nota de Apresentação
Foreword
Portuguese Economic Development in the European Context: Determinants and Policies V
NOTA DE APRESENTAÇÃO
A Conferência “Desenvolvimento Económico Português no Espaço Europeu:

Determinantes e Políticas” constituiu uma iniciativa do Banco de Portugal que
concretizou uma preocupação que oportunamente manifestei no sentido de que «o Banco
tem o dever de apoiar mais o debate informado dos problemas económicos e sociais do
país mesmo que transcendam o domínio estrito da política monetária». Procuramos seguir,
assim, o exemplo de outros Bancos Centrais e, consequentemente, a esta Conferência
outras se seguirão com o mesmo objectivo e formato.
Procurou configurar-se esta iniciativa com alguns traços relativamente inovadores. Em

primeiro lugar, o Banco de Portugal procurou incentivar economistas oriundos da
academia a reflectirem sobre os desafios do desenvolvimento económico em Portugal. Por
outro lado, para além de uma reflexão de grande rigor técnico, pretendeu-se que os
estudos apresentados fossem úteis do ponto de vista de política económica, devendo ter um
carácter aplicado e considerando Portugal como um “case-study”.
Os estudos que estão na base da Conferência provieram quer de um concurso público

dirigido aos Departamentos de Economia das Universidades Portuguesas, quer de convites
directos a alguns economistas. Seguiu-se um processo de selecção dos estudos, tarefa que
foi da responsabilidade do Comité Científico, presidido por Luís Campos e Cunha (Vice-
Governador do Banco de Portugal e Universidade Nova de Lisboa), e com a participação
de Pedro Duarte Neves (Banco de Portugal e Universidade Católica de Portugal), José
Ferreira Machado (Universidade Nova de Lisboa), Isabel Horta Correia (Banco de
Portugal e Universidade Católica de Portugal) e Pedro Portugal (Banco de Portugal e
Universidade Nova de Lisboa).
Da qualidade e conteúdo dos trabalhos seleccionados dá conta esta publicação, onde são
divulgados os estudos apresentados na primeira Conferência “Desenvolvimento
Económico Português no Espaço Europeu: Determinantes e Políticas”, que se realizou a
24 e 25 de Maio, na Fundação Calouste Gulbenkian, em Lisboa.
Vítor Constâncio
Governador
Portuguese Economic Development in the European Context: Determinants and Policies VII
FOREWORD
The Conference “Portuguese Economic Development in the European Context:

Determinants and Policies” was promoted by the Banco de Portugal, in line with my
concern that the “Bank should contribute more actively to the debate on the economic and
social problems of the country, even when these go beyond the strict scope of monetary
policy”. Against this background, and following the tradition of other Central Banks, we
are envisaging the organisation of further Conferences of the same type.
This initiative presents a number of rather innovative features. The Banco de Portugal
encouraged a number of academics to reflect on the challenges relating to the economic
development of Portugal. In addition to promoting a debate of technical quality, it was
deemed desirable that the research submitted should be useful from an economic policy
viewpoint and should take Portugal as a case-study.
The research presented at the Conference resulted both from a call for papers addressed to
the Departments of Economics of Portuguese Universities and from direct invitation to a
number of economists. The papers submitted were subject to a selection process carried out
under the responsibility of a Scientific Committee chaired by Luís Campos e Cunha
(Deputy-Governor of the Banco de Portugal and Universidade Nova de Lisboa), with the
participation of Pedro Duarte Neves (Banco de Portugal and Universidade Católica de
Portugal), José Ferreira Machado (Universidade Nova de Lisboa), Isabel Horta Correia
(Banco de Portugal and Universidade Católica de Portugal) and Pedro Portugal (Banco
de Portugal and Universidade Nova de Lisboa).
Attesting the quality of the research produced, this publication puts together the papers
presented in the first Conference “Portuguese Economic Development in the European
Context: Determinants and Policies”, that took place on 24 and 25 May, at the Calouste
Gulbenkian Foundation, in Lisbon.
Vítor Constâncio
Governor
Índice
Contents
Contents V
Nota de Apresentação / Foreword
Nota de Apresentação ................................................................................................... V

Foreword ...................................................................................................................... VII
Vítor Constâncio
Conferência do Banco de Portugal sobre “Desenvolvimento Económico

no Espaço Europeu: Determinantes e Políticas” / Conference Held by the Banco
de Portugal on “Portuguese Economic Development in the European Context:
Determinants and Policies”
Conference Held by the Banco de Portugal on “Portuguese Economic Development

in the European Context: Determinants and Policies” - A Personal Summary ............ 3
José A. Ferreira Machado
Sessão 1 – Convergência Real Portuguesa / Session 1 – Portuguese Real

Convergence
Why Growth Rates Differ in the Long Run: Capital Deepening, Productivity Growth
and Structural Change in Portugal, 1910-1990............................................................. 13
Pedro Lains (Instituto de Ciências Sociais, Universidade de Lisboa)
Portuguese Economic Growth Re-Examined: an Anti-Fado Manifesto ....................... 33

Miguel Lebre de Freitas (Universidade de Aveiro)
Productivity Convergence: Portugal and the European Union ..................................... 61

Pedro Pita Barros (Universidade Nova de Lisboa e CEPR)
Sessão 2 – Geografia Económica e Infra-estruturas / Session 2 – Economic

Geography and Infra-structures
Public Investment in Transportation Infrastructures and Economic Performance

in Portugal ................................................................................................................... 85
Alfredo Marvão Pereira (The College of William and Mary)
Jorge M. Andraz (Universidade do Algarve)
Production Technologies and Technical Efficiency: Evidence from Portuguese

Manufacturing Industry ................................................................................................ 127
Ana Faria (Universidade do Minho)
Paul Fenn (Nottingham University Business School)
Alistair Bruce (Nottingham University Business School)
VI Contents
Transport Policies in Light of the New Economic Geography:

The Portuguese Experience .......................................................................................... 147
António Carlos Fernandes Teixeira (CORE, Université Catholique de Louvain e
Faculdade de Economia do Porto)
Acessibilidade Económica e Bem-Estar: Evidência da Península Ibérica.................... 189

Armando Pires (Instituto Superior de Economia e Gestão)
Sessão 3 – Capital Humano e Crescimento / Session 3 – Human Capital and Growth
Education and Earnings in Portugal ............................................................................. 219

Pedro Telhado Pereira (Universidade da Madeira, CEPR e IZA)
Pedro Silva Martins (University of Warwick)
Evaluating Efficiency in the Portuguese Health and Education Sectors....................... 253

Miguel St. Aubyn (Instituto Superior de Economia e Gestão)
Qualification Requirements and Educational Attainment in Portugal, 1985-1997....... 303

Maria Clementina Santos (Faculdade de Economia do Porto e CETE)
Manuel Mendes de Oliveira (Faculdade de Economia do Porto e CETE)
Sessão 4 – Instituições, Reformas e Crescimento / Session 4 Institutions,

Reforms and Growth
Firms Financial Markets and the Law: Institutions and Economic Growth
in Portugal .................................................................................................................... 327
José Albuquerque Tavares (Universidade Nova de Lisboa)
Labor Markets in Portugal: Recent Performance and Challenges for Development

in the European Context ............................................................................................... 395
Daniel Traça (INSEAD)
A Justiça e seu Impacto Sobre as Empresas Portuguesas ............................................. 445

Célia da Costa Cabral (Universidade Nova de Lisboa)
Armando Castelar Pinheiro (BNDES e Fundação Getúlio Vargas)
Fixed-Term Contracts, Employment Flows, and Productivity ..................................... 479

José Manuel Varejão (Faculdade de Economia do Porto)
Conferência do Banco de Portugal “Desenvolvimento
Económico Português no Espaço Europeu: Determinantes e
Políticas - Uma Síntese Pessoal
Conference Held by the Banco de Portugal on “Portuguese

Economic Development in the European Context:
Determinants and Policies” – a Personal Summary
CONFERENCE HELD BY THE BANCO DE PORTUGAL ON “PORTUGUESE
ECONOMIC DEVELOPMENT IN THE EUROPEAN CONTEXT:
DETERMINANTS AND POLICIES”
- A PERSONAL SUMMARY
José A. Ferreira Machado
1. Introduction
On the 24 and 25 May 2002 the Banco de Portugal held a conference with the main
purpose of encouraging economists from the academy to reflect upon the challenges of
economic development in Portugal.
The main subject of the conference was the so-called “real convergence” of the
Portuguese economy. The focus was on “how and where should the economy converge”
rather than on “whether the economy is converging”, and what role do government policies
play in this process.
Economic development is a very wide and open subject. Therefore, the conference
encompassed a general subject and some more specific subjects.
The general topic dealt with the analysis of overall growth and development and also
with the “establishment of facts” on Portuguese real convergence. The specific subjects
suggested by the hosts were as follows: Polarisation or dispersion of growth in Europe;
Laws and institutions in the development process; Public finances and growth; Factor
markets and growth; Human capital and growth.
This article summarises the addresses delivered at the conference as well as the main
indications of economic policy they contain. This synthesis is personal and, therefore, is
probably subjective. Also, it is not intended to do justice to the technical details of
contributions. Above all, and paraphrasing a well-known definition of culture, it represents
what the author recalls after having forgotten what he had learnt in reading the different
contributions.
2. Convergence
Economic growth is important. First, for the obvious reason that through growth,
citizens may aspire to a better standard of living. A differential of one percentage point
(p.p.) in the average rate of growth of output would double the average standard of living in
approximately two generations. But there is also a relative component in growth, i.e.
Portugal’s performance compared to other economies: are we getting poorer or wealthier
than citizens, for instance, in Southern European countries?
The article by Pedro Lains, “Economic growth in Portugal in the long run: investment,
productivity growth and structural changes in Portugal, 1910-1990,” attempts to answer
these questions. Among the four poorest countries in Europe-15 — Portugal, Spain, Greece
and Ireland —, Portugal was the country which converged more quickly in the course of the
20th century. In particular, in the period 1913-1998, real output per capita grew, at an
average of 2.79 per cent in Portugal, against 2.2 percent in Spain, 2.29 per cent in Greece
4 José Ferreira Machado
and 2.19 per cent in Ireland. In contrast, the nine wealthier countries grew only by 2.06 per
cent.
However, convergence has not been uniform over time: in the period between the two
wars (1913-1950) the annual convergence rate stood at 0.19 per cent; in the so-called
“golden age” — the period from 1950 to 1973 — the Portuguese economy converged at an
annual rate of 1.85 per cent; in the post-revolution period, the annual convergence rate
declined further to 0.44 per cent. Nevertheless, even during this period, growth in Portugal
stood above that in Spain and Greece and was only exceeded by that in Ireland.
The path of the economy in the course of the 20th century was, rather surprisingly, very
positive. What does it tell us about the future? Can that pace of growth be extrapolated? Or,
on the contrary, does the deceleration recorded in the last quarter of the last century already
anticipate a profound waning of convergence?
The contribution by Pedro Pita Barros, “Convergence in productivity: Portugal and the
European Union”, gives a pessimistic perspective on those issues. Analysing how aggregate
developments detected by Lains were reflected on the structure of Manufacturing, Barros
shows that productivity in Manufacturing has grown at declining rates and, more
importantly, a concentration of industrial output persists in sectors that make little
investment in research and development. As these sectors are considered by the author as
being of low growth in the long run, the maintenance of the specialisation pattern would
impair growth dynamics.
On the contrary, according to de Freitas (“Portuguese Economic Growth: A manifest
Anti-Fado”), there is some hope. For better or for worse, the past evolution cannot be
automatically extrapolated. As in the past, it will depend on the choices made concerning
policy and institutions. Illustrating his argument, Lebre de Freitas shows that in OECD and
EU counties there is no inverse relationship between the initial income level per capita and
subsequent growth: countries which are relatively poorer do not grow more rapidly and,
thus, convergence in this set of relatively wealthy countries does not necessarily occur.
What is behind the different growth rates in OECD countries? What is the margin for
economic policies? What policies should be favoured? The remaining contributions to the
conference attempted to answer these issues. But, as an introduction, Lebre de Freitas
anticipates — on an empirical basis — the significant role of factors such as the quality of
human resources (measured by the availability of skilled labour force), the quality of
institutions (reflected in the efficiency of justice and in bureaucracy, corruption, respect for
private property and the credibility of commitments made by the State) and, finally, the
flexibility of labour legislation.
3. Opening the black box
More growth requires more and better inputs or a more efficient use of those prevailing.
The conference provided important ideas on how this could be achieved.
3.1. Transport infrastructures
Transport infrastructures are important for development in as far as accessibility to

consumer, input or knowledge markets is an important determinant of decisions on the
location of economic activities.
Conference Held by the Banco de Portugal on “Portuguese Economic Development in the European Context: 5
Determinants and Policies - A Personal Summary
The contribution by Armando Pires (“Economic Accessibility and Welfare: Evidence in

the Iberian Peninsula”), presents indices of the economic accessibility of the different
peninsular regions, reflecting the trade costs incurred. This work reveals the importance of
transports and accessibility and provides a pessimistic diagnosis: given the current trade
pattern, the Portuguese regions record very low accessibility indices.
There seems to be room for important improvements in accessibility; given the
externalities involved, this is, in turn, a privileged area for government intervention in the
economy. The importance of this intervention as a development factor has long been
recognised in Portugal, and a substantial effort of public investment in transport
infrastructures has been made since the late 1980s.
Transport infrastructures not only directly promote growth by facilitating a more
efficient use of resources, but also have indirect effects via increasing employment and
investment. Taking all these effects into account, Pereira and Andraz (“Public Investment
in Transport Infrastructures and the Economic Performance in Portugal”) estimate that the
investment made in transport infrastructures in the 1980s and 1990s has recorded an annual
rate of return of around 16 per cent, clearly above that expected for private investment. This
investment has had thus an important impact on economic growth. Breaking down
investments, authors found stronger effects on output from investments in ports and in the
national and municipal road network.
Besides its effect on growth potential, investment in accessibilities is frequently
justified by the correction of regional disparities. António Teixeira (“Transport Policies in
the Light of the New Geographical Economy: The Portuguese Experience”), analyses
investment in transport infrastructures in this perspective. The reduction in transportation
costs induced by these investments does not necessarily lead to the regional dispersion of
industrial activities. There is a critical threshold which has to be overcome, otherwise there
might be an intensification of industrial polarisation. Investment made in Portugal over the
last decades of the last century was not sufficient to overcome that threshold. However, the
author estimates that the pursuance of the prevailing expansion plan of the road network
will lead, in the future, to a more balanced distribution of economic activities.
It is interesting to note that, using a different methodology, the mentioned article by
Armando Pires also indicates how a reduction in trade costs leading to what he calls “a
more complete Iberian integration” will not bring greater geographical equality. On the
contrary, it will tend to benefit mainly those regions which are already more advanced, inter
alia, Lisboa e Vale do Tejo.
These effects, perhaps paradoxical in the light of common sense, should be seen in
perspective, since there is no evidence that sustained economic growth benefits from a
uniform geographical distribution of economic activities. The decrease in the regional
inequality is, strictly speaking, an extra economic purpose.
3.2. Human Capital
The connection between education and economic growth has been obvious for
economists since the beginning of their Science. Even Adam Smith had argued that public
money should be invested in education on the basis that benefits from education are felt by
not only those who benefit from it but also by society as a whole (c.f., Gylfason, 1999,
page. 21 and citations contained therein).
We have already seen in Lebre de Freitas how quality of human resources measured by
an index of skilled labour availability is an important explanatory factor of different growth
paths in OECD countries. Also in Lains it is estimated that, for the period 1973-1990, the
contribution of human capital (measured by the average number of years of education of
the labour force) for an annual output growth of 3.9 per cent stood at 1.6 per cent, almost
the same level as that of the accumulation of physical capital.
Overall, the level of education of the population increases growth potential. If this
happens is because, at an individual level, more educated employees tend to be more
productive. As more productive employees are better paid, it is possible to measure part of
the effects of education through its effect on wages. Typically, however, this measure of
private benefits will underestimate overall effects of education given the externalities
mentioned in the introduction of this section. Pereira and Martins (“Education and Wages
in Portugal”), acknowledged that employees in Portugal benefit from a high rate of return
of education: on averages an extra year of education gives rise, in employees otherwise
comparable, to an increase of wages of around 11 per cent. This is a high rate by
international patterns (on average in developed countries it reaches a value of 8 per cent)
and probably reflects the already mentioned relative shortage of skilled employees. This
interpretation seems likely if we notice that the rate of return is particularly high for higher
education (18 per cent) and, within it, for degrees in Engineering.
Typically, human capital is measured by variables related to education: years of
education, rates of attendance and also expenditure in education. These measures have two
obvious problems. First, they try to measure output through input without reflecting the
quality of education provided at school. Second, they ignore that factors such as health may
also improve the stock of human capital.
Notwithstanding that the Portuguese population in the past few years has recorded a
marked convergence to EU average levels, several indicators point to a state of health
below that in other countries. Similarly, the indicators of quality of education such as
secondary education graduation rates and the performance of students in international tests
present low levels. To grow faster, Portugal needs to increase the stock of human capital.
On the other hand, public expenditure plays a key role in direct financing of human
capital formation corresponding to virtually all the expenditure in education and to almost
two-thirds of the expenditure in health. Therefore the question is whether it is necessary to
make heavier investments or whether, on the contrary, funds should be more efficiently
used. St. Aubyn (“Assessment of Efficiency in Portugal in Health and Education Sectors”),
analyses precisely this question giving a clear answer: the improvement of human capital in
these two dimensions does not necessarily require higher investments, but rather structural
changes making its use more efficient.
3.3. Labour Market
As Traça remarks in his contribution “Labour Markets in Portugal: Recent Performance

and Challenges for Development in the European Framework”, a fundamental trend for
labour markets in industrialised countries is the increase in the volatility of labour demand.
In a context in which employment requirements will change considerably in terms both of
location (company, industry or region) and of skills, it is crucial to have a smooth
functioning of the labour market in order to avoid unemployment and ensure an efficient
allocation of work to different industries and firms.
Traça identifies two critical factors for success: flexibility and adaptability. Flexibility
refers to the ability of real wages to adjust, by reacting to the market signals so as to avoid
mismatches between wage and productivity developments and unemployment. In this
respect, past experience of wage setting mechanisms in Portugal — translated into a low
unemployment rate — provide positive expectations.
The situation is less favourable concerning “adaptability”. The low unemployment rate
conceals high long-term unemployment and a reduced labour flows both among jobs and
between unemployment and employment (vide Blanchard and Portugal, 2001). In the core
of this stagnation is an extremely protectionist labour legislation and the low efficiency of
the matching between the unemployed and employers.
The evidence of the mismatch between employees and employers in terms of
qualifications provided by Santos and Oliveira (“Qualifications Required and
Qualifications Provided in Portugal, 1985-1997"), supports the view of a stagnant labour
market and of the poor efficiency of active employment policies. The mismatch between
qualifications of employees and market requirements is the source of significant
productivity losses which, according to the estimates of Santos and Oliveira, may reach 5
per cent in the case of over-qualification for the function carried out.
It is interesting to note that market mechanisms seem to find ways to operate, even in
adverse legal contexts, so as to respond to employer and employee requirements. A good
example is the evolution of temporary contracts from their primitive function — a response
to temporary employment requirements — to become an important factor of labour market
flexibility, in particular concerning a more efficient employee-employer matching (vide
José Varejão, “Temporary Contracts, Employment Flows and Productivity”).
The aforementioned work by Daniel Traça isolates three fundamental elements to
promote the adaptability of the labour force. Two are related to the educational system —
both formal and lifelong — and the other to institutional changes. First, the education of
labour force, intended to ensure the ability of learning new tasks. Then, the easiness of
hiring and firing which ensures the easy flow of resources to sectors in relative expansion.
Finally, active employment policies which improve the matching and ensure the
professional training of the unemployed.
The need for “adaptability” — understood as the ability to react prompt and efficiently
to market signals — is not confined to the labour market. The adoption of “flexible
production technologies”, which enable adjustments in the output mix at low cost, may
have significant effects on overall productivity as it is documented in Faria and Bruce
(”Technological Flexibility and Efficiency: Evidence from Portuguese Manufacturing
Industry using a stochastic frontier approach”).
3.4. Institutions
Today there is a consensus among economists on the importance of institutions for

economic growth. However, which institutions are important and why are they important is
less consensual. The most important channel through which institutions may influence
growth is the impact on costs and uncertainties associated with economic transactions.
Typically, economic agents have imperfect and asymmetric information, are involved in
transactions that involve a great number of agents making the co-operation hard. In such a
world, institutions — legal or other, such as mere rules of conduct — are important, as they
reduce the costs of obtaining information, negotiation and implementation of contracts.
José Tavares (“Companies, Financial Markets and Laws: Institutions and Economic
Growth in Portugal”) considers that the degree of development of the Portuguese
institutions vis-à-vis those in other countries in areas such as the legal system, the internal
organisation of companies and the financial system, may be highly responsible for the low
level of income per capita and the modest convergence rate.
It is, however, at the level of legal and judicial systems that higher gains in terms of
growth potential may be recorded. The aggregated indicators considered by Tavares —
Rule of Law, Risk of Breach of Contracts, Risk of Expropriation, Access to Justice,
Efficiency of the Judicial System, Corruption and Compliance with Contracts — registered
levels below the EU average and the “Asian tigers”. The differences are mainly pronounced
concerning the indices of efficiency of the judicial system and compliance with contracts.
In particular, the length of judicial procedures in Portugal is the longest of the sample. As
an illustration, the procedure of collection of a cheque without provision takes almost twice
the time of the EU average.
The inefficiency of the judicial system is also confirmed by the business survey carried
out by Célia Cabral and Armando Bacelar (“Justice and its Impact on Portuguese
Companies”). The survey reveals that, for the companies surveyed, the major problem of
the Portuguese judicial system is its lack of celerity reflected, in particular, in time-
consuming legal proceedings. The assessment is also very negative with regard to access
costs. In short, the large majority of companies (88 per cent) considers the judicial system
as “bad” or “very bad”, i.e. slow and expensive.
4. Concluding remarks and policy implications
Sustained economic growth is not mainly the result of exogenous factors such as
technology or resources. If we compare the evolution in the course of the second half of the
20th century of countries initially as similar as Eastern Germany and Western Germany,
Austria and the Czech Republic, China and Taiwan or Northern Korea and Southern Korea,
we immediately realise the importance of economic regimes, institutions and policies. An
important conclusion to be drawn from the conference is, therefore, the importance of
choices made by societies, i.e. by all of us as citizens.
Most policies proposed are of an “horizontal” nature, i.e. directed to the fundamentals
of economic growth. Among the intervention areas discussed in the conference, I would
point out four: infrastructures, educational system, labour market and judicial system.
i. It is necessary to pursue the effort of improving transport infrastructures. However, I

would emphasize two points. In the early 1980s little had been done and, therefore, it
was possible to obtain the high rates of return estimated in Marvão and Andraz.
Presently, it is absolutely necessary to be more selective. But genuinely productive
investments pay themselves through additional tax revenues and, therefore, do not
generate pressures on the public debt.
ii. The situation of the educational system — either formal or vocational training or
even life-long training — is a cause of concern. The problem is not a result of the lack
of investment in education, since it has accompanied the evolution of more advanced
countries. It is the return of that investment that must be questioned, because no clear
results are being obtained (c.f., St. Aubyn and Pereira and Martins). The problem does
not seem to lie in the contents, although some authors have emphasized the need for
higher levels of exigency in subjects like English, sciences and mathematics (e.g., St.
Aubyn). Still on the contents of formal education, stress was also laid on the need to
ensure sound general competencies, of a wide spectrum, which facilitate the adjustment
to a constantly evolving market requirements (c.f., Traça).
iii. As outlined by several contributions, the problem of the educational system lies
mainly on its the failure to recognize and reward the merit of schools, teachers, and
students (e.g., St Aubyn). Such a new culture would imply an increase in school
autonomy and a strengthening of competition mechanisms among them. The key idea
should be “experimentation and evaluation”, i.e. to refuse global “top-down” reform
projects, but instead, allow competition — with evaluation and accountability —
among several curricular and governance models.
iv. It is necessary to evaluate the results of public investment in vocational training.
v. The labour market needs reforms which liberalise the employment protection
legislation, namely on collective dismissal for economic reasons and on procedural
impediments to individual dismissal (c.f., Traça, 2002). During this liberalisation
process it is important not to destroy the few flexibility elements existing in the current
framework, such as temporary contracts, without having created alternatives (vide,
Varejão).
vi. The suggestions for legal and judicial reforms arising from the diagnosis carried out
(vide, Tavares and Cabral and Pinheiro) indicate that, in general, Portugal does not
seem to need more laws or new ones, but rather a firm and prompt enforcement of the
existing legislation. The excessive trend to legal formalism results in a loss of
efficiency affecting growth without clear benefits in terms of citizens’ rights.
The implementation of the proposals presented in this article naturally involves (material,
social and political) costs and the — perhaps uncertain — results have a medium and long-
term horizon. They assume a wide consensus at the political level as to the diagnosis and
therapy to be adopted. But, only in this way may Portuguese acquire a renewed dynamics
of sustainable development.
References
Gylfason, T. “Principles of Economic Growth”. Oxford U. P., 1999.

Blanchard, O. and P. Portugal. “What hides behind an unemployment rate: comparing the
Portuguese and US labour markets”. American Economic Review, 91(1), 2001.
Sessão 1 - Convergência Real Portuguesa
Session 1 - Portuguese Real Convergence

WHY GROWTH RATES DIFFER IN THE LONG RUN: CAPITAL DEEPENING,
PRODUCTIVITY GROWTH AND STRUCTURAL CHANGE IN PORTUGAL,
1910-1990
Pedro Lains
Instituto de Ciências Sociais, Universidade de Lisboa
pedro.lains@ics.ul.pt
December, 2001
Paper prepared for submission to the Conference on

‘Desenvolvimento económico português no espaço europeu: determinantes e políticas’,
Banco de Portugal, 2002.
Abstract
This paper has two main purposes. The first is to define and describe the main cycles of
growth and convergence of the Portuguese economy, during the twentieth century
The second purpose of the paper is to explore the causes of economic growth and
slowdown over the century, within a growth accounting framework. We conclude that the
acceleration of growth after World War II was due to capital deepening and total factor
productivity growth and that the downturn after 1973 can be ascribed to the decline in factor
productivity growth.
The paper proceeds by exploring the causes of the post-1973 slowdown in factor
productivity. We propose an explanation that is based on the effects of structural change in
the industrial sector, which was induced by the increasing integration within the European
Union market.
14 Pedro Lains
1. Introduction
Throughout most of the nineteenth century, the levels of income per capita of the poor
economies of the western European periphery diverged from those of the first
industrializers. Contrarily, over the twentieth century, there was a convergence of incomes
per capita within the continent. Convergence, however, occurred with different degrees of
intensity and it was mostly concentrated in the period from 1950-73. This period has been
studied more deeply, particularly in a cross-country comparative perspective. Yet, our
understanding of the causes behind changes in convergence rates can be increased by
paying attention to changing patterns of economic growth within a single country, in a long
run perspective.1
Within the European periphery, Portugal had a particularly good performance over the
twentieth century.2 That performance can be partially explained by the fact that the country
was spared from the direct effects of World War I. Portuguese economic growth was also
less affected than other European countries by the contraction of international trade and
capital flows that followed the New York 1929 crash. Truly, emigration and revenues from
colonial trade were severely diminished through the impact of the Great Depression in
Brazil and Africa.3 But the fact is that, during the 1930s, Portugal faired better than Spain,
which was ravaged by civil war. Portugal was also not directly affected by World War II,
contrarily to what happened to Greece, which also suffered the effects of a civil war in the
late 1940s. After World War II, Portugal engaged in the emerging international institutional
framework, which regulated and fostered the resurgence of the international economy, and
Spain and Ireland followed about one decade later.4
Traditional explanations of growth and slowdown of the Portuguese economy have put
more emphasis on internal factors and, in particular, on economic policy options, which
would have shifted in important ways. To start with, the financial indiscipline that
characterized the Republican period (1910-26) would have been responsible for alleged
economic stagnation. The emergence of the Estado Novo, in 1933, although it enhanced
monetary and financial stability did not lead to higher levels of economic growth,
according to some authors, because of agrarian and industrial policy options, which would
have shifted domestic resources towards the ‘wrong’ sectors.5 Growth resumed after World
War II because, contrarily to the previous period, the dictatorship government imposed a
‘strategy aimed at economic growth and structural change’.6 The higher growth of the post
World War II period is also traditionally attributed to the fact that Portugal was a founding
member of the EFTA, which implied a shift towards open trade policies. Again, the
economy did not expand as much as it could have done, because opening up policies were
not backed by a more interventionist stance by the government, which kept the budget
balanced, for most of the years to the end of the dictatorship, in 1974.7 The slowing down
1
See Prados and Sanz (1996), on Spain, and Ó Gráda and O’Rourke (1996), on Ireland. See also van Ark (1996b).
2
On Portuguese growth performance over the twentieth century see Lains (1994). See also Mateus (1998).
3
Lains (1998).
4
See Leitão (2001).
5
See Lains (2003, forthcoming).
6
Marques (1988, pp. 23-26). See also Rosas (2000, Chap. 2).
7
Pereira de Moura (1973)
Why Growth Rates Differ in the Long run: Capital Deepening, Productivity Growth 15
and Structural Change in Portugal, 1910-1990
of economic growth that followed is frequently attributed to the revolution and the
nationalization spree in 1975. Contrarily, the European Union, which Portugal joined in
1986, and privatization would have set the economy on the good direction again. The
relation between changes in trends in economic policy and changes in trends in economic
growth and convergence is however weaker than it is often posited.8
Some economists have questioned the dominance of internal factors in shaping the
pattern of growth of the Portuguese economy. Silva Lopes (1996), recognizes that both
internal and external factors were relevant for ‘the acceleration of economic development,
macroeconomic stability and increasing openness of the economy [during 1950-73]’ but he
stresses that ‘it was above all because of foreign stimuli that the Portuguese economy
expanded as it did and became more open to foreign relations’. The same author holds that
economic growth slowdown after 1973 was also mainly a consequence of ‘external
factors’.9 In order to explain the change in the rhythm of growth after 1973, Marques
Mendes (1993) argues that ‘it was the reduction by half of the growth rate of the European
Communities during the 1970s and 1980s that provides the fundamental explanation for the
slowdown in reducing the income gap between Portugal and the centre’. Moreover,
according to the same author, the ‘complete halt in convergence can only be explained by
the fact that Portugal run into balance of payments problems and has suffered terms of trade
losses during the [1971-92] decades’.10
This paper deals with the causes of Portugal’s long-run economic performance during
the twentieth century. We want to detect the causes for changes in rates of growth and
convergence over the century. We use parameters from augmented Solow models in order
to estimate the effect of inputs on growth and the size of changes in total factor productivity
growth over the twentieth century. The remaining of the paper is organized as follows: the
next section sets down the main periods of growth and convergence of the Portuguese
economy; section 3 analysis the contribution of the growth of factor inputs and productivity
on overall economic growth, within a growth accounting framework; section 4 discusses
the causes of slowdown of economic growth after 1973; section 5 presents the main
conclusions of the paper.
2. The comparative performance of the Portuguese economy
The evolution of the Portuguese economy during the twentieth century has some
common points with the evolution of the European economy. In particular, Portugal was
affected by the international financial disequilibrium in the period after World War I and
took part in the general economic expansion in the second post-war period, which ended in
1973. But trends of economic growth and fluctuations in Portugal and the rest of Europe
also show many important differences. Graph 1 depicts an index for the growth of
Portugal’s real income per capita and for the growth of an unweighted average income per
8
See Costa Lobo (2000).
9
Silva Lopes (1996, pp. 17 and 23).
10
Marques Mendes (1993, p. 13).
16 Pedro Lains
capita for nine European economies.11 Table 1 shows annual growth rates between peak
years of the two GDP series.
Table 1
Growth of real income per capita in Portugal and Europe, 1910-1990
(peak-to-peak annual growth rates; per cent)
Portugal Average 9
1910-1934 1.57 1913-1929 1.35
1934-1947 1.15 1929-1939 1.28
1947-1973 5.03 1939-1973 2.67
1973-1990 2.32 1973-1990 2.05
1910-1990 2.77 1913-1990 2.08
‘Average 9’ is based on an unweighted average index for the following European countries:
UK, France, Belgium, the Netherlands, Germany (West Germany to 1991), Italy,
Denmark, Norway and Sweden. Peak years defined in relation to a log-linear time trend.
Sources: Batista et al. (1997), Pinheiro et al. (1997) and Maddison (1995 and 2001).
GRAPH 1 ABOUT HERE

During World War I, both the Portuguese and the European economies were in a
depressive cycle, and income per capita reached a trough at the end of the war. From then
on, the two indices increased, but the European index peaked in 1929, whereas Portugal’s
income growth peaked in 1934 (although there was a blip in the series in 1927). A period of
stagnation followed this peak and it lasted down to the end of World War II. This
periodization contrasts sharply to what one could expect from the stabilization program that
followed the advent of the Estado Novo. Stagnation in Portugal’s income series lasted
throughout the 1930s, but growth resumed shortly after, and another peak in the income per
capita series was reached in 1947.12 Portuguese economic growth was comparatively high
during World War II, whereas the European economy, as represented by our average index
for nine countries, was negatively affected by the war and it hit a trough in 1945. However,
the recovery started earlier in Europe, where economic growth resumed immediately after
1945. In contrast, the Portuguese economy remained virtually stagnant from 1947 to 1950.
From then on, economic growth expanded consistently in the nine more developed
European economies, as well as in Portugal, until a new and coincident peak was reached in
1973. After 1973, there was an inflexion of the index for the average of nine European
11
Namely: the United Kingdom, France, Germany (West Germany to 1991), Belgium, Netherlands Italy,
Sweden, Denmark and Norway.
12
Our income per capita series is a linked index of the series from Batista et al. (1997) and Pinheiro et al. (1997),
which are based on direct evaluations of output. The alternative series from Nunes et al. (1989) is an indirect
estimate, based on the evolution of government revenue and expenditure, and imports, and fluctuates
accordingly. The indirect estimate implies a steep decline of income per capita from 1910 to 1921 and a steep
recovery thereafter. Yet, it also shows a decline in the trend growth rate after 1932. See Lains and Reis (1991).
countries, rather than a period of slowdown, to 1986, followed by rapid growth, to 1998,
which was also the case of Portugal.13
Table 2 shows growth rates according to Maddison (1995)’s phases of economic
development in the twentieth century. We may observe there that Portuguese income per
capita increased at a faster pace than the index for the average of Europe in every phase,
except during 1929-38 and 1973-86. The table also shows growth rates for Spain and
Greece. Within this group of countries, Portugal had a better performance in the interwar
period and that was mainly due to the fact that international and civil wars affected Spain
and Greece. During the second post war period, growth rates in these three countries were
pretty much similar, with Greece expanding at a slightly higher rate. After 1973 the
Portuguese economy faired better. Finally, Table 2 shows data for Ireland, which depicts a
different pattern of growth throughout the twentieth century.
Table 2
Growth of real income per capita in the European periphery, 1913-1998
(Maddison’s phases of development; annual growth rates between 3-years averages;
per cent)
Portugal Spain Greece Ireland Average 9

1913-1929 1.35 1.65 2.45 0.33 1.39
1929-1938 1.28 -3.53 1.50 0.87 1.16
1938-1950 1.56 1.48 -2.72 0.94 1.00
1950-1973 5.47 5.63 5.99 2.98 3.55
1973-1986 1.52 1.31 1.75 2.47 2.01
1986-1998 3.45 2.65 1.39 5.42 1.88
1913-1950 1.40 0.31 0.51 0.66 1.21
1950-1973 5.47 5.63 5.99 2.98 3.55
1973-1998 2.40 1.92 1.59 3.81 1.95
1913-1998 2.79 2.20 2.29 2.19 2.06
Notes and sources: see Table 1
Table 3 reports income convergence rates for the same growth periods (see also Graph
2).14 As shown in that table, Portugal’s rate of convergence, during the period from 1950 to
1973, was quite exceptional, at 1.94 per cent per year. Over the periods from 1913 to 1950
and from 1973 to 1998, the Portuguese economy also converged but only slightly.
13
Portugal had one of the highest variances of income per capita growth within the 15 European Union countries,
in 1965-94. See Pinto Barbosa et al. (1999, pp. 154-9).
14
We consider absolute rates of convergence, which do not take into account differences in growth potential or in
steady state growth rates, which are contemplated by estimates of conditional convergence, as defined by Barro
and Sala-i-Maritn (1995). Aguiar and Figueiredo (1999) show a positive and significant rate of conditional
convergence for the Portuguese economy in the long-run (1870-1990), taking into account the initial income
level and degree of openness of the economy.
18 Pedro Lains
Convergence after the 1973 oil crisis, however, was concentrated in the years between 1986
and 1998. The Portuguese rate of convergence during the 1950-73 was inferior to that of
Spain and Greece, whereas convergence in the period from 1986 to 1998 was higher.
Ireland was as, again, an exceptional case.
GRAPH 2 ABOUT HERE

Following the period of higher growth and convergence, from 1950-73, Portugal’s
income per capita growth entered a new period of slower growth, which lasted to the
present times. However, after 1973, the trend growth rate of the Portuguese economy was
higher than the trend growth before 1950. Graph 3 shows clearly the higher trend growth
after 1973. Portugal fits what Crafts and Mills (1996) termed the ‘reverse Janossy
hypothesis’. In other words, despite the slowdown in Portuguese economic growth after
1973, the rate at which the economic increased after 1973 was higher than the rate of
economic growth before the 1950-73 period. The Janossy hypothesis implies that
reconstruction from war damage had an important role in the high growth levels observed
in the 1950-73 period and that the rates of economic growth would return to the levels
previous to the war. That did not happen in most European economies, including Portugal.
Such a result implies that during the period of high growth there was a ‘greater
accumulation of technological capability’, as well as infrastructures and economic
institutions, which helped the increase in the trend growth for the Portuguese economy.15
Table 3
Convergence of real incomes per capita in the European
periphery, 1913-1998
(Maddison’s phases of development; annual growth rates between 3-years averages;
per cent)
Portugal Spain Greece Ireland

1913-1929 -0.04 0.26 1.04 -1.04
1929-1938 0.12 -4.64 0.33 -0.29
1938-1950 0.55 0.47 -3.69 -0.06
1950-1973 1.85 2.01 2.36 -0.546
1973-1986 -0.49 -0.69 -0.26 0.45
1986-1998 1.54 0.76 -0.48 3.48
1913-1950 0.19 -0.89 -0.69 -0.54
1950-1973 1.85 2.01 2.36 -0.55
1973-1998 0.44 -0.03 -0.36 1.82
1913-1998 0.72 0.14 0.23 0.13
Notes: convergence defined according to:

φ = [(y i / y9) ( t + 1) / (y i / y9) ( t ) ] [ 1 / ( t +1 - t ) ]
15
Crafts and Mills (1996, pp. 416-7).
where y i is income per capita for the 4 countries in the table and y9 is the average for the United
Kingdom, France, Germany (West Germany to 1991), Belgium, the Netherlands Italy, Sweden, Denmark
and Norway.
Source: see Table 1.
GRAPH 3 ABOUT HERE
Portuguese economic growth slowdown after 1973 can be attributed both to domestic
and to external causes. A simple counterfactual static exercise can help us in determining
how much of the slowdown can be attributed to the slowdown of the European economy,
and how much to domestic factors. In fact, if domestic factors had not changed after 1973,
the rate of convergence to the level of the average for the nine European countries can be
assumed as constant, at 1.85 per cent per year. In that case, Portugal’s income per capita in
1998 would amount to 90.6 per cent of the average for the nine more developed countries,
instead of the actual ratio of 64.8 per cent. In other words, if the rate of convergence had
not declined, Portugal’s rate of economic growth during 1973-98 would have been 3.84 per
cent, instead of the actual growth of 2.40 per cent. This simple exercise implies that about
equal parts of the decline in 3.07 percentage points of the rate of growth of the Portuguese
economy, after 1973, from 5.47 per cent to 2.40 per cent per year, can be ascribed to the
decline in Portuguese rate of convergence (i.e. 1.44 p.p. = 3.84 - 2.40 per cent), and to the
decline in the rate of growth of the ‘European’ economy (i.e. 1.63 p.p. = 5.47 - 3.84 per
cent).16
In the appropriate comparative framework, the slowdown of the Portuguese economy
after 1973 appears less spectacular and is at least partially explained. In fact, in the decades
following 1973, the trend rate of growth of the economy was higher than it had been before
the period of rapid growth. Moreover, as shown by a comparison of rates of convergence,
Portugal’s slowdown was in a large part due to the decline in the rate of European
economic growth. Such conclusions imply that domestic factors explain about half of
Portugal’s economic growth slowdown.17
3. Accounting for economic growth
Neo-classical growth theory attributes the sources of output growth to the accumulation
of human and physical capital and to exogenous technological change. The sources of
growth are measured through a production function with constant elasticities. According to
Maddison (1995 and 1996), growth accounting models explains fairly well the catching-up
process of western European income levels (United Kingdom excepted) to that of the USA,
which is the country with the highest average productivity level in the twentieth century.
The model shows that most European economies converged because they had higher
16
A similar conclusion may be reached following Dowrick and Nguyen (1989, Table 6) growth rates adjusted for
the catch-up effect. In fact, the adjusted growth rate for Portugal was 30 per cent of the actual growth in 1950-
60 (1.32 vs. 4.39), 69 per cent in 1960-73 (4.40 vs. 6.41) and 49 per cent in 1973-85 (0.70 vs. 1.42).
17
The domestic component of the variance of the rate of growth of the Portuguese economy declined after 1973,
according to Pinto Barbosa et al. (1999, pp. 157-9), due to the increase in trade and to the better tuning of
macroeconomic policies within the EU. This is in accordance with our findings.
20 Pedro Lains
growth rates of both capital stock and total factor productivity, in 1950-73, than the USA.18
That was also the case of Spain, in the 1965-90 period.19
Table 4 sets down the evidence on the growth of inputs for Portugal. The data shows
that both human and physical capital expanded more rapidly after 1947. In the case of
human capital, which is measured as the average years of schooling of the active
population, it increased by 2.08 per cent per year, in the 1910-34 period, 1.14 per cent, in
1934-47, 2.47 per cent, in 1947-73, and 4.83 per cent, in 1973-90. The rate of growth of
physical capital doubled twice between 1910 and 1973, from 1.25 per cent per year, in
1910-34, to 3.89 per cent, in 1934-47, and 7.73 per cent, in 1947-73. After 1973, the rate of
growth of capital stock declined but it still remained higher than it was before World War
II.
Table 4
Growth of factors and GDP, 1910-1990
(peak-to-peak annual growth rates; per cent)
Labour Human capital Capital GDP
1910-1934 1.00 2.08 1.25 2.17

1934-1947 1.31 1.14 3.89 2.09
1947-1973 0.70 2.47 7.73 5.17
1973-1990 0.05 4.83 5.21 3.92
Notes: ‘Labour’ is total employment estimated as 95% of active population, to 1925, and total number of hours
worked of employed population, thereafter. The weekly number of hours worked declined from 48.5 hours in
1925 to 40 hours, in 1990.
‘Human capital’ is the average years of schooling of active population (according to Barro and Lee, 1993), based
on Census data;
‘Capital’ is the stock of capital based on the growth of gross domestic capital formation (residential capital
excluded).
Sources: Computed from Mateus (1998, Electronic Data Set, File ‘cn_h_sl’, Sheet L). See also Mateus (1998,
Apêndice estatístico).
Table 5 shows the growth accounting for twentieth century Portugal, which is based on
an ‘average production function’20 by Nehru and Dhareshwar (1994). The first striking
conclusion we may draw from the table is that total factor productivity growth did not have
a paramount role in accounting for Portuguese economic growth during the century. The
highest contribution of the growth of factor productivity was in the 1910-34 period,
whereas in 1934-47 it was slightly negative. The comparatively small contribution of
productivity growth in Portugal in the period to 1973 contrasts with what happened in the
rest of Western Europe. Portuguese economic growth was more dependent on capital
18
See Maddison (1995, pp. 40-9). See also Denison (1967), Dowrick and Nguyen (1989), Crafts and Toniolo
(1996) and Crafts (2000).
19
See Prados and Suarez (1996, p. 359).
20
Temple (1999, p. 120).
deepening. That happened particularly in the period from 1934-47, but also in 1947-73 and
1973-90, when capital growth accounted, respectively, for 49.9 and 44.3 per cent of
domestic output growth.21 The contribution of human capital growth was relatively small in
the years to 1973 and it increased to 41.0 per cent during the last period in the table. The
contribution of total factor productivity declined after 1973. This is in accordance to what
happened elsewhere in Europe.22
Table 5
Growth accounting for Portugal: sources of growth and output growth, 1910-1990
Annual growth rates As percent of output growth

Human Human
Labour Capital TFP Output Labour Capital TFP
capital capital
1910-1934 0.33 0.70 0.42 0.72 2.17 15.4 32.1 19.2 33.3
1934-1947 0.44 0.38 1.30 -0.02 2.09 20.8 18.2 62.0 -0.10
1947-1973 0.23 0.82 2.58 1.53 5.17 4.5 15.9 49.9 29.7
1973-1990 0.02 1.61 1.74 0.56 3.93 0.5 41.0 44.3 14.2
Notes: Sources of growth are based on factor growth rates from Table 4 weighted by factor shares of 1/3,
according to Nehru and Dhareshwar (1994). See Mateus (1995b, Tab. 9) and Mateus (1998, Apêndice estatístico).
Sources: see Table 4
Portugal’s growth accounting shows that the country’s growth experience was more
akin to what happened in most of the Asian ‘tigers’, in the second post war period.23 This
form of ‘extensive growth’ was also common to some eastern European countries such as
Czechoslovakia and East Germany and was in opposition to the ‘“intensive growth” model
which was predominant in western Europe during the post-war period.’24 After 1973, the
contribution of capital growth to total output growth declined only slightly and the decline
in the rate of growth of total output can be ascribed mainly to the decline in the contribution
of total factor productivity growth. If Portugal’s total factor productivity growth had not
declined during 1973-90, by 1 percentage point, the rate of total output growth during
would have been 4.90 per cent per year (i.e. 3.93 + 0.97 per cent), which is close to the rate
of the two decades prior to 1973.
Afonso (1999) provides an alternative model where output growth is a log-linear
function of investment per worker, imports of machinery per worker, and exports per
worker. The author also adds as an exogenous variable average TFP growth of the Europe
Union (12 members), in order to capture the convergence effect.25 According to this author,
21
For growth accounting exercises with similar conclusions for Portugal, see César das Neves (1994), Silva
Lopes (1996), Amaral (1998) and Mateus (1998).
22
van Ark and Crafts (1996, pp. 5-6).
23
See Mateus (1995b).
24
van Ark (1996b, p. 298).
25
See Afonso (1999, pp. 59-64).
22 Pedro Lains
in the 1960-73 period, the growth of the Portuguese economy was led by the growth in
capital stock and total factor productivity. The contribution of these two factors of growth
added to 93.4 per cent of output growth. Afonso (1999) concludes further that the growth
of total factor productivity was mainly due to the growth in the labour ‘efficiency’, which
was positively affected by the four exogenous variables in his model. On the other hand,
the contribution of the growth of capital ‘efficiency’ is negligible. For the 1974-85 period,
the author finds out a reduction in the explanatory power of capital deepening and labour
productivity growth, and a negative contribution of capital productivity. For the period after
1986, capital and total factor productivity are again the two main sources of growth,
although at a lower rate of total output growth.26 The findings by Afonso (1999), regarding
the declining contribution of factor productivity growth, after 1973, are in accordance with
the findings based on the growth account exercise presented in Table 5.
Levine and Renelt (1992) propose an alternative model to estimate elasticities of
income growth with respect to of a series of exogenous variables, including initial income
levels, for a sample of 103 countries in the 1960-85 period. Crafts and Toniolo (1996) use
one of Levine and Renelt’s equations in order to ‘consider what new growth theory might
suggest for the speeding up and slowing down of European growth’ in the three
Maddison’s phases of development for the twentieth century.27 The chosen equation is
based on a regression where income per capita growth rates is explained by the initial
income per capita in relation to the US level, the investment ratio, secondary and primary
enrolment ratio, the ratio of government expenditure to GDP, and population growth.28 The
Levine-Renelt model predicts considerably well the growth of the European economies
during 1923-38 and 1950-73, but it underestimates growth for 1973-89. The major
differences between the periods before and after 1973 are the higher negative effect of the
initial income per capita level, which has decreasing importance over the periods, as the
average income of the sample of European countries got closer to the US level. Secondly,
the government expenditure share also had a higher negative effect.29
26
Afonso (1999, pp. 74-82).
27
Crafts and Toniolo (1996, pp. 17-8). The Levine and Renelt (1992) model is an augmented Solow model of the
Mankiw, Romer and Weil (1992) type. See Temple (1999).
28
Levine and Renelt (1992) have also tried with export share growth, but it proved to be statistically non-
significant. See below.
29
See Crafts and Toniolo (1996, Table 1.11).
Table 6
Growth factors according to the Levine-Renelt model: Portugal, 1910-1990
Contribution to growth
Levels
(annual growth rates; per cent)
1910-34 1934-47 1947-73 1973-90 1910-34 1934-47 1947-73 1973-90
Constant -- -- -- -- 2.01 2.01 2.01 2.01
Initial income 0.245 0.302 0.208 0.448 -1.08 -1.33 -0.92 -1.98
Investment share 0.050 0.083 0.209 0.301 0.47 0.77 1.95 2.80
Second. enrol. ratio 0.018 0.057 0.177 0.622 0.02 0.07 0.21 0.75
Primary enrol. ratio 0.582 0.798 1.231 1.373 1.04 1.43 2.20 2.46
Government share 0.110 0.100 0.099 0.135 -0.70 -0.64 -0.63 -0.86
Population growth 0.803 1.089 0.201 0.770 0.06 0.09 0.02 0.06
Forecast growth -- -- -- -- 1.82 2.40 4.84 5.25
Actual growth -- -- -- -- 1.57 1.15 5.03 2.32
Notes: the contribution of each factor is taken from the parameters of the following equation (see Levine and
Renelt (1992, Tab. 5, col. ii)
gyp = 2.01 – 0.69* inyp + 9.31* inv + 1.21 sec + 1.79* pri - 6.37* gov + 0.08 gpo
( 0.83) (0.12) (2.08 ) (1.17) (0.58) (2.03) (0.18)
N = 103; R2 = 0.68; * = statistical significant at the 0.05 level
Sources: ‘Initial income level’ (inyp): relative income level Portugal/USA in the beginning of each period from
Summers and Heston (1988). The value of the observation is the ratio in the table multiplied by the United States
1950 GDP per capita level (i.e. $US 6,401).
‘Investment’ and ‘Government shares’ (inv and gov): GDP shares in current market prices. From Batista et al.
(1997, pp. 93-4), to 1952; and from Pinheiro et al. (1997, pp. 197-200), for 1953-90. ‘Primary enrolment ratio’
(pri): Ratio of enrolment in public (to 1940) and private (from 1941) primary schools as percent of 5-9 (to 1940)
and 6-9 (from 1941) age groups. ‘Secondary enrolment ratio’ (sec): Ratio of enrolment in general public, general
private (from 1917), professional public (from 1929) and professional private (from 1941) secondary schools as
per cent of 10-19 (to 1940) and 10 to 17 (from 1941) age groups. Sources: Luciano, 2001 (to 1940) and Teixeira
(1999, pp. 147-9) (from 1941) and Valério (2001, p. 55). All shares and ratios are averages for the periods
indicated.
‘Growth of income per capita’ and ‘Population growth’ (gyp and gpo): same as Table 1.
See also Crafts and Toniolo (1996, p. 18).
The Levine-Renelt model can be used in order to determine to what extent Portugal’s
growth performance was in accordance to a world ‘norm’. The results for Portugal are
shown in Table 6. We conclude from the data in that table that the Levine-Renelt model is a
relatively good predictor for Portugal’s income per capita growth in 1910-34 and 1947-73,
although in the second period it slightly underestimates the actual growth rate (i.e. 4.84 vs.
5.03 per cent). However, the model does not account for the slowing of Portugal’s
economic growth during the interwar period and in the period after 1973. In fact, despite
the reduction in the initial income gap and the increase in the government share, which
have negative coefficients, the model predicts an increase in the annual growth rate of the
Portuguese economy, from 4.84 to 5.25 per cent, whereas actual growth declined from 5.03
to 2.32 per cent. The reason for the best performance predicted by the model is that
Portugal’s investment and school enrolment ratios remained considerably high after 1973.
24 Pedro Lains
Clearly, in comparison to the European experience, Portugal’s initial income per capita
had a lower negative effect, which is due to the fact that the gap was larger in Portugal in
relation to the US. The investment effect in Portugal was comparable to that of Europe,
whereas the sum of the human capital effects was lower in Portugal, as was the government
share effect. The estimates for 1973-90 depict Portugal as an outlier, as the forecast income
per capita growth rate is 5.25 per cent per, whereas the actual growth was only 2.32 per
cent. This indicates that the observed reduction in Portugal’s income growth after 1973
cannot be attributed to the performance of neither the investment ratio, neither the
investment in human capital as measured by the school enrolment ratios. It is important to
note that the growth predicted by the Levine and Renelt model for Portugal during 1973-90
is close to the growth rate of the Portuguese economy, if the 1947-73 rate of convergence
was maintained in the post-1973 period, which was pointed out previously (i.e. 4.89 per
cent).
The coefficient for exports is not statistically different from zero in any of the equations
of the Levine and Renelt model. This result seems to contradict the generally held
assumption that foreign trade is a major factor of growth in small open economies, but is in
accordance with further evidence on Portugal.30 The openness to trade of the Portuguese
economy, which was quite considerable, throughout the second half of the twentieth
century, occurred in two phases. The first phase followed membership of EFTA and the
ratio of foreign trade to GDP increased from about 17 per cent to 30 per cent, from 1960 to
1973. In the next period to 1986, the ratio remained constant. The second phase followed
the entering of the EEC and, from 1986 to 1994, the ratio increased from 30 per cent to
close to 55 per cent.31 In 1994, Portugal ranked as the fourth most open economy in the
European Union.32 The fact that the increase in foreign trade was more rapid in the 1986-94
period which had slower growth than in the period 1950-73, is indicative of the small
explanatory effect of trade in Portuguese growth. In fact, according to Marques Mendes
(1993), the effect of European integration in Portugal’s economic growth was relatively
small. He estimates that the participation in EFTA and the 1972 trade agreement with EEC
explains between 2 and 2.5 per cent of Portugal’s growth of per capita income; and that the
gains from joining the European Union accounted for 10.1 per cent of the income per capita
growth.33
Both the growth accounting framework and the Afonso (1999) and Levine and Renelt
(1992) models point to the conclusion that trends in investment ratios in respect to both
physical and human capital explain the increase in growth rates after World War II, but
they fail to explain the slowdown after 1973. In fact, there were significant investments in
human and physical capital in the 1950-73 period, in Portugal, which either increased in the
30
Empirical tests on the export-led growth model have generally refuted a direct causality link between exports
and growth. According to Levine and Renelt (1992) findings, trade and growth are linked through investment.
Pereira and Xu (2000) find out that the link is through investment and employment. See also Pessoa (1998) on
the negligible effect of openness on Portuguese economic growth throughout 1960-90.
31
The ratio is defined as the average of exports and imports over GDP, from Silva Lopes (1996, Graph 4.1).
32
See Pinto Barbosa et al. (1999, p. 149).
33
Marques Mendes (1993, pp. 16-21). The European Union effect is measured through the impact of structural
funds alone (see also Gaspar and Leite 1995). Aguiar and Figueiredo (1999) have concluded that Portugal’s
level of foreign trade ratio, relative to the average of seven more developed European countries, affected
positively and significantly the rate of convergence of the Portuguese economy, over the long-run (1870-
1990). Yet, they do not provide estimates for shorter periods.
following period or only marginally declined. Yet, after 1973 there was a sharp decline in
the overall rate of growth of the Portuguese economy. As mentioned in the previous
section, only half of the decline in Portugal’s income growth rate after 1973 can be
attributed to the decline in the rate of growth of the European economy. We thus need to
explain further domestic factors of growth. .
4. Structural change and economic growth slowdown
The single most important factor in Portuguese economic growth slowdown after 1973
was the decline in the rate of growth of total factor productivity. In this section we analyse
the extent to which the productivity fall is related to changes in the structure of output.
Table 7 shows the performance of total productivity in the three sectors of the Portuguese
economy during the two development phases, before and after 1973. As shown there, the
growth of total factor productivity fell from 2.64 per cent per year, in 1952-73, to 0.31 per
cent, in 1973-91. The performance of factor productivity in the industrial and the services
sectors was the main cause behind the fall in the growth of total factor productivity, as
productivity growth in the agricultural sector increased slightly after 1973.34 Taking into
account the behaviour of total factor productivity in the three sectors of the economy, we
may conclude that the decline in the rate of growth of the Portuguese economy after 1973
was due to the decline in the performance of the industrial and the services sector.35
Table 7
Sectoral sources of growth and output growth, 1952-91
(annual growth rates; per cent)
Sources of growth Output
Labour Capital TFP growth
GDP
1952-73 0.37 2.80 2.64 5.80
1974-91 0.86 1.60 0.31 2.76
Agriculture
1954-73 -0.33 0.94 0.47 1.09
1974-87 -1.14 0.66 2.23 1.76
Industry
1954-73 1.02 4.37 3.00 8.38
1974-87 0.72 1.99 -0.90 1.81
Services
1954-73 0.80 2.79 2.51 6.10
1974-87 2.10 0.96 0.10 3.16
Source: Neves (1994, pp. 72-73). The estimates in this table are not fully comparable to those of Table 5
above, most of all because they do not take into account the role of human capital.
34
Agricultural labour productivity growth during 1973-90 is associated with the decline in agricultural labour
force, by –2.8 per cent per year. Both the decline in labour force and the increase in labour productivity were
below rates for most western European countries. See Lains (1994, p. 939) and van Ark (1996a).
26 Pedro Lains
In order to explain the fall in industrial factor productivity growth, we need to take into
account the major distinctive features of growth in the periods before and after 1973.36 The
high levels of industrial growth during the period from 1960-73 were due to the expansion
of external demand, induced by European growth and to Portugal’s participation in EFTA,
as well as to the overall favourable performance of the economy and, in particular, of
domestic demand. According to Silva Lopes (1996), the growth of the industrial sector
output in the period to 1973 was enhanced by growth inducing government policies,
including the protection from foreign competition granted to some branches of industry,
fiscal incentives, public investment in social overhead capital and in key capital intensive
industrial sectors, as well as wage and price controls and low interest rates.37 The joining of
the EFTA meant the opening up to external competition, but the Portuguese government
managed to negotiate gradual and selected reductions in tariff and other forms of domestic
protection, whereas Portuguese industrial exports took advantage of the opening up of
foreign markets.38
After 1973, there was a shift of the industrial sector towards the increase in the output
of some more labour intensive sectors. In fact, the contribution of the foodstuffs and
textiles sectors in the industrial sector increased, from 1950-73 to 1973-80, and even more
to 1980-90. On the contrary, the contribution of more capital intensive sectors to total
output growth declined. Foodstuffs and textiles accounted for half of total industrial
growth, in 1973-80, and over 2/3, in 1980-90. The more capital-intensive sectors accounted
for 42 per cent of industrial growth in 1973-80 and 30.7 per cent in 1980-90. It is worth
noting that the performance of total factor productivity in the industrial sector was more
favourable in the 1973-80 period than it was in 1980-90.39 This indicates that the growth of
the foodstuffs and textile sectors had a negative impact on industrial productivity growth,
as compared to the impact of the more capital-intensive sectors.40
According to Peres Lopes (1994), Portugal’s industrial labour productivity in relation to
the United Kingdom’s was comparatively higher in the ‘traditional’ sectors, namely,
textiles, wearing apparel, leather and footwear, wood products, paper and electrical
appliances. Moreover, relative productivity levels in these sectors declined over 1977-90,
whereas it increased in the remaining industrial sectors.
To fully understand the causes of the fall of factor productivity growth in the industrial
sector, after 1973, we have thus to explain what caused the observed shifts in the structure
of the output. One of the main reasons for that shift was the increasing openness of the
industrial sector to foreign trade. According to Pinto Barbosa et al. (1999), in 1993, the
structure of Portugal’s industrial sector was in tune with the structure of revealed
comparative advantages. Yet, according to the same authors, labour productivity in the
35
The decline in total factor productivity in the services sector after 1973 was due to the decline in labour
productivity, which is related to the sharp increase in the rate of growth of employment in the sector (see Lains
1994). That increase is related, on the one hand, to the incorporation of immigrant workers from the colonies,
and to the post-1975 employment protection legislation.
36
Lains (1994).
37
See also Centeno (1995) and Mateus (1998, pp. 194-8).
38
Silva Lopes (1996, pp. 87-8). See also Macedo et al. (1988), Lains (1994), Amaral (1998) and Confraria
(1999).
39
Neves (1994, pp. 72-73).
40
Lains (1994, p. 945).
industrial sectors with better export performance and higher growth rates of output, was
below the average of the sector.41 The fact that the Portuguese industrial sector adapted to
the changes in the structure of demand imposed by the increasing integration with the
European Union economy, led to an increase of the weight of industrial sectors with higher
levels of comparative advantage. However, that was achieved at lower levels of labour and
capital productivity. Such structural transformation led to a reduction in the average
productivity level of the Portuguese industrial sector.
5. Conclusions
Economic historians have shown in the last decades that there is a wide diversity of
economic growth experiences, across countries. Different experiences stem from the fact
that the conditions for economic growth can be very different from country to country.
Such conclusion has a corollary that has not been sufficiently exploited in long-term
economic growth studies, which is that conditions for economic growth also vary
substantially across time. The Portuguese experience of growth during the twentieth
century is a good example of that.
Many authors have recognized that the external conditions for Portuguese economic
growth changed throughout the century. Those changes are associated with transformations
in the international and, in particular, in the European economy, which can be described as
Maddison’s phases of development. Yet, less attention has been paid to the fact that the
domestic conditions for the growth of the Portuguese economy have also changed
substantially. Although the consequences of changes in industrial, monetary and fiscal
policies, as well as overall political conditions, have been largely accounted for, less
attention has been paid to the possibility that economic conditions for growth have also
changed in a substantial way.
The Portuguese economy went through an intensive process of structural transformation
during the 1950-73 period. Whereas in 1950, Portugal was largely an agrarian economy,
with about 50 per cent of the population employed in the agricultural sector, by 1973 the
growth of the industrial and the services sectors had substantially transformed the
economy. The sheer shift of labour from agriculture to the other sectors of the economy
was a source of growth, as the labour productivity in agriculture was about half that of
industry and services. This paper has put together evidence showing that high levels of
investment in human and physical capital were the main instruments for rapid economic
growth to 1973. The fact that Europe was expanding rapidly was also of major help. After
1973, though, the overall conditions for growth were less favourable, as most of the gains
from structural change had already been reaped. The fact that the European economy
slowed down imposed further restrictions on rapid growth. However, after two decades of
growth, the Portuguese economy was substantially different. Although the potential effect
of structural change on economic growth had died out, the fact was that the various forms
of capital and capacities acquired in the period of rapid growth pushed the economy into a
stage of higher trend growth rates. During this latter stage, the Portuguese economy is
41
See Pinto Barbosa et al. (1999, pp. 282-4). The authors do not infer a causal relationship between export ratios
and labour productivity. See also Crespo Faustino (1995) and Pita Barros (1997).
28 Pedro Lains
presently in a process of adaptation to the average structure of the European economy. That
is a slow but probably sustained process.
References
Afonso, O. (1999), Contributo do comércio externo para o crescimento económico
português, 1960-1993. Lisbon, Conselho Económico e Social.
Aguiar, A. and O. Figueiredo (1999). ‘Abertura e convergência da economia portuguesa,
1870-1990’. Estudos de Economia 19, pp. 209-32.
Amaral, L. (1998). ‘Convergência e crescimento económico em Portugal no pós-guerra’.
Análise Social 33, pp. 741-76.
Amaral, L. (2001). ‘Education and human capital accumulation in Portugal, 1930-1973’
(Lisbon, mimeo).
Barbosa, A. Pinto et al. (1999), O impacto do euro na economia portuguesa. Lisbon,
Publicações Dom Quixote.
Barro, R. and J. W. Lee (1993). ‘International comparisons of educational attainment’.
Journal of Monetary Economics 32, pp. 363-94.
Barro, R. and X. Sala-i-Maritn (1995). Economic growth. New York, McGraw-Hill.
Barros, P. Pita (1997), ‘Crescimento económico’. In P. Pita Barros and M. St. Aubyn.
Problemas actuais da política económica portuguesa: crescimento, desemprego,
participação na União Económica e Monetária. Lisboa, Conselho Económico e
Social, pp. 9-53.
Batista, D., C. Martins, M. Pinheiro and J. Reis (1997). ‘New estimates for Portugal’s
GDP, 1910-1958’. História Económica 7.
Centeno, M. (1995). ‘Poupança pública versus poupança privada: que relação de longo
prazo?’. Boletim Económico 1 (1), pp. 67-72.
Confraria, J. (1999). ‘Portugal: industrialization and backwardness’. In J. Foreman-Peck
and G. Federico (eds.). European industrial policy. The twentieth-century
experience. Oxford, Oxford University Press, pp. 268-94.
Crafts, N. (2000), ‘Development History’. Working Papers no. 54 (London School of
Economics).
Crafts, N. and T. C. Mills (1996). ‘Europe’s Golden Age: an econometric investigation of
changing trend rates of growth’. In B. van Ark and N. Crafts (eds.). Quantitative
aspects of post-war European economic growth. Cambridge, Cambridge
University Press, pp. 415-31.
Crafts, N. and G. Toniolo (eds.) (1996). Economic growth in Europe since 1945.
Cambridge, Cambridge University Press.
Crafts, N. and G. Toniolo (1996). ‘Post-war growth: an overview’. In N. Crafts and G.
Toniolo (eds.). Economic growth in Europe since 1945. Cambridge, Cambridge
University Press, pp. 1-37.
Denison, E. E. (1967). Why growth rates differ: postwar experience in nine Western
countries. Washington (D.C.), The Brookings Institution.
Dowrick, S. and D-T. Nguyen (1989). ‘OECD comparative economic growth, 1950-1985:
catch-up and convergence’. American Economic Review 79, pp. 1010-30.
Faustino, H. Crespo (1995). ‘O cluster europeu de Portugal em termos de comércio intra-
sectorial: análise ao nível dos principais produtos para o período 1983-1992’.
Estudos de Economia 15, pp. 391-426.
Gaspar, V. and A. Nogueira Leite (1995). ‘Coesão e convergência: efeitos económicos dos
fundos estruturais’. Boletim Económico 1 (3), pp. 49-55.
Lains, P. (1994). ‘O Estado e a industrialização em Portugal, 1945-1990’. Análise Social
29, pp. 923-58.
Lains, P. (1998). ‘An account of the Portuguese African empire, 1885-1975’. In P. O’Brien
and L. Prados (eds.). The costs and benefits of European imperialism (1415-1974).
Revista de Historia Económica 16, pp. 235-63.
Lains, P. (2003). ‘New wine in old bottles. Output and productivity trends in Portuguese
agriculture, 1850-1950. European Review of Economic History 7 (forthcoming).
Lains, P. and J. Reis (1991). ‘Portuguese economic growth, 1833-1985: some doubts’.
Journal of European Economic History 20, pp. 441-53.
Leitão, N. Andresen (2001). ‘Portugal’s European integration policy, 1947-1972’. Journal
of European Integration History 7, pp. 25-35.
Levine, R. e D. Renelt (1992). ‘A sensitivity analysis of cross-country growth regressions’.
American Economic Review 82, pp. 942-63.
Lobo, M. Costa (2000). ‘Portugal na Europa: uma leitura política da convergência
económica’. In A. Barreto (ed.). A situação social em Portugal, 1960-1999.
Lisbon, Imprensa de Ciências Sociais, pp. 611-43.
Lopes, L. Peres (1994). ‘Manufacturing productivity in Portugal in a comparative
perspective’. Notas Económicas 4, pp. 57-76.
Lopes, J. da Silva (1996). A economia portuguesa desde 1960. Lisbon, Gradiva.
Macedo, J. B. de, C. Corado e M. L. Porto (1988). ‘The timing and sequencing of trade
liberalization policies: Portugal, 1948-1986’. Working Papers no. 114 (Faculdade
de Economia, Universidade Nova de Lisboa).
Maddison, A. (1995). Monitoring the world economy, 1820-1992. Paris, OECD.
Maddison, A. (1996). ‘Macroeconomic accounts for European countries’. In B. van Ark
and N. Crafts (eds.). Quantitative aspects of post-war European economic growth.
Cambridge, Cambridge University Press, pp. 27-83.
Maddison, A. (2001). The world economy. A millennial perspective. Paris, OECD.
Mankiw, N., D. Romer and D. N. Weil (1992), ‘A contribution to the empirics of economic
growth’. Quarterly Journal of Economics 107, pp. 407-37.
Marques, A. (1988). Política económica e desenvolvimento em Portugal (1926-1959).
Lisbon, Livros Horizonte.
Mateus, A. (1995a). ‘Os factores de crescimento da economia portuguesa: capital e
integração’. In Proceedings of the Conference Para onde vai a economia
portuguesa?. Lisbon (ISEG).
Mateus, A. (1995b). ‘O sucesso dos “Tigres asiáticos”. Que lições para Portugal?’,
Working Papers no. 254 (Faculdade de Economia, Universidade Nova de Lisboa).
Mateus, A. (1998). Economia portuguesa. Crescimento no contexto internacional (1910-
1998). Lisbon, Verbo.
Mendes, A. J. Marques (1993). ‘The development of the Portuguese economy in the
context of the EC’. In J. Silva Lopes (ed.). Portugal and EC membership
evaluated. London, Pinter Publishers, pp. 7-29.
Moura, F. Pereira de (1973). Por onde vai a economia portuguesa? Lisboa, Seara Nova.
Nehru, V. and A. Dhareshwar (1994). ‘New estimates of total factor productivity growth
for developing and industrial countries’. Policy Research Working Papers no.
1313 (World Bank).
30 Pedro Lains
Neves, J. César das (1994). The Portuguese economy: a picture in figures. XIX and XX
centuries. Lisbon, Universidade Católica Editora.
Neves, J. César das (1996). ‘Portuguese postwar growth: a global approach’. In N. Crafts
and G. Toniolo (eds.). Economic growth in Europe since 1945. Cambridge,
Cambridge University Press, pp. 329-54.
Nunes, A., E. Mata e N. Valério (1989). ‘Portuguese economic growth, 1833-1985’.
Journal of European Economic History 18, pp. 291-330.
Ó Gráda, C. and K. O’Rourke (1996). ‘Irish economic growth, 1945-1988’. In N. Crafts
and G. Toniolo (eds.). Economic growth in Europe since 1945. Cambridge,
Pereira, A. M. and Z. Xu (2000). ‘Export growth and domestic performance’. Review of
International Economics 8, pp. 60-73.
Pessoa, A. (1998), ‘Catch-up tecnológico, investimento e convergência real: um exercício
de contabilidade do crescimento aplicado à economia portuguesa’. Working
Papers no. 83 (Faculdade de Economia, Universidade do Porto).
Pinheiro, M. et al. (1997). Séries longas para a economia portuguesa. Pós II Guerra
Mundial vol. I - Séries estatísticas. Lisbon, Banco de Portugal.
Prados de la Escosura, L. and J. G. Sanz (1996). ‘Growth and macroeconomic performance
in Spain, 1939-1993’. In N. Crafts and G. Toniolo (eds.). Economic growth in
Europe since 1945. Cambridge, Cambridge University Press, pp. 355-87.
Rosas, F. (2000). Salazarismo e fomento económico, 1928-1948. O primado do político na
história económica do Estado Novo. Lisbon, Editorial Notícias.
Summers, R. and A. Heston (1988). ‘A new set of international comparisons of real product
and price levels estimates for 130 countries, 1950-1985’. Review of Income and
Wealth 34, pp. 1-25.
Teixeira, A. (1999). Capital humano e capacidade de inovação. Contributos para o estudo
do crescimento económico português, 1960-1991. Lisbon, Conselho Económico e
Social.
Temple, J. (1999). ‘The new growth evidence’. Journal of Economic Literature 37, pp.
112-56.
Tortella, G. (1994). ‘Patterns of economic retardation and recovery in south-western
Europe in the nineteenth and twentieth centuries’. Economic History Review 47,
pp 1-21.
Valério, N. (ed.) (2001). Estatísticas históricas portuguesas. Lisbon, Instituto Nacional de
Estatística.
van Ark, B. (1996a). ‘Sectoral growth accounting and structural change in post-war
Europe’. In B. van Ark and N. Crafts (eds.). Quantitative aspects of post-war
European economic growth. Cambridge, Cambridge University Press, pp. 84-164.
van Ark, B. (1996b). ‘Convergence and divergence in the European periphery: productivity
in eastern and southern Europe in retrospect’. In B. van Ark and N. Crafts (eds.).
Quantitative aspects of post-war European economic growth. Cambridge,
van Ark, B. and N. Crafts (1996). ‘Catch-up, convergence and the sources of post-war
European growth: introduction and overview’. In B. van Ark and N. Crafts (eds.).
Quantitative aspects of post-war European economic growth. Cambridge,
van Ark, B. and N. Crafts (eds.) (1996). Quantitative aspects of post-war European
economic growth. Cambridge, Cambridge University Press.
Graph1 - Growth of income per capita: Portugal and Average 9, 1910-98

(1990 GK dollars; semi-logarithmic scale)
10.0
Portugal
9.5
Average 9
9.0
8.5
8.0
7.5
7.0
1910 1915 1920 1925 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995
Sources and notes: see Table 1.
Graph 2 -Portuguese income per capita as per cent of Average 9, 1910-1998

(1990 G-K dollars)
70
65
60
55
50
45
40
35
30
1910 1915 1920 1925 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995
Sources and notes: see Table 1.

32 Pedro Lains
Graph 3 - Segmented trend for Portugal’s income per capita growth, 1910-1998
9 .5
9 .0
8 .5
8 .0
7 .5
7 .0
10 20 30 40 50 60 70 80 90
Sources: See Table 1.

Note: The segmented trend is estimated according to Crafts and
Mills (1996, pp. 418-9) with break years in 1919, 1939,1950 and 1973.
PORTUGUESE ECONOMIC GROWTH RE-EXAMINED:
AN ANTI-FADO MANIFESTO *
Miguel Lebre de Freitas

Universidade de Aveiro
This paper provides an empirical assessment of the Portuguese economic growth,

both in a historical and in a comparative perspective. We first test whether
productivity in Portugal has been converging to the EU average, using the stochastic
approach to convergence. We then stress that convergence is not an automatic
process. To this aim we show that there is no systematic tendency for poor countries
to grow faster than rich countries, even within OECD. We also investigate the extent
to which the information content of some policy induced variables helps to explain the
variance of growth rates across OECD economies. We find positive correlation
between economic growth and measures of institutional development, labour skills
and labour market flexibility.
* Prepared for the conference "Desenvolvimento Económico Português no Espaço Europeu: Determinantes e
Políticas”, promoted by Banco de Portugal. We thank the conference Scientific Committee for helpful
comments. Financial support from Banco de Portugal is also acknowledge.
34 Miguel Lebre de Freitas
There are many countries, not essentially different either in the degree of security
which they afford to property, or in the moral and religious instruction received by the
people, which yet, with nearly equal natural capabilities, make a very different progress in
wealth.
THOMAS MALTHUS (1766-1834)1
I. Introduction
In the last forty years, the Portuguese economy exhibited an outstanding growth
performance. The episode of fast economic growth allowed the country to reduce
consistently its income gap vis-à-vis the industrialised world. Some may argue that such
achievement is not impressive. Since Portugal departed from a low standpoint, higher
growth rates would be expected anyway. This reasoning has a long tradition in economic
thinking. Following David Hume (1758), economists have been arguing that transfer of
technology and decreasing returns provide poor economies with an impetus to “catch up”.
A fact that has received large consensus in the economic profession, however, is that there
is no systematic tendency for poor countries to grow faster than rich countries. Moreover, if
any important lesson came about with the recent developments in the theory of economic
growth, this is that "policy matters". Economic growth is nor automatic nor immune to
government actions and in no way can be extrapolated from the past.
In line with the thematic of the conference, this paper first addresses the question as to
whether Portugal has been converging to the EU average. We find that productivity in
Portugal has indeed converged, but the pace of convergence has not been uniform along
time. By showing that there is no systematic tendency for poor countries to grow faster than
rich countries, even within OECD, we stress the Thomas Malthus claim that similar
economies may exhibit distinct growth performances. We then investigate the extent to
which the information content of some policy induced variables may help to predict
productivity growth in OECD nations. The aim of the exercise is not to develop any new
theory of economic growth. Simply, by focusing on the role of policy, we stress the case
against hazard and we point out some directions for reform. Following the topics covered
by the conference, we focus on institutions, labour market flexibility, government finance
and the quality of human resources. Physical capital accumulation and labour participation
are excluded from the analysis so as to capture the overall effect of the policy, regardless as
to whether it acts through factor accumulation or efficiency change. Restricting the analysis
to OECD economies, we gain in two dimensions. First, we stress the case against the
believe that convergence is kind of automatic in particular clubs. Second, we reduce
substantially the complexity of the analysis, as one need not to discriminate the role of
those attributes that are equally hated by industrialised economies.
1 Thomas Malthus (1817). The quotation is borrowed from Gylfason (1999).

Portuguese Economic Growth Re-examined: an Anti-fado Manifesto 35
The paper proceeds as following. In section 2, we present some comparative data and
we discuss alternative measures of convergence. In section 3, we test whether Portuguese
productivity has been converging to the EU average and the extent to which the pace of
convergence has been affected by major regime shifts. In section 4, we test the simple
convergence hypothesis in the OECD sample. In section 5, we run some regressions so as
to investigate the extent to which some policy induced variables help to discriminate the
different growth performances among OECD economies. Section 6 concludes.
II. Growth accounting
After a secular trend of divergence, the Portuguese economy appears to have engaged in
a convergence track towards the industrialised world2. With real per capita GDP (expressed
in comparable units) growing at an average rate of 4% per annum in the period 1960-1992,
Portugal is ranked 11th in a sample of 119 countries (see Table 1). Using data from the
European Commission, Portugal is ranked third in per capita GDP growth and first in
growth of GDP per worker, among 22 OECD economies in the period 1960-2000. In the
1980-2000 sub-period, Portugal is ranked third in terms of per capita GDP growth and
second in growth of GDP per worker.
This fast growth experience allowed standards of living in Portugal to approach those in
more advanced nations. Column I of Table 2 shows that per capita GDP, in current PPS3,
rose from 41.5% of the EU average in 1960 to 73.8% in 2000. This evolution has not been
uniform along time. As shown in the table (see also Figure 1), the relative income gap has
widened during the oil shocks (1973-85)4. The same happened in Greece and Spain, but not
in Ireland5.
GDP per capita (PCGDP) provides an appropriate measure to evaluate whether
standards of living are approaching. But it should be taken into account that this is a rather
complex variable, as it is influenced by different factors, such as demographic changes,
labour participation and productivity. To abstract from demographic changes, which are
mainly exogenous to policy (at least in medium term), one may look instead at GDP per
2 De Long (1988), points to a fall in per capita GDP from 61% of the US level in 1870 to 29% in 1913. Neves
(1994: 105) examines the series of per capita GDP in Portugal relative to 16 industrialised countries, along
1830-1985. Visual inspection of the data suggests a secular trend of divergence along the nineteen century
(briefly interrupted in 1860-1880) and up to the first decade of last century. A short convergence episode took
place in the 1920s, followed by two more decades of divergence. Although a slight recovery may be identified
in the 1950s, only in the decade after did the Portuguese economy meet fast economic growth. In 1985, relative
GDP was already at the same level as that of 1830.
3 PPS (Purchasing Power Standard) is a measure of purchasing power parity produced by the Eurostat, according
to the methodology established by the UN International Comparison Project.
4 Visual inspection of Figures 1 and 2 suggests that the series of relative PCGDP inherits some persistence from
the population series. In the case of Portugal, population movements reflect the massive emigration that took
place during the ultramar war, the repatriation of Portuguese citizens from the ex-colonies in the aftermath of
the revolution and also statistical smoothness, because population data are interpolated from census.
5 Periodicity in Table 2 was chosen so as to allow the reader to check the figures at the time of main regime
changes. Since data are not filtered, strict comparisons across sub-periods are misleading.
working age person (labelled GDPWAP in column II of Table 2)6. This variable measures
what a society gets out of its pool of human resources, irrespectively as to whether people
with working age are employed, unemployed or even out of the labour force. Hence,
policies raising economic efficiency, education and capital accumulation will impact on this
variable, regardless as to whether the channel is labour productivity or incentives to work.
Column III, displays data on GDP per worker (GDPPW), relative to the EU average. As
shown in the table, GDPPW in Portugal has been approaching consistently the EU average
(see also Figure 1), contrasting with less obvious paths in Spain and Greece. This measure
may be informative to evaluate technology and the quality of the inputs being used7, but tell
us nothing about unemployment and participation, which are, in large extent, endogenous to
policy. For example, while relative GDPPW evolved at similar rates in Portugal and Ireland
along 1986-2000 (1.6% per annum), growth rates of relative GDPWAP differ substantially
(2.8% in Ireland and 1.1% in Portugal). The difference is accounted for by "employment
deepening", which, in the case of Ireland, was an important outcome of a major policy shift,
after decades of high unemployment8.
Columns IV and V of Table 2 display information on relative endowments of physical
and human capital. In the case of Portugal, observation of columns III and IV suggests that
the sharp rise in capital per worker since 1973 was the return to a balanced growth path,
following a period of insufficient capital accumulation. But measurement errors, specially
in the beginning of the sample, call for caution in the interpretation of this data9.
Educational attainment in column V is measured by the average years of schooling,
relative to four European countries. After a fast recovery from very low levels in the 1960s
and the 1970s, education attainment in Portugal turns out to have a disappointing evolution
in the 1980s (see also Figure 3). Complementary information in Figure 4 also raises some
concerns about the quality of the Portuguese educational system.
A last comment on Table 2 is motivated by the Greek interesting picture. This country
is apparently better endowed in terms of physical and human capital than both Portugal and
Spain. This advantage has not translated, however, into higher productivity. The figures
6 Ireland provides an interesting case to distinguish PCGDP and GDPWAP. Because of a baby boom in the
1980s, the ratio of “working age to total population” (relative to EU) rose at an average rate at 0.9% per annum
in the period 1986-2000. This purely demographic effect translates into a faster growth of relative PCGDP
(3.7% a year) than relative GDPWAP (2.8%). In the case of Portugal, the difference between growth rates in
columns I and II accounts for both a demographic effect (0.4% per annum) and also a 0.5% real appreciation
effect (note that GDP per capita in column I is measured at current prices).
7 With some caution, of course, as it measures production "per employee", rather than "per hour worked". The
European Commission computes labour productivity measuring employment in terms of full time equivalents,
but the data available are not fully comparable across countries.
8 The existence of a low standpoint has been often ignored in the assessment of the recent Irish growth
performance. A similar path would have been difficult to achieve in Portugal, because in this country
unemployment has remained low and the participation rate has been traditionally high (see Freitas, 2000, for a
discussion).
9 The European Commission follows the standard procedure of cumulating net investment flows at constant
prices to estimate the capital stock. The method involves, however, the specification of the capital stock in a
base year. The European Commission postulates a capital-income ratio equal to 3 for all countries in 1960.
This may led to a significant bias, specially in the beginning of the sample. For recent years, data are more
reliable, because they are less dependent on the initial assumption.
suggest that Greece is now facing a similar situation to that of Ireland in the mid-eighties
and accords to the view that factor accumulation is not sufficient for economic growth.
III. Testing Portugal-EU convergence
The observation that growth rates in Portugal have been higher than in EU for a
particular period of time does not provide evidence of convergence. If productivity shocks
are independent, there will be no tendency for productivity levels to converge in the long
run or even to follow parallel paths. In that case, the two economies will be drifting apart,
even though in a particular period of time the respective time series looked like being
approaching each other. For convergence to take place, productivity in Portugal shall
evolve each moment in time so as to approach an equilibrium differential vis-à-vis the EU
level. Hence, an obvious way to assess Portugal-EU convergence is to investigate the
persistence of shocks to relative productivity: if income disparities follow a stationary
process, the null of non-convergence is rejected10.
In this section, Portugal-EU convergence is assessed using Dickey Fuller tests. A drift
and a time trend are included, so as to allow for different balanced growth paths and
transition dynamics11. Structural breaks are also allowed for, in order to capture eventual
changes in the convergence path. This is done using the “crash and trend” model, with the
following (a priori defined) structural breaks: the oil shock cum revolution (1974), EC
accession (1986) and the Maastricht discipline (1993).
Columns I and V of Table 3 display the results of simple ADF tests on the series of
relative GDPWAP and relative GDPPW12 13. In both cases, the null of no-convergence is
not rejected. Estimates with structural breaks are in columns II-IV and VI-VIII. Results in
columns II and VI suggest that the oil shock had an asymmetric impact, hitting the
Portuguese economy harder than the EU region as a whole. When such break is allowed
for, the null of no-convergence is rejected. The pace of convergence declined from 3.0%
before the shock to 0.8% after the shock, when measured by GDPWAP and from 2.1% to
1.3% when measured by GDPPW (details on the de-trending method in the table).
Specifications with two breaks (columns III-IV and VII-VIII) do not in general allow for
rejection, except in column VIII. This points to a positive effect of the Maastricht
discipline, but the robustness of this result has still to be proven, when more data become
available.
10 See, for example, Carlino and Mills (1993). The implied assumption is that productivity shocks have a uniform
long run impact across countries. Bernard and Durlauf (1995) also examined a less restrictive notion of
convergence, requiring only the existence of common stochastic trends. St. Aubyn (1999), allowed for a
smooth transition from non (stochastic) convergence to convergence, using a Kalman filter.
11 Barros and Garoupa (1996) also used Dickey-Fuller tests to investigate Portugal-EU convergence in the period
from 1951 to 1993. These authors did not specify, however, a drift and a time trend, preventing the asymptotic
distribution of relative incomes to be approximated by the sampling distribution. This may have caused a
significant estimation bias (see Bernard and Durlauf, 1996, for a discussion).
12 Of course, since Portugal is not being subtracted from the EU average, some endogeneity remains. We trust,
however, the implied bias to be negligible.
13 Results for PCGDP are not qualitatively different from those obtained with GDPWAP.
IV. Miracles and disasters
A well established result in the growth literature is that there is no general tendency
for poor countries to grow faster than rich countries. As pointed out by De Long (1988),
even if (cross-section) convergence looks like holding in particular samples14, results are in
general fragile to small sample modifications. This means that one should not trust too
much forces pushing for convergence.
To illustrate this, we refer to figures 5 and 6. Figure 5 graphs on the x-axis the 1960
level of GDPWAP and on the y-axis the growth of this variable along 1960-80. The figure
suggests a strong tendency for economies to converge, with initially poorer countries
growing faster than richer countries. A common textbook interpretation for this evidence is
that industrialised nations, because they have similar technology levels, labour skills,
investment rates, and population growth rates, are expected to enjoy convergence. Baumol
(1986) goes a bit farther (pp 1077): “It seems not to have mattered much whether or not a
particular country had free markets, a high propensity to invest, or used policy to stimulate
growth. Whatever its behaviour, that nation was apparently fated to land close to its
predestined position (...)” and (pp 1081): “(...) the convergence of productivity levels in
industrialised countries condemned those with high 1870 productivity levels to relative
slow growth since then”. No doubt, these conclusions carry a lot of determinism.
Figures 5 and 6 provide an illustration of the De Long Critique in the time dimension:
when the sample period changes to 1980-2000, no evidence of convergence is found. This
is not caused by any particular outlier, but rather by few of them, including “growth
miracles” such as Ireland and Luxembourg and “growth disasters”, such as Greece and
New Zealand15.
Table 4 displays the results of some formal tests (details in the table). Different samples
and productivity measures are allowed for, so as to stress the case against unconditional
convergence16. As shown in the table, the regression slopes (betas) tend to be significant in
1960-1980 but not in 1980-2000 (exception for GDPPW). The case against convergence is
reinforced by the observation that the "beta test" tends to be biased towards rejection
(Friedman, 1992, Quah, 1993). Similar conclusions hold using the Carree and Klomp
(1997) T3 statistic.
Observation of Figure 7 suggests that, if convergence appeared to have dominated the
data in the sixties and the seventies (eventually capturing the role of transition dynamics
that followed the disruption caused by WWII and also the technological diffusion that was
14 Baumol (1986) found significant evidence of convergence among the Maddison 16, in the period 1870-1979.
For OECD economies in the post-WWII period see, for example, Dowrick and Nguyen (1989), Mankiw et al.
(1992), Barro and Sala-i-Martin (1992), Carre and Klomp (1997).
15 Fuente (1998) uses the residuals of simple regressions like those in figures 5 and 6 to evaluate the relative
performance of each economy, after controlling for the "convergence" effect. As shown below, however, this
method may suffer a significant estimation bias.
16 The terminology is due to Mankiw et al. (1992) and Barro and Sala (1992). If countries differ in their long-run
per capita income levels, there will be no general tendency for poor countries to grow faster than rich countries
(unconditional convergence). Conditional convergence will occur if each country tends to grow faster, the
greater the gap between its initial per capita income level and its own long term potential.
made possible by the establishment of the New Economic Order17), in the eighties
idiosyncrasies are likely to have played a major role in explaining the cross-section
variation of growth rates.
V. Some policy oriented calculations
To understand why some OECD nations have grown faster than others, it is necessary to
learn on what is driving the motion. To deal with this question, we run some ad-hoc (Barro
type) regressions, testing the significance of policy induced variables potentially related to
growth18. Investment rates and labour participation are excluded from the analysis, so as to
capture the overall effect of the policy. The reason is that some important links between
policy and growth are mediated through factor accumulation and labour participation. If
one controlled for these variables, crucial effects of the policy would not be detected.
The choice of the explanatory variables is largely driven by the thematic of this
conference. A specific question to be addressed is the degree to which good institutions
matter. To capture the role of institutions, we use the Sachs and Warner (1997) General
Institutional Quality Index (INST), which in turn is an average of 5 sub-indexes, measuring
the "rule of law", "bureaucracy", "corruption", "expropriation risk" and "government
repudiation of contracts". These 5 sub-indexes shall not be used separately, because they
tend to be highly correlated.
Another topic of interest is the role of public finance. To capture the distortions and
inefficiencies associated to big governments and high taxation, we include "total
expenditures in percentage of GDP" (GOVX). For the composition of government
spending, we use "government fixed capital formation in percentage of GDP" (GINV)19.
Both variables are period averages20.
As for the role of factor markets, we focus on labour market flexibility. The reason is
that capital markets in this sample are mostly free of controls. In contrast, employment
protection varies considerably across countries and is currently a major problem in Portugal
(see Blanchard and Portugal, 2001). Employment protection is measured by the OECD
index of "overall strictness against dismissals", late 1980s (EMPROT).
The last topic of the conference is the role of human capital. In the empirical literature,
the evidence for this variable has been mixed, eventually due to excessive emphasis on
schooling, rather than training (Temple, 1999). In this paper we proxy the quality of the
human resources by a measure of "availability of skilled workers", based on a international
survey (World Competitiveness Yearbook, 1991).
17 Empirical support to the role of trade as vehicle for convergence can be found, for example, in Ben-David
(1993, 1996). Mankiw et al. (1992) stress the role of transition dynamics in explaining convergence among
OECD nations after WWII.
18 For a structural approach focusing on the Portuguese economy, see Duarte and Simões (2001).
19 Note that the information content of this variable is somehow mixed, because the same level of infra-structures
can be provided with different arrangements involving the government and the private sector.
20 Two problems arise. First, these variables are silent in respect to the time path of government expenditures.
Second, and more important, these variables are not entirely exogenous. To keep simplicity, however, we
decided not to implement any corrective action.
Although inflation is no longer a major issue in Portugal, one may want to stress the
benefits of price stability. For this reason, average inflation (INFL) is also included among
the explanatory variables. The direct link is through efficiency, as inflation acts as a tax on
working capital. Since this variable is usually associated to macroeconomic imbalances,
however, it may also capture the effects of related syndromes.
A more difficult task is to control for trade openness. Since New Zealand was the only
country in this sample not satisfying the Sachs and Warner (1995) criteria for “open
economy” in 1980 (see description in Table 5), the variable OPEN collapses into a "New
Zealand dummy".
In estimation, two problems arise. One is that some variables have missing values. The
other is that, because of multicolinearity, the significance of each variable may depend on
the particular combination of variables included in the regression21. To deal with these
problems, instead of presenting the result of a single equation, we experiment different
combinations of explanatory variables and sample sizes. This allows the reader to evaluate
the uncertainty concerning the estimates.
Various regression estimates are reported in Table 5. Results for the simple
(unconditional) convergence equation are in columns I and VI. In both cases, the 1980
productivity level is not significant, confirming the earlier result that the simple
convergence hypothesis does not hold in this sample (remarkably, this is true even in
column VI, after removing three of the outliers identified in Figure 6). In contrast, the
initial productivity level becomes significant, even with Luxembourg in, once two
variables, SKILL and INST are controlled for.
Interpreting the results, one should take into account that some indicators used may be
capturing the effect of more general syndromes, to which these variables are related. In all
equations, the signs of the estimated coefficients are as expected. The New Zealand dummy
is significant whenever this country is included (II-V), meaning that the poor performance
of this country is not fully accountable by the other regressors. We cannot tell, however, if
this reflects trade restrictions or any other "Kiwi effect"22. INFL gets a significant
coefficient in some equations (V, XI), but is disturbed by the presence of both INST and
GOVX. In general, the variables capturing the role of public finance have the expected sign
but low significance. In contrast, SKILL, EMPROT and INST tend to be significant and to
exhibit reasonably stable coefficients, even across samples. In equation IV, we see that
these three variables, together with the initial income (capturing either decreasing returns or
technological catch up) and the New Zealand dummy, account for 77% of the cross-country
variation of productivity growth (for partial associations, see Figures A1-A4, in Appendix
1).
21 See Levine and Renelt (1992). In this paper, this is not a major problem because, by focusing on OECD
economies, we reduce considerably the range of potential explanatory variables. For larger samples, the
methodology proposed by Sala-i-Martin (1996) provides a superior, although demanding, alternative.
22 Using the stochastic approach to convergence, Greasley and Oxley (2000) failed to reject the non-convergence
null for New Zealand vis-à-vis some trading partners. The authors emphasised the role of excessive protection
in explaining why this country has been in a divergence track for almost a century.
As a final exercise, we use the estimation results from equation IV to characterise

particular growth experiences (the method follows Dowrick and Nguyen, 1989). Figure 8
reproduces Figure 6, with growth rates redefined in terms of deviations from that of
Holland23. Two trend lines are also displayed. The slashed line corresponds to the simple fit
with two variables, like in Figure 6. The thick line depicts the unbiased "convergence”
effect, as implied by equation IV. It gives the differential growth that would be observed in
a country having attributes similar to that of the benchmark economy, conditional on its
initial income.
Vertical distances in respect to the thick line in Figure 8 measure the relative effect of
the remaining explanatory variables and the country residual. The quantitative contributions
of these variables are displayed in Table 6. Interpreting, the Irish “miracle” appears to be
largely accounted for by availability of skilled workers and labour market flexibility. In
Portugal and Spain no variable apart from the “catch-up” effect pushes for convergence
towards the benchmark economy. In Portugal, the quality of institutions, the quality of
human resources and employment protection account together for 1.48% loss of relative
economic growth (1.38% in Spain). Interpretation of these figures should be kept, however,
in reasonable limits, as the exercise carries a lot of counter-factual.
Comparing the actual growth rates with the predicted rates in Table 6, we verify that
this small set of variables does a quite good job in discriminating growth performances. The
major exception is US, which actual growth rate goes well beyond the one implied by the
quality of the attributes considered. The correspondent residual is the “measure of our
ignorance”.
VI. Concluding remarks
A well established result in the growth literature is that there is no systematic tendency
for poor countries to grow faster than rich countries. In the last century, the large majority
of developing economies remained poor and incapable of reducing the income gap in
respect to more advanced nations. A dozen of countries, however, including Portugal, have
done consistently well.
In this paper we stress that growth performances depend on policy actions. If the
Portuguese economy has achieved impressive growth rates, this shall not be de-linked from
the strategic option to participate in major international movements in the 1960s. EFTA
membership in 1959 and GATT participation in 1962 mark the policy move towards
openness, breaking with a period of economic slowdown, characterised by inward oriented
economic policies and limited competition. Progressive integration in the global economy
proceeded with the 1971 EFTA-EC agreement, EC accession in 1986, participation in EMS
in 1992, the adoption of the Single Market in 1993, and the decision to join the first wave
of EMU.
23 Holland is roughly the median country, both in terms of initial income and economic growth. Choosing a
different benchmark would affect the intercept of the adjusted line, but not the slope.
The benefits from openness came in various fronts (see, for example, Macedo, 1990,
Lopes, 1996). On one hand, increasing competition in international markets is likely to
have impacted through efficiency and greater exposure to external productivity shocks. On
the other hand, and most important, these external commitments required the
implementation of complementary policy actions, that helped to sustain economic growth.
These include the elimination of Condicionamento Industrial (an inheritance from the
1930s restricting competition), financial liberalisation, sound public finance, privatisation,
dismantling of monopolies, adoption of a consistent macroeconomic policy framework and,
broadly speaking, the implementations of the Acquis Communautaire in a number of areas.
Cunha and Martins (2001), called the disciplinary device implied by external commitments
a “virtuous pressure”.
Much has still to be done, though. The results obtained in this paper suggest that more
attention should be given to structural matters. Inflexibility, market distortions, unfair
competition, taxation, bureaucracy, corruption, privilege and status quo are in the core of
the policy agenda all over the world and we believe the same should hold in Portugal. For
these tasks to be addressed, however, a redefinition of priorities on the fiscal side has to be
considered. Otherwise it will be impossible to proceed with consolidation of the public
finance and at the same time to keep in shape with the policy agenda.
A major challenge for policymakers in the current circumstances is the declining
incidence of the “virtuous pressure”. Perhaps this explains some loss of impetus in
important areas of reform (as an example, the different paths followed by Portugal and
Spain in respect to EMPROT, figure 9; for more, see OECD, 2001, pp 114-115). To
overcome the natural resistance of the established groups, a greater pedagogy is likely to be
required in the near future. To banish fateful prospects is part of the task. Fate brings
indifference and indifference provides an appropriate social framework for interest groups
to defend the status quo. The Anti-Fado Manifesto is not only a testimony against the
believe that some mysterious forces inside the EU will push income levels in Portugal
towards the EU average, irrespectively of the domestic policy actions. It is also a cry
against catastrophic thinking that tends to emerge in the Portuguese society in periods of
economic slowdown. For both, I end up expressing my militancy with a quite dramatic
quotation:
Morra o fado!24 Morra! - Pim25.
24 Vasco Santana, in Telmo (1933).

25 Almada Negreiros (1915).
References
Baumol, W. (1986), “Productivity Growth, Convergence and Welfare: What the Long Run
Data Show”, Americam Economic Review 76 (5), 1072-1085
Barro, R. J. (1991), “Economic growth in a cross-section of countries”, Quarterly Journal
of Economics 106:2, 407-43.
Barro, Robert and J.W. Lee (1993), "International comparisons of educational attainment”,
Journal of Monetary Economics 32, 363-394.
Barro, R. and X. Sala-i-Martin (1992), “Convergence”, Journal of Political Economy, 100
(2), 223-251.
Barros, P. and Garoupa, N. (1996), “Portugal-European Union Convergence: Some
evidence”, European Journal of Political Economy 12 (3), 545-53.
Ben-David, D. (1993), “Equalising Exchange: trade liberalisation and income
convergence”, Quarterly Journal of Economics, 63 83), 653-680.
Ben-David, D. (1996), “Trade and convergence among countries”, Journal of International
Economics, 279-298.
Bernard, A. and S. Durlauf (1995), “Convergence in International Output”, Journal of
Applied Econometrics 10, 97-108.
Bernard, A. and S. Durlauf (1996), “Interpreting tests of the convergence hypothesis”,
Journal of Econometrics 71, 161-173.
Blanchard, O. and P. Portugal (2001), "What hides behind an unemployment rate:
comparing Portuguese and U.S. labour markets", American Economic Review 91
(1), 187-207.
Carlino, G. and Mills, L. (1993), “Are US regional incomes converging?: a time series
analysis”, Journal of Monetary Economics 32, 335-346,
Carree, M., and L. Klomp. (1997), “Testing the convergence hypothesis: a comment”,
Review of Economics and Statistics 79, 683-686.
Cunha, L. and F. Martins (2001), “The Portuguese economic Integration in the EU: some
lessons to be drawn”, mimeo, Banco de Portugal.
De Long, J. (1988), “Productivity Growth, Convergence and Welfare: Comment”,
Americam Economic Review 78 (5), 1138-1154.
Dowrick, S. and D-T. Nguyen (1989), “OECD comparative economic growth 1950-85:
catch-up and convergence”, The American Economic Review 79(5), 1010-1030.
Duarte, A. and M. Simões (2001), Principais factores de crescimento da economia
portuguesa no espaço europeu”, proceedings of the fourth conference of CISEP –
Centro de Investigação sobre Economia Portuguesa, ISEG, Lisboa.
Freitas, M. (2000), "Quantity versus quality: the Irish growth accounting", Boletim
Económico, Banco de Portugal, Março, 59-70.
Freitas, M. (1992), “Sobre o valor real do Escudo: uma abordagem empírica", Economia,
Vol XVI, Nº3, 325-356.
Fuente, X. (1998), “Convergence equations and income dynamics: the sources of OECD
convergence, 1970-95”, Centre for Economic Policy and Research Discussion
Paper No 1794.
Friedman, M. (1992), “Do old falacies ever die?”, Journal of Economic Literature XXX,
2129-2132
Greasley, D. and L. Oxley (2000), “Outside the Club: New Zealand’s economic growth,
1870-1993”, International Review of Applied Economics 14 (2), 173-192.
Gylfason, T. (1999), Principles of Economic Growth, Oxford University Press, Oxford UK.
Hume, D. (1758), Essays and Treatie’s on several subjects, London: A. Millar.
Levine, R. and D. Renelt (1992), “A sensitivity analysis of cross-country growth
regressions”, American Economic Review 82(4), 942-63.
Lopes, J. (1996), A Economia Portuguesa desde 1960, Gradiva.
Macedo, J. B. (1990), "External liberalisation with ambiguous public response", in Bliss. C.
and J.B. Macedo (eds), Unity with Diversity in the European Economy: The
Community's Southern Frontier, Cambridge University Press.
MacKinnon, J. (1991), “Critical values for cointegration tests”, in Engle, R. and C. Granger
(eds), Long Run Economic Relationships, Oxford Academic Press.
Maddison, A.(1995), Monitoring the world development 1820-1992, Paris OECD.
Mankiw, G., D. Romer and D. Weil (1992), “A contribution to the empirics of economic
growth”, Quarterly Journal of Economics, 107 (2), 407-38.
Malthus, Thomas (1817), Essay on Population, 5th edn., London.
Negreiros, J. A. (1915), "Manifesto Anti-Dantas", Textos de Intervenção, Orpheu, nº3,
Lisboa.
Neves, J. C. (1994), The Portuguese Economy: a Picture in Figures, Universidade Católica
Portuguesa, Lisboa.
Quah, D. (1993), “Empirical cross-section dynamics in economic growth”, European
Economic Review 37, 426-434.
OECD, 1999, Employment Outlook. Paris, June.
OECD, 2001, Economic Surveys: Portugal. Paris.
Rappoport and Reichlin (1989), “Segmented trends and non-stationary time series”, The
Economic Journal, 99, 168-177.
Rodrick, D. (1997), “TFPG controversies, institutions and economic performance in East
Asia”, Working Paper NBER 5914.
Sachs, J. and A. Warner (1995), “Economic Reform and the Process of Global Integration”,
Brooking Papers of Economic Activity 1-118.
Sachs, J. D. and Warner, A. M. (1997) Fundamental sources of long-run growth. American
Economic Review, papers and proceedings, May.
Sala-i-Martin, X. (1996), “I just run four million regressions”, NBER Working Paper No
6252.
St. Aubyn (1999), “Convergence across industrialised countries (1890-1989): new results
using time series methods”, Empirical Economics 24, 23-44.
Telmo, C. (1933), A Canção de Lisboa. Tobis Portuguesa.
Temple, J. (1999), “The new growth evidence”, Journal of Economic Literature XXXVII,
112-156.
The International Institute for Management Development (1991), The World
Competitiveness Yearbook , Lawsanne, Switzerland.
Table 1
Per capita GDP, annual average growth rates, 1960-92
Rank Country Growth rate
1 Korea 6.6
2 Singapore 6.2
3 Taiwan 6.1
4 Hong Kong 6.1
5 Botswana 5.6
6 Malta 5.4
7 Japan 5.2
8 Thailand 4.6
9 Cyprus 4.5
10 Malaysia 4.3
11 Portugal 4.0
12 Indonesia 3.9
13 Lesotho 3.7
14 Greece 3.7
15 Spain 3.6
16 Ireland 3.5
17 Seychelles 3.5
18 Tunisia 3.3
19 Cape verde 3.2
20 Italy 3.2
Source: Sala-i-Martin (1996).

Ranking is computed over a sample of 119 countries.
Table 2 –
Growth accounting
I II III IV V
Per capita GDP per GDP per Capital-labour Education
GDP working age Employed ratio attainment
person
PPS current 1995 PPS 1995 PPS 1995 PPS
EU4=100
EU15=100 EU15=100 EU15=100 EU15=100
Portugal
1960 41.5 45.59 36.7 36.9 1960 29.8
1973 59.9 3.1 66.34 3.2 48.3 2.3 34.8 -0.5 1975 40.6 2.1
1985 54.8 -0.7 61.53 -0.6 52.2 0.6 47.0 2.5 1985 53.0 2.7
2000 73.8 2.0 73.0 1.1 66.3 1.6 62.0 1.9 1990 53.8 0.3
Spain
1960 60.7 59.64 63.6 64.0 1960 59.8
1973 78.8 2.2 78.91 2.4 82.9 2.2 64.4 0.1 1975 73.7 1.4
1985 72.9 -0.6 74.35 -0.5 96.6 1.3 85.2 2.4 1985 85.4 1.5
2000 81.1 0.7 81.76 0.6 92.0 -0.3 88.0 0.2 1990 89.2 0.9
Greece
1960 44.5 49.43 55.6 56.0 1960 77.4
1973 72.7 4.2 81.03 4.2 98.5 4.9 95.8 4.6 1975 86.6 0.8
1985 65.8 -0.8 75.48 -0.6 85.0 -1.2 106.7 0.9 1985 98.6 1.3
2000 67.9 0.2 67.51 -0.7 80.4 -0.4 116.7 0.6 1990 104.1 1.1
Ireland
1960 64.7 75.11 78.8 79.3 1960 102.5
1973 62.1 -0.3 71.83 -0.4 78.0 -0.1 86.3 0.7 1975 103.9 0.1
1985 68.8 0.9 81.37 1.0 95.3 1.7 108.3 1.9 1985 105.4 0.1
2000 118.4 3.7 123.9 2.8 121.3 1.6 90.6 -1.2 1990 104.5 -0.2
Note: Average annual growth rates for each sub-period are in italic. Data on education
attainment follows a distinct periodicity, due to lack of data.
Source: Columns I-IV: own calculations, using data from the European Commission.
Column V: education attainment is measured by the average schooling years in the
population over age 15 (Barro and Lee, 1993, TYR15). EU4 is a non-weighted average of
UK, Germany, France and Italy.
Figure 1.
Per capita GDP, GDP per working age person and GDP per worker (Portugal,
constant PPS)
0.00
G D P per capita
G D P per working age person
-0.20 G D P per worker
-0.40
EU=0.00
-0.60
-0.80
-1.00
-1.20
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
Source: Own calculations using data from the European Commission. 1999
Figure 2:
Population, employment and GDP (Portugal)
0.02
0.034
0.019
0.032
0.018
0.03
0.017
EU=1.00
EU=1.00
0.028
0.016
0.026
0.015
0.024
0.014
0.022
Rel. Population (left scale)
0.013
Rel.Employment (left scale)
Rel. GDP (right scale)
0.02 0.012
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
Source: Same as Figure 1.

Figure 3
Percentage of "no schooling" in the population over age 15
50
45 PRT
IRL
40 GRE
SP
35 POL
HUN
30 US
25
20
15
10
0
1960 1965 1970 1975 1980 1985 1990
Source: Barro and Lee (1993) data set.
Figure 4
International Test Scores: Math, 1993-98, age of pupils: 13
65
60
55
50
45
40
35
30
Denmark
Germany
Colombia
Ireland
Romania
Belgium
Portugal
Cyprus
Hungary
Norway
Spain
Yugoslavia
Australia
Bulgaria
Canada
Iran, I.R. of
Japan
Korea
Thailand
Austria
France
Greece
Sweden
United Kingdom
Netherlands
United States
Czech Republic
Slovak Republic
Hong Kong
Singapore
New Zealand
Switzerland
South Africa
Russian Federation

Table 3
Stochastic convergence: Portugal-EU
I II III IV V VI VII VIII
Dependent: Dependent:
Relative GDP per working age population (logs) Relative GDP per worker (logs)
Discontinuities none 1974 1974 1974 none 1974 1974 1974

1986 1993 1986 1993
ρ 0.74 0.49 0.37 0.49 0.66 0.25 0.21 0.08

ADF -2.7 -4.8 ** -4.6 -4.2 -2.7 -4.4 * -4.1 -4.8 *
Impetus of catch-up:
First segment 0.030 0.021 0.021

10.0 9.5 11.6
(+) Break 1 -0.022 -0.008 -0.008

-7.3 -3.5 -4.1
(=) Second segment 0.008 0.013 0.013
(+) Break 2 0.007

2.1
(=) Third segment 0.020
Adjusted R-squared 0.544 0.556 0.515 0.174 0.335 0.306 0.361

S.E. of regression 0.0173 0.0171 0.0179 0.1295 0.0194 0.0198 0.0190
Lags used 5 1 1 1 0 1 1 1
LM(SC,2) 1.13 1.20 4.45 1.15 2.94 2.21 4.32 2.93
Included observations: 35 39 39 39 40 39 39 39
Notes to Table 3: t-ratios in italic. (**) and (*) denote significance of the unit root test at 1% and 5%, respectively.
LM(SC, 2) denotes the Breusch-Godfrey tests for the null of no serial correlation of order 2. Critical values for the
unit root test without discontinuities are from MacKinnon (1991). The critical values for the unit root null with
structural breaks are from Rappoport and Reichlin (1989): 4.73 and 4.08 (models with two segments) and 5.45
and 4.76 (models with three segments).
0 when t < i
Method: Let z t denote the dependent variable, Di =  and i the time of the structural break. For
1 when t ≥ i
columns III, IV, VII and VIII we run
∆zt = α 0 + α1D1 + α 2 D2 + β 0t + β 1D1t + β 2 D2t + (ρ − 1)zt −1 + δ∆zt −1 + ut . Estimates in columns II and VI are
computed in a similar way, imposing α 2 = β 2 = 0. Column I and V display the results of simple unit root tests.
Whenever the unit root null is rejected, the “impetus of catch-up” is also displayed. This is estimated by non-linear
least squares, imposing α i = α~i (1 − ρ ) + β i (ρ − δ ) and β i = β i (1 − ρ ) , so as to obtain directly in the output the
~ ~
~ ~ ~
coefficients and t-statistics of the equivalent representation: z = α~ + α~ D + α~ D + β t + β D t + β D t + u~ ,
t 0 1 1 2 2 0 1 1 2 2 t
[ −1
]
in which u~t = 1 − (ρ + δ )L + δL2 ut is a zero-mean covariance-stationary process. The transformation requires
the roots of the characteristic equation λ2 − (ρ + δ )λ + δ = 0 to lie inside the unit circle of the complex plan,
which in equations above is always satisfied (see Freitas, 1992, for further details).
Figure 5
Growth rates versus initial levels, GDP per working age population, 1960-1980
7.00%
Change in GDP per working age population , 1960-1980
6.00%
GRE
JAP
5.00%
PT
SP
4.00%
IRL
3.00%
LUX US
UK
2.00% CH
y = -0.0286x + 0.1019
2
R = 0.8319
1.00%
NZ
0.00%
1.3 1.5 1.7 1.9 2.1 2.3 2.5 2.7 2.9 3.1 3.3
GDP per working age population in 1960 (logs)
Sample: 22 OECD countries: Portugal (PT), Luxembourg (LUX), Belgium (BEL), Denmark (DNK), Germany
(GER), Greece (GRE), Spain (SP), France (FR), Ireland (IRL), Italy (IT), Holland (NDL), Austria (AUT), Finland
(FIN), Sweden (SWE), United Kingdom (UK), United States (US), Japan (JP), Canada (CAN), Switzerland (CH),
Norway (NOR), Australia (AUS), New Zealand (NZ).
Figure 6
Growth rates versus initial levels, GDP per working age population, 1980-2000
0.045
LUX
0.040
Change in GDP per working age population, 1980-2000
IRL
0.035
0.030
0.025
PT
JAP US
0.020
SP
0.015
0.010
NZ
CH
y = -0.0084x + 0.0452
0.005 R2 = 0.0394 GRE
0.000
2.4 2.6 2.8 3 3.2 3.4 3.6
GDP per working age population in 1980 (logs)
Source, Sample: Same as Figure 5.
Figure 7
Sigma convergence
0.43
GDP per capita

GDP per working age population
0.38 GDP per employed
Coefficient of Variation
0.33
0.28
0.23
0.18
0.13
1960
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000

Sample: Same as Figure 5.
Table 4
Tests for unconditional convergence
N β λ t σ0 σΤ T3
GDP per working age population

1960-1980 G22 22 -0.029 0.044 6.6 ** 0.39 0.20 7.8 **
EU15 15 -0.026 0.036 6.0 ** 0.35 0.19 5.5 **
1980-2000 G22 22 -0.008 0.009 0.8 0.20 0.23 -1.2
EU15 15 -0.006 0.006 0.4 0.19 0.25 -1.7
Per capita GDP
1960-1980 G22 22 -0.026 0.036 5.5 ** 0.39 0.22 5.8 **
EU15 15 -0.023 0.031 5.0 ** 0.36 0.21 4.4 **
1980-2000 G22 22 -0.014 0.016 1.4 0.22 0.23 -0.3
EU15 15 -0.013 0.015 0.9 0.21 0.24 -0.6
GDP per employed
1960-1980 G22 22 -0.030 0.047 6.5 ** 0.45 0.21 9.1 **
EU15 15 -0.028 0.041 3.8 ** 0.38 0.21 4.7 **
1980-2000 G22 22 -0.020 0.026 2.6 * 0.21 0.17 1.7 *
EU15 15 -0.016 0.020 1.6 0.21 0.19 0.6
Samples: G22 refers to the sample described in Figure 5. EU15 refers to all European Union
countries. Notes: (**) and (*) denotes significance at 1% and 5%, respectively. N refers to the
number of countries; β is the measure of beta-convergence, obtained running
(xiT − xi 0 ) T = α − βxi 0 + ε i , where: xiτ is the value of the dependent variable at hand (in
logs) in country i at time τ; 0 and T refer to the first and last period of the sub-sample
(
considered. The implied speed of convergence, λ, is obtained from β = − 1 − e−λT T . The t – )
statistic tests the significance of λ. στ is the cross-section standard deviation of xiτ . The last
column displays the values of the Carree and Klomp (1997) statistic,
σ 02 / σ T2 − 1
T3 = (N )0.5
( )
, which has a standard normal distribution under the null of no
0.5
2 1 − (1 + Tβ ) 2
sigma-convergence.
Table 5
Regression variables explaining growth between 1980 and 2000
I II III IV V VI VII VIII IX X XI
C 0.05 0.03 0.08 0.07 0.11 0.05 0.09 0.09 0.09 0.10 0.12
1.6 1.0 2.7 3.9 4.0 2.1 3.8 4.0 4.1 4.2 3.8
GDPWAP80 -0.008 -0.030 -0.042 -0.042 -0.037 -0.011 -0.039 -0.038 -0.041 -0.041 -0.035
-0.9 -2.6 -5.6 -5.8 -5.0 -1.3 -4.2 -4.3 -4.8 -4.8 -4.1
SKILL 0.00033 0.00033 0.00034 0.00032 0.00034 0.00035 0.00032 0.00033 0.00031
1.6 2.7 3.0 2.4 2.7 2.9 2.6 2.7 2.3
EMPROT -0.0029 -0.0029 -0.0026 -0.0029 -0.0025 -0.0036 -0.0030 -0.0026
-2.7 -2.8 -2.4 -1.8 -2.1 -2.8 -2.6 -2.1
INST 0.0044 0.0035 0.0039 0.0039 0.0040 0.0038 0.0039
2.0 1.7 3.3 3.0 3.2 3.0 3.0
GOVX -0.00012 -0.00017
-0.7 -1.2
GINV 0.0008 0.0015
0.4 1.1
INFL -0.0002 -0.0011 -0.0011
-0.3 -2.5 -2.3
NZD 0.02 0.02 0.02 0.02
1.6 2.8 3.2 2.1
R-squared 0.039 0.421 0.772 0.771 0.723 0.099 0.752 0.748 0.741 0.716 0.656
Adjusted R-squared -0.009 0.285 0.675 0.695 0.630 0.043 0.617 0.643 0.633 0.629 0.550
S.E. of regression 0.00832 0.00701 0.00387 0.00375 0.00413 0.00637 0.00403 0.00389 0.00394 0.00396 0.00437
Included observations 22 22 21 21 21 18 18 18 18 18 18
Notes to Table 5: t-statistics in italic.
Dependent: Average annual change in log of GDPWAP between 1980 and 2000 (European Commission).
Explanatory variables:
GDPWAP80: Log of GDPWAP in 1980 (European Commission).
SKILL – Availability of technically skilled labour (World Competitiveness Yearbook, 1991).
EMPROT – Index of employment protection (overall strictness against dismissals, late 1980s, OECD, 1999).
INST - Institutional quality index (Sachs and Warner’s, 1997). Average 5 sub-indexes, each based on survey data compiled during the 1980s from Political Risk
Services, measuring (1) the rule of law (2) the bureaucratic quality (3) the corruption in government (4) the risk of expropriation (5) the government repudiation
of contracts.
GOVX - Ratio of total government expenditures to GDP, average 1980-2000. Levels were adjusted for the 1995 accountancy change (own calculations using
data from the European Commission).
GINV - Government investment divided by GDP, average 1980-2000. Levels were adjusted for the 1995 accountancy change (own calculations using data from
the European Commission).
INFL - Average CPI inflation rate along 1980-2000 (European Commission).
NZD: Fraction of years during the period 1980-1990 in which the country is rated as an open economy according to the criteria in Sachs and Warner (1995). This
is based on the satisfaction of 5 requirements at the same time: (1) average tariff rates below 40 percent; (2) average quota and licensing coverage of imports of
less than 40 percent; (3) a black market exchange rate premium that averaged less than 20 percent during the decade of the 1970s and 1980s; and (4) no extreme
controls (taxes, quotas, state monopolies) on exports.
Samples:
22: Portugal, Luxembourg, Belgium, Denmark, Germany, Greece, Spain, France, Ireland, Italy, Holland, Austria, Finland, Sweden, United Kingdom, United
States, Japan, Canada, Switzerland, Norway, Australia, New Zealand.
21: All but Luxembourg.
18: Excludes Luxembourg, Switzerland, Australia, New Zealand.
Table 6
Sources of differential growth
New United
P ortugal S pain G reece Ireland
Zealand S tates
E qu atio n IV
Actual growth rate 2.34 2.08 0.72 3.96 0.92 2.26
Predicted 2.39 1.79 1.06 3.51 0.92 1.50
O f w hich:
Catch up effect 2.27 1.57 1.23 1.40 0.53 -1.03
SK ILL -0.17 -0.29 -0.24 0.67 -0.21 0.09
EM PRO T -0.49 -0.23 0.17 0.43 0.40 0.84
INST -0.82 -0.86 -1.70 -0.59 -0.06 0.00
NZD 0.00 0.00 0.00 0.00 -1.33 0.00
Benchm ark growth rate 1.60 1.60 1.60 1.60 1.60 1.60
Residual -0.05 0.30 -0.33 0.45 - 0.76
Source: Own calculations, using equation IV from Table 5. The table displays (a) the actual growth rate along
1980-2000, gi, (b) the predicted rate,gˆ i = αˆ − βˆx i 0 + γˆz i , where x i 0 is the (log) GDP per working age person
in 1980 and z i is the vector of explanatory variables, and (c) the residuals εˆ i = g i − gˆ . The predicted growth
rate is break down by the effects of the explanatory variables, redefined in terms of deviations from the
benchmark economy, Holland. That is, gˆ i = gˆ NED − βˆ ( x i ,0 − x NED , 0 ) + γˆ ( z i − z NED ) .
Figure 8
Actual growth rates versus “catch-up” effect
0.03
IRL
0.02
Change in GDP per working age person, 1960-1980
0.02
(deviations from the benchmark)
0.01
0.01 PRT
NOR US
SP
UK
0.00
IT
-0.01
NZ CHE
-0.01
GRE
-0.02
2.500 2.600 2.700 2.800 2.900 3.000 3.100 3.200 3.300 3.400 3.500
GDP per working age person in 1980 (logs)
Notes: Referring to the notes in Table 6, the figure displays deviations from the benchmark
growth rate, ( g i − g NED ) against initial income, xi 0 . The dashed line depicts the fitted values
of the simple unconditional model (as in figure 6). The thick line depicts the catch-up effect, as
implied by equation IV, βˆ ( x i ,0 − x NED ,0 ) .
Figure 9
Employment strictness of protection against dismissals
5.0
4.5 Late 80s

Late 90s
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
SP
ITA
POL
GER
GRC
FR
AUT
PRT
CAN
NOR
CH
BEL
DNK
IRL
NDL
UK
US
JP
AUS
FIN
HUN
SWE
NZL
Source: OECD (1999).
Appendix 1
Figure A1 - Partial association between GDPWAP growth and 1980 GDPWAP
-0.10
2.400 2.600 2.800 3.000 3.200 3.400 3.600
-0.11
-0.12
-0.13
-0.14
y = -0.0425x
-0.15 R2 = 0.8705
-0.16
Figure A2 - Partial association between GDPWAP growth and SKILL
0.0350
0.0300
0.0250
0.0200
0.0150
y = 0.0003x
R2 = 0.4114
0.0100
0.0050
-
40.00 45.00 50.00 55.00 60.00 65.00 70.00 75.00 80.00 85.00
Figure A3 - Partial association between GDPWAP growth and EMPROT
0.0100
0.0050
-0.0050
-0.0100
y = -0.0029x
-0.0150 R2 = 0.4842
-0.0200
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Figure A4 - Partial association between GDPWAP growth and INST
0.0500
0.0450
0.0400
0.0350
0.0300
0.0250
y = 0.0039x
R2 = 0.6293
0.0200
0.0150
5 6 7 8 9 10
Source (all figures): Own calculations, using the results of regression IV, Table 5 (the method
follows Barro, 1991).
PRODUCTIVITY CONVERGENCE:
PORTUGAL AND THE EUROPEAN UNION
Pedro Pita Barros1

Universidade Nova de Lisboa and CEPR (London)
Abstract
Convergence within the European Union is a long-standing policy goal of both
national Governments and the European Commission.
We examine whether, or not, relative productivity of EU-periphery economies, in
particular Portugal, has converged to average EU levels. We consider 27
manufacturing industries, over a 18 year period (1978-1996). Portuguese and
Spanish industries have, on average, performed better than the Greek industry.
Economic factors do influence the growth rate of relative productivity of Portuguese
industries, namely export intensity and value-added content. No such effects are
detected for Spanish and Greek industries. In addition, the Single Market Programme
of 1992 and the accession to the European Union did not have a measurable direct
impact on convergence speed.
The paper also addresses the pattern of absolute manufacturing productivity growth.
It is found that employment contraction has played a role in raising value added per
worker, though the dominant effect is due to pure value added growth. The
predominance of low-research intensity industries in labour productivity growth
raises some concerns for the future.
Correspondence address:
Faculdade de Economia,
Universidade Nova de Lisboa,
Campus de Campolide
1099-032 Lisboa
Email: ppbarros@fe.unl.pt
Fax: 213 886 073
1
I benefited from the comments and suggestions of Pedro Neves and an anonymous referee. All errors and
opinions expressed are the sole responsibility of the author.
62 Pedro Pita Barros
CONVERGÊNCIA NA PRODUTIVIDADE: PORTUGAL

E A UNIÃO EUROPEIA
RESUMO
Pedro Pita Barros

Fevereiro 2002
O processo de convergência real da economia portuguesa para níveis médios comunitários
tem recebido atenção mediática e no debate político. Na maior parte dos casos, a discussão
centra-se em aspectos de muito curto prazo (saber se Portugal cresceu mais ou menos que a
média comunitária num determinado ano, por exemplo).
Essa forma de colocar a questão da convergência real é errada, uma vez que este, por
natureza, é um processo que exige tempo. O outro aspecto normalmente ignorado é a
identificação das forças que originam essa convergência real e que factores a favorecem. O
presente estudo contribui para uma melhor compreensão deste último aspecto. Existem muito
factores, internos e externos à empresa (às instituições, em geral), que influenciam
potencialmente a sua produtividade. Optou-se por analisar factores associados com a abertura
da economia aos mercados internacionais e o seu efeito sobre a convergência da produtividade
por trabalhador na indústria transformadora portuguesa. Esta análise é importante pois permite
avaliar que indústrias mais têm contribuído para o crescimento da produtividade, se têm
mantido o seu grau de importância (ou se há substituição de umas indústrias por outras) e que
tipo de indústrias são.
Não foi abordada a evolução da produtividade nos serviços e na administração pública.
Esta opção deveu-se quer à disponibilidade de informação estatística para realização da análise
quer à menor relevância da disciplina competitiva exercida pelos mercados internacionais
nesses sectores. A análise cobre o período 1978-1996.
Como principais resultados encontram-se efeitos positivos importantes associados com a
intensidade exportadora, revelando que a presença activa em mercados internacionais está
associada para uma mais rápida convergência na produtividade. Em média, para o período
coberto, também se detectou uma correlação positiva entre as indústrias com maior
contribuição para o valor acrescentado e convergência da produtividade.
Esta inferência é confirmada quando se decompõe o crescimento da produtividade em efeito
de crescimento do valor acrescentado e efeito de variação de emprego.
A decomposição da evolução da produtividade, por períodos de tempo (1978/1986,
1986/1992, 1992/1996), revela, porém, que esse modelo de convergência real ao nível da
produtividade da indústria transformadora se está a esgotar (os efeitos detectados apresentam
uma tendência decrescente).
Apesar da principal indústria promotora do crescimento da produtividade do trabalho em
Portugal ter deixado de ser a têxtil, passando as indústrias alimentar e de maquinaria eléctrica
a ter maior preponderância, há uma concentração da geração de valor acrescentado em
sectores pouco intensivos em investigação e desenvolvimento, que são tipicamente apontados
como de fraco potencial de crescimento a longo prazo.
Em suma, encontram-se os seguintes resultados:
• a associação positiva entre crescimento da produtividade e exposição ao ambiente de
concorrência internacional;
• a importância das indústrias tradicionais, que não têm perdido peso relativo enquanto
geradoras de valor acrescentado por trabalhador; e
• diferenças de evolução entre as várias indústrias da indústria transformadora.
Productivity Convergence: Portugal and the European Union 63
1. Introduction
Convergence within the European Union is a long-standing policy goal of both national
Governments and the European Commission. This concern has led to programmes in the
European Union (EU) aimed at improving faster the living standards of the so-called less-
favoured regions.
The EU-peripheral countries have been the main recipients of such funds (though by no
means the only ones). Portugal is one of the economies involved in this catching-up
process. At the aggregate level, the usual indicator GDP per capita shows convergence
over time,2 which may be seen as a result of these efforts, at least partially.
For the convergence process to be sustainable over the long-run, it must be rely on
productivity convergence as well. Although productivity convergence is a clear objective of
economic authorities (both at national and European Union levels), it is not clear that
economics forces lead to such convergence. Nonetheless, one can find several arguments
for it to occur.
On the one hand, technological differences are a natural consequence of economic isolation
and they play a role in the integration process following economic liberalisation. This is
especially true for less advanced economies. Thus, economic integration should have led to
productivity convergence at the industry level, through diffusion of better technologies and
business practices and through productive specialisation.3
On the other hand, classical trade theory also predicts that as barriers to trade decrease
factor prices tend to be equalised, which under competitive input markets means that
productivities must also converge (Samuelson, 1971). Cross-industry differences can be
used in another way to shed some light over the possible drivers of manufacturing
productivity convergence: composition effects vs. technological progress across the board.
Economic integration, read freer trade among European Union countries, may lead to
economic specialisation. One implication from economic theory is that expansion should
occur in industries with a higher relative price.
Related questions are those associated with geographical specialisation across the European
Union. Both classical international trade and agglomeration theories suggest that
geographic relocation of economics activities should occur in the aftermath of economic
integration episodes. The neo-classical theory predicts inter-industry specialisation, guided
by comparative advantage. The new trade theories focus on increasing returns, product
differentiation and imperfect competition. Together with market size, these elements
provide forces towards clustering of economic activities. Reduction in transport costs
should lead to concentration of economic activities near core, larger, markets, unless
periphery markets retain some other advantage (lower labour costs, higher productivity). In
terms of available empirical evidence, Brulhart (2000) found increased specialisation on the
period 1972 – 1996, though no evidence is found that the Single Market Programme
triggered agglomeration of economic activities. Other studies finding increasing
specialisation in the European Union are due to Brulhart (1998) and Amiti (2000). See also
Helg et al. (1995) and De Nardis et al. (1996), who found both dispersing and
concentrating sectors in the European Union. Here, we will not address the issue of
2
See, for example, Barros and Garoupa (1996).
3
This latter effect can be traced back to the implications of the Ricardo-Viner type of models.
geographical specialisation. Our focus is on the productivity effects, measured at the

industry level.
In this paper, we look at a precise question: is there convergence to European average
levels of productivity in the Portuguese manufacturing industry? A related question is also
addressed: is the Portuguese experience different from that of the other peripheral EU
economies? Thus, we compare the evolution of Portuguese productivity to that of Spain
and Greece.4
Real convergence of the Portuguese GDP per capita to European Union average level has
been discussed since long. The typical analysis looks only at the GDP per capita. However,
over the long run, standards of living can only converge to some (higher) reference level if
productivity increases.
Labour productivity, measured by output per worker can be seen as the weighted sum of
each industry productivity, where the weights are the labour share of each industry. This
means that economy-wide increases in labour productivity can be achieved by either
improvement in relative productivity or by contraction (expansion) of those sectors with
lower (higher) labour productivity.
Thus, a first step in the analysis is to check whether, at the manufacturing level, we can
observe any relative productivity improvement. Relative productivity is defined as the ratio
of labour productivity of country i in industry j to the average labour productivity over the
reference countries in industry j, and to establish if such evolution was dictated by
productivity increases at the sectoral level or by change in industries' sizes.
Some related studies on industry convergence have been done previously. Carree, Klomp
and Thurik (2000) report an analysis of convergence of 19 OECD economies, by industry,
in the period 1971-1992, focusing on β-convergence and σ-convergence. They find that the
spread in the speed of convergence across industries is large. Daveri (2000) looks at the
impact of information technology on growth, for a sample of OECD countries, and finds
significant differences across countries. He does not deal directly with convergence issues.
We complement their work in that we look to a particular set of countries and a particular
potential element behind convergence, economic integration. Works on industry
convergence (Bernard and Jones, 1996a, 1996b; Gouyette and Perelman, 1997) revealed
that manufacturing industries have slow or no convergence patterns while services seem to
be the main engine of convergence in value added per worker. There is also findings
pointing towards considerable dispersion in convergence patterns across industries. It is by
now well documented that income and productivity differences within the OECD area have
decreased over the last decades (Pilat, 1998).
There are two strands of literature to which our work relates. First, regional convergence.
Convergence between regions has received attention from several researchers. Close to
ours is Esteban (2000), who reports regional differences as the result of productivity
differences across regions (and not due to distinct industrial structures).5
4
Due to data availability, we are not able to include Ireland in our analysis. The Irish case seems to be a special
one, and several accounts of the ``Irish miracle" can be found. See, for example, Freitas (2000).
5
Other useful studies are, among others, Barro and Sala-i-Martin (1991), Neven and Gouyette (1995), Paci
(1997), Quah (1994).
Second, the role of technology, namely information and communication technologies, to

growth. These studies have focused almost exclusively on the US, and the jury on the issue
is still out.6
The mechanisms underlying convergence of GDP per capita between countries can only be
understood if movements at a lower level of aggregation are known. To have economy-
wide productivity growth one needs to have, at least in some activities, productivity at the
microeconomic level. We attempt to find it at the manufacturing industries level.
The paper is organised as follows. Section 2 reports the database used. Section 3 presents
the estimates of productivity convergence and the potential role of several economic
variables. Section 4 explores the pattern of absolute productivity growth. Finally, Section 5
concludes.
2. Data and methodology
Evaluation of productivity convergence is based on the evolution of value added per

worker relative to a reference group. The reference group is constituted by France,
Germany, Italy and the United Kingdom. Relative productivity is computed on an industry-
by-industry basis. We consider an average of value added per worker.7
The following relationship is estimated: Let Yijt be the log of the relative productivity
measure of country i, industry j and period t; define ∆Yijt as the associated change; the
equation
 ∆Yijt = κij Yijt-1 + εijt (1)
gives the coefficient κij, which is negative in the case of convergence and positive
otherwise.8 The term εijt is a random noise, assumed to be i.i.d. All data were obtained from
the OECD STAN Database (OECD, 1998). The data range from 1978 to 1996 and cover 27
industries, though complete coverage for the 27 industries is only available up to 1994 for
all variables. Estimation of this equation provides the basis of our analysis.9
The raw data used from the STAN database are production, value added, number engaged,
export and import data. Production is national accounts compatible production in current
prices; Value added is the contribution of each industry to national GDP. It is available in
both current and 1990 constant prices; Number engaged includes number of employees as
well as self-employed, owner proprietors and unpaid family workers; export and import
data are converted values from the SITC commodity classification to the ISIC industry
classification, though with some degree of error given the difficulties in matching
definitions from these two classifications. A list of the industries is presented in the
6
See Jorgenson and Stiroh (1999), Oliner and Sichel (2000), Schreyer (2000) and the references therein for the
US debate on this issue.
7
The results are essentially the same if a weighted average is considered, with the weights defined by the number
of workers in each of these four countries. Details available upon request.
8
The model has also been used by Ben-David (1993). Under certain conditions it can be shown that this model is
.
a particular case of Barro and Sala-i-Martin (1992) model
9
The procedure is essentially a unit-root test to determine convergence. The use of Augmented Dickey-Fuller tests
is limited by the small number of observations available on the time dimension.
appendix. The relevant variables have been converted to USD using the exchange rate
values for 1990.
3. Productivity convergence
As a first step, it is useful to have a look at the aggregate, whole manufacturing, level
picture. Figure 1 presents the evolution of a relative labour productivity index,10 for each
country, for the period 1978-1996. The value of 1976 was normalised to 100 for each
country. It becomes clear that Spanish manufacturing has experienced an improvement in
relative productivity, while Portugal shows a similar progress, with exception of the mid-
eighties, and Greece, if anything, has regressed. The natural question is whether the
aggregate level trend hides significant variation across industries, or it is replicated for most
industries.
10
Relative productivity refers to a comparison with average productivity in the reference group for whole
manufacturing.
Figure 1: Productivity convergence in manufacturing
120.00
110.00
100.00
90.00
80.00
70.00
60.00
78
19
1
Portugal Spain Greece
Figure 2: Average annual increase (%) in relative productivity vs initial

value (1978) - Portugal
1.5
1
productivity
0.5
0
0 0.5 1 1.5 2 2.5
-0.5
-1
Relative
Relative productivity
productivity in 1978
in 1978

value (1978) – Greece
1.5
1
productivity
0.5
0
0 0.5 1 1.5 2 2.5
-0.5
-1
Relative
in 1978

value (1978) - Spain
1.5
1
productivity
0.5
0
0 0.5 1 1.5 2 2.5
-0.5
-1
Relative
in 1978
Figures 2 – 4 present the average annual increase in relative productivity against the initial
value of relative productivity.
From these Figures it becomes clear that Spanish manufacturing has performed more
favourably than the Portuguese and Greek ones. This occurs across the board. The
comparison of Portugal with Greece is also clear. Portuguese manufacturing seems on
average to be performing better than Greek manufacturing, though the latter has a better
initial relative productivity.
These two main features can be made more precise by estimation of equation (1) for each
country (assuming κij = κi), presented in Table 1. In addition, we allow for industry
characteristics to influence convergence speed in manufacturing productivity. This
amounts, in our approach, to take κij as a function of fundamental factors. Several variables
are included. The first one is country identity. Country dummies for Greece and Spain were
included (Portugal is kept as the baseline case). This captures characteristics common to all
industries but specific to the country.
Economic variables that may influence the convergence speed are the openness to trade
(either by imports or exports) of each industry, and the relative importance of the export
market. The main difference between these two variables lies in the role of imports. Access
to foreign markets is expected to demand increases in relative productivity. This is
measured by the exporting intensity, defined as the ratio exports per worker. The effect is
captured by a negative sign in the second variable.11 However, the discipline exerted by
imports may be imposing significant productivity increases to manufacturing industries. To
allow for this, the degree of openness to international trade of the industry, measured by the
sum of imports and exports over total production, is included in the analysis. If both
variables are statistically significant, then imports and exports have different channels of
influence over incentives for productivity gains. The degree of openness measure has been
widely used. Frankel and Roemer (1999) find a positive effect of trade on labour
productivity, though precision of estimates is not high. On the other hand, Irwin and Tervio
(2000) find no such link. These studies rely on cross-country variations to identify the
effects, while here we rely on cross-industry differences.12 Alcalá and Ciccone (2001)
criticize the theoretical foundations of the use of the degree of openness. They show that
productivity gains, at the economy level, may be positively or negatively associated with
the degree of openness. Using an adjusted measure, they find a positive effect from more
openness to trade on productivity. Their criticism relies on the role of the price of non-
traded goods and services. Since our analysis is performed at the manufacturing industry
level, and uses cross-industry differences, instead of cross-country ones, to identify effects,
their reasoning is not applicable to our setting.
A significant fraction of international trade within the European Union is of the intra-
industry type. To account for the role of intra-industry trade as a possible engine for
productivity convergence, a standard intra-industry index was included. The index is
defined by:
2 min{imports ijt , exportsijt }
IT = (2)
imports ijt + exportsijt
11
Remember that a more negative number is associated with a higher convergence speed.
12
See also Miller and Upadhyay (2000).
We also allow for some role to be played by the capacity of each industry to generate value
added (measured by what we call the value-added content – value added divided
production at the industry level).
A comment is in order. Many other economic elements may influence convergence speed
and are unaccounted for. This is fundamentally due to data limitations. The OECD STAN
database, version 1998,13 does include some other information, namely on investment and
labour compensation. Since we do not have a measure of the capital stock for a single year
on an industry by industry basis for each country, the investment effort cannot be properly
used in the analysis. On the other hand, labour compensation is most probably endogenous
to productivity gains obtained in each industry, and it is therefore excluded from the
analysis.
The role of technology, in particular of information and communication technologies
(ICT), is treated in Barros (2001). The ICT nature of the industry seems to not matter for
sustained convergence in productivity. In a related analysis, though under a different
approach (and at a coarser level of aggregation), Godinho and Mamede (2000) discuss the
role of technology convergence. They essentially argue that convergence has been smaller
than what could have been expected, as there is still room for significant improvements on
this matter. Our analysis should be seen as complementary to those previous efforts.
Dummy variables for each industry were also included, one at a time, when considering the
full sample. The ones that survived a simple test of individual statistical significance were
then jointly added to the empirical specification. Finally, a common time trend was
included to capture other motives for productivity convergence, shared by all industries in
the three countries. Table 1 reports the results.
Table 1
Basic convergence speed results
Dependent variable: ∆Yijt

All Portugal Greece Spain
-0.0098 -0.0102 0.0022 -0.0421
t-stat (-3.064) (-2.63) (0.28) (-4.65)
P-value [0.002] [0.009] [0.777] [0.000]
Log Likelihood 1062.06 352.345 291.208 454.152
Observations 1302 434 434 434
13
A new version of the OECD STAND database has not yet become available. It is expected to be released by the
end of 2001.
Table 2
Economic forces and convergence results
Const. -0.0071 -0.0009 -0.0173 -0.0077
(-0.62) (-0.06) (-0.70) (-0.15)
Degree of Openness 0.0013 0.0064 0.0007 -0.0331
(0.61) (1.38) (0.16) (-1.10)
Export intensity -0.0002 -0.0008 -0.0001 -0.0003
(-1.97) (-3.80) (-0.76) (-1.25)
Value-added content -0.0485 -0.0600 -0.0460 -0.1060
(-2.15) (-1.87) (-0.69) (-1.10)
Time trend 0.0018 0.0016 0.0025 0.0052
(2.47) (1.70) (1.40) (1.86)
Intra-industry trade -0.0049 -0.0012 0.0150 -0.0168
(-0.36) (0.94) (0.46) (-0.52)
Log Likelihood 1067.43 358.800 292.278 462.525
Likelihood test 10.74 12.91 2.14 16.75
of joint significance maintain reject maintain reject
Note: t-statistics in parenthesis; the critical value of a χ2(5) at the 5% level is 11.07.
Convergence speed seems to have no systematic relation, in general, with the economic
variables considered. Only for Portugal, the variable export intensity is associated with a
faster productivity convergence speed. That is, Portuguese industries relying more on
external markets have experienced higher relative productivity growth. No analogous
finding can be presented for the other two countries. Greek manufacturing shows no
convergence at all, on average, while Spanish convergence speed is faster than that of
Portuguese manufacturing, although we cannot relate it to export or import discipline.
It may seem that there is an apparent contradiction between these values and Figures 2 – 4.
The key to consistency is simple. The starting level of relative productivity is considerably
higher for Spain than for Portugal. This implies that the same growth rate results in a higher
absolute growth in relative productivity. Further, even the same absolute growth will cover
a larger fraction of the gap to average levels the smaller the gap is. A property of the model
is geometric convergence – the same convergence speed implies a smaller growth rate as a
country approaches the reference value. Thus, roughly similar growth rates means, for
countries with distinct initial relative productivities, different convergence speeds.
As to individual industries (see Table 3), Paper & Products have, on average, a higher
convergence speed than other industries, while Wearing Apparel converges slower to EU
average levels. Another interesting issue that can be addressed in this framework is whether
EU-accession has had any impact on the speed of productivity convergence. This is taken
up in the last column of Table 3. From it, we can readily conclude that manufacturing
productivity convergence was not boosted by membership of the European Union. This
holds true for the three Southern Europe countries.14
14
These inferences are robust to inclusion/exclusion of other variables. Details available upon request.
Table 3
Productivity convergence and European Union Membership

Const. -0.0127 -0.0236 -0.0147
(-1.08) (-2.51) (1.09)

Trend 0.0016 0.0014 0.0027
(2.12) (1.91) (1.79)
Degree of Openness 0.0003 0.0006
(0.09) (0.16)
Value-added content -0.0285 -0.0314
(-1.01) (-1.10)
Export intensity -0.0002 -0.0001 -0.0002
(-1.80) (-1.55) (-2.06)
Intra-industry trade -0.0042 0.0015 0.0006
(-0.33) (0.12) (0.04)
Greece 0.0143 0.0160 (0.0053)
(1.64) (1.81) (0.41)
Spain -0.0233 -0.0263 -0.0390
(1.70) (-1.95) (-1.73)
Wearing Apparel 0.0293 0.0269
(2.18) (1.96)
Paper & Products -0.0798 -0.0810
(-1.97) (-2.02)
EU Portugal -0.0198
(-1.35)
EU Greece 0.0181
(1.09)
EU Spain 0.0393
(1.54)
Log Likelihood 1071.26 1073.43 1076.31
Note: t-statistics in parenthesis.
A related problem is whether the Single Market Programme of 1992 resulted in faster speed
of convergence. Since our variable of interest is relative productivity, the impact of the
Single Market Programme on convergence speed is a priori ambiguous. It is expected that
it spurs productivity both at peripheral and reference (central) economies. It may well be,
on first impact, relatively more important in central economies, though over time, if
convergence exists, the effect should be negative. It turns out that no statistical significant
result is obtained related to the role of the Single Market Programme of 1992 in the
convergence of productivity.15
15
Estimates available from the author upon request.
Table 3
Industry convergence – Portugal

Const. 0.034 0.037
(1.31) (8.50)
1978 Relative Productivity -0.379 -0.226
(-0.30) (-0.24)
Degree of openness 1978 0.174
(0.43)
Export intensity 1978 -0.010
(-0.83)
Value-added content 1978 0.576
(0.02)
Relative importance of industry value added -0.026
(-0.22)
Intra-industry trade 1978 0.013
(0.67)
Log likelihood 70.223 68.760
Note: t-statistics in parenthesis
The main message is that, although there may exist convergence in relative productivity of
the overall manufacturing industry to European levels, the productivity of the individual
industries has evolved without any relation to the available aggregate economic variables.
As no precise picture emerges for convergence speed determinants, we now look at direct
differences in growth rates determinants (of relative labour productivity). We estimate the
following relationship:
 ∆Yijt = κ Yijt-1 + Xijt' β + νijt (3)
where xi ∈ X denotes economic determinants of growth rate differentials in labour

productivity, β is the vector of parameters to be estimated and νijt is a random term with the
usual properties. In the previous analysis, the determinants xi were assumed to influence κij,
while here we assume a common convergence factor but allow for distinct growth rates in
relative productivity across industries and over time.
This linear structure can be interpreted, in terms of convergence speed, as implying a
marginal effect decreasing in the initial value of relative productivity. That is, the same
increment in, say, export intensity has a stronger effect upon convergence speed for lower
levels of relative productivity.
Table 4
Economic forces and differentials in growth rates

Const. -0.0532 -0.0560 -0.0288 -0.0167
(-3.73) (-1.93) (-0.96) (-0.83)
Degree of Openness -0.0041 -0.0118 -0.0027 0.0060
(-1.07) (-1.96) (-0.60) (-0.66)
Export intensity 0.0003 0.0013 -0.0001 0.0002
(2.91) (3.60) (-0.26) (-1.84)
Value-added content 0.0978 0.1060 0.0250 0.0487
(3.55) (2.07) (0.36) (1.46)
Time trend -0.0014 -0.0015 -0.0013 -0.0014
(-2.05) (-1.24) (-1.05) (-1.31)
Intra-industry trade 0.0094 -0.0048 -0.0014 -0.0037
(0.82) (-0.23) (-0.06) (-0.21)
Relative productivity (t-1) -0.0294 -0.0318 -0.0325 -0.0428
(-5.38) (-2.78) (-2.60) (-3.52)
Log Likelihood 1082.43 360.767 303.082 458.547
Likelihood test 40.14 16.84 25.34 8.79
of joint significance reject reject reject maintain
Note: t-statistics in parenthesis; the critical value of a χ2(6) at the 5% level is 12.59
The results in Table 4 qualify in an important way the previous estimates, especially for
Portugal. For Greece and Spain, we find again no evidence of a systematic relationship
from economic variables to growth rates in relative productivity. Nonetheless, the
“convergence parameter” (the coefficient on the previous period relative productivity level)
is negative, statistically significant and roughly of the same magnitude in all three
countries.16 The main difference between Greece and Spain lies in the existence of a
negative evolution in growth rates common to all Greek industries.
More interesting are the economic effects that we find for Portuguese manufacturing,
Industries with a higher export intensity and/or a higher value-added content have had a
higher relative productivity growth. This implies a non-linear dependence of convergence
speed on these economic variables.
The estimates also indicate a weak, marginally statistically significant, positive relation
between openness to international trade and the growth rate in relative productivity. The
driving force behind this effect lies in the existence of a sector (other manufacturing, nec)
with a very high value for trade openness but with a small relative productivity growth.
Removal of these observations results in a non-significant relation, while the other
qualitative results remain unchanged.17
16
Roughly similar means that testing whether each point estimate is equal to any of the other two (country)
estimates does not reject the null hypothesis of equality.
17
Details available upon request from the author.
4. Productivity growth
It is now worthwhile to probe further into the absolute growth of value added per worker
and its differences across industries. The contribution of each industry to total
manufacturing value added per worker can be obtained as the sum of two elements: the
pure value added effect and the employment effect. The pure value added effects is
computed as the difference in value added per worker, weighted by the associated
employment share at the end of the period. The employment contribution is computed as
the change in the industry share (in total manufacturing employment), weighted by the
initial value added per worker. Formally, the change in manufacturing value added between
periods 0 and 1 can be written as:
 
L0j  VA1j VA0j  VA1j  L1j L0j 
VA1
L1
−
VA0
L0
= ∑ L  L
0 1
j
− 
L0j 
∑+  −
L1j  L1 L0 
 (4)
j j  
The first term is the pure value added effect while the second one is the employment
effect.18
Our interest lies in the cross-industry differences. Table 5 shows that five industries
concentrate roughly half of the contribution to overall growth in manufacturing
productivity.
18
To make the interpretation more natural, we normalise the change in value added to 100.
Table 5
Productivity growth in Portuguese manufacturing
Value Added Employment Total

contribution contribution contribution
1978/1996 1978/1996 1978/1996
Textiles 18.53% -5.33% 13.20%

Electrical Machinery 10.15% 2.14% 12.29%
Food 14.02% -3.22% 10.80%
Paper & Products 9.08% -0.19% 8.89%
Non-Metallic Products, nec 7.56% -0.79% 6.76%
Metal Products 3.87% 0.83% 4.69%
Beverages 4.15% 0.40% 4.55%
Printing & Publishing 3.86% 0.50% 4.36%
Wearing Apparel 2.36% 1.49% 3.85%
Iron & Steel 4.77% -1.42% 3.34%
Transport Equipment 3.64% -0.87% 2.77%
Footwear 0.43% 2.24% 2.67%
Wood Products 2.74% -0.43% 2.31%
Other Manufacturing 1.49% 0.76% 2.24%
Glass & Products 2.72% -0.51% 2.21%
Non-Electrical Machinery 2.24% -0.07% 2.17%
Pottery, China etc 0.88% 1.29% 2.17%
Other Chemicals 1.84% 0.26% 2.10%
Petroleum Refineries 2.00% -0.22% 1.78%
Plastic Products, nec 1.47% 0.21% 1.68%
Industrial Chemicals 3.71% -2.23% 1.48%
Rubber Products 1.23% -0.25% 0.98%
Non-Ferrous Metals 0.99% -0.12% 0.88%
Furnitures & Fixtures 0.46% 0.12% 0.58%
Professional Goods 0.28% 0.30% 0.58%
Leather & Products 0.74% -0.21% 0.52%
Tobacco 1.78% -1.63% 0.15%
Total 106.97% -6.97% 100.00%
It is also clear that growth in manufacturing productivity was not due to mere reshuffling of
economic activities. Most industries have positive pure value added effects, while
employment effects have been relatively minor. Another salient feature is that four out of
the main five contributing industries reduced their relative size in total manufacturing.
It is also interesting to look at the time evolution of these contribution effects. Table 6
presents the breakdown over three periods. The critical points selected (1986 and 1992)
coincide with EC accession and the Single Market Programme dates. These figures
highlight the period 1978 – 1996 was not an homogeneous one.
Table 6
Manufacturing productivity over time
Total contribution (%)

1996/1978 1996/1992 1992/1986 1986/1978
Textiles 13.20% -4.41% 11.54% 25.29%

Electrical Machinery 12.29% 18.86% 15.75% 5.33%
Food 10.80% 19.94% 10.12% 5.85%
Paper & Products 8.89% 7.59% 8.84% 9.71%
Non-Metallic Products, nec 6.76% 7.54% 7.79% 5.40%
Metal Products 4.69% 10.24% 2.48% 3.25%
Beverages 4.55% 3.80% 9.57% 0.67%
Printing & Publishing 4.36% 4.43% 4.12% 4.53%
Wearing Apparel 3.85% -1.66% 2.43% 8.40%
Iron & Steel 3.34% 1.78% 6.09% 1.92%
Transport Equipment 2.77% 3.46% 6.07% -0.49%
Footwear 2.67% 0.82% 1.78% 4.56%
Wood Products 2.31% -0.61% 5.11% 1.66%
Other Manufacturing 2.24% 1.61% 3.88% 1.22%
Glass & Products 2.21% 2.44% 2.56% 1.77%
Non-Electrical Machinery 2.17% 2.99% 1.95% 1.87%
Pottery, China etc 2.17% 2.42% 2.50% 1.73%
Other Chemicals 2.10% 3.00% -2.72% 5.71%
Petroleum Refineries 1.78% 2.91% 1.22% 1.59%
Plastic Products, nec 1.68% 6.85% -2.31% 1.99%
Industrial Chemicals 1.48% 1.80% -2.07% 4.34%
Rubber Products 0.98% 5.06% -1.28% 0.46%
Non-Ferrous Metals 0.88% 1.44% 1.09% 0.34%
Furniture & Fixtures 0.58% -0.15% 1.29% 0.42%
Professional Goods 0.58% 1.11% 0.75% 0.11%
Leather & Products 0.52% -0.14% 0.75% 0.73%
Tobacco 0.15% -3.09% 0.69% 1.64%
Average annual growth 3.11% 2.78% 3.06% 3.32%
First of all, growth of manufacturing productivity as a whole has been positive but
decreasing in pace.19 The main contributor industry, textiles, experienced a decreasing
relevance. In the last time period under review, its contribution to manufacturing
productivity growth even become negative, essentially due to the employment effect. On
the other hand, the electric machinery industry has been increasing in relative importance,
both through the pure value added effect and the employment effect. Also the food and the
metal products industries have gained relevance as contributors to overall manufacturing
productivity. At the EU level, Aiginger et al. (2001) classified food, beverages, leather,
wearing apparel, publishing and printing, as low research intensity and low productivity
growth industries. Thus, the predominance of such industries in Portuguese manufacturing
raises some concerns over the possibility of achieving a rapid convergence in labour
productivity.
19
The same finding is present in Aiginger et al. (2001), both for total manufacturing and for total economy
productivity growth, though they use slightly different time periods, and use data from EUROSTAT up to 2000
(projections for the last two years).
5. Final Remarks
In this paper, we addressed the issue of productivity convergence of EU peripheral

economies to average levels of a group of reference countries.
A question addressed was whether, or not, a higher degree of economic integration
motivated a higher convergence speed in labour productivity. Proxying the notion of higher
economic integration by the degree of openness, the export intensity, the degree of intra-
industry specialisation, and the realisation of the Single Market Programme and EU
accession for the three peripheral countries, we found no significant impact common to all
three countries. The only significant effects occurred in Portuguese manufacturing. Firms
that rely more heavily on export markets did had a faster convergence speed. The growth
rate of relative productivity was also higher in industries with a larger value-added content.
Import discipline and intra-industry trade (as a proxy for product differentiation and scale
economies) were non-significant determinants of productivity convergence.
The main message is that productivity convergence seems to be occurring in Portuguese
manufacturing. Export intensive industries seem to perform better, which corroborates the
usual view that international markets discipline favours productivity growth.
On the other hand, the major milestones in the European market construction had not, per
se, a separate impact. In the sense they promote international economic integration, the
European policies favoured productivity convergence.
We also explored the pattern of absolute productivity growth. Several findings are worth
pointing out. Manufacturing productivity has increased in the period under review but at a
decreasing rate. The relative weight of traditional industries, like textiles and food products,
is still high. Part of the productivity gains were obtained through employment contraction,
namely in the main contributor industries to manufacturing value added. Decomposing
labour productivity growth into value added and employment contributions shows the main
productivity growing industries are low research intensity ones, which raises some doubts
over the possibility of a faster convergence process at the manufacturing level. The
predominance of traditional industries in the growth of value added per worker in
manufacturing should be of some concern. The recent years, after the end of the period for
which we have comparable data, suggests that such worries may have materialised.
Several questions are still open. Namely, whether this pattern of convergence extends to
services, or not; the role of technological innovation as a driver of productivity growth; and
whether, or not, the more recent years have continued the identified trends. These questions
are left for future research.
References
Aiginger, K., 2001, Speed of change and growth of manufacturing, Chapter 4 in Peneder,
M. (coordinator), Structural change and economic growth, WIFO.
Aiginger, K., G. Hutschenreiter, H. Hollenstein, N. Knoll, H. Leo, and R. Wieser, 2001,
Innovation and productivity of European manufacturing, WIFO.
Alcalá, F. and A. Ciccone, 2001, Trade and productivity, CEPR Discussion Paper nº 3095,
December.
Amiti, M., 1999, Specialization patterns in Europe, Weltwirtschaftliches Archiv, 135: 573
– 593.
Brulhart, M., 1998, Trading places: industrial specialization in the European Union, Journal
of Common Market Studies, 36(3): 319 – 346.
Brulhart, M., 2000, Evolving geographical specialization of European manufacturing
industries, University of Lausanne.
Carree, M., L. Klomp and R. Thurik, 2000, Productivity convergence in OECD
manufacturing industries, Economics Letters, 66(3): 337-345.
Barro, R. and X. Sala-i-Martin, 1991, Convergence across states and regions, Brookings
Papers on Economic Activity 1, 107 - 182.
Barro, R. and X. Sala-i-Martin, 1992, Convergence, Journal of Political Economy, 100: 223
– 251.
Barros, P.P. and N. Garoupa, 1996, Portugal – European Union convergence: some
evidence, European Journal of Political Economy, 12: 545 – 553.
Ben-David, Dan, 1993, Equalizing exchange, trade liberalization and income convergence,
Quarterly Journal of Economics, 108: 653 – 679.
Bernard, A. and C. Jones, 1996a, Comparing apples to oranges: productivity convergence
and measurement across industries and countries, American Economic Review,
86: 1216 - 1238.
Bernard, A. and C. Jones, 1996b, Productivity across industries and countries: time series
theory and evidence, Review of Economics and Statistics, 78: 135 - 146.
Daveri, F., 2000, Is growth an information technology story in Europe too?, mimeo,
Università di Parma and IGIER.
De Nardis, S., A. Goglio, and M. Malgarini, 1996, Regional specialization and shocks in
Europe: some evidence from regional data, Weltwirtschaftliches Archiv, 132: 197
– 214.
Dollar, D., W.~Baumol and E.~Wolff, 1988, The factor price equalization model and
industry labor productivity: an empirical test across countries, in Empirical
methods for international trade, edited by R. Feenstra, MIT Press, Cambridge,
MA.
Esteban, J., 2000, Regional convergence in Europe and the industry mix: a shift-share
analysis, Regional Science and Urban Economics, 30: 353 - 364.
Frankel, J. and D. Romer (1999), Does Trade Cause Growth?, American Economic
Review, June, 89(3), pp. 379-99.
Freitas, M.L., 2000, Quantity and quality: the growth accounting in Ireland, Economic
Bulletin – Banco de Portugal, March, 59 – 70.
Godinho, M.M. e R. P. Mamede, 2000, Critical factors in the catching-up of European
LFRs: The Portuguese case, ISEG-UTL mimeo.
Gouyette, C. and S. Perelman, 1997, Productivity convergence in OECD service industries,

Structural Change and Economic Dynamics, 8: 279 - 295.
Helg, R., P. Manasse, T. Monocelli and R. Ravelli, 1995, How much (a)symmetryin
Europe? Evidence from industrial sectors, European Economic Review, 39: 1017
– 1041.
Irwin, D. and M. Tervio (2000), Does Trade Raise Income: Evidence from the Twentieth
Century, NBER Working Paper No 7745.
Jorgenson, D. and K. Stiroh, 1999, Information technology and growth, American
Economic Review, 89: 109 – 115.
Leamer, E. and J. Levinsohn, 1995, International trade theory: the evidence, in Handbook
of International Economics, Vol.III, edited by G. Grossman and K.~Rogoff,
Elsevier Science.
Miller, Stephen M. and Mukti P. Upadhyay (2000), The Effects of Openness, Trade
Orientation, and Human Capital on Total Factor Productivity, Journal of
Development Economics, December, 63(2): 399 – 423.
Neven, D. and C. Gouyette, 1995, Regional convergence in the European Community,
Journal of Common Market Studies, 33: 47 – 65.
OECD, 1998, The OECD STAN Database for Industrial Analysis.
Oliner, S. and D. Sichel, 2000, The resurgence of growth in the late 1990s: is information
technology the story?, Journal of Economic Perspectives.
Paci, R., 1997, More similar and less equal - Economic growth in the European regions,
Weltwirtschaftliches Archiv, 133: 609 – 634.
Pilat, D., 1998, What drives productivity growth?, The OECD Observer, nº 213.
Quah, D., 1994, Empirics for economic growth and convergence, European Economic
Review, 40: 1353 – 1375.
Samuelson, P., 1971, An exact Hume-Ricardo-Marshall model of international trade,
Journal of International Economics, 1: 1 – 18.
Schreyer, P., 2000, The contribution of information and communication technologies to
output growth: a study of the G7 countries, OECD - DSTI/DOC(2000)2.
Appendix
The following industries were considered:

Food
Beverages
Tobacco
Textiles
Wearing Apparel
Leather & Products
Footwear
Wood Products
Furniture & Fixtures
Paper & Products
Printing & Publishing
Industrial Chemicals
Other Chemicals
Petroleum Refineries
Rubber Products
Plastic Products, nec
Pottery, China etc
Glass & Products
Non-Metallic Products, nec
Iron & Steel
Non-Ferrous Metals
Metal Products
Non-Electrical Machinery
Electrical Machinery
Transport Equipment
Professional Goods
Other Manufacturing
Sessão 2 - Geografia Económica e Infra-estruturas
Session 2 - Economic Geography and Infrastructures

PUBLIC INVESTMENT IN TRANSPORTATION INFRASTRUCTURES
AND ECONOMIC PERFORMANCE IN PORTUGAL
Alfredo M. Pereira
Department of Economics
The College of William and Mary
Williamsburg, VA 23187
Email: ampere@wm.edu
Jorge M. Andraz
Faculdade de Economia
Universidade do Algarve, Campus de Gambelas
Faro, Portugal
Email: jandraz@ualg.pt
Abstract - This paper uses a VAR approach to investigate the effects of aggregate
and disaggregate measures of public investment in transportation infrastructures on
private investment, employment, and output in Portugal. Estimation results suggest that
public investment in transportation infrastructures crowds in private investment and
employment and, therefore, has a strong positive effect on output. Indeed, we estimate
that one euro invested in public investment increases output in the long-term by 9.5
euros. This figure suggests that public investment pays for itself 3.3 times in the form of
tax revenues over the life span of the public capital asset. Furthermore, this figure
corresponds to a rate of return of 15.9%, which is clearly higher than the rate of return
expected on private investment activities. A close look at the effects of different types of
public investment is very informative, since it shows which types of public investments
are the most productive. In terms of marginal productivities, the highest effects on
private investment come from public investment in ports, airports and national roads. In
terms of job creation, the highest effects come from public investment in ports, municipal
roads, and national roads. Finally, in terms of the effects on output the largest effects
come from investment in ports followed by national roads, municipal roads, airports,
and railroads. The results in this paper are very important from a public policy
perspective. This is because they suggest that public investment in transportation
infrastructures has been a powerful instrument to promote long-term growth and that the
strategy followed by the Portuguese authorities of investing in public infrastructures is
justified both from a long-term development perspective as well as from a public
budgetary perspective.
JEL Classification: C32, E62, H54, O52.

Keywords: infrastructures, economic performance, Portugal.
(*) This paper is part of a research project on the impact of infrastructure investment in the Portuguese economy
sponsored by the Fundação Luso-Americana/Portuguese-American Foundation. This paper was prepared in the
context of the conference on “Desenvolvimento Económico Português no Espaço Económico Europeu:
Determinantes e Políticas” organized by the Banco de Portugal/Bank of Portugal. We would like to thank an
anonymous referee for this conference for very useful comments and suggestions.
86 Alfredo Marvão Pereira, Jorge M. Andraz
I. Introduction
One of the nagging aspects in the Portuguese economic performance is the relative
backwardness of the Portuguese economy vis-à-vis the European Union partners. From the
1970s until around the late 1980s the Portuguese GDP per capita in purchasing power parity
was just approximately around 55% of the EU average. The magnitude and persistence of the
relative backwardness of the Portuguese economy has been explained by the lack of domestic
long-term growth fundamentals. Historically serious distortions in the financial markets lead to
lagging private investment while a narrow domestic tax base hindered the development of a
modern infrastructure. These difficulties justified the EU structural funds programs after 1989.
The cornerstone of these structural transfer programs was the development of a modern
transportation infrastructure network. Therefore, over the last decade, the strategy of
development in Portugal has been largely based on transportation infrastructure development.
Interestingly enough, despite the central role of public infrastructure development and the
intuitive knowledge of the relative scarcity of public infrastructures in Portugal, no information
was available on the actual impact of this development strategy. While the impact of
infrastructure development on private investment, employment and output has been assumed to
be positive and important, there has been complete ignorance as to what the actual effects might
be. In particular, no estimates exist of the rates of return on different types of infrastructure
investment. Therefore no information exists on relationship between the rates of return on
public investments and the rate of return on private investment projects. This information,
however, is crucial to determine the appropriateness of the development strategy followed in
Portugal.
The most important reason for the absence of estimates of the rates of return to public
infrastructure investment in Portugal, as indeed for most other countries of comparable or lower
levels of development, has been the absence of the most basic data on public investment itself.
This is because of the highly decentralized nature of the institutions in charge of the different
types of public investment as well as the constant shifting in jurisdictions in public investment
activities. In the case of Portugal, however, the problem of the absence of a data set has now
been solved. The authors concluded recently the construction of a detailed database of public
investment in transportation infrastructures, under the auspices of the Portuguese Ministry of
Planning [see Pereira and Andraz (2001)].
In this paper we estimate empirically the impact of infrastructure investment in the area of
transportation on aggregate economic performance in Portugal. We focus on aggregate public
investment as well as on different types of transportation infrastructure - roads and highways,
ports, airports, and railways - to evaluate the effects of such public investments on private
investment, employment and, ultimately, on output. We seek to estimate the marginal products
and the rates of return of public investment in different types of transportation infrastructures.
Although this paper focuses on the Portuguese case and deals with issues that are of great
importance for policy making in Portugal, its interest is not merely parochial. Indeed, the issue
of the effects of public investment on private sector performance has been at the center of the
policy debate in many countries, in many regions of the world. In particular, in the European
Union, the development strategy of the less development countries, like Greece, Ireland, and
Portugal, has been based largely on public investment projects. For these countries, public
investment on infrastructures, through EU structural programs, has been the instrument of
choice to induce real convergence of the domestic economy to the EU standards of living.
Public Investment in Transportation Infrastructures and Economic Performance in Portugal 87
Furthermore, in the near future, the eastward expansion of the EU will bring into the fold
countries with similar problems. For these Eastern European countries, economy recovery
seems to depend, in large scale, on the reconstruction of obsolete infrastructures. For these
countries joining the EU and, thereby, embarking in large public infrastructure projects seems
to be the expected vehicle for vanquishing their relative backwardness.
The empirical evaluation of the effects of public capital formation on private output was
brought to the limelight by the work of Aschauer (1989a, 1989b). Using a single-equation
static production function approach based on aggregate measures of public capital, Aschauer
(1989a, 1989b) suggests that public capital has been a powerful engine for growth in the United
States. In fact, his results suggest that public investment would pay for itself close to three
times in the form of additional tax revenues over the duration of the public capital assets [see
Reich (1991)]. Subsequent analysis applying the same methodology to regional and sector-
specific data in the United States as well as international data, however, failed to replicate such
large effects. Indeed, it often even failed to find meaningful positive results [see Gramlich
(1994) and Munnell (1992) for detailed surveys of the literature and Hulten and Schwab (1993)
for a detailed presentation on the infrastructure debate].
The work of Aschauer inspired an important body of literature on the impact of
infrastructure development for other countries. This includes contributions that are country-
specific and others that have a multi-country focus. In the first case one could mention, for
example, the work of Otto and Voss (1996) for Australia, Seitz (1994) for Germany, Sturm
and de Haan (1995) for Holland, Merriman (1990) for Japan, Shah (1992) for Mexico, Pereira
and Roca (1999) for Spain, Berndt and Hansson (1992) for Sweden, and Lynde and Richmond
(1993) for the UK. In the second case one could mention, for example, the work of Aschauer
(1989c), Evans and Karras (1993), Ford and Poret (1991), and Mittnik and Newman (1998),
all focusing on developed OECD countries. The magnitude and significance of the empirical
results varies greatly among countries. Furthermore, international comparisons are rendered
very difficult by the use in the literature of different measures of public capital, different levels
of aggregation, and different methodologies.
The approach used in Aschauer (1989a, 1989b) and much of the literature that followed
focuses on measuring the effects of public investment on private output using a single equation,
static production function approach. In this approach, private output is regressed on public
capital and private inputs – employment and capital. This approach has been criticized on
econometric grounds. It has been observed that the estimation of static, univariate production
functions in levels (or log-levels) is based on non-stationary variables. Therefore, OLS
estimates are spurious in the absence of cointegration. Moreover, OLS estimates suffer from
simultaneity bias. Even if this bias is corrected, conclusions about causality still cannot be
drawn. [See Jorgenson (1991) and Munnell (1992) for comprehensive discussion of these
econometric problems.]
In this paper, we follow Pereira (2000) and adopt a vector auto-regressive/error correction
mechanism approach. This multivariate time series approach allows us to address the
aforementioned econometric criticisms in a rigorous and comprehensive manner. It also brings
a more precise conceptual focus to the debate about whether or not public capital is productive.
In fact, the static single-equation framework, so often used in the literature, excludes the
presence of feedbacks, in particular dynamic feedbacks, among the relevant variables. This
exclusion is of paramount importance for it is likely that feedbacks exist. If they do, a zero
elasticity of private output with respect to public capital, as obtained from a single-equation
static production function approach, is neither a necessary nor a sufficient condition for public
investment to be ineffective in influencing output.
Dynamic feedbacks are essential to a conceptual understanding of the relationship between
public investment and aggregate economic performance. Indeed, public investment affects
output directly as an additional input in the production function. Moreover, as a positive
externality to aggregate production, public investment should, ceteris paribus, lead to higher
aggregate production. Public investment also affects aggregate production indirectly via its
effects on the use of private inputs, capital and labor. It is conceivable that a greater
availability of public capital could reduce the demand for private inputs (a substitution effect).
Higher availability of public capital, however, also increases the marginal productivity of
private inputs. This lowers the marginal costs of production, thereby potentially increasing the
level of aggregate production (a scale effect).
In turn, the evolution of private inputs and aggregate output can conceivably affect the
evolution of public investment. Indeed, increasing aggregate output provides the government
with a growing tax base and the potential for greater public investment. Furthermore, declining
employment has often led to short-term policy packages that involve increased public
investment. There is, therefore, a real possibility that reverse causality exists. By this we mean
that it is possible that the evolution of aggregate output and private inputs may be leading the
evolution of public investment.
This paper is organized as follows. In Section II, we present the data set used in our
analysis. In particular, we present in some detail the new public investment data for Portugal
(see also the Appendix). We also report preliminary empirical results including univariate and
cointegration analysis and report on the specification of the vector auto-regressive/error
correction mechanism models. In Section III, we introduce and discuss some methodological
issues in the identification and measurement of the effects of innovations in public investment.
In Section IV, we analyze the effects on economic performance - output, employment, and
private investment - of aggregate and disaggregate measures of public investment through the
use of orthogonalized impulse response functions. Finally, in Section V, we provide a summary
and some concluding remarks.
II. Data and Preliminary Empirical Results

A. Data: sources and description
We use annual data for the period 1976 to 1998. We consider output (gdp), employment
(emp), private investment (inv), in addition to public investment in transportation
infrastructures (pinv). The data on output, employment, and private investment in presented in
Table 1. This data was obtained from the Bank of Portugal/Banco de Portugal (1997),
Commission of the European Communities (1999), and Ministry of Finance/Ministério das
Finanças (2000). Output and private investment are measured in millions of constant 1995
Portuguese escudos while employment is measured in full-time equivalent employees.
The data for public investment in transportation infrastructures (pinv) is obtained from
Pereira and Andraz (2001). This database is the result of a long and meticulous investigation,
sponsored by the Portuguese Ministry of Planning. This database includes data on public
investment, both at current and constant 1995 prices deflated by the GDP as well as the private
investment deflators. It includes public investment in national roads, municipal roads,
highways, ports, airports, and railways. It covers the period from 1974 to 1998, despite some
failures of information regarding the two first years, due to lost data at the source. Since this
database has not been published before and is used in this article for the first time, it is provided
here in Table 2 and we discuss some of its main features below. For the same reason we also
included in an Appendix to this paper the executive summary of Pereira and Andraz (2001). All
of the data is in 1995 Portuguese escudos deflated using the GDP price deflator. The use of the
private investment deflator would lead to only marginal changes in the empirical results in this
paper.
To talk about the main features of the public investment data in Portugal one has
immediately to recognize the existence in the second half of the sample period of EU sponsored
structural transfer programs in the form of Community Support Frameworks for Portugal. The
first Community Support Framework program covered the period from 1989 to 1993 and the
second covered the period from 1994 to 1999. Therefore, our sample includes 13 years prior
the programs and 10 years of with the programs.
In what follows we consider an aggregate measure of public investment in transportation
infrastructures, as well as six disaggregated measures pertaining to public investment on roads,
ports, airports and railways. We present the evolution of each type of investment as a
percentage of the GDP and as a percentage of private investment in Tables 3 and 4,
respectfully, and in Figures 1-7. We present the evolution of the composition of public
investment in transportation infrastructures in Table 5 and Figure 8.
The first type of public investment (pinv1) is core infrastructure investment in national
roads. It averages 0.48% of the GDP for the sample period. It experiences a strongly increasing
trend during the sample period, from 0.34% of the GDP in the early years of the sample to
0.76% by the end of the sample period. The second type of public investment (pinv2) is core
infrastructure investment in municipal roads. It averages 0.40% of the GDP for the sample
period and shows less of a variation in that it averages 0.35% in the first part of the sample and
0.45% in the second. The third type of public investment (pinv3) is core infrastructure
investment in highways. It represents an average of 0.21% of the GDP over the sample period,
although the average in the early years is just 0.13% and in the second part of the sample is
0.32%. The fourth type of public investment (pinv4) is core infrastructure investment in ports.
It represents on average 0.12% of the GDP and has experienced a decline from 0.15% in the
1970s and 80s to about 0.08% in the last decade. The fifth type of public investment (pinv5) is
infrastructure investment in airports, and has remained stable over the sample period at about
0.05% of the GDP. Finally, the sixth type (pinv6) is core infrastructure investment in railways.
It averages 0.29% of the GDP for the period and it only shows an upward trend in the last few
years of the sample.
Overall, aggregate public investment (pinv) averages 1.55% of the GDP for the sample
period. It changes, however, from an average of 1.24% for 1976-88 to an average of 1.96% for
1989-1998. The data suggests that the increase through the 1990s in the overall figures is due
mostly to increases in public investment in national roads (pinv1) and highways (pinv3) and,
more recently, in railroad investment (pinv6).
All of the considerations above suggest that the data fully reflect the conventional wisdom
that the EU structural transfer programs brought a greater dynamism to the public investment in
infrastructures. They are also very informative about the effects of the EU Community Support
Frameworks in terms of the composition of public investment in transportation infrastructures.
In fact, core investment in national roads (pinv1) is one of the greatest beneficiaries of these
programs. Its share on total public investment increased from 27.4% in the 1980s to 33.7% in
the 1990s. Core investment in highways (pinv3) also increased its share, from 10.1% to 16.5%.
Core investment in railways (pinv6), was also positively affected by EU programs. During the
period between 1989 and 1998, it represents 19.3% of total public investment after having
accounted for about 17% until 1988. On the other side of the spectrum are the other types of
public investment whose shares decreased during the period covered by EU programs. The
share of public investment on municipal roads (pinv2) declined from 29.2% to 23.3%, while the
share of investment on airports (pinv5) declined from 3.7% to 2.9%. However, the greater
losses occurred in public investment on ports (pinv4). From a share of 12.1% in the period
before 1989, it represents only 2.9% of total public investment during the 1990s.
Besides the changes in magnitude and composition of public investment before and after the
Community Support Frameworks, it is also possible to detect some changes from the first
program (1989-93) and second (1994-98). The shares of public investment on national roads
(pinv1) and on railways (pinv6) show an increasing trend during the two structural programs.
Their shares to total public investment during the second program are higher than the average
share for the 1990s. The share of public investment on highways (pinv3), decreased from
17.4% to 15.7% to total public investment during the second program. Public investment on
municipal roads (pinv2) shows a continuous decreasing pattern during the 1990, from 29.2%
before the structural programs, to 25.9% and 20.6% during the first and the second programs,
respectively. Public investment on ports (pinv4), whose share to total was 12.1% in the period
until 1988, suffered a sharp decline during the first program to 5.3%. During the second
program, its share still decreased to 3.5%. Finally, the share of public investment on airports
(pinv5) decreased to 3.4% during the first program, and to 2.4% thereafter.
We conclude this discussion of the public investment data with some brief international
comparisons. International data comparisons are very difficult. This is mostly due to the fact
that the definition of the public investment data and its scope vary greatly across countries.
Furthermore, detailed disaggregated public investment data sets are not readily available for
most countries. Despite these cautionary words we provide a tentative comparison of aggregate
public investment in transportation infrastructures in Portugal, Spain and the United States. For
the Spanish data sources and specific definitions see Pereira and Roca (1999) and for the US
case see Pereira (2000). We present the evolution of public investment in transportation
infrastructures as a percentage of the GDP for these three countries in Figure 9.
There are two aspects that are worth mentioning. First, aggregate public investment in
transportation infrastructures in Portugal is of the same order of magnitude as in Spain. In both
countries public investment in transportation infrastructures tends to be somewhat above the
levels for the US, in particular after the late 1980s. This is an obvious implication of the EU
structural programs, which have been in effect for both countries since 1989. Second, the
upward trend that can be detected after the late 1980s in both Portugal and Spain is much less
pronounced in the Spanish case after 1993. This can also be explained by the characteristics of
the EU structural programs. Indeed, the EU structural programs for Spain became less
important after 1994, with the inception of the second Community Support Framework.
B. Univariate and cointegration analysis
We start by using the Augmented Dickey-Fuller (ADF) t-test to test the null hypothesis of a
unit root in the different variables. We use the Bayesian Information Criterion (BIC) to
determine the optimal number of lagged differences to be included in the regressions, and we
include deterministic components, a constant and/or a trend, in the regressions if they are
statistically significant.
The results of the ADF t-tests applied to the different variables in log-levels, are presented
in the top part of Table 6. In all the cases, the t-statistics are lower, in absolute levels, than the
5% critical values. Therefore, the ADF tests cannot reject the null hypothesis of a unit root in
these variables. In turn, the results of ADF t-tests applied to the first differences of the log-
levels, i.e., the growth rates of the original variables, are presented in bottom part of Table 6.
All critical values are greater, in absolute value, than the 5% critical value. Therefore, we can
reject the null hypothesis of unit roots in the growth rates of the variables. We take this
evidence as an indication that stationarity in first differences is a good approximation for all the
time series under consideration.
We also test the null hypothesis of a unit root in the different variables using the Phillips-
Perron test, which takes into consideration the possible existence of structural breaks in the
evolution of the variables. This is an important step since due to the different EU structural
programs structural breaks are likely to exist. We follow the same strategy as above in the
determination of optimal lags and deterministic components in the tests. The test results are
reported in Table 7. The results from the Phillips-Perron unit roots tests completely confirm
the previous unit root test results. Again the strong evidence is that stationarity in first
differences is a good approximation for all the time series under consideration.
It should be pointed out that this empirical evidence is consistent with the conventional
wisdom in the macroeconomics literature that aggregate public investment, output,
employment, and private investment are stationary in first differences. Although our public
investment series is more disaggregated, the same pattern of stationarity in first differences is
not surprising.
We now test for cointegration among output, employment, aggregated private investment,
and aggregated public investment as well as each one of the six public investment variables. We
use the standard Engle-Granger approach to test for cointegration. We have chosen this
procedure over the often-used Johansen approach for two reasons. First, since we do not have
any priors that suggest the possible existence of more than one cointegration relationship, the
Johansen approach is not strictly necessary. More importantly, however, for small samples,
Johansen's tests are known to induce strong bias in favor of finding cointegration when it does
not exist. [See, for example, Gonzalo and Lee (1998).] Therefore, our relatively small sample
size suggests that the standard Engle-Granger approach will lead to more accurate results.
Following the standard Engle-Granger approach, we perform four tests in each case. This is
because it is possible that one of the variables will enter the cointegrating relationship with a
statistically insignificant coefficient. We do not know, a priori, whether or not this will happen.
If it does happen, however, a test that uses such a variable as the endogenous variable will not
pick up the cointegration. Therefore, a different variable is endogenous in each of the four
tests. We apply the ADF t-test to the residuals from the regressions of each variable on the
remaining variables. In all of the tests, the optimal lag structure is chosen using the BIC, and a
deterministic component is included if it is statistically significant.
The results of the cointegration tests at the aggregate level are reported on the top part of
Table 8. The value of the t-statistics is lower, in absolute value, than the 5% critical values for
at least three of the four cases considered. Moreover, all the test statistics are lower, in absolute
value, than the 1% critical values. Thus, the ADF tests cannot reject the null hypothesis of a
random walk, and we cannot reject that the variables are not co-integrated at this aggregated
level.
Cointegration tests were also performed with aggregate output, employment and private
investment, together with each of the six different types of public investment. Results are also
reported in Table 8. For all six public investment variables, the value of the t-statistics is lower,
in absolute value, than the 5% critical values for all but four of the twenty-four cases
considered. We take this as strong evidence that, consistently with the results at the aggregate
level, the variables are not cointegrated at the more disaggregated level.
C. VAR specifications and estimates
We have now determined that all of the variables have the same order of integration and, in
particular, that they are stationary of first order. We have also determined that the variables do
not seem to be cointegrated, either at the aggregate level or at the more disaggregated level.
Accordingly, we follow the standard procedure in the literature and determine the specifications
of the VAR models using growth rates of the original variables (denoted by ggdp, gemp, ginv,
gpinv, etc).
We estimate seven VAR models. All VAR models include aggregate output, employment,
and private investment. In addition, each of the seven VAR models includes a different public
investment variable - one for aggregate public investment and one for each of the six different
types of public investment. This means that, consistently with our conceptual arguments, the
public investment variables are endogenous variables throughout the estimation procedure. For
the sake of brevity, the details on the model selection for the different VAR models are not
reported here. They are available from the authors upon request.
The specifications of the different VAR models are determined using the BIC. The test
results are reported in Table 9. The VAR specification has two dimensions, which were
determined jointly - the specification of the deterministic components and the consideration of
the possibility of structural breaks. In all cases a first order specification were selected. A
higher order was not considered due to relative small size of sample. The BIC selects a
specification with constant and trend for the disaggregated models for national roads, municipal
roads, highways, and ports. For the aggregate model, as well as the disaggregated models for
airports and for railways, the BIC selects a specification with only a constant.
In order to consider the possible structural changes due to the two Community Support
Frameworks, different VAR specifications were considered. One could possibly distinguish
three periods in which there might have been structural changes: the period before 1989, the
period of first program, i.e., 1989-93 and the period of the second program, i.e., 1994-98.
Therefore, we consider three alternatives in terms of the VAR specification. The first is the case
of no structural break/no dummies. The second is the case of one structural break/one dummy
distinguishing the periods before and after the EU structural programs. Finally, we consider the
possibility of two structural breaks/two dummies reflecting the possibility of the three different
periods mentioned above. We find that the BIC criterion leads to the selection of VAR with two
structural breaks/two dummies for aggregate public investment as well as for each one of the
six types of public investment. This suggests that in addition to considering the differences
before and after the EU structural programs, there are also important changes associated with
each of the EU structural programs.
III. Identifying and Measuring the Effects of Innovations in Public Investment
We use the impulse-response functions associated with the estimated VAR models to
examine the effects of the different types of public investment on the performance of output,
employment and private investment variables. In this context our methodology allows dynamic
feedbacks among the different variables to play a critical role. This is true in both the
identification of innovations in the public investment variables and the measurement of the
effects of such innovations.
A. Identifying innovations in the public investment variables
While the public investment variables are endogenous in our econometric framework, the
key methodological issue for the determination of the effects of public investment on the other
variables is the identification of innovations in the public investment variables that are truly
exogenous. This means that we need to identify shocks to public investment variables that are
not contemporaneously correlated with shocks in the remaining variables. These exogenous
shocks are not subject to the reverse causation problem. In dealing with this issue we draw
from the approach typically followed in the literature on the effects of monetary policy on the
economy [see, for example, Christiano, Eichenbaum and Evans (1996, 1998), and Rudebush
(1998).]
Ideally, the identification of shocks to public investment which are uncorrelated with shocks
in other variables would result from knowing what fraction of the government appropriations in
each period is due to purely non-economic reasons. The econometric counterpart to this idea is
to imagine a government policy function which relates the rate of growth of public investment
to the information in the relevant government information set; in our case, the past and current
observations of the growth rates of the output, employment and private investment variables.
The residuals from this policy function reflect the unexpected component to the evolution of
public investment and are uncorrelated with other innovations.
In the central case, we assume that the relevant information set for the government policy
function includes past values but not current values of the other variables. This is equivalent in
the context of the standard Choleski decomposition to assuming that innovations in public
investment lead innovations in the other variables. This means that we allow innovations in
public investment to affect the other variables contemporaneously, but not the reverse.
We have two reasons for making this our central case. First, it seems reasonable to believe
that the private sector reacts within a year to innovations in public investment decisions.
Second, it also seems reasonable to assume that the public sector is unable to adjust public
investment decisions to innovations in the private-sector variables within a year. This is due to
the time lags involved in information gathering and public decision-making. Nevertheless, to
determine the robustness of our central case results, we also consider all the possible
alternatives in terms of the definition of which observations of the private sector variables are
included in the government information set. This is equivalent to considering all the possible
orderings of the variables within the Choleski decomposition framework. We report the
corresponding range of results for the variance decomposition in Table 10 and for the impact
indicators in Table 11.
It should be pointed out that the sensitivity analysis efforts could conceivably be generalized
in two different directions. First, we could consider the effects of innovations in the private
sector variables, for example a supply shock, under our current sensitivity analysis framework.
To do so, however, would require a great deal of assumptions as to the ordering of the private
sector variables. Our approach has the advantage of providing a measure of the effects of
innovations in public investment variables on private sector variables that is independent of the
ordering of the private sector variables. We can, therefore, remain agnostic about the issue of
the order of these variables. Second, we could generalize the sensitivity analysis framework to
consider non-recursive or signal extraction schemes. This would reflect, however, an
econometric more than an economic concern. It would only be justified if we had less strong
priors about what the central case should be and it would entail alternatives of less clear
economic interpretation. Because of these reasons we have not pursued either path in this
paper.
B. Policy functions
The policy functions for aggregate public investment as well as the different types of public
investment are reported in Table 12. These policy functions relate the evolution of the public
investment variables to the evolution of the private sector variables lagged one year, according
to the selected VAR specification. For the aggregate model, there is no feedback from the
other variables to public investment. This means that public investment is truly an exogenous
variable.
It is interesting to note that the exogeneity of public investment in Portugal is in contrast
with the findings for the US, for example. In fact, Pereira (2000) shows that changes in public
investment in the US are positively correlated with lagged changes in output and negatively
correlated with lagged changes in employment. Therefore, changes in private-sector variables
affect the evolution of public investment in the US, which is not an exogenous variable. The
exogeneity of public investment decisions in Portugal, however, is easily explained by the fact
that for long public investment decisions have been closely related with the Portuguese
participation in the EU. This is particularly true after 1989, when the bulk of the public
investment in transportation infrastructures in Portugal has been conducted under the two
Community Support Frameworks. These programs are typically negotiated between the
recipient economies and the EU, focusing on long-term goals and deliberately avoiding short-
term considerations.
It should be pointed out that while public investment seems to be an exogenous variable at
the aggregate level, the aggregate results hide some important effects on the evolution of
different types of public investment. This means that although the magnitude of public
investment seems to be truly exogenous there may be some effects from the economy on the
composition of public investment.
In fact, the policy functions suggest that changes in the evolution of public investment in
national roads (pinv1) respond positively to changes in output while the evolution of public
investment in municipal roads (pinv2) depends positively on the evolution of private
investment. In turn, the evolution of investment in highways (pinv3) depends positively on the
evolution of employment and negatively on the evolution of private investment. Public
investment in ports (pinv4) depends positively on lagged changes in output and negatively on
lagged changes in private investment. Finally, the evolution of public investment in airports
(pinv5) and in railroads (pinv6) does not seem to respond to lagged economic performance.
C. The impulse-response functions
We consider the effects of one-percentage point, one-time random shocks in the rates of
growth of the different types of public investment on output, employment, and private
investment. We expect these temporary shocks in the growth rates of the different types of
public investment to have temporary effects on the growth rates of the other variables. They
will, however, have permanent effects on the levels of those variables. The accumulated
impulse response functions are reported in Figures 10-16.
There are a few interesting points worth mentioning in terms of these accumulated impulse-
response functions. Let us start by acknowledging that all accumulated impulse-response
functions converge within approximately a five-year period. This is not inconsistent with the
idea that public investment takes time to build before it really impacts the private sector
performance. This is because our measures of public investment are aggregate measures, which
are made of spending from a series of overlapping public investment projects. This being the
case, in any given year a substantial part of the observed public investment corresponds to
projects that have been concluded that year.
It should also be noted that the convergence path of the private sector variables is not only
relatively fast but also very smooth. In turn, the convergence path of the public investment
variables, although fast, is less smooth in the early years. This pattern can easily be understood
if one considers that the initial exogenous shock to public investment variables is followed by
an endogenous adjustment in public investment in response to the private sector variables. This
endogenous adjustment is dictated in the context of the VAR model by the policy functions
presented in Table 12 and discussed above.
These policy functions suggest a negative recursive pattern in the evolution of public
investment in addition to their response to private sector variables. This negative recursive
pattern dominates in the early years while the effects on the private sector variables are
relatively small. This explains the dip in the impulse-response function in the early years. In
later years, however, the positive feed back from the evolution of the private sector variables
seems to dominate, although in some cases it is not strong enough to bring the accumulated
long-term change in public investment to its initial level on impact. Hence, for aggregate
public investment (pinv), for example, the long-term change in public investment associated
with a 1.0 percentage point change on impact is 1.2 approximately, while for national roads
(pinv1) is about 1.0, and for municipal roads (pinv2) is about 0.8.
D. Measuring the effects of innovations in public investment variables
In this paper we estimate the long-term accumulated elasticities of the different variables
with respect to each type of public investment. Long-term is defined as the time horizon over
which the growth effects of innovations disappear, i.e., the accumulated impulse-response
functions converge. These elasticities represent the total percentage point changes in the
different variables for each long-term accumulated percentage point change in public
investment once all the dynamic feedback effects among the different variables have been
considered.
We report the long-term accumulated marginal productivity of public investment in terms of
the other variables in Tables 13 to 15. These figures measure the change, in million euros, in
output and private investment for every million euros accumulated change in public investment.
In Table 14, we report the effects in terms of the number of jobs created in the long-term per
one million euros in public investment.
We obtain the marginal product of public investment reported in Table 15, by multiplying
the output to public investment ratio for the last ten years by the elasticity of output with respect
to public investment. The choice of the output to public investment ratio for the last ten years is
designed to reflect the relative scarcity of public investment of the different types. We consider
the relative scarcity at the margin of the sample period without letting these ratios be overly
affected by business cycle factors or by the different priorities established by EU structural
programs.
It should be noted that we use the term marginal product in a way that departs from the
conventional definition of the word. This is because these elasticities and marginal products are
not based on ceteris paribus assumptions. In this paper, the term marginal product includes all
of the dynamic feedbacks among the variables. Therefore, the marginal product that we
calculate is a total marginal product. That is, it measures both the direct effects of public
investment on output, and the indirect effects of public investment on output through changes in
the evolution of inputs. Of course, this is the relevant concept from the standpoint of policy
making.
Finally, the annual rates of return, also reported on Table 15, are calculated from the
marginal product figures by assuming a life horizon of twenty years for all types of public
capital assets. That is, the rate of return applied to one euro over a twenty-year period yields
the value of the accumulated marginal product. These rates of return are adjusted to
accommodate for a public capital depreciation rate of 5%, which is implicit in the life horizon
of twenty years for the public capital assets.
IV. On the Economic Effects of Public Investment
A. Aggregate effects of public investment in transportation infrastructure
The effects on employment and private investment of public investment at the aggregate
level are reported on the top part of Tables 13 and 14. The results from the impulse response
analysis at the aggregate level suggest that in Portugal, public investment in transportation
infrastructure crowds in both private investment and employment. Indeed, when we estimate
the effects of shocks to aggregate public investment in transportation infrastructures on the
evolution of the other variables, we find that the elasticity of private investment with respect to
aggregate public investment is 0.639, which corresponds to a marginal product of 8.1. This
means that at the aggregate level, public investment crowds in private investment and that one
euro of additional public investment will induce, in the long-term, an accumulated total of 8.1
euros of private investment. In turn, the elasticity of employment with respect to aggregate
public investment is 0.079. This figure suggests that 230 additional jobs will be created in the
long-term for each additional one million euros in public investment in transportation
infrastructures.
In turn, the long-term effects of innovations in investment in transportation infrastructures
on output are reported on the top part of Table 15. We find that aggregate public investment
has a positive effect on output with an elasticity of 0.183, which corresponds to a marginal
product of 9.5. This implies that the increase of one euro in public investment leads to a total
accumulated increase of 9.5 euros in output.
One possible way of interpreting the marginal product figures is by calculating the
corresponding average rate of return. The estimated annual rate of return over a twenty-year
period of public investment in transportation infrastructures is 15.9%. This figure suggests that
the rate of return of public investment in transportation infrastructures is well above the range
one would expect for the rate of return on private investment. From this perspective, the
reliance on public investment in transportation infrastructures as the cornerstone of a
development strategy in Portugal seems to have been justified.
Another possible way of interpreting this figure is by calculating the value of the tax
revenues generated by this increase in output. Since tax revenues tend to hover around the 35%
of the GDP, then the marginal product of public investment in transportation infrastructures
suggests that over the life expectancy of the public capital assets, the public sector would
collect 3.3 euros. Therefore, the public sector collects an additional 3.3 euros in tax revenues
for each euro spent in public infrastructure. According to this evidence, the public investment
assets in transportation infrastructures pay for themselves over their life span and still generate
additional funds, which can be used for other public activities.
B. Effects of public investment in different types of transportation infrastructure
In the discussion above, we have established empirically that public investment in

transportation infrastructure makes a positive and significant contribution to private-sector
performance. We are ready to determine which types of public investment are the most
productive. The positive crowding in effects of public investment in transportation
infrastructures on private investment and employment observed at the aggregate level are also
present at the disaggregated level. All types of public infrastructure in transportation affect
private investment and employment positively in the long-term. Not surprisingly, the same
pattern can be found in terms of the effects on output.
The effects of public investment in transportation infrastructures on private investment are
reported in Table 13. In terms of the effects of public investment on private investment, the
strongest effect comes from public investment in national roads (pinv1) with an elasticity of
0.766. It is followed, by the investments in municipal roads (pinv2), ports (pinv4), and railways
(pinv6) with elasticities of 0.254, 0.281 and 0.264, respectively. Finally, public investment in
highways (pinv3) and in airports (pinv5) display the lowest elasticities, 0.110 and 0.079,
respectively.
In terms of marginal productivities, a better measure of relative scarcity, the highest

marginal effects on private investment come from public investment in ports (pinv4) and
airports (pinv5) with marginal products of 84.4 and 39.1 respectively. The marginal products
of public investment in national roads (pinv1), municipal roads (pinv2), and railroads (pinv6)
are still relatively large – 29.6, 14.1, and 18.8, respectively. The lowest effects on private
investment come from public investment in highways (pinv3) with a marginal product of 9.2.
The effects of public investment in transportation infrastructures on employment are
reported in Table 14. The strongest effect comes now from shocks to ports (pinv4) with an
elasticity of 0.070, followed by municipal roads (pinv2), with an elasticity of 0.054, national
roads (pinv1), with an elasticity of 0.045. In turn, the elasticities of public investment in
highways (pinv3) and railways (pinv6) are substantially smaller, 0.009 and 0.012, respectively.
Finally, the effect on employment of public investment in airports (pinv5) is only marginally
different from zero.
In terms of job creation, one million euros invested in ports (pinv4) will create, in the long-
term, about 4800 new jobs. This number reduces sharply to 692, 404, 204, and 164 new jobs
per million euros invested in municipal roads (pinv2), national roads (pinv1), railways (pinv6),
and highways (pinv3) respectively. Finally, public investment in airports (pinv5) actually
eliminates about 500 jobs per million euros.
The effects of public investment in transportation infrastructures on output are reported in
Table 15. The effects of shocks to the different public investment variables on output are all
positive. In terms of the long-term accumulated elasticities, the strongest effect comes from
shocks to public investment in national roads (pinv1) with an elasticity of 0.198. This is
followed by the effect of shocks in public investment in municipal roads (pinv2), with an
elasticity of 0.098, in ports (pinv4), with an elasticity of 0.087, and railways (pinv6), with an
elasticity of 0.062. In turn, the elasticities of output with respect to public investment in
highways (pinv3) and airports (pinv5) are the smallest, respectively 0.024 and 0.009.
Let us now consider the marginal product figures. These figures are a better measure of the
relative effects of different types of public investment. This is because they reflect the relative
scarcity of the different types of public investment at the margin of the sample period. The
marginal product figures suggest that all types of public investment are productive. Although
there is a wide range of effects, four of the six types of transportation infrastructure have
marginal products within a relatively small range, between 18.5 and 31.4. This is the case of
public investment in national roads (pinv1), municipal roads (pinv2), airports (pinv5) and
railroads (pinv6), with marginal products of 31.4, 21.3, 19.2, and 18.5, respectively. The two
extremes are given by public investment in highways (pinv3) with a marginal product of just
8.2 and public investment in ports (pinv4) with a marginal product of 107.1.
Another way of interpreting these results is by considering the rates of return on the
different types of public investment. Again, all rates of return for all different types of public
investment in infrastructures are above the expected ranges for private investment. Over a
twenty-year period, the average rate of return to public investment in ports (pinv4), is 30.8%,
and is the highest. It is closely followed by the rate of return to public investment in national
roads (pinv1) of 23.0%, municipal roads (pinv2) of 20.9%, airports (pinv5) of 20.0%, and
railroads (pinv6) of 19.7%. The lowest rate of return, although still high, is for public
investment in highways (pinv3) with 15.0%.
It is important to highlight the importance of considering both the direct and the indirect
effects of innovations in public investment. The explicit consideration of the indirect effects of
public investment on private investment and employment allows us to highlight the mechanisms
through which the different types of public investment tend to affect output. Indeed, the strong
effects of public investment in ports (pinv4) on output, seems to be related to strong effects on
both employment and private investment. The converse is true for public investment in
highways (pinv3) in which case the less strong effects seem to be related to less strong effects
on also both private investment and employment. In turn, the effects on output of public
investment in municipal roads seems to be due mostly to the important effects on employment
while the effects of public investment in airports (pinv5) seem to be mostly related to the effects
on private investment.
C. Comparison with the international evidence
The comparison of the results in this paper with the international evidence available in the
literature is not easy. This is primarily because the international literature has used a variety of
econometric techniques, which makes similar terms, like elasticity or marginal product not
always comparable with the way such terms are used in this paper. Also, most of the literature
on the effects of public infrastructures considers public investment as an exogenous variable
and focuses on the effects of public investment on private output and is not designed to address
the impact on private inputs. Furthermore, the definitions of public investment used in the
literature vary wildly.
Although comparisons are difficult they are not impossible. The results in this paper are
most directly comparable with the results in Pereira and Roca (1999) for Spain and in Pereira
(2000) for the US. Pereira and Roca (1999) consider for Spain the effects of public capital in
transportation infrastructures. The empirical results suggest a marginal product of private
investment with respect to public investment of 10.2 and that one million euros in public
investment create 129 jobs in the long-term. Moreover, the results indicate that the marginal
product of public investment in Spain is 5.5. This corresponds to a rate of return of 8.9%.
Accordingly, the results obtained in this paper for Portugal, 230 new jobs created per million
euros in public investment and a rate of return of 15.9%, tend to be higher than the ones for
Spain.
In turn, Pereira (2000) finds that public investment, although under a much broader
definition, crowds in private investment with a marginal product of 0.8 while it seems to have a
negligible effect on private employment. The results in this paper for Portugal show much
larger figures for the marginal effects of public investment in transportation infrastructures on
private investment – about ten times, while the effects on employment in Portugal are
substantial - 230 jobs per one million euros in public investment. More importantly, Pereira
(2000) suggests that the marginal product of public investment in the US is 4.5. This
corresponds to a rate of return of 7.8%, compared to a rate of return of 15.9% in the Portuguese
case. Again, the results in this paper tend to be substantially higher than the results for the US.
This is understandable given the relatively greater scarcity of public infrastructures in the
Portuguese economy.
International comparisons in terms of the disaggregated effects of different types of public
investment are even more difficult. Again, probably the closest comparisons can be made with
Pereira (2000). In Pereira (2000) there is a core infrastructure variable, which represents
highways and streets and that closely resembles the aggregate of national roads, municipal
roads, and highways. Another variable in Pereira (2000), core infrastructure in ports, airports,
etc, seems to be close to the aggregate of ports and airports in this paper. The results indicate
that the marginal products of these two types of public investment are 1.97 and 19.79,
respectively. The correspondent rates of return are 3.4% and 16.1%, respectively. In this paper,
the range of rates of return is from 15.0% to 23.0% for public investment in roads and
highways, and 30.8% and 20.0% for public investment in ports and airports, respectively.
Again, the figures for Portugal tend to be substantially higher than the ones for the US.
An important feature of the empirical results in this paper is that in Portugal public
investment in transportation infrastructures would more than pay for itself in the form of added
tax revenues over the life span of the public investment assets. This is reminiscent of the
supply-side Laffer-curve effect found for the United States by the early literature. Indeed, the
seminal contribution of Aschauer (1989a) has been interpreted as suggesting that [see, for
example, Reich (1991)]. This result was disputed by subsequent research for the United States
case. For example, Pereira (2000) suggests that the marginal product of public investment
would just pay for itself over time. Furthermore, Pereira and Roca (1999) show that the same is
true for Spain while the results in Mittnik and Newman (1998) for Canada, France, Germany,
Japan, The Netherlands, and the United Kingdom, in a time series context not incompatible
with the approach in this paper, seem to imply the same. Interestingly enough, however, the
same type of result seems to resurface in the case of Portugal. This leaves open the question as
to whether a supply-side Laffer-curve effect while not present in more developed economies
could be a fixture of less developed countries.
V. Summary and Concluding Remarks
This paper analyzes empirically the effects of public investment in transportation

infrastructure on economic performance in Portugal. To do so, we use a new data set on public
investment in transportation infrastructures in Portugal for the period 1976-98, recently
published by Pereira and Andraz (2001). We follow a VAR approach, which is consistent with
the argument that the analysis of the effects of public investment on output, employment and
private investment variables requires the consideration of dynamic feedback effects among the
different variables.
We can summarize the empirical results as follows. We find that in the long-term,
aggregate public investment in transportation infrastructures crowds in private investment as
well as employment. More importantly, we find that it has a positive effect on output. Indeed,
we estimate that one euro invested in public investment increases output in the long-term by 9.5
euros. This figure suggests that public investment pays for itself 3.3 times in the form of tax
revenues over the life span of the public capital asset. Furthermore, the marginal product figure
corresponds to a rate of return of 15.9%. This rate of return is clearly higher than the rate of
return expected on private investment activities.
The importance of the effects of public investment in transportation infrastructures at the
aggregate level opens the door to the next stage of our analysis: the study of the effects of
different types of public investment on economic performance. Consistent with the aggregate
results, we find that all types of public investment crowd in the other variables. Nevertheless, a
close look at the effects of different types of public investment on the remaining variables
suggests that the disaggregation of public investment is very informative, since it shows which
types of public investments are the most productive. In terms of marginal productivities, the
highest effects on private investment come from public investment in ports, airports and
national roads. In terms of job creation, the highest effects come from public investment in
ports, municipal roads, and national roads. Finally, in terms of the effects of output the largest
effects come from investment in ports followed closely by national roads, municipal roads,
airports, and railroads.
The results in this paper are very important and timely from a public policy perspective in
Portugal. From a retroactive perspective, the empirical evidence suggests strongly that public
investment in transportation infrastructures has been a powerful instrument to promote long-
term growth in Portugal. Moreover, it suggests that the strategy followed by the Portuguese
authorities of investing in public infrastructures has been justified both from a long-term
development perspective as well as from a public budgetary perspective.
More importantly, from a prospective perspective, the results in this paper provide broad
guidelines for the country's future development strategies. This is very important due to the still
relative backwardness now. As a matter of fact, Portuguese GDP is still at about 75% of EU
average while sources of outside financing are being reduced and the country faces a great
budgetary restraint in the context of EMU. This requires greater attention to relative benefits
and much more fine-tuned development policies. It is also important to highlight the fact that
given current budgetary constraints in the context of the Stability and Growth Programs, the
tendency for achieving budgetary consolidation through reduction in public investment is a
mistake from the standpoint of long-term growth. It is also a mistake from the standpoint of
long-term budgetary situation.
Although the results in this paper are important from the perspective of policy making in
Portugal, its interest is far from parochial. In fact, there is a number of Eastern European
waiting to join the EU. These countries have levels of development and infrastructure
scarcities that are not unlike the Portuguese case by the end of 1980s. Furthermore, there are
already structural transfer programs in place to smooth the transition of these countries into the
EU and they are expected to benefit from large EU structural funds upon accession, much like
Greece, Ireland, Portugal, and Spain currently do. From this paper we learn that the general
strategy of investing in public infrastructure may be very effective in promoting real
convergence of these economies to EU standards. Furthermore, given the difficulties of data
gathering one would encourage data collection and coordination of policies and implementation
agencies from early stages is critical to provide info to help design basic programs.
Despite all the considerations above it is appropriate to conclude on a cautionary note.
Although we have established empirically the importance of public investment in transportation
infrastructure for economic development in Portugal, we have done so with a relatively small
data set. This places some limitations on the statistical analysis in the paper. More
importantly, maybe, is the fact that establishing that public investment has been important in the
past does not establish that it will be important in the future. Indeed, one could easily
conjecture a pattern of decreasing marginal returns to public investment. Finally, even if we
could legitimately conjecture, based on the relatively high rates of return we estimated, that
these public investments will continue to be important, we did not address the issue of which
types of investment are the most important. Indeed, just showing that public investment in
infrastructures is productive does not mean that it is more productive than private investment or
investment in human capital, for example.
References
Aschauer, D. (1989a): “Is Public Expenditure Productive?, ” Journal of Monetary Economics,

23, pp. 177-200.
_________ (1989b): “Does Public Capital Crowd Out Private Capital?,” Journal of Monetary
Economics, 24, pp. 171-188.
_________ (1989c): “Public Investment and Productivity Growth in the Group of Seven,”
Economic Perspectives, 13(5), pp. 17-25.
Banco de Portugal (1997) Séries Longas para a Economia Portuguesa, Lisboa, Portugal.
Bernd, E. and B. Hansson (1992). "Measuring the Contribution of Public Infrastructure in
Sweden," Scandinavian Journal of Economics 94, pp. 151-168.
Christiano, L., M. Eichenbaum, C. Evans (1996): “The Effects of Monetary Policy Shocks:
Evidence from the Flow of Funds,” Review of Economics and Statistics 78(1), pp. 16-
34.
_________ (1998): “Monetary Policy Shocks: What Have we Learned and to What End?,”
NBER Working Paper 6400.
Commission of the European Communities (1999): European Economy No 69 - Statistical
Annex, Brussels, Belgium.
Evans and Karras (1993): “Is Government Capital Productive? Evidence From a Panel of
Seven Countries,” Journal of Macroeconomics, 16(2), pp. 271-279.
Ford, R. and P. Poret (1991). "Infrastructure and Private Sector Productivity," OECD
Economic Studies 17, pp. 63-89.
Gonzalo, J, and T. Lee (1998). "Pitfalls in Testing for Long-Run Relationships," Journal of
Econometrics 86, pp. 129 - 154.
Gramlich, E. (1994): “Infrastructure Investment: A Review Essay,” Journal of Economic
Literature, XXXII(3), pp. 1177 - 1196.
Hulten C. and R. Schwab (1993): “Infrastructure Spending: Where Do We Go From Here?,”
National Tax Journal, XLVI, pp. 261 - 274.
Jorgenson, D. (1991): “Fragile Statistical Foundations,” The Public's Capital, pp. 6-7.
Lynde and Richmond (1993): “Public Capital and Long-Run Costs in U.K. Manufacturing,”
The Economic Journal, 103, pp. 880-893.
Merriman, D. (1990): “Public Capital and Regional Output: Another Look at Some Japanese
and American Data,” Regional Science and Urban Economics, 20, pp. 437-458.
Ministério das Finanças (2000), The Portuguese Economy, Direcção Geral de Estudos e
Previsão, Lisbon, Portugal.
Mittnik, S, and T. Newman (1998): “Dynamic Effects of Public Investment,” Institute of
Statistics and Econometrics Christian Albrechts, University at Kiel, Olshauserstr,
Germany.
Munnell, A. (1992): “Infrastructure Investment and Economic Growth,” Journal of Economic
Perspectives, 6, pp. 189 -198.
Otto, G. and G. Voss (1996): “Public Capital and Private Production in Australia,” Southern
Economic Journal, 62(3), pp. 723-738.
Pereira, A. (2000): “Is All Public Capital Created Equal?,” Review of Economics and Statistics,
82 (3), pp. 513-518.
Pereira A. and J. Andraz (2001): “Investimento Público em Infra-estruturas de Transporte em
Portugal Continental,” Ministério do Planeamento, Portugal.
Pereira, A. and O. Roca (1999): “Public Capital Formation and Regional Development in
Spain,” Review of Development Economics, 3(3), pp. 281-294.
Reich, R. (1991): “The Real Economy,” The Atlantic Monthly, pp. 35-52.
Rudebusch, G. D. (1998): “Do Measures of Monetary Policy in a VAR Make Sense?,”
International Economic Review, 39(4), pp.907-931.
Seitz, H. (1994): “Public Capital and the Demand for Private Inputs,” Journal of Public
Economics, 54, pp. 287-307.
Shah, A. (1992): “Dynamics of Public Infrastructure, Industrial Productivity and Profitability,”
The Review of Economics and Statistics, LXXIV(1), pp. 28-36.
Sturm, J. and J. de Haan (1995). Is Public Expenditure Really Productive? New Evidence for
the USA and the Netherlands," Economic Modeling 12, pp. 60-72.
Table 1
Data set for Portugal
Years Output Employment Private investment Public investment
1976 9.023.294 3.624 2.033.225 113.212

1977 9.519.412 3.671 2.265.797 103.305
1978 9.788.215 3.770 2.427.918 123.781
1979 10.341.059 3.862 2.376.059 134.398
1980 10.813.624 3.943 2.577.030 122.724
1981 10.991.600 3.939 2.728.533 159.585
1982 11.224.467 3.965 2.770.359 162.976
1983 11.205.164 3.878 2.579.592 137.623
1984 10.991.821 3.937 2.137.447 114.935
1985 11.305.010 3.932 2.071.879 127.688
1986 11.768.296 3.900 2.293.706 144.316
1987 12.518.244 4.006 2.677.251 155.751
1988 13.451.070 4.096 3.102.701 178.492
1989 14.144.498 4.236 3.236.613 198.425
1990 14.759.562 4.279 3.501.980 247.907
1991 15.104.553 4.335 3.624.977 295.401
1992 15.483.749 4.359 3.801.235 269.922
1993 15.311.404 4.295 3.585.621 306.278
1994 15.651.492 4.449 3.701.556 324.773
1995 16.102.000 4.416 3.880.582 311.933
1996 16.584.869 4.445 4.067.238 335.804
1997 17.260.447 4.530 4.545.614 433.520
1998 17.866.194 4.740 4.952.094 404.936
Units: Output, private investment, and public investment – millions of 1995 escudos.
Labor – thousand workers.
Sources: Output, employment and private investment: the Bank of Portugal/Banco de Portugal (1997), Commission
of the European Communities (1999), and Ministry of Finance/Ministério das Finanças (2000)
Public investment: Pereira and Andraz (2001).
Table 2
Public investment in transportation infrastructures
Years Aggregate National Municipal Highways Ports Airports Railways

public
roads roads
investment
1976 113.212 28.538 20.900 16.216 7.110 3.365 37.083

1977 103.305 20.773 24.829 12.310 8.476 2.595 34.322
1978 123.781 26.976 33.694 16.570 13.350 1.535 31.657
1979 134.398 30.532 49.580 9.823 14.878 2.411 27.174
1980 122.724 38.032 29.789 16.553 24.567 1.359 12.424
1981 159.585 41.197 49.274 25.381 25.381 3.442 14.911
1982 162.976 40.038 50.851 19.051 24.853 6.567 21.616
1983 137.623 43.451 41.734 9.760 21.032 5.078 16.568
1984 114.935 32.908 43.160 1.283 17.412 6.839 13.333
1985 127.688 35.674 43.418 11.159 15.543 4.469 17.425
1986 144.316 45.205 39.258 10.393 16.489 10.473 22.497
1987 155.751 52.960 43.753 13.288 12.834 8.702 24.215
1988 178.492 55.360 51.542 18.243 12.324 11.686 29.336
1989 198.425 56.892 57.549 35.603 10.185 11.041 27.156
1990 247.907 78.109 56.034 56.980 13.917 10.923 31.945
1991 295.401 84.248 67.958 67.331 18.409 8.936 48.518
1992 269.922 100.680 74.192 24.851 13.544 5.718 50.937
1993 306.278 100.777 83.302 43.571 13.326 5.207 60.095
1994 324.773 119.079 66.098 59.462 12.297 5.336 62.500
1995 311.933 128.160 54.440 57.847 12.730 6.146 52.610
1996 335.804 139.065 71.566 37.089 9.983 8.126 69.975
1997 433.520 126.036 90.559 75.553 12.059 10.502 118.811
1998 404.936 118.974 93.988 52.518 16.675 14.217 108.563
Units: Millions of 1995 escudos.

Source: Pereira and Andraz (2001).
Table 3
Public investment as a share of GDP (%)
Averages
Public investment 1976-88 1989-98 Sample
1976-80 1981-85 1986-88 1989-93 1994-98
Average
pinv: Aggregate public 1.21 1.26 1.27 1.76 2.16 1.24 1.96 1.55
investment
pinv1: National roads 0.29 0.35 0.41 0.56 0.76 0.34 0.66 0.48
pinv2: Municipal roads 0.32 0.41 0.36 0.45 0.45 0.36 0.45 0.40
pinv3: Highways 0.15 0.12 0.11 0.31 0.34 0.13 0.32 0.21
pinv4: Ports 0.14 0.19 0.11 0.09 0.08 0.15 0.08 0.12
pinv5: Airports 0.02 0.05 0.08 0.06 0.05 0.05 0.05 0.05
pinv6: Railways 0.30 0.15 0.20 0.29 0.49 0.22 0.39 0.29
Table 4
Public investment as a share of Private Investment (%)
Averages
Public investment 1976-88 1989-98 Sample
1976-80 1981-85 1986-88 1989-93 1994-98 Average
pinv: Aggregate public 5.13 5.72 5.95 7.40 8.56 5.55 7.98 6.60
investment
pinv3: Highways 0.62 0.52 0.51 1.29 1.35 0.55 1.32 0.89
pinv4: Ports 0.57 0.84 0.53 0.39 0.30 0.67 0.35 0.53
pinv5: Airports 0.10 0.22 0.39 0.24 0.20 0.21 0.22 0.22
pinv6: Railways 1.25 0.69 0.94 1.22 1.91 0.96 1.57 1.23
Table 5
Shares of total public investment (%)
Averages
Public investment 1976-80 1981-85 1986-88 1989-93 1994-98 1976-88 1989-98 Sample
Average
pinv3: Highways 12.1 8.9 8.7 17.4 15.7 10.1 16.5 12.9
pinv4: Ports 11.3 14.8 8.9 5.3 3.5 12.1 4.4 8.7
pinv5: Airports 1.9 3.9 6.5 3.4 2.4 3.7 2.9 3.4
pinv6: Railways 24.4 12.0 15.9 16.3 22.2 17.6 19.3 18.3
Figure 1: Aggregate public investment in transportation infrastructures

(pinv)
as % of GDP and private investment
3.00 12.00
2.50 10.00
(% of Private Investment)
2.00 8.00
(% of GDP)
1.50 6.00
1.00 4.00
0.50 2.00
0.00 0.00
76
78
80
82
84
86
88
90
92
94
96
98
19
19
19
19
19
19
19
19
19
19
19
19
Years
Share of Private Investment Share of GDP

Figure 2: Public investment in national roads (pinv1)

0.90 4.00
0.80 3.50
0.70 3.00
0.60
(% of GDP)
2.50
0.50
2.00
0.40
1.50
0.30
0.20 1.00
0.10 0.50
0.00 0.00
76
78
80
82
84
86
88
90
92
94
96
98
19
19
19
19
19
19
19
19
19
19
19
19
Years
Share of GDP Share of Private Investment
Figure 3: Public investment in municipal roads (pinv2)

0.60 2.50
0.50 (% of Private Investment)

2.00
0.40
(% of GDP)
1.50
0.30
1.00
0.20
0.10 0.50
0.00 0.00
76
78
80
82
84
86
88
90
92
94
96
98
19
19
19
19
19
19
19
19
19
19
19
19
Years

Figure 4: Public investment in highways (pinv3)

0.50 2.00
0.45 1.80
0.40 1.60
0.35 1.40
(% of GDP)
0.30 1.20
0.25 1.00
0.20 0.80
0.15 0.60
0.10 0.40
0.05 0.20
0.00 0.00
76
78
80
82
84
86
88
90
92
94
96
98
19
19
19
19
19
19
19
19
19
19
19
19
Years
Figure 5: Public investment in ports (pinv4)

0.25 1.20
1.00 (% of Private Investment)

0.20
0.80
(% of GDP)
0.15
0.60
0.10
0.40
0.05 0.20
0.00 0.00
76
78
80
82
84
86
88
90
92
94
96
98
19
19
19
19
19
19
19
19
19
19
19
19
Years

Figure 6: Public investment in airports (pinv5)

0.10 0.50
0.09 0.45
0.08 0.40
0.07 0.35
(% of GDP)
0.06 0.30
0.05 0.25
0.04 0.20
0.03 0.15
0.02 0.10
0.01 0.05
0.00 0.00
76
78
80
82
84
86
88
90
92
94
96
98
19
19
19
19
19
19
19
19
19
19
19
19
Years
Figure 7: Public investment in railways (pinv6)

0.80 3.00
0.70
2.50
0.60
2.00
(% of GDP)
0.50
0.40 1.50
0.30
1.00
0.20
0.50
0.10
0.00 0.00
76
78
80
82
84
86
88
90
92
94
96
98
19
19
19
19
19
19
19
19
19
19
19
19
Years

Figure 8: Composition of public investment in transportation infrastructures
1.00
0.80
0.60
0.40
0.20
0.00
1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
Years
National roads Municipal roads Highways Ports Airports Railw
Figure 9: Aggregate public investment in transportation infrastructures

as % of GDP in Portugal, Spain and in the U.S.A.
3.00
2.50
2.00
(%)
1.50
1.00
0.50
0.00
76
78
80
82
84
86
88
90
92
94
96
98
19
19
19
19
19
19
19
19
19
19
19
19
Years
Spain U.S.A. Portugal

Table 6
Testing for the null hypothesis of unit roots using the ADF test
Deterministic Order Test statistic Critical values

components (BIC) 5% 1%
Variables: log levels
gdp: Output CT 2 -3.0118 -3.60 -4.38

emp: Employment CT 0 -1.6349 -3.60 -4.38
inv: Private investment CT 1 -2.0732 -3.60 -4.38
pinv: Aggregate public investment CT 0 -1.8699 -3.60 -4.38
pinv1: National roads CT 0 -1.9461 -3.60 -4.38
pinv2 : Municipal roads CT 1 -3.2594 -3.60 -4.38
pinv3: Highways CT 0 -3.4357 -3.60 -4.38
pinv4: Ports C 0 -2.1310 -3.00 -3.75
pinv5: Airports CT 2 -2.7925 -3.60 -4.38
pinv6. Railways CT 2 -0.9072 -3.60 -4.38
Variables: growth rates
gdp: Output C 3 -4.6470 -3.00 -3.75

emp: Employment C 0 -3.9231 -3.00 -3.75
inv: Private investment N 0 -2.2950 -1.95 -2.66
pinv: Aggregate public investment C 0 -4.8404 -3.00 -3.75
pinv1: National roads C 0 -5.5086 -3.00 -3.75
pinv2 : Municipal roads C 1 -4.1368 -3.00 -3.75
pinv3: Highways N 0 -4.9659 -1.95 -2.66
pinv4: Ports N 0 -3.3923 -1.95 -2.66
pinv5: Airports N 0 -4.7955 -1.95 -2.66
pinv6. Railways CT 1 -5.8943 -3.60 -4.38
Table 7
Testing for the null hypothesis of unit roots using the Phillips-Perron test
Deterministic Order Test statistic Critical values

components (BIC) 5% 1%
Variables: log levels
gdp: Output C 1 -0.1828 -12.5 -17.2

emp: Employment CT 0 -8.3216 -17.9 -22.5
pinv1: National roads CT 0 -11.7611 -17.9 -22.5
pinv2 : Municipal roads CT 1 -17.4691 -17.9 -22.5
pinv3: Highways CT 0 -13.6992 -17.9 -22.5
pinv4: Ports C 0 -8.3041 -12.5 -17.2
pinv5: Airports CT 2 -9.1929 -17.9 -22.5
pinv6. Railways CT 2 -6.0600 -17.9 -22.5
Variables: growth rates
gdp: Output C 0 -10.1261 -12.5 -17.2

emp: Employment C 0 -21.0370 -12.5 -17.2
inv: Private investment N 0 -8.4164 -7.3 -11.9
pinv: Aggregate public investment C 0 -23.0084 -12.5 -17.2
pinv1: National roads C 0 -23.3204 -12.5 -17.2
pinv3: Highways N 0 -23.1967 -7.3 -11.9
pinv4: Ports N 0 -16.1431 -7.3 -11.9
pinv5: Airports N 0 -22.5798 -7.3 -11.9
pinv6. Railways CT 1 -113.4204 -17.9 -22.5
Table 8
Testing the null hypothesis of no cointegration
Variables Deterministic Optimal Lag Test Statistic Critical Values

Components (BIC) 5% 1%
gdp: Output CT 0 -1.4347 -4.16 -4.65

emp: Employment C 0 -3.9402 -4.11 -4.73
gdp: Output CT 0 -1.3356 -4.16 -4.65

emp: Employment C 0 -3.9385 -4.11 -4.73
pinv1: National roads N 2 -3.5147 -3.74 -4.30
gdp: Output CT 0 -1.4802 -4.16 -4.65

emp: Employment C 0 -4.3649 -4.11 -4.73
gdp: Output CT 0 -2.1859 -4.16 -4.65

emp: Employment C 0 -4.0019 -4.11 -4.73
pinv3: Highways CT 1 -5.3735 -4.16 -4.65
gdp: Output CT 0 -1.7135 -4.16 -4.65

emp: Employment C 2 -5.2319 -4.11 -4.73
pinv4: Ports C 0 -3.3592 -4.11 -4.73
gdp: Output CT 0 -1.3165 -4.16 -4.65

emp: Employment C 0 -4.0143 -4.11 -4.73
pinv5: Airports N 2 -2.9691 -3.74 -4.70
gdp: Output CT 0 -2.0721 -4.16 -4.65

emp: Employment C 0 -4.0632 -4.11 -4.73
pinv6. Railways CT 0 -3.7113 -4.16 -4.65
Table 9
VAR specification (BIC)
Public investment Model Deterministic No dummy One dummy Two dummies

order components (1989) (1989,1994)
pinv: Aggregate public investment 1 N -24.79172 -24.89442 -25.11975

1 C -25.05496 -25.46241 -25.69166
1 CT -25.09007 -25.58873 -25.69035
pinv1: National roads 1 N -24.85156 -24.94396 -25.52282

1 C -25.07617 -25.52986 -26.10160
1 CT -25.24993 -26.45380 -26.77999
pinv2 : Municipal roads 1 N -23.64226 -23.68089 -23.73939

1 C -23.92128 -24.30213 -24.36215
1 CT -24.11397 -24.51930 -24.59061
pinv3: Highways 1 N -20.04118 -20.08434 -20.13513

1 C -20.32396 -20.75705 -20.81506
1 CT -20.37029 -20.87376 -21.01449
pinv4: Ports 1 N -23.77836 -23.84164 -23.89150

1 C -24.06601 -24.45286 -24.51007
1 CT -24.31672 -24.93027 -25.28925
pinv5: Airports 1 N -21.64709 -21.68944 -21.81481

1 C -21.96905 -22.52255 -22.66662
1 CT -22.11673 -22.50320 -22.62232
pinv6. Railways 1 N -22.95706 -23.30818 -23.35772

1 C -23.47973 -23.91985 -23.97603
1 CT -23.57522 -23.87059 -23.92901
NB: In bold face is the selected specification.

Table 10
Variance decomposition: percentage of long-term variation in the variables
due to variations in public investment
Public
Variable Output Employment Investment
Investment
pinv: Aggregate
public investment
central case 37.6% 18.7% 37.2% 88.7%

range of variation [9.8%;37.6%] [8.8%;19.2%] [6.9%;37.2%] [69.2%;88.7%]
pinv1: National
roads
central case 33.5% 3.3% 43.9% 76.7%

pinv2: Municipal
roads
central case 11.2% 11.0% 5.8% 64.1%

pinv3: Highways
central case 7.8% 3.9% 17.0% 53.7%

pinv4: Ports
central case 16.7% 32.1% 12.8% 65.7%

pinv5: Airports
central case 0.6% 7.1% 6.1% 90.5%

pinv6: Railways
central case 12.4% 10.8% 22.3% 90.5%

Table 11
Long-term accumulated elasticities of private sector variables
with respect to public investment
Variable Output Employment Investment
pinv: Aggregate public

investment
central case 0.18264 0.07860 0.63871

range of variation [0.105;0.183] [0.045;0.079] [0.356;0.639]
pinv1: National roads
central case 0.19807 0.04524 0.76549

range of variation [-0.133;0.202] [-0.206;0.049] [-0.259;0.772]
pinv2: Municipal roads
central case 0.09839 0.05441 0.25396

pinv3: Highways
central case 0.02416 0.00865 0.11013

pinv4: Ports
central case 0.08736 0.07025 0.28102

range of variation [-0.057;0.087] [0.005;0.070] [-0.075;0.281]
pinv5: Airports
central case 0.00937 -0.00438 0.07858

range of variation [-0.014;0.030] [-0.005;0.009] [0.002;0.137]
pinv6: Railways
central case 0.06247 0.01221 0.26418

range of variation [0.014;0.090] [-0.010;0.031] [0.080;0.341]
Table 12
Policy functions for different types of public investment
GPINV Constant Trend D1989 D1994 GGDP(- GEMP(- GINV(- GPINV(-

1) 1) 1) 1)
pinv: Aggregate 0.01652 ---- 0.08141 0.00836 1.06899 0.91162 0.43275 -0.31825
public investment
(0.22) (1.02) (0.10) (0.38) (0.37) (0.55) (-1.18)
pinv1: National 0.08439 -0.01174 0.20328 0.08463 3.98963 0.98707 -0.78891 -0.18631
roads
(0.79) (-1.10) (1.76)** (0.53) (1.59)* (0.43) (-1.16) (-0.96)
pinv2: Municipal 0.43479 -0.02938 0.16555 0.28084 -5.38487 -0.20668 2.93549 -0.34957
roads
** * **
(2.37) (-1.62) (0.86) (1.04) (-1.28) (-0.05) (2.63) (-1.72)
pinv3: Highways -0.85095 0.24811 -2.78687 -3.82736 -0.53118 53.91315 - -0.29846
15.56848
(-0.74) (2.13)** (-2.25)** (-2.21)** (-0.02) (2.19)** (-1.61)* (-2.25)**
pinv4: Ports 0.19952 -0.05471 0.54009 0.82740 9.67703 0.24936 -1.66084 -0.12838
(1.00) (-2.69)** (2.68)** (2.98)** (2.40)** (0.07) (-1.46)* (0.56)
pinv5: Airports 0.60145 ---- -0.60197 -0.10877 -8.35899 -4.37766 2.86747 -0.38275
(1.97)** (-1.91)** (-0.32) (-0.77) (-0.45) (1.00) (-1.57)*
pinv6: Railways 0.12489 ---- 0.17389 0.17051 -3.33996 -3.75863 1.32406 -0.12572
(0.75) (1.00) (0.87) (-0.54) (-0.69) (0.79) (-0.46)
NB: t-statistics in parenthesis.

* Significant at 10% level.
** Significant at 5% level
Figure 10: Accumulated impulse response function for aggregate public investment
(pinv)
1.4
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Year s
Output Employment Private investment Aggregate public investm
Figure 11: Accumulated impulse response functions for national roads (pinv1)
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Years
Output Employment Private investment National roads

Figure 12: Accumulated impulse response functions for municipal roads (pinv2)
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Years
Output Employment Private investment Municipal ro
Figure 13: Accumulated impulse response functions for Highways (pinv3)
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Years
Output Employment Private investment Highways (

Figure 14: Accumulated impulse response functions for ports (pinv4)
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Years
Output Employment Private investment Ports (pinv4)
Figure 15: Accumulated impulse response functions for airports (pinv5)
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
-0.2
Years
Output Employment Private investment Airports (p

Figure 16: Accumulated impulse response functions for railways (pinv6)
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Years
Output Employment Private investment Railways (p

Table 13
Effects of public investment on private investment
Public investment variable Marginal productivity

Elasticities

investment
central case 0.63871 8.12
pinv3: Highways
pinv4: Ports
pinv5: Airports
pinv6: Railways

Table 14
Effects of public investment on employment
Number of jobs
Public investment variable (per million of
Elasticities
Euros)
pinv: Aggregate public investment
central case 0.07860 230
pinv3: Highways
pinv4: Ports
pinv5: Airports
central case -0.00438 -500
pinv6: Railways

Table 15
Effects of public investment on output
Public investment variable Elasticities Marginal Rates of

productivity return

investment
central case 0.18264 9.54 15.9%
central case 0.19807 31.41 23.0%
central case 0.09839 22.32 20.9%
pinv3: Highways
central case 0.02416 8.24 15.0%
pinv4: Ports
central case 0.08736 107.14 30.8%
pinv5: Airports
central case 0.00937 19.18 20.0%
pinv6: Railways
central case 0.06247 18.47 19.7%

PRODUCTION TECHNOLOGIES AND TECHNICAL EFFICIENCY:
EVIDENCE FROM PORTUGUESE MANUFACTURING INDUSTRY
Ana Fariaa*, Paul Fennb and Alistair Bruceb

a
Department of Economics, University of Minho, Braga, Portugal
b
Nottingham University Business School, Nottingham University, United Kingdom
December 2001
Abstract
This paper aims to test whether a given type of process innovation, namely
flexible production technologies (FPTs), contributes to increased firm efficiency.
Using one-year firm data from Portuguese manufacturing industry and applying a
parametric stochastic frontier approach, individual technical efficiencies are obtained
and their determinants simultaneously estimated, using a single-step procedure
recently proposed by Battese and Coelli (1995). The results support the hypothesis
that technological flexibility, measured through the use of FPTs, is important in
explaining differences in efficiency. Furthermore, given the specifications of the
stochastic frontier function, the null hypothesis that Portuguese firms are fully
technically efficient is rejected.
Keywords: Technical efficiency, flexible production technologies, stochastic frontier.

JEL Classification: L23, L60, O30
_______________________________________________________________________
* Corresponding author: Ana Faria. Address: University of Minho, Campus de Gualtar, 4710-057 Braga, Portugal
Telephone: +351 253 604510. Fax: +351 253 676375. E-mail: apfaria@eeg.uminho.pt
128 Ana Faria, Paul Fenn, Alistair Bruce
1. Introduction
The impact of technological change on productivity has been an issue of major

importance in economic literature as the perpetual adoption of new technology by
economic agents plays a critical role in economic growth of firms and nations.
Despite the acknowledged importance of technological innovation to productivity
growth, a large number of empirical studies at the aggregate, industry or firm level have
failed to observe a positive impact specifically of new information technologies1 on
productivity. This result has been referred to as the “productivity paradox” and led to a
series of empirical studies. Whilst the earlier consensus was that the explanation for this
paradox was due to difficulties relating to the measurement of technological change and the
lack of adequate data (OECD, 1991)2, more recent studies have demonstrated a strong
positive impact of computers and information-related technologies on productivity (Siegel
and Griliches, 1992; Siegel, 1997; Lehr and Lichtenberg, 1998). However, other recent
studies have failed to confirm this relationship. For example, Berndt and Morrison (1995)
investigated the relationship between industry performance and investments in high-tech
office and information technologies for two-digit U.S. manufacturing industries during the
period 1968-86. They found that increases in the share of high-tech capital in the total
physical capital stock were negatively correlated with multifactor productivity and tended
to be labour-using.
Another important result was recently provided by Power (1998) who investigated the
impact of different vintages of technology on productivity and productivity growth in U.S.
manufacturing. Power’s results indicated that plant heterogeneity was more important in
explaining productivity differentials than sunk costs or capital reallocation. Given these
results, Power acknowledges the importance of future research to attempt to identify the
occurrence and implications of different types of investment, where, for example, if
investment is expansionary, then increases in investment need not be associated with
increases in productivity.
At any point in time, firms face a wide range of options about their investments in
technologies and innovations. In the 1980s, flexible production technologies (FPTs)
emerged as potentially viable technologies for competing in industries that were
traditionally characterised by high-volume repetition manufacturing, but which have more
recently been subjected to greater competition and environmental volatility. The new FPTs
were thus designed to provide firms with the capability of changing levels of production as
well as producing a larger variety of products, in the face of frequent demand variations, at
minimum cost. In other words, these technologies provided firms with flexibility in their
production process. As suggested by Milgrom and Roberts (1990), these technologies are
largely associated with modern manufacturing which is characterised by flexibility, speed,
economies of scope and core competencies. In fact, the relevance of this form of process
innovation has been well illustrated in the literature which clearly identifies it as a new
technological paradigm, given the pervasive effects it has upon the way production
processes are organised (Carlsson, 1984; Dosi, 1988; Lesourne and Barré, 1991).
1
These are microelectronic-based information technologies (IT) which have diffused rapidly throughout the
economies and many people have hypothesised that IT may also constitute a ‘general purpose technology’
(GPT), which can perform generic functions and thus have a major impact on economic growth (see, for
example, Bresnahan and Tratjenberg, 1995).
2
On this issue see also Griliches (1991).
Production Technologies and Technical Efficiency: Evidence from Portuguese Manufacturing Industry 129
Variations in productivity can be explained through differences in production

technology, differences in efficiency of the production process and/or differences in the
environment. By viewing productivity and efficiency as two different concepts, this work
aims to provide new empirical evidence on the relationship between technological
innovation and productivity by assessing the impact of new production technologies on
estimated measures of the efficiency of firms.
To investigate the effect of technological flexibility on the efficiency of firms we
employ stochastic frontier analysis where inefficiency is measured as the deviation of the
output of the productive unit from an idealised frontier function computed for the whole
industry. As such, efficiency of a production unit is given by a comparison between
observed and optimal values of its output or inputs (Lovell, 1993). Furthermore, efficiency
is said to be technical when the optimum is defined in terms of production possibilities, or
allocative when the optimum is defined in terms of the behavioural goal of the production
unit. Therefore, this work extends the literature that defines the performance of firms as a
function of the state of technology and economic efficiency.
The objective of this paper is twofold. First, it aims to provide measures of technical
efficiency of Portuguese manufacturing firms by estimating a stochastic frontier production
function. Second, it aims to investigate the impact of the new FPTs on technical
inefficiencies. In particular, we investigate the impact of eight different types of flexible
production technology adopted by Portuguese manufacturing firms.
The empirical analysis produces two key results. First, estimates show that different
technologies have different impacts on the efficiency of firms; that is, results support the
hypothesis that technological heterogeneity is important to understand differences in
efficiency. Second, the hypothesis that Portuguese firms are fully technically efficient is
rejected. This paper is organised as follows. Section 2 briefly discusses the theoretical
implications of endogenous technological flexibility for firms’ efficiency and the main
contributions of this study. Section 3 describes the data employed in the estimation and
provides an illustration of the diffusion of these technologies across industries. Section 4
defines the econometric model and section 5 presents the empirical results. Some
conclusions and policy implications are discussed in section 6.
2. Endogenous Firm Flexibility
The introduction of the concept of flexibility in economic theory is usually associated

with Stigler’s work “Production and Distribution in the Short-run” (1939). By assuming
that there is a continuum between short and long runs, Stigler argued that short-run
marginal costs can be constant within certain ranges of output and, therefore, the plant and
equipment can be adapted to changes in output rates. According to Stigler, a flexible firm is
one that has a relatively flat average cost curve and, therefore, the cost of deviating from
the minimum cost level of production is less than in an inflexible firm. Thus, flexibility is
least when the marginal cost curve is steep and average costs rise precipitously around the
minimum marginal cost; conversely, flexibility increases as average costs are flatter and
marginal costs less steep. Firms must then choose between being more efficient over a
larger range of output but less efficient at capacity output, that is, firms must resolve a
flexibility-efficiency trade-off.
In the 1980s, changes in demand patterns (e.g., greater product diversification, reduced
product life cycles) led to an increasing need for flexibility in the manufacturing process. At
the same time, technological advances in microelectronics and the machine tool industry
made manufacturing flexibility an important alternative to the mass production strategy.
Two major groups of benefits are associated with technological flexibility, namely a short-
run responsiveness to day-to-day problems and a long-run accommodation of market
changes. In turn, these benefits are key aspects in achieving increased productivity and cost
reduction, e.g., through improved machine utilisation, operational control, reduced labour
and reduced inventory (Maleki, 1991). As such, given the characteristics of the new
production technologies, they can be seen as potentially capable of resolving Stigler’s
flexibility-efficiency trade-off. Indeed, in this work firm flexibility is measured by the use
of FPTs, which is common to other studies either theoretical or empirical (Ungern-
Sternberg, 1990; Röller and Tombak, 1990; Vives, 1993; Norman and Thisse, 1999).
However, one should note that endogenous firm flexibility is not an exclusive characteristic
of technology. It may also be identified with labour characteristics, in which case numerical
flexibility (i.e., change in the number of employees) and functional flexibility (i.e., the
capacity of employees to perform different tasks). Also, flexibility can be a characteristic of
the organisation, in which case flexibility refers to the organisation’s capability to adapt to
changes either within the firm or in its environment. In this study, we consider flexibility as
a characteristic of technology given its importance to the theory of the firm and also due to
the potential economic effects that the new FPTs may have on the activities of firms (Faria,
2001). Another important characteristic of the FPTs is their complementarity with other
technologies and forms of organisation (Milgrom and Roberts, 1990). In other words, a
firm must implement a set of other complementary technologies as well as reorganising the
firm’s structure and strategy (e.g. network information technologies, just-in time
production) in order to reap the full benefits of flexible production technologies. As such,
an important characteristic of FPTs is their pervasive effect since they are not restricted to
manufacturing activities, but also extend to marketing, engineering and organisation. The
effect of complementarity between these technologies will also be investigated in the
empirical model.
There are several reasons that justify the relevance of an empirical investigation on the
effects of the use of FPTs on the efficiency of firms. First, technological innovation has
long been recognised as the most important force driving the process of growth. However,
there is some mixed evidence on the relationship between new technologies based in
microelectronics and productivity, as presented in the previous section. In fact, the limited
contribution of many empirical studies in explaining the ‘residual’ of aggregate production
functions is largely because they tend to treat all forms of technical change in the same
manner (Bresnahan and Trajtenberg, 1995; Power, 1998). This paper contributes to the
existing literature by providing empirical evidence using firm-level data on the
characteristics of production technology. Second, this paper is also a contribution to the
literature on industry dynamics, which focuses on the importance and implications of plant
heterogeneity, and in particular on the role of the production technology, in explaining
various economic phenomena (Mills and Schumann, 1985). Third, two of the most
expected benefits from the adoption of flexible production technologies, as put forward by
the engineering literature, are productivity increase and cost reduction (Maleki, 1991).
However, empirical evidence does not always support these positive effects of new
technologies on firms’ performance. Indeed, the degree of automation is not necessarily
positively related to productivity, since automation may serve other purposes. For example,
it may be necessitated by new products or the requirements of serving particular markets, or
by conditions in the work environment and the labour market. Moreover, a company that
has managed to reduce the unit cost of products no longer in demand can in a narrow sense
demonstrate high productivity, but with declining sales and profits (Carlsson et al., 1994).
This study is therefore a contribution to the empirical literature that aims to assess the
relationship between those variables by using a different econometric approach. Finally,
this empirical investigation is also a contribution to the literature that investigates the
sources of variation in efficiency. As Lovell (1993) observes, there are two important
elements that merit investigation, namely the behaviour of production under uncertainty
conditions and the internal organisation of firms. The new FPTs represent a shift in the way
production is organised, where the issue of demand uncertainty takes a central role.
3. Data
The data used in this analysis are from the Indinova survey, which was conducted by the
Portuguese Ministry of Industry in 1990 and aimed to collect information on both
innovation and production technologies in Portuguese firms. The survey covered firms with
10 or more employees in both mining and manufacturing industries. The Indinova sample
was obtained from the database of the Portuguese Ministry of Employment, which covers
virtually all the firms in the country, and was stratified by size class. The questionnaire was
sent to 3000 firms and 1026 answers were received, which represents a response rate of one
third. Since this work only considers manufacturing industry, our final sample has 1003
firms, as 23 firms belonging to the mining industry were excluded.
The questionnaire consisted of 14 groups of questions related to (i) characteristics of the
establishment, including ownership, strategy, employment and investment; (ii) innovation
of products and processes and R&D sourcing and financing; and (iii) production
technologies. With respect to production technologies, respondents were asked to indicate
their use of a selection of technologies, which covered the areas of design and engineering,
fabrication and assembly, automated materials handling, integration and control. It should
be noted that the Indinova survey does not provide data on the intensity of use of each
technology, hence our measure of technology usage is based on the number of flexible
production technologies that a firm reports using. Nevertheless, an important advantage of
this data set is that it enables us to test for inefficiency using truly microeconomic data. In
fact, it has been found that empirical tests, which rely on microeconomic data, provide
clearer evidence of inefficiency than studies that make use of more aggregate data, since
there is a loss of information in the aggregation process (Schmidt and Lovell, 1979).
The Indinova survey showed that 254 firms in the sample had adopted at least one
flexible production technology by the year 1989, a 25% overall adoption rate.3 Figure 1
shows the percentage of adopters by type of technology. It can be seen that the technologies
3
The incidence of adoption of FPTs in Portuguese manufacturing industry contrasts with the incidence of adoption
in American and Canadian manufacturing industries. In 1989, the American and Canadian surveys of
manufacturing technology (SMT) indicated that 74% of American and 58% of Canadian establishments used at
least one advanced manufacturing technology. This gap in the rates of adoption is to be expected since Portugal
was still initiating its diffusion process by the time the Indinova survey was conducted (see Faria, 2001).
with a higher incidence of adoption are NC machines, CNC machines and Network
systems, with 13%, 9% and 5 % adoption rates respectively.
16%
14%
12%
% Adopters a
10%
8%
6%
4%
2%
0%
NC CNC ROBOT CAD CAM NETW. MHS AIS
Technology
Figure 1. Adopters (%) by technology.
The percentage of firms in the sample that adopt the remaining technologies varies
between 3% (CAM technologies) and 0.5% (AIS systems). The higher incidence of
adoption of NC and CNC machines can be explained by the fact that these technologies
were the first to be introduced in the market, thus they are in higher position of the
diffusion curve. Furthermore, the lower adoption rate of Robots, MHS and AIS
technologies4 can also be explained by the high investment costs usually associated with
these technologies.
The diffusion process of new technologies is largely determined by the industry
characteristics to which the firm belongs. Figure 2 shows the percentage of adopters by
industry at two-digit level of aggregation.5 It can be seen that the Non-electrical and
Electrical Machinery (38) has a higher proportion of adopters, 8%, and in the second place
are the Textiles and Clothing industries (32) with 5% of adopters.
4
A similar adoption pattern is verified in the U.S. and Canada (Faria, 2001).
5
The two-digit industries considered are: Food and Beverages (CAE-31), Textiles and Clothing (CAE-32), Paper
(CAE-33), Wood and Cork (CAE-34), Chemical (CAE-35), Non-metallic Mineral (CAE-36), Basic Metals
(CAE-37), Non-electrical and Electrical Machinery (CAE-38), Miscellaneous (CAE-39).
By contrast, in the Paper (34) and Chemical (35) industries the proportion of adopters is
around 3%. Indeed, as expected, the adoption rate of FPTs varies greatly among industries.
30%
25%
20% % of adopters in each

industry
15% Industry weight in the
sample
10%
5%
0%
31 32 33 34 35 36 37 38 39
2 - digit industry
Figure 2. Adopters (%) by Industry
Figures 3,4 and 5 show the percent of plants using a technology in 1989 broken out by
major industry group. The following observations can be made. First, as noted in the
previous figure, it is clear that the adoption of these technologies differs across industries
with industries such as Basic Metals (37) and Miscellaneous (39) having an extremely low
number of adoptions. By contrast, industries such as Food and Beverages (31), Textiles and
Clothing (32), Chemical (35) and Non-electrical and Electrical Machinery (38) clearly have
a higher incidence of adoption of FPTs. Second, one should also note that some
technologies are not even adopted by some industries. This is very clear with respect to
industries Basic Metals (37) and Miscellaneous (39). Even among industries, which have a
higher proportion of adopters, some of the technologies have not been adopted at all. For
example, AIS systems have not been adopted by Food and Beverages (31), Paper (34) and
Non-metallic Mineral Products (36) industries. Whereas, the Wood and Cork (33) as well
as the Non-metallic Mineral Products (36) industries have not adopted any Robots. Finally,
the Paper industry (34) did not adopt any CAD technology, although it adopted CAM
technology. This is an interesting observation because supports the idea that these two
technologies are not necessarily complementary, as frequently referred in the literature. As
such, these figures illustrate well the fact that different industries have different
technological regimes (Dosi, 1988), thus different adoption rates.
0.14
0.12
NC
0.1 CNC
% of Adopters
Robot
0.08
CAD
CAM
0.06
Netw.
0.04 MHS
AIS
0.02
0
31 32 33
2 -digit industry
Figure 3. Technology Adopters by Type of Technology and Industry
1.4
1.2
NC
1
CNC
Robot
% Adopters
0.8
CAD
CAM
0.6
Netw.
MHS
0.4
AIS
0.2
0
34 35 36
2 - digit industry

0.25
NC
0.2
CNC
Robot
% Adopters
0.15 CAD
CAM
0.1 Netw.
MHS
AIS
0.05
0
37 38 39
2 - digit industry
Besides inter-industry differences, the literature on technology diffusion has pointed out
other variables as being key determinants in explaining different adoption rates among
firms, as for example, characteristics of firms, prices of competing technologies, and
market characteristics. Faria (forthcoming) estimates a model of adoption of FPTs for the
same sample of firms. An interesting result was that certain firm characteristics, e.g. age of
the firm, its R&D capabilities and its location, were more important in explaining the
adoption of FPTs than mere firm size effects. Besides this result, two other main results
were found, namely that (i) differences in the technological regimes across industries also
played a key role in explaining adoption; and (ii) higher demand uncertainty faced firms
had a positive effect on the decision to adopt FPTs. Tables 1 and 2 of the Appendix.
provide a brief description of the technologies considered in this study and the descriptive
statistics of variables used for the subsequent econometric analysis.
4. Econometric Model
To estimate firm-level technical efficiency and investigate its determinants, we use the
parametric stochastic frontier approach suggested by Battese and Coelli (1995). This
procedure is a development of the Aigner, Lovell and Schmidt (1977) and Meeusen and van
den Broeck (1977) stochastic frontier model, which models the technical inefficiency
effects in terms of other explanatory variables. The advantage of this procedure is that it
also models the determinants of firms’ inefficency, besides estimating them. As such, by
using a simultaneous estimation of the models, it avoids a problem of endogeneity of the
regressors in the second step and the inconsistency of the estimator (Kumbhakar and
Lovell, 2000). One should note that a similar problem would occur if the model were
estimated by the alternative approach of DEA. Furthermore, another advantage of using this
econometric modelling relative to DEA is that it considers statistical noise, which is more
adequate considering the type of data used in this study, although it comes at the cost of
having to impose a functional specification on the technology of the production function.
To investigate the effect of flexible production technologies on technical inefficiency of
firms, we used the general translogarithmic functional form for the stochastic frontier
production model for the sample of firms involved
Yi = β 0 + β1K i + β 2 Li + 0.5 β 3 K i2 + 0.5 β 4 L2i + β 5 KLi + vi − ui (1)
where the subscript i represents the i-th firm and Y represents output, and K and L the
inputs capital and labour, respectively. The gross value of production as reported in the
survey is our measure of output.6 For capital, the fixed assets figures given in the database
are used which describe the value of land, buildings, plant and equipment, whereas the
measure of labour is the number of employees as reported in the database; all these
variables are in logarithms. The firms can be allocated to different industrial sectors within
manufacturing. To allow for different input requirements across industry groups a set of 8
dummy variables for two-digit industry groups was included in the regression. The
industries are those as described in the previous section, and Non-electrical and Electrical
Machinery industry is treated as the reference industry given that it has the higher adoption
rate of FPTs.7 This stochastic frontier production function is characterized by an error term
that has two components, a symmetric error term to account for random errors, vi, assumed
to be i.i.d. N(0, σv2) and independent of a non-negative unobservable random error term, ui,
accounting for technical inefficiency of production, such that, for the given technology and
levels of inputs, the observed output falls short of its potential output, thus, ui is obtained by
truncation (at zero) of the N(ziδ, σu2) distribution. The model for the technical inefficiency
effects in the stochastic frontier equation (1) is then defined by
8
u i = δ 0 + ∑ δ j z ji (2)
j =1
where the z1, z2, …z8 variables are dummy variables representing different types of FPTs
and the δjs associated parameters to be estimated. In order to assess the effect of
complementarity between FPTs and other technologies, as discussed in section 2, we shall
re-estimate the model and define each dummy variable in the inefficency model as an
interactive term with the dummy variable TECHM. This dummy variable equals 1 if the
firms reports having adopted information technologies in the various departments which
6
Estimation of these models use either gross output or value added as the dependent variable. Yet, we should note
that value added as the dependent variable is more appropriate when materials also enter in the production
fucntion since separability between materials and labour and capital inputs is assumed in this case (Caves and
Bailey., 1992).
7
The dummy variable approach is admittedly a crude way to account for differences in production across
industries, but it was the best solution given the limitations of our data with respect to the number of
observations, which did not allow us to obtain estimates by industry in the inefficiency model.
favour the efficiency of FPTs. Battese and Coelli (1995) are followed and maximum
likelihood estimation is used for simultaneous estimation of the parameters in the frontier
production function and in the technical inefficiency effects function. The likelihood
function is parameterized in terms of the variance parameters, σ2 = σu2 + σv2 and γ = σu2/σ2,
thus γ is bounded between 0 and 1, which allows to test the null hypothesis, γ = 0, of an
absence of inefficiency, signifying that all of the variance can be attributed to statistical
noise.
5. Empirical Results
Maximum-likelihood estimates of the parameters in the translog stochastic frontier

production function, given the specifications for the technical inefficiency effects defined
by equations (1) and (2), and generalised likelihood-ratio tests for parameter restrictions,
are given in Tables 1 and 2. As presented in Table 2, all estimates show that the null
hypothesis of no technical inefficiency effects in Portuguese manufacturing firms is
rejected at the 1% level. This means that the variances of the inefficiency terms are
significant and that the error structure specified in the estimated model is preferred to the
typical error term employed in a mean response function.
Table 1
Maximum Likelihood Estimates for Parameters of
Stochastic Frontier Production Function and Technical Inefficiency Model
Production Function Parameters Inefficiency Parameters
Estimate
Variable Estimate Variable
Eq. (1) Eq.(2)
Constant 1.18*** Constant -3.08*** -2.52***
(0.26) (1.04) (0.80)
Capital 0.23*** NC 1.47*** 0.91**
(0.07) (0.51) (0.40)
Labour 0.92*** CNC -1.84** -0.61
(0.16) (0.82) (0.64)
Capital2 0.05*** ROBOT -0.82 -0.34
(0.02) (1.01) (1.00)
Labour2 -0.0088 CAD -0.30 0.61
(0.0549) (1.00) (0.99)
Capital*Labour -0.028 CAM -0.41 0.19
(-0.027 (1.00) (0.97)
Food and Beverages 0.52*** NETW -0.46 -3.87***
(0.09) (0.89) (0.97)
Textiles and Clothing -0.32*** MHS -6.42** -4.41**
(0.07) (2.91) (2.09)
Paper -0.003 AIS -4.66** -3.13*
(0.084) (2.35) (1.8)
Wood and Cork 0.06
(0.10)
Chemical 0.18*
(0.11)
Non-metallic mineral -0.24***
(0.09)
Basic Metals 0.19
(0.17)
Miscellaneous -0.40*
(0.23)
Variance parameters σ2 = 2.36 (0.43)

γ = 0.87 (0.03)
Loglikelihood = –951.477
Nº obs. 870
Standard errors are in parenthesis. ***, **, * indicates that the coefficient is
statistically significant at the 1, 5, and 10% level, respectively.
Table 2
Generalised Likelihood-Ratio Tests of Hypotheses for Production Function
Equation Null Hypothesis χ2 Statistica

1 H0: γ = δ0 = δ1 = … = δ8 = 0 81.92***
1 H0: β3 =β4 = β5 =0 15.20***
2 H0: γ = δ0 = δ1 = … = δ8 = 0 75.19***
***, **, * on the value of the chi-squared statistic indicates that the null hypothesis is rejected at the 1, 5, and 10%
level, respectively.
a
For the hypotheses involving the γ parameter the critical value is obtained from Table 1 in Kodde and Palm
(1986) using 10 degrees of freedom.
Also, the ML estimate for γ is 0.87, which is consistent with the conclusion that the true
γ-value is greater than zero and suggests that around 21% of the combined variance is
attributable to managerial inefficiency, and 83% to measurement error. LR tests indicate
that the translog specification is preferred to the Cobb-Douglas specification, therefore only
the first specification is presented in the Table. The coefficients associated with capital and
labour are significant and have the expected sign. Whereas the second order coefficient
associated with capital is highly significant, indicating a positive contribution of higher
intensity of capital to the productivity of Portuguese firms. The elasticity of frontier output
with respect to the k-th input variable for the translog model were estimated accordingly to
Battese and Broca (1997). These elasticities were estimated to be 0.20 and 0.90 for capital
and labour.8 As for returns to scale, results suggest that the technology used in production
exhibits constant returns-to-scale; indeed, the sum of the coefficients of labour and capital
is around 1. The estimates of the industry dummies suggest that it is important to control
for inter-industry differences with respect to input requirements and that some industries
(Textiles and Clothing, Non-metallic Mineral and Miscellaneous) exhibit lower
productivity levels relative to the reference industry.
It is the impact of FPTs on the efficiency of firms that is of particular interest to this
study. As such, with respect to the estimates of the δ-parameters in the inefficiency model,
we should note the following results. First, with respect to equation (1), all the coefficients
have a negative sign, except for NC machines, suggesting a positive effect of these
technologies on the efficiency of firms. As for significance, t-statistics indicate that
coefficients associated with NC, CNC, MHS and AIS machines are significant either the
1% or 5% level. Hence, these results seem to corroborate the hypothesis that FPTs have a
positive effect on the efficiency of firms except for the NC machines, meaning that the
latter have a negative impact on firms’ technical efficiency.
The negative effect of NC machines can be explained by the fact that these technologies
are considered the least productive within the group of FPTs, as they are designed to
produce high variety of products but low volumes. Furthermore, NC machines are
frequently seen as the first generation of CNC machines, which are far more flexible and
efficient machines. Also, we should note that Carlsson,Taymaz and Tryggestad (1994)
found a negative impact of NC machines on productivity. Finally, we should note that the
8
These elasticities were estimated at the means of the input variables and they differ little between the translog
and Cobb-Douglas functional form.
size of the estimate for the coefficient of the NC machines is lower than the size of the
other statistically significant δ-parameters representing the different FPTs, which reflects
the relative importance of the NC machines. The fact that the remaining technologies are
not statistically different from zero may be the result of non-explored learning-by-doing
effects due to the early phase of the diffusion process of the new technologies in Portuguese
manufacturing firms.
Another hypothesis investigated in this work is the complementarity effect of the
adoption of new technologies related to the organisation of firms. Equation (2) in Table 1
shows the ML estimates for parameters in a stochastic frontier production model with
inefficiency effects model where each technology dummy variable enters in the regression
as an interactive term with the dummy variable TECHM.9 These estimates show that the
coefficients for δ1 (NC*TECHM), δ6 (NETW*TECHM), δ7 (MHS*TECHM) and δ8
(AIS*TECHM) are significant with a negative sign, except for δ1. Thus, these results seem
to support the hypothesis that there exists an interactive effect between these two major
groups of technologies.
Finally, the estimation of the stochastic frontier production function allowed us to
predict the mean efficiency of the firms in the sample. Indeed, results show that the mean
efficiency is of 0.67, meaning that Portuguese firms could improve their productivity by 34
percent with the same quantity of inputs.10 Figure 6 illustrates the distribution of the
estimated individual technical efficiencies. The predicted technical efficiency differs
substantially among the firms, ranging between a minimum of 2.3% and a maximum of
89%. Of the 870 firms, approximately 60% have an efficiency score between 70% and
80%. This result is similar to other recent evidence on technical inefficiency that shows a
substantial dispersion of efficiency values among agents (Jaforullah, 1999; Ajibefun and
Abdulkadri, 1999).
9
For convenience only the estimates of the inefficiency model are presented.
10
Mendes (1999) estimates the technical efficiency for a sample of firms in two industries, namely the
Manufacture of Automobile Components and the Manufacture of Jewellery in Portuguese manufacturing
industry. Although a direct comparison between his study and ours is not possible given that different samples
are used, it is interesting to note his results. Hence, Mendes’ estimates of mean technical efficiency are 0.72 for
the Manufacture of Automobile Components industry, and 0.52 for the Jewellery industry, with individual
values ranging between 0.47-0.85 and 0.12-0.64, respectively, which seems to indicate a similarity between his
results and ours.
350
300
250
Number of firms
200
150
100
50
0
0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
Efficiency
Figure 6. Frequency Distribution of Technical Efficiencies
6. Conclusions
This paper had two main purposes, namely to obtain estimates of technical efficiency of
Portuguese manufacturing firms and to assess the effect that the new FPTs have on the
efficiency of firms. The interpretation of results should be cautious given the limitations of
our data set. First, one should consider that the sample was obtained by a survey and some
sample selection bias may be present. As such, these results can only be generalised to the
sample of firms involved. Second, the use of dummy variables to control for differences in
input requirements across industries is a crude solution. However, our data is truly
microeconomic which is the most adequate type of data to be used in the estimation of
these models.
Having said that, the estimates of the stochastic production frontier with inefficiency
effects model indicate two main results. First, Portuguese firms exhibit various degrees of
technical inefficiency, and for the sample of firms considered the mean inefficiency level
was around 34%. Second, FPTs seem to have a positive impact on the technical efficiency
of firms. An interesting finding was the negative impact of NC machines on technical
efficiency: this is a result that had already found support in previous studies, although using
a different methodology. The fact that NC machines contribute to increased technical
inefficiency can be the result of two effects. First, it can be due to the fact that these
machines are described as a more flexible technology within FPTs generally, and thus less
productive. Second, given that the NC machines may be seen as the first generation of CNC
machines, we might expect the earlier technology to be less efficient than the later. Finally,
the estimates support the hypothesis that some complementarity effects exists between
FPTs and the adoption of other technologies by the firm which have a positive impact on
the technical efficiency of firms.
Some policy implications can be drawn from this empirical analysis. First, we should
consider that one of the major factors explaining international differences in the rate of
technical change is differences in the rate of investment in machinery and equipment,
which influence the degree to which the national productive potential embodies recent
technical advances (Pavitt, 1998:560). In the light of the empirical evidence provided by
this paper, the new flexible production technologies have a strategic role to play in the
development of efficient firms and consequently in the rate of technical change of Portugal
because their adoption contributes to reduce inefficiency. Given that Portuguese firms are
mainly imitators, policies should support the diffusion of the new technologies that would
promote the appearance of a large enough number of first echelon firms in the market that
would be operating on the frontier. This, in turn, has three main advantages. First, it
supports firms that emphasise strategies of product development as the prime source of
competitive advantage, in contrast to emphasising low costs. Second, it eventually favours
the appearance of local suppliers, which then are in better position to respond to the
necessities of local users. Third, by promoting the development of local suppliers on can
also expect a more rapid diffusion of the new technologies, therefore reducing Portugal’s
disadvantage with respect to the supply of technology. Another policy implication is that
this diffusion policy should also aim to promote the capabilities needed to operate with the
new technologies among potential users. In this respect, the supply of engineers and other
technicians who have the skills to implement the new technologies and develop networks of
information and co-operation between firms and other agents involved in the innovative
process are of major importance. Finally, government should be aware that technology
diffusion policies should be both technology and industry-specific. Different technologies
require different technical expertise, but their user characteristics may also be different
across industries, thus requiring different supporting schemes. As such, what the previous
paragraphs are suggesting is that in order for Portugal to accelerate its diffusion process,
both supply and adoption innovation policies are needed.
As to future research, a clearer picture of the contribution of the new technologies to
efficiency of firms could be obtained by estimating a cost and/or a profit function. FPTs can
bring other benefits besides productivity increases and cost reductions, such as quick
response capability to demand variations either in volume or in product range. Indeed, a
firm may simultaneously be experiencing both an increase in costs after adopting the new
technologies, but also experiencing an increase in sales. As such, this study calls for the
pertinence of further work using an enriched data set.
Acknowledgements
We would like to express appreciation to George Battese, Tim Coelli and anonymous
referees for various comments and suggestions. Financial support from FCT, Portugal,
2001, and Fundação Calouste Gulbenkian, 2001 is also gratefully acknowledged.
Appendix
Table 1.
Flexible Production Technologies
TECHNOLOGY DESCRIPTION
NC machines Numerically controlled machines
CNC machines Computer controlled machines
Robot Pick and place robot which transfer items from place to place
MHS (Material Handling Robots and automated guided vehicles: computer controlled equipment
System) providing for the automatic handling and storage of materials, parts and
finished products
Network and Control Individual controller units intended to supervise a manufacturing unit
CAD Computer Aided Design
CAM Computer Aided Manufacturing
AIS Automated Inspection Systems
Appendix
Table 2.
Descriptive Statistics
Standard Number of
Variable Mean Minimum Maximum
Deviation Observations
Nº of Employees 201.34 446.48 10 5,853.00 1003
Gross Value
1,504.82 6,649.96 1.50 154,000.00 892
Production a
Capital a 1,140.25 6,847.14 0.10 131,017.70 959
TECHM 0.56 0.4958 0 1 1003
NC 0.14 0.3415 0 1 1003
CNC 0.09 0.2831 0 1 1003
ROBOT 0.01 0.1132 0 1 1003
CAD 0.02 0.1291 0 1 1003
CAM 0.03 0.1590 0 1 1003
NETW 0.05 0.2258 0 1 1003
MHS 0.01 0.0994 0 1 1003
AIS 0.05 0.0705 0 1 1003
a
Values in million escudos.
b
Values in thousand escudos.
References
Aigner, D., Lovell, C.A.K., and Schmidt, P. (1977). “Formulation and Estimation of
Stochastic Frontier Production Function Models”, Journal of Econometrics, Vol. 6,
21-37.
Ajibefun, I.A. and Abdulkadri, A.O. (1999). “An Investigation of Technical Inefficiency of
Production of Farmers under the National Directorate of Employment in Ondo State,
Nigeria”, Applied Economics Letters, Vol. 6, 111-114.
Battese, G. and Coelli, T. (1995). “A Model for Technical Inefficiency Effects in a
Stochastic Frontier Production Function and Panel Data”, Empirical Economics,
Vol. 20, 325-332.
Battese, G. and Broca, S. (1997). “Functional Forms of Stochastic Frontier Production
Functions and Models for Technical Inefficiency Effects: A Comparative Study for
Wheat Farmers in Pakistan”, Journal of Productivity Analysis, Vol. 8 (4), 395-414.
Berndt, E. and Morrison, C. (1995). “High-tech Capital Formation and Economic
Performance in US Manufacturing Industries, An Exploratory Analysis”, Journal of
Econometrics, Vol. 65, 9-43.
Bresnahan, T. and Trajtenberg, M. (1995). “General Purpose Technologies: Engines of
Growth?”, Journal of Econometrics, Vol.65, 83-108.
Carlsson, B. (1984). “The Development and Use of Machine Tools in Historical
Perspective”, Journal of Economic Behaviour and Organisation, Vol. 5(1), 91-114.
Carlsson, B., Taymaz, E. and Tryggestad, K. (1994). “Factory Automation and Economic
Performance: A Micro-to-Macro Analysis” in O. Granstrand ed. Economics of
Technology, Amesterdam: North-Holland, 37-63.
Caves, R. and Bailey, S. (1992). Industrial Efficiency in Six Nations. Cambridge: MIT
Press.
Dosi, G. (1988). “Sources, Procedures, and Microeconomic Effects of Innovation”, Journal
of Economic Literature, Vol. XXVI, 1120-1171.
Faria, A. (2001). Technological Flexibility and the Firm: Evidence from Portuguese
Manufacturing Industry, Ph.D. Dissertation, Nottingham University.
Faria, A. (forthcoming). “The Adoption of Flexible Production Technologies”, Economics
of Innovation and New Technology.
Griliches, Z. (1991). “Productivity and Technological Change: Some Measurement Issues”,
in Technology and Productivity: The Challenge for Economic Policy, Paris: OECD,
229-232.
Jaforullah, M. (1999). “Production Technology, Elasticity of Substitution and Technical
Efficiency of the Handloom Textile Industry of Bangladesh”, Applied Economics,
Vol. 31, 437-442.
Kodde, D. and Palm, F. (1986). “Wald Criteria for Jointly Testing Equality and Inequality
Restrictions”, Econometrica, Vol. 54, 1243-1248.
Kumbhakar, S.and Lovell, C.A.K. (2000). Stochastic Frontier Analysis, Cambridge:
Cambridge University Press.
Lehr, W. and Lichtenberg, F. (1998). “Computer Use and Productivity Growth in US
Federal Government Agencies, 1987-92”, The Journal of Industrial Economics, Vol.
XLVI (2), 257-279.
Lesourne, J. and Barré, R. (1991). “On the Emergence of a New Techno-Economic

System”, in Technology and Productivity: The Challenge for Economic Policy,
Paris: OECD.
Lovell, C.A.K. (1993). “Production Frontiers and Productive Efficiency”, in H.O. Fried,
C.A.K. Lovell, and S.S. Schimdt eds., The Measurement of Productive Efficiency,
New York: Oxford University Press, 3-67.
Maleki, R. (1991). Flexible Manufacturing Systems: The Technology and Management,
Prentice-Hall, USA.
Meeusen, W., and van den Broeck, J. (1977). “Efficiency Estimation from Cobb-Douglas
Production Functions with Composed Error”, International Economic Review, Vol.
18 (2), 435-444.
Mendes, J.M.Z. (1999). “A Eficiência Técnica das Empresas – Uma Aplicação Estocástica
a dois Subsectores da Indústria Transformadora Portuguesa”, Documentos de
Trabalho, CISEP, Working Paper nº3, Lisboa.
Milgrom, P. and Roberts, J. (1990). “The Economics of Modern Manufacturing:
Technology, Strategy and Organisation”, The American Economic Review, Vol. 80
(3), 511-528.
Mills, D. and Shumann, L. (1985). “Industry Structure with Fluctuating Demand”, The
American Economic Review, Vol.75, 758-767.
Norman, G. and Thisse, J. (1999). “Technology Choice and Market Structure: Strategic
Aspects of Flexible Manufacturing”, The Journal of Industrial Economics, Vol.
XLVII (3), 345-372.
Pavitt, K. (1998). “The Inevitable Limits of EU R&D Funding”, Research Policy, Vol. 27,
559-568.
Power, L. (1998). “The Missing Link: Technology, Investment and Productivity”, The
Review of Economics and Statistics, Vol. LXXX (2), 300-313.
Röller, L. and Tombak, M. (1990). “Strategic Choice of Flexible Production: Technologies
and Welfare Implications”, Journal of Industrial Economics, Vol. 38(4), 417-432.
Schmidt, P. and Lovell, C.A. Knox (1979). “Estimating Technical and Allocative
Inefficiency Relative to Stochastic Production and Cost Frontiers”, Journal of
Econometrics, Vol. 9, 343-366.
Siegel, D. (1997). “The Impact of Computers on Manufacturing Productivity Growth: A
Multiple-indicators Multiple-causes Approach”, Review of Economics and Statistics,
Vol. 79, 68-78.
Siegel, D. and Griliches, Z. (1992). “Purchased Services, Outsourcing, Computers, and
Productivity in Manufacturing”, in Z. Griliches ed., Output Measurement in the
Service Sectors, Chicago: Chicago University Press, 429-458.
Stigler, G. (1939). “Production and Distribution in the Short Run”, Journal of Political
Economy, Vol. 47(3), 305-27.
Ungern-Sternberg, T. (1990). “The Flexibility to Switch Between Different Products”,
Economica, Vol. 57, 355-369.
Vives, X. (1993). “Information, Flexibility, and Competition”, Journal of the Japanese and
International Economies, Vol. 7, 219-237.
TRANSPORT POLICIES IN LIGHT OF THE NEW ECONOMIC GEOGRAPHY
The Portuguese experience
A.C. Fernandes Teixeira

CORE, Université Catholique de Louvain
1
and Faculdade de Economia do Porto
Abstract
The purpose of this paper is to qualify the empirical relevance of a new economic
geography model and to analyze the policy implications of this result with reference to the
Portuguese transport policy from 1985 to 1998. It namely asks whether the reduction of
transport costs is likely to contribute to spatial polarization. In light of new data, it results
that the Portuguese transport policy is not working towards spatial equity due to intricate
mechanisms at work. An assessment of the implications of the envisaged transport network
in 2010 for regional development and spatial equity based on a simulation of the model is
also made. This simulation proves right another major prediction of the new economic
geography: if transport costs are lowered sufficiently the spreading of industry occurs.
KEYWORDS: economic geography, agglomeration, Portuguese transport policy, panel data

JEL classification: R3; R11; H54; C33
1
Address: CORE, 34 Voie du Roman Pays 1348 LLN, Belgium. Tel.: + 3210474361. Fax: +3210474310. Email:
teixeira@core.ucl.ac.be.
The author is especially grateful to P.Ph. Combes and J. Thisse for early support and discussions throughout. Comments from
an anonymous referee of the Banco de Portugal's Development of Portuguese Economy Conference, and participants of the
CORE workshop, UCL, are gratefully acknowledged. The author would like also to express his appreciation to the president of
the IEP (Portuguese Ministry of Public Works). This paper is supported by the Portuguese Ministry for Science and
Technology.
148 António Carlos Fernandes Teixeira
1. Introduction
Besides some distinguishable features, the idea that increasing returns to scale cause industries to
agglomerate geographically emerges as a common element in the analyses of the New Economic
Geography (henceforth NEG). In such a context, regardless of whether the economies of scale were internal
or external to the firms, transport costs have a major impact on the spatial distribution of economic
activities. The resulting trade-off between transport costs and fixed production costs, known since the work
of Lösch, gives firms an incentive to concentrate production. However, the existence of transport costs also
gives firms an incentive to locate near markets and suppliers, which means that, due to the presence of
market-size effects, they will locate near each other.
Changes in transport cost, alongside some “historical accidents”, induce a circular causation
mechanism that can produce agglomerations. However, this mechanism is opposed by centrifugal forces
generated by product market and factor-market competition (such as the bidding up of local land and wage
costs). The same centrifugal and centripetal forces that explain the agglomeration of economic activities in
early stages of integration can explain the spread of industry to less developed regions when integration
goes further, i.e., when transport costs are sufficiently lowered (Puga, 1999).
A logical corollary of this argumentation is that NEG constitutes a natural conceptual framework able
to encompass all transportation cost effects throughout the entire economy. Theoretically, this has been
widely documented by some of the most prominent scholars. But, according to one of them, we know that
economic argumentation succeeds at least as much on its aesthetics as its empirical support (Krugman,
1991, p.34). Hence, a natural question arises: what is the relevance of the NEG models?
A way to answer this question is to analyze the evolution pattern of the transport costs and its economic
implications, within a NEG conceptual framework. But since the transport costs are largely determined by
the transport infrastructures (e.g. Limão and Venables, 2001), this is equivalent to analyzing the impact of
the transport infrastructures on the basis of the tools provided by the NEG.
If this pattern of research seems to lead naturally to policy analysis, little work has been done in this
direction. Indeed, the economic implications of infrastructures have been a “matter” for the macroeconomic
literature about public capital and productivity2. However, a common feature of this macroeconomic
literature consists in the consideration of transport infrastructures as a local production input (overlooking
their network nature) as well as a pure nonrival public good (overlooking their congestion constraints).
In reality, transport systems (namely, the transport road system) form a spatially interconnected
network and, consequently, the conventional perpetual inventory method of measuring capital stocks is
generally not appropriate to encompass all transportation infrastructure3. Transport systems are also
primarily related to the traffic capacity of all links of the network. That is to say, neglecting congestion
constraints is a critical source of misunderstanding in the economic analysis of transport infrastructures
(Fernald, 1999).
The NEG setting constitutes then a macroeconomic's complementary analysis since it captures the
spatially network nature of transport infrastructures. Additionally, it may also provide a tractable approach
for congestion constraints.
At least to some extent, the policy analysis' scarcity within NEG models is due to the particular features
embodied in the imperfect competition à la Dixit-Stiglitz (see Fujita et al., 1999, for a detailed survey of
this literature). In particular, as highlighted by Hanson (1998, 2001), the “iceberg” transport cost problem
and the multiple meaning that a parameter of the function may have in equilibrium are particularly
troublesome.
To overcome these limits, a resurgence in alternative economic geography models is at work. In a
seminal work, Combes and Lafourcade (2001) have developed a tractable spatial competition framework,
introducing explicit strategic interaction between agents and a transport cost function that has a unique
meaning in equilibrium and discriminates the heterogeneous impact of the transport costs across industries.
2
A substantial amount of work has been produced (Gramlich, 1994, called it a “bubble of speculative research”), showing a
strong correlation between infrastructure and productivity in several economies. Some authors argue that infrastructure provides
highly valuable services to private sector. Other authors argue that the public capital is endogenous, so causality runs from
productivity to public investment, or that the correlation is completely spurious, reflecting a misspecification of trend.
3
See, for instance, Chandra and Thompson (2000) for an assessment of the importance of the network nature of the transport
systems.
Transport Policies in Light of the New Economic Geography: The Portuguese Experience 149
In addition, the work of Combes and Lafourcade proves the strength of the fundamental mechanisms of the
NEG beyond the Dixit-Stiglitz-Krugman's setting for a French data set – which includes a detailed transport
cost matrix and a sample of firms with more than 20 employees – for 1993.
The first objective of our work is then to retest the strength of the fundamental mechanisms of the NEG
within a spatial competition model à la Cournot (Combes and Lafourcade, 2001) for Portuguese data, using
a slightly different econometric approach. Moreover, since we have a well-specified model, it can guide us
in our interpretation of empirical results and policy analysis. The second objective of our work is then to
analyze the Portuguese Transport Policy in the light of the NEG.
The choice of Portugal is not a coincidence: seen as a “good pupil” in the Community classroom – it is
now a member of the Euro's “inner circle” and has undergone a fast process of catching up –, this country
has been using its transport policy as a crucial instrument in economic policy. Indeed, Portuguese
authorities believe that a better transport system promotes national growth and a more balanced
geographical distribution of economic activities.
Is transport policy working towards spatial equity? We have found strong evidence that this policy has
been working in the opposite direction (even if this result is mitigated by some spatial exceptions, due to the
nonlinearity between transport costs and spatial distribution of economic activities).
The insights resulting from our work may also be seen as an evaluation key for European4 policies. In
fact, Portugal has been a full recipient of European structural policies, a large part of which has been
dedicated to transport infrastructure. The Portuguese experience could be then projected in the European
enlargement process. Given the similarities of the candidates' economies with that of Portugal in the middle
of the 80's – namely, in the infrastructure grounds –, our results should constitute a basis for reflection on
the future allocation of funds to the Central and Eastern European Countries.
A new data set, including, among other things, all firms and workers operating in Portugal and a multi-
regional transport costs matrix (accounting for congestion issues), will be used for a period of time
exclusively determined by the theory. A Panel Data and adequate econometric approaches, accounting for
regional unobserved characteristics, endogeneity issues and spatial dependence, will be used in order to
estimate the structural model.
The paper is organized as follows: in Section 2, we present the data and discuss their issues. We also
present summary statistics on the spatial distribution of economic activity in Portugal in geographical form
and summary statistics on the Portuguese Transport Policy. In Section 3, we present the model. In Section
4, we discuss the model's empirical specification and the estimation methods. In Section 5, we analyze the
results of the econometric specification. We also present some simulations of the model in order to obtain
insights about future location. Section 6 is devoted to some final remarks.
2. Portuguese Geography and Data Issues
If a common Portuguese who fell asleep in the middle of 1980's was to wake up in the late of 1990's, he
would have enormous difficulties in recognizing his own country. Indeed, a country characterized by a high
level of political, social and economic instability became a case-study on economic convergence grounds:
during the 14 years of membership up to 1 January 2000, the PIB per capita jumped from 54 % to 75 % of
the European average. The scope of this paper is to investigate whether this spectacular economic growth
affected the spatial pattern of economic activities change during this period, and if so, how is this change
related to transport policy. To orient the empirical analysis, we describe for the period 1985-1998 some
stylized facts about the spatial distribution of activities and how the transport costs have evolved over time
as a result of the Portuguese transport policy.
Our choice of the period of time over which to carry out our analysis is deliberate. The NEG has
predictions for inter-regional differences in economic outcomes on the basis of agglomeration and
dispersion forces triggered by significant modifications in transport cost. Therefore, according to the
theory's predictions, these costs should be measured from a point in time at which they were relatively high
and ending at a point in time sufficiently distant to allow a substantial reduction in them. Hence, an
appropriate starting point appears to be 1985, the year chosen by Portugal to sign the treaty of accession to
4
To our knowledge, this is the first work aiming at analyzing such a crucial European policy in a “cohesion country” using a
structural NEG model.
the EEC, joining the community with the third wave of new members (January 1, 1986). Besides its
symbolism, 1985 gives us a general view of the road infrastructure existing before the first important wave
of European funds (1988-1991). In addition, the National Road Plan was also approved in 1985. The best
end point is as close as possible to the present. By using symmetric arguments (in order to capture the
impact of belonging to the European Union's “inner circle”), we have defined 1998 as our second temporal
reference for testing the theory. Similarly, it was in 1998 that the National Road Plan was updated.
Next, we faced a “classical” problem, which is related to the choice of the spatial units. Highly
aggregated regions such as NUTS II cannot adequately account for agglomeration economies (which, by
definition, are local phenomena) and encompass substantial regions heterogeneous within themselves. In
contrast, spatial units should respect “minimum market size” conditions (Boldrin and Canova, 2000), which
can be violated in the case where the absence of firms in some particular area is a common problem. This
will be the case if we split the Portuguese territory into its 275 mainland concelhos (a small regional level in
the Portuguese administrative system), because such a division does not capture the regional-sectoral
interdependencies present in the theory. We solved this trade-off through a solution of compromise between
spatially disaggregated levels and “minimum market size” conditions by choosing the district as our spatial
unit of analysis, thereby preserving computation feasibility. The district is a higher administrative regional
level, which is composed by several adjacent concelhos. The Portuguese mainland is divided in eighteen
districts, having an average dimension of 60x60 Km. This is equivalent to splitting the United States
territory into 2,060 geographical units.
In order to account for the number of firms and the number of workers, we use the yearly survey
collected by the Ministry of Employment for all the existing companies operating in Portugal (except family
businesses without paid employees). This survey consists of data on every worker as well as some basic
information on each company such as location, sector of activity and number of employees.
Looking now at the spatial distribution of Portuguese industrial workers in 1985 displayed in map 1,
with a simple indicator such as the relative density across regions5, we immediately ascertain two facts: a
strong “coastalization” and two major “centers” of activity, based around Lisbon and Porto. The relative
employment densities vary from 13.38 times the Portuguese mainland average in Lisbon to 0.0397 times the
Portuguese mainland average in Bragança.
5
Employment is calculated as the average annual number of workers per square kilometer. The indicator used is expressed
relative to the weighted average for the Portuguese mainland.
More than 60 % of the total of industrial workers are located in Lisbon and Porto. It means that the
Portuguese economy is strongly concentrated in space, because these two districts only account for 5.7 % of
Portuguese land area6.
During this period, the development of the roads network has been an important focus. Naturally, the
extent of the impact of the European community funds on this fact is far from being negligeable. This is not
surprising since the national authorities share with community authorities the view that a better road system
may promote national growth and a more balanced spatial distribution of economic activity7 (EC, 1999,
2000).
6
This spatial concentration is magnified if we add to Porto its adjacent districts (Aveiro, Braga) : now, these four districts,
corresponding to 11.9 % of Portuguese land, contain more than 76 % of the industrial workers.
7
This view is not exclusive to Portuguese and European authorities. Other countries have also undertaken enormous highway
construction initiatives, with the purpose of stimulating economic growth and spatial equity of activities. For example, in 1998,
Accordingly, 13.5% of structural funds were allocated to transport infrastructures (62.2% of transport
infrastructure amounts were allocated to road networks). In addition, the European Investment Bank also
contributed for the “road investment effort”, since 29% of the loans to date have been primarily related to
transports. Of course, the Portuguese counterpart – that is, public financing of transport infrastructures with
an increasing tendency for more involvement from the private sector – contributes to a real “transport
policy” exception. This can be easily deduced from the following figures.
President Clinton signed into law the Transportation Equity Act. The renewing with tradition of the New Deal era, it set aside
203 billions of dollars to improve the nation's highway infrastructure.
Figure 1 - Road Network Investment
450
400
350
Relative Growth
300
250
200
150
100
1985 1988 1991 1994 1997
Period
EU-15 Portugal
Figure 2 - Motorway Lenght
700
600
Relative Growth
500
400
300
200
100
1985 1990 1992 1993 1994 1995 1996 1997 1998
Period
EU-15 Portugal
Sources: European Commission (1999) Portuguese Ministry of Public Works (2000) ECMT, OECD (1999)
However, what really matters to private firms is the (generalized) transport costs they will face. Hence,
how does this dramatic investment in infrastructure affect the transport costs? Let us first define a multi-
regional transport costs matrix in such a way that it resumes the lowest cost itineraries among all districts.
Basically, the lowest cost itinerary between two districts can be formulated as a minimum pathfinder
problem: minimization of a set of parameters – total distance and time costs – subject to a feasible set of
possibilities – the existing routes and their characteristics (physical components and congestion constraints)
– for joining two points. Applying this computational process (see Appendix 1) to both years, it is shown
that the transport costs have fallen dramatically in the last 18 years (45 % on average) among the district
capitals. This is due to new construction, extension, reconstruction, renewal, and major repairs in the road
system and to the technological progress verified in transport sector (as well as political decisions
concerning that sector).
Map 2 shows the transport costs reduction relative to the average for the Portuguese mainland during
the period in consideration. Important transport costs savings can be observed in the districts presenting
very low employment densities in 1985 (e.g. V. Real presents a transport cost savings of 1.094 times the
Portuguese mainland average).
This means that substantial improvements in road infrastructure were made where demand for it was
certainly low, i.e., where industrial activity presents relatively low levels. These basic statistics clearly
express the concern of the Portuguese authorities about structuring the economic landscape. In light of the
NEG, such a drastic reduction in the regional transport costs – reinforced by the fact the Portuguese road
network carries 85,7 % of the national intra-trade (against 67,8 % of the European intra-trade average) –
should imply changes both in spatial distribution and the efficiency of economic activities.
It is now worthwhile to look at 1998. As we have mentioned, Portugal has undergone a fast catching up
process – so it is natural that all districts would have now absolute higher indices of economic activity.
However, we are concerned by the relative evolution of the spatial pattern of activities. Looking at maps 3
and 4, which contain the relative employment densities and the relative growth employment densities over
the period, we ascertain two major facts. Firstly, Lisbon and Porto present an interesting pattern of
evolution. Indeed, still being the largest industrial regions, they become less (Lisbon) or relatively less
(Porto) industrialized over time. This result suggests that a compositional shift from industry to services
(high-transportation cost goods) was particularly important in these districts or/and that their centrifugal
forces dominate the centripetal ones.
Secondly, maps 3 and 4 show that the evolution of the spatial pattern of activities might be described by a
“sequential (re)location process”. Large industrial regions (Braga, Aveiro and Leiria) have become more
industrialized over time relative to the national average. Intermediate industrial regions (Setubal, Faro,
Santarém, V. Castelo, Coimbra and Viseu) have become relatively more industrialized than the national
average but less than the large regions. Small industrial regions (C. Branco, Guarda, Évora, Portalegre,
Bragança and Beja) have become relatively less industrialized over time – in comparison with large and
intermediate regions –, (re)emphasizing their peripheral condition. These basic descriptive statistics show
that the spatial inequality has been reinforced over time.
It is no w necessary to rely these stylized facts on the NEG theory and the question to be answered is
whether the reduction of transport costs is likely to contribute to spatial polarization. In order to accomplish
our goal, we will adopt a spatial competition model à la Cournot. A market structure à la Cournot will
enriches the analysis, introducing explicit strategic interaction between agents. This theoretical setting will
also allow us to use a transport cost function that has an unique meaning in equilibrium, discriminates the
heterogeneous impact of the transport cost across industries (goods) and accounts for congestion issues.
3. A simple model of inter-regional trade
On the demand side, consumers have the same preferences across regions. The utility function of a
representative consumer is Cobb-Douglas
U = ∏ (q sj ) γ
s
(1)
s
where (q sj ) is the individual consumption of goods s in region j, and γ s is its share in the consumer's
expenditure (parameters are normalized such as
∑γ s
s
= 1 ). Given the consumers’s income in region j,
γ sRj
R j , demand in region j is thus given by Q sj = s
, where Pjs is the price of good s in region j.
p j
Turning to the supply side, a representative firm in a sector s located in the region j has a Cobb-
Douglas technology
y sj = (l sj ) λ ∏ (k s´ s φ s ' s
s
j ) , (2)
s'
where l sj is the sector specific labor demand and (k sj 's ) s '=1...S ' the intermediate input demand for goods
s’. It is assumed that ∑φ
s'
s´ s
+ λs = 1 , implying a constant marginal cost. The firm's cost minimization
program is
 Min l s , ( k s ' s ) ( w s l sj + ∑ p sj ' k sj ' s + f js

 j j s ' = 1 ... S '
 , (3)
j ) = (l j )
s λs
∏ ( k sj ' s ) φ
s 's
 s.t . y sj ( l sj , k 1j s ,..., k Ss
 s'
where w s is the sector wage on a national basis8 and f js is the sector-regional fixed cost of production,
inducing economies of scale.
The representative firm’s labor and input demands in sector s and region j are therefore:
c sj y sj c sj y sj
l =λ
s
j
s
and k s 's
j =φ s 's
, (4)
ws p sj
where c sj is the marginal cost, given by
(w s ) λ ∏( p s ' φ s 's
s
j )
c =
s s' . (5)
∏ (φ s 's )φ
j
(λ s ) λ
s s 's
s'
The regional demand for good s in region j, D sj , is the sum of the final and intermediate consumption of
consumers and firms in all sectors
R sj
D = Q + ∑n k =
s
j
s
j
s'
j
s'
j
, (6)
s' p sj
where R sj = γ s R j + ∑ φ s 's c sj ' n sj ' y sj ' is the regional expenditure devoted to sector s in region j and
n sj corresponds to the sector-regional number of firms. The market j equilibrium condition for good s is
8
Sectoral immobility and a nominal national basis for wages lead to unemployment in all sectors and regions.
thus given by D s =
j ∑n i
s
y ijs , where yijs denotes the good s produced in the region j and exported to
i
region i.
The profits of a representative firm in sector s and region j are :
π sj = ∑ pis − c sj − t sji )Y jis − f js , (7)

i
where t sji denotes the transaction cost for shipping one unit of good s from region j to region i.
Given the location of production, it is now asked what quantity can maximize the firms' profit in
a Cournot setting (e.i., each firm chooses strategically the quantity produced for each market)9. From the
first order conditions, it is easy to derive the explicit expression for the interior equilibrium price and
quantities in the short-run:
∑n t s s
i ji + ∑ nis cis
p sj = i i
(8)
N s −1
p sj − c sj p is − c sj − t sji
y sj (= y sjj + y sji ) = R sj + Ris (9)
( p sj ) 2 ( p is ) 2
where N2 is total sectoral firms. Finally, regional incomes, ( R j ) j are endogenous in the model: they
are given by the sum of renters’ (retired people...) income, worker's wage and firm's profits, which are
assumed to be locally distributed to workers:
R j = w l j + ∑ ( w s ' n sj ' l sj ' + n sj 'π sj ) , (10)
s'
where w is the renters' average income and l is the number of renters in region j.
The short-run equilibrium is thus characterized by equation [9] to [10], checking for possible corner
solutions10. In the long-run equilibrium, the number of firms is endogenously determined by the zero profit
conditions, observed for all regions and sectors.
The theoretical framework presents “centripetal” and “centrifugal” forces. The dispersion forces (intra-
regional competition) are embodied both in prices and in market-share mechanisms. The more numerous
the firms in a region, the lower their price and the lower their market-shares in the local market,
respectively. This gives them incentives to locate in regions where competition is low. However, another
mechanism (inter-regional competition) is at work. Given the assumptions made, the total production is
higher in regions that have a larger number of firms because regional incomes are endogenous - the
expenditure being, thus, higher in those regions. The fact that individual firm's market shares may be lower
in competitive regions is compensated by a higher individual firms' production in those markets.
The tension between these two opposing mechanisms, and their strength, will determine the spatial
configuration of the economy. In a two-sector two-region framework, Combes (1997) shows that
agglomeration (inter-regional competition effects) is likely to dominate over dispersion mechanisms (intra-
regional effects), leading to higher profits in the region where firms are more numerous in the short run11.
Since new firms are allowed to enter into the market, they will locate in the region with higher profits,
which corresponds to the competitive zone. This attraction will work as long as the asymmetry in the
number of firms between zones is not too large. The attraction of new entrants may be magnified by lower
9
As usual in economic geography models, it is considered that firms do not internalize the effect of their strategy on regional
incomes.
10
When transaction costs are high and/or asymmetries in the number of firms located in each region important, firms may not
necessarily produce for all markets.
11
Including supplier networks, the model argues that firms like even more to be close to each other not only because of linkages
working through the demand for goods from consumers but also because of direct input-output linkages among themselves.
transport costs and higher economies of scale. To a certain extent, an asymmetric equilibrium can be
achieved in the long run, i.e., one region benefiting from the location of more firms.
A multi-regional model is more sophisticated but, unfortunately, also more complex. The structure of
the spatial equilibria is determinant (i.e., in terms of how many manufacturing centers exist and what share
of manufacturing activity each activity contains) just for an existing fixed number of firms, regionally
immobile. In this case, a reduction of the transport costs (until a critical value) implies that the highest
profits are found in the large region. This is due both to the endogenous mechanisms and to an increase of
exports to other regions (promoted by low transport costs). In the same logic, intermediate regions present
higher profits than small regions. However, when the value of the transport costs is sufficiently lowered
(going behind its critical value), the large industrial region becomes less profitable for location than
intermediate industrial region.
In the long run, it is then argued that the impact of a general reduction on the transport costs may imply
a sequential (re)location process: large regions will attract new entrants and some firms previously located
in intermediate and small regions; intermediate regions will attract some new entrants and some firms
previously located in small regions and so on. Small regions will preserve at least firms producing for the
local market. This process will be in equilibrium when profits equalize12.
4. Empirical specification
We turn now to translating the theory into an implementable empirical specification. Since Eq.[9]
corresponds to the final model equilibrium condition, it should be the basis for the estimation. However, the
model cannot be directly applied to the data. This is due particularly to the absence of trade flows among
districts. In order to overcome this obstacle, and using Eq.[4], the model equilibrium condition is (re)written
in such a way that the regional sectoral employment, at to the left of the equation, might be expressed as a
function of well-specified variables. Moreover, the empirical implementation of the model faces two more
problems: (i) the multi-regional sectoral transport cost, t sji , is unknown (what we get from the data is t ji ,
a multi-regional unitary transport cost of the physical shipment of products) and (ii) the non-existence of
some key variables in their regional dimension (in fact, regional prices, p j , or marginal costs, c j , do not
appear in official data).
In order to solve these problems, it is proposed first a linear specification for the transport costs,
establishing that v s t ji = t sji . Since the v s parameter will be the object of the estimation, the first problem is
solved. A sort of approximation process is then proposed, consisting of writing the true variable values on
the basis of their national level plus a regional disturbance. For instance, the equilibrium regional price is
written as
p sj = p s (1 + δ sj ) , (11)
where δ sj represents the relative gap between national and regional prices. The trick is to (re)write the
model as a JxS system for the (δ sj ) j =1... J ( regions), s =1... S (sec tos )
unknowns. The next step is to plot the
corresponding expression of any δ s
j
on the relations meanwhile established (e.g. regional prices),
approximating the original value of the variables. Sectoral regional production, as defined by the Eq. [4], is
now a linear function of vs and of well-known variables (see Appendix 2):
12
It is then argued that full, partial or no agglomeration long-run equilibria may emerge, depending on the transaction and
economies of scale parameters, as in Krugman (1993).
  
( p s B s − P s (w s )λ ) s s λ2
s
s  Bs  s 2 ss P ( w ) ' −
yj = s
∑ Ri 1 + v s ( p ) X i − ∑ φ s' s X ss
j 
( p s )2 B s i  p s B s − P s (w s )λ
2
 bs i 
  
  
Bs s s λ2
 s 2 ss P (w )  
2 X ss
j
' + ∑ v s ' (1 +v s (
( p ) X i −
s ' s s ' s s '
∑ φ X j  − 2 X j ).s (12)
s
p s B s − P s (w s ) λ
s'≠ s 2 b 
 s' 
  
S' I  S ''  I
Lsj = ∑ v s' ∑  Eis ( X is' s − ∑ b s'' s X sj' s'' ) − Fis t ijα / s, s')  + ∑ Gis (13)
s'=1 i =1  s '' =1  i =1
where
Ris (2 P s (w s ) λ − p s B s )λs P s
s
s Ris λs P s
Ei = , Fis = s ,
( p s B s ) 2 ( w s )1− λ ( p s B s ) 2 ( w s )1− λ
s
Ris ( p s B s − P s ( w s ) λ λs P s
s
1 if s' = s s
α s, s ' )  and Gi = .
s' ≠ s ( p s B s ) 2 ( w s )1− λ
s
1 if
s
Because all elements of the Eq.[13] are known - except the parameters ( v ), which we will estimate -, it
can be summarized in a more convenient way for empirical work as:
s = 1...S
l sj = α sj + β s x sj + ε sj (14)
j = 1...J
where l sj (sectoral regional employment) is the dependent variable, α sj and β sj = ( β sj ,..., β js ) are 1x1 and
1xk vectors of constants, respectively; x sj = ( x sj1 ,..., x sjk ) is a 1xk vector of exogenous variables and ε sj
is an error term. The exogenous variables are then composite values, in the sense that they are determined
from a combination of different parameters (e.g., regional revenues, wages and technological parameters) as
defined by the model13. The regional dimension will be expressed through the j index (i.e., j=1..18
corresponds to the Portuguese districts) and the index s (s=1...25) serves for industrial sectors.
How might the parameters β sj be interpreted? If the spatial competition model holds, then all β '
should be positive and significant. Otherwise, neither the Cournot setting nor the transport costs measures
used are correct. The magnitude of the coefficients will depend on the extent of the transport costs
contribution for the total production cost function, so the β sj are allowed to vary across industries. Hence,
a change in transport costs might affect differently the spatial allocation of economic activity. The way the
coefficient values evolve over time is related to both the evolution of regional transport costs and regional
trade; for instance, a significant reduction in the coefficient value should imply an increase in the regional
trade, entailing all the spatial consequences described in the model14.
The sectoral classification we use (SIC-2) has 25 industrial branches and span all industrial activity15.
We obtain the remain relevant data from several publications published by the National Institute of
Statistics, INE, for both years16. In order to get the sectoral share of the total expenditure in personal and
13
To better understand the role played by the theory as the organizer of the data and the way how these data was constructed, the
matricial computation written in Matlab is available upon request.
14
The β s enables us also to compare different types of shipments.
j
15
Two sectors were excluded for the simple reason that they are operating as monopolies.
16
A large amount of the information was drawn from the INE web site: http://infoline.ine.pt.
intermediate consumptions, we use the Industrial firms statistics and the Integrated counts system of firms.
For the determination of the sectoral wages we also use primary information published by the INE,
Industrial firms statistics, matched with other mentioned sources. The Input-output table is the reference for
the technical coefficients matrix computation. For the determination of the sectoral prices, we use the
annual Consumer price (final goods) and the Division and subsection (intermediate goods) indices
published by the INE, as well as for the consumption parameters. The annual regional revenues, including
wages and aggregate regional result of firms before tax, were obtained from Regional counts, INE.
All exogenous variables of Eq.[14] are directly derived from the theoretical model.
Unfortunately, their interpretation is not obvious since they are the result of a combination of different
parameters defining the model (e.g. regional revenues, wages, technological parameters...).
Interdependencies, however, do not imply a very high correlation between variables, which is in
consonance with the general properties of panel data approach. It gives more variability and less collinearity
among the variables since the cross-section dimension adds a lot of variability as well as more information.
Thus, the variation in the data can be broken down into variation between regions of different sizes and
characteristics, and variation within regions (i.e., across sectors).
Table 1 - Correlation matrix
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24)(25)
(1) 1,0
(2) 0,04 1,0
(3) 0,05 0,03 1,0
(4) 0,09 -0,01 -0,05 1,0
(5) 0,30 0,02 -0,06 0,41 1,0
(6) 0,26 0,01 0,06 0,37 0,88 1,0
(7) 0,03 -0,02 -0,03 -0,02 0,03 0,02 1,0
(8) 0,05 -0,03 -0,04 -0,03 0,06 0,04 -0,03 1,0
(9) 0,31 0,04 -0,06 0,41 0,97 0,82 0,03 0,05 1,0
(10) 0,30 0,02 -0,05 0,41 0,97 0,80 0,02 0,05 0,96 1,0
(11) 0,06 0,06 -0,05 0,20 0,34 0,32 -0,02 -0,01 0,33 0,34 1,0
(12) -0,07 -0,02 0,07 -0,07 -0,10 -0,10 -0,02 -0,03 -0,11 -0,11 -0,07 1,0
(13) 0,28 0,02 -0,06 0,43 0,94 0,86 0,02 0,05 0,94 0,94 0,32 -0,10 1,0
(14) -0,03 -0,02 -0,04 -0,02 0,06 0,03 0,06 -0,03 0,05 0,04 -0,01 -0,04 0,03 1,0
(15) 0,19 0,03 -0,06 0,28 0,68 0,64 0,01 0,03 0,67 0,67 0,21 -0,08 0,65 0,01 1,0
(16) 0,17 -0,02 -0,04 0,08 0,24 0,20 0,07 0,07 0,23 0,23 0,05 -0,04 0,22 -0,02 0,15 1,0
(17) 0,22 -0,02 -0,06 0,17 0,46 0,41 -0,02 0,01 0,46 0,45 0,13 -0,07 0,43 -0,01 0,30 0,08 1,0
(18) 0,33 0,03 -0,06 0,41 0,99 0,89 0,03 0,06 0,98 0,98 0,34 -0,10 0,95 0,05 0,68 0,24 0,46 1,0
(19) 0,21 0,04 0,02 0,51 0,56 0,12 0,04 0,04 0,25 0,45 0,41 -0,14 0,18 0,05 0,45 0,23 0,43 0,12 1,0
(20) 0,23 0,05 0,03 0,35 0,42 0,36 0,51 -0,05 0,23 0,61 0,21 -0,24 0,51 0,06 0,31 0,34 0,31 0,31 0,41 1,0
(21) 0,32 -0,06 0,06 0,40 0,23 0,14 0,06 0,06 0,34 0,23 0,31 0,16 0,31 -0,02 0,25 0,31 0,34 0,14 0,31 0,23 1,0
(22) 0,15 0,07 -0,04 0,23 0,54 0,32 0,04 -0,12 0,16 0,45 0,31 -0,43 0,41 0,04 0,31 0,51 0,36 0,31 0,36 0,31 0,34 1,0
(23) 0,16 0,06 0,03 0,37 0,27 0,41 0,01 0,13 0,15 0,17 0,41 0,19 0,31 -0,09 0,14 0,34 0,37 0,37 0,37 0,54 0,36 0,13 1,0
(24) 0,23 -0,04 0,06 0,35 0,29 0,51 0,12 0,14 0,12 0,61 0,41 0,17 0,19 0,03 0,31 0,31 0,31 0,19 0,46 0,39 0,34 0,12 0,15 1,0
(25) 0,25 0,05 -0,07 0,26 0,12 0,31 0,31 0,09 0,34 0,32 0,31 -0,31 0,17 0,41 0,23 0,61 0,21 0,37 0,51 0,45 0,41 0,14 0,13 0,4 1,0
A number of estimators are possible for Eq.[14], relying on different stochastic assumptions which
concern the regressors and the components of the error term. In a panel context, the problem specifying the
stochastic nature of the disturbances becomes more complicated conceptually. For example, some of the
omitted variables may reflect factors which are peculiar to both regions and sectors for which observations
are obtained. Other omitted variables may reflect sector-specific differences which tend to affect the
observations for a given sector. Other variables may represent factors which are peculiar to the specific
region. Considering those other causal variables leads us to the following component structure for
disturbances:
ε sj = λ s + u j + e sj (15)
i.e., the error term contains both sector-specific λs and region-specific uj effects; idiosyncratic shocks
are represented by e sj .
Preserving the identity on the regression slope coefficients hypothesis, Eq.[14] becomes :
l sj = α sj + β ' x sj + u j + λ s + e sj (16)
The regressors, or a subset of them, may be assumed (or not) to be correlated with the region-specific
and sector-specific effects, i.e., with uj and λ s . Assuming that the idiosyncratic shocks are serially
uncorrelated the most popular estimators are the fixed effects estimator and the random effects estimator.
The within estimator assumes that the uj and the λ s are fixed effects for each region and sector. The
GLS estimator assumes that the uj and λs are random variables and uncorrelated with the disturbance
term.
5. Results
5.1 Proving the model
Turning now to the empirical results themselves, we present the results for the OLS, within and GLS
regression in columns (1) through (3) of table 2. The period is 1998 and T-statistics are reported in
parentheses.
Table 2 - Transport Cost Coefficient Estimation – Industry (a)
OLS Within GLS TSLS TSLS
Method
(1) (2) (3) (4) (5)
Extraction 0.069 (1.172) 0.174* (3.205) 0.1803* (3.235) 0.192* (3.432) 0.188* (3.351)
Food and beverages 0.017 (0.466) 0.095* (2.907) 0.093** (2.728) 0.103* (2.871) 0.096* (2.787)
Textile -0.032 (-0.452) 0.054** (2.351) 0.051** (2.210) 0.061** (2.412) 0.045* (2.245)
Clothing -0.052 (-0.018) 0.012** (2.312) 0.009** (2.120) 0.102** (2.560) 0.091* (2.491)
Leather and footwear -0.006 (-0.017) 0.073** (2.189) 0.068 ** (1.984) 0.102** (2.321) 0.069** (2.014)
Wooden and cork 0.219 ** (1.535) 0.479* (3.778) 0.500 * (3.799) 0.720* (3.321) 0.522* (3.927)
Paper and cardboard 0.296* (4.118) 0.384* (6.115) 0.402 * (6.152) 0.450* (5.231) 0.411* (6.266)
Printing and publishing 0.028 (0.966) 0.073* (2.813) 0.068 ** (2.551) 0.081* (2.671) 0.068** (2.545)
Petroleum refining 0.131 (0.893) 0.245** (2.341) 0.203** (2.213) 0.321* (2.981) 0.301** (2.451)
Chemical products 0.103 (1.227) 0.333* (4.284) 0.343 * (4.402) 0.421* (3.912) 0.353* (4.551)
Rubber and plastic materials 6.390* (14.864) 5.893* (15.867) 5.749 * (14.980) 4.012* (14.782) 5.669* (14.312)
Other non-metal products 2.435* (13.753) 2.335* (15.738) 2.733 * (15.111) 2.515* (15.211) 2.385* (15.091)
Basic Metallurgy 0.046** (0.838) 0.182* (3.779) 0.178 * (3.587) 0.201* (3.231) 0.183* (3.659)
Metal products 0.321** (1.432) 0.398* (4.120) 0.326* (3.521) 0.399* (3.451) 0.301* (3.910)
Machines and equipment 0.744** (6.862) 0.812* (8.713) 0.815* (8.468) 0.910* (8.521) 0.819* (8.486)
Office machinery, equipment for treatment

0.113* (1.432) 0.252* (3.633) 0.232* (3.250) 0.291* (3.123) 0.230* (3.212)
of information
Electronic machinery 0.376* (2.502) 0.635* (4.822) 0.658 * (4.853) 0.720* (4.139) 0.677* (4.978)
Communication 0.260* (1.302) 0.406** (2.311) 0.376 ** (2.078) 0.521* (2.512) 0.368** (2.031)
Medical-surgical, Optical (2.580)
0.071 (1.724) 0.133** (2.541) 0.136 ** (2.521) 0.234* (2.231) 0.140**
and Watch making
Table 2 (cont.)
Car and long vehicle 0.741** (2.234) 0.822* (3.121) 0.762* (3.251) 1.010* (3.910) 0.901* (3.912)
Other transport material 0.531** (2.523) 0.592* (4.210) 0.501* (4.298) 0.611* (4.561) 0.591* (4.091)
Furniture 0.752* (4.211) 0.922* (5.231) 0.081* (4.651) 1.222* (5.641) 1.021* (5.023)
Table 2 - Transport Cost Coefficient Estimation – Industry (a)
Method
(1) (2) (3) (4) (5)
Recycling 0.362 (1.228) 0.382** (3.120) 0.231 (2.165) 0.431** (3.678) 0.301** (1.991)
Electricity, gas, water 0.069 (1.201) 0.081** (2.123) 0.072 (1.980) 0.091** (2.371) 0.059** (1.923)
Building 0.091 (2.134) 0.102** (2.540) 0.099** (2.050) 0.131** (2.670) 0.098** (1.921)
Spatial lag variable 0.391* (3.513)

Adj. R2 0.85 0.95 0.935
Signifance of:
sectoral effects
[~F(S-1, S(J-1)-k)] 0.209 (0.891)
regional effects
[~F(J-1, SJ-J-k)] 7.255 (0.000)
LM Test
2 78.40 (0.000)
[~ ℵ 1]
Hausman Test
2 56.383 (0.000)
[~ ℵ (K) ]
Instrumental set Full (b) Full (b´)
Sargan Test (c)
2 6.0276 (0.9912)
[~ℵ 1]
Moran's I test (d) 11.812 (0.000)
Nº of Observations 450 450 450 450 450
Notes: * Significant at 1 % and ** significant at 5%.

(a) Estimation period is 1998. All standard errors robust to arbitrary heteroskedasticity and autocorrelation.
(b) Instrumental variables: explanatory variables for 1985 plus regional sectoral production for 1998.
(b’) Instrumental variables: the same as for (b) plus the spatial weighted explanatory variables for 1998.
(c) Sargan Test is a test of the exogeneity of the instrument set.
(d) Moran's I test (extended to TSLS residuals) is a test of spatial dependence.
Looking at the OLS parameters estimates, we can see that only three variables present a wrong sign (but the
negative coefficient of Textile, Clothing and Leather and Footwear are not significant). Moreover, this
regression, while showing a reasonable fit and presenting correct signs and plausible magnitudes of the
estimated parameters, clearly suggests the presence of specific effects. Indeed, the F-test for a common
intercept clearly indicates that it should vary across regions (the computed F comes out as F=7.253, which
is greater than the critical F(17,407) value of 1.62. On the contrary, the hypothesis of a constant intercept
across sectors is not rejected (the computed F comes out as F=0.209, which is less than the critical
F(24,400) value of 1.30). Since the sector specific effect was expected to capture sectoral features common
to all regions, the result is quite counter-intuitive. But, as we have already emphasized, final variables are
the result of a matricial computation, involving different economic dimensions, which certainly determines
much of their variability. On the basis of these statistical results, we will ignore, from now on, the sector
specific effects.
The results of the within estimation, displayed in column (2), clearly show a positive and significant
sign for all β even if five of them are only significant at 10%. Looking now at GLS estimates reported in
'
column (3), we ascertain fairly sizeable changes in both the estimated standard errors and in the estimators
when compared with the within specification. In particular, the coefficients of Food and Beverage, Printing
and Publishing have lost their 5% of significance. A clear quasi-generalized slight reduction also appears in
the magnitude of the coefficients, even if the coefficient's rank of this magnitude remains unchangeable. A
LM test statistic, for testing the GLS specification and whether the individual components exist, comes out
as LM=78.40, which is greater than the critical χ 2 value of 3.84, implying the rejection of the hypothesis
that individual components do not exist.
The preceding tests show that the OLS model is rejected in turn by both the within and the GLS
models. However, the Hausman test of the within against the GLS estimator gives a χ 2 25 statistic 56.383
against a critical value (for α = 0.10 ) of 44.3, suggesting that the fixed effects is a reasonable approach for
our study, since we are confident that differences between units can be viewed as parametric shifts of the
regression function.
Turning now to the empirical results for 1985, in columns (1) through (3) of table 3 we present the
OLS, the within and the GLS results for the general specification of the model presented above. Again, the
F-test for a common intercept clearly indicates that it should vary across regions (the computed F comes out
as F=4.30) but not across sectors (the computed F comes out as F=0.19).
Table 3 - Transport Cost Coefficient Estimation for - Industry (a)

Method
(1) (2) (3) (4) (5)
Extraction 0.101 (1.092) 0.201* (2.681) 0.190* (2.315) 0.201 (3.921) 0.194* (3.532)
Food and beverages 0.067 (0.396) 0.109* (2.761) 0.103 ** (2.208) 0.150 (2.976) 0.099* (2.537)
Textile -0.032 (-0.359) 0.054** (2.891) 0.051** (2.110) 0.061** (2.653) 0.045* (2.251)
Clothing -0.052 (-0.009) 0.012** (2.190) 0.009** (2.211) 0.102** (2.109) 0.091* (2.411)
Leather and footwear 0.030 (-0.015) 0.089** (2.013) 0.078 (1.987) 0.101 (2.432) 0.093** (2.141)
Wooden and cork 0.313 (1.491) 0.562* (2.945) 0.520 * (3.045) 0.610 (3.654) 0.597* (3.982)
Paper and cardboard 0.345 (4.028) 0.411* (5.152) 0.487 * (5.451) 0.554 (5.543) 0.448* (6.312)
Printing and publishing 0.082** (0.896) 0.068* (2.332) 0.069** (2.019) 0.091 (2.222) 0.083* (2.542)
Petroleum refining 0.243 (0.873) 3.182** (2.091) 2.198 (2.190) 3.412 (2.291) 2.101** (2.961)
Chemical products 0.193* (1.077) 0.382* (4.091) 0.352 * (4.056) 0.391 (4.122) 0.391* (4.098)
Rubber and plastic materials 6.950* (13.942) 5.876* (14.647) 5.921 * (13.101) 6.011 (14.925) 5.992* (14.222)
Other non-metal products 2.829** (12.053) 2.530* (14.348) 2.211 * (11.671) 2.724 (14.453) 2.335* (15.919)
Basic Metallurgy 0.086** (0.718) 1.003* (3.494) 0.189 * (3.098) 0.991 (3.864) 0.982* (3.991)
Metal products (1.091) (3.320) (3.211) (3.413) (3.871)
Machines and equipment 1.044** (5.162) 0.917* (7.163) 0.932* (7.168) 0.912 (7.621) 0.823* (8.927)
Office machinery, equipment for treatment
0.921* (1.032) 0.320* (3.073) 0.325* (3.106) 0.391 (3.452) 0.291* (3.871)
of information
Electronic machinery 0.591** (1.902) 0.867* (4.092) 0.683 * (4.065) 0.701 (4.369) 0.693* (4.981)
Communication 0.460* (1.002) 0.452** (2.029) 0.393 ** (1.965) 0.671 (2.829) 0.462* (2.267)
Medical-surgical, Optical
0.097 (1.324) 0.190** (2.067) 0.192 (1.021) 0.161 (2.871) 0.153** (2.099)
and Watch making
Table 3 (cont.)
Car and long vehicle 0.814** (1.923) 0.912* (3.121) 0.812* (3.091) 1.012* (3.019) 0.932* (3.121)
Table 3 - Transport Cost Coefficient Estimation for - Industry (a)

Method
(1) (2) (3) (4) (5)
Other transport material 0.623** (2.001) 0.671* (4.210) 0.623* (2.912) 0.701* (4.231) 0.614* (3.911)
Furniture 0.912* (3.009) 1.210* (5.231) 1.011* (3.321) 1.301* (5.098) 1.012* (4.672)
Recycling 0.421 (2.184) 0.671** (2.167) 0.541 (1.976) 0.610** (3.101) 0.312** (2.125)
Electricity, gas, water 0.072 (0.911) 0.091** (2.123) 0.082 (1.013) 0.097** (2.054) 0.062** (1.781)
Building 0.150 (1.001) 0.160** (2.540) 0.172** (1.021) 0.171** (2.706) 0.152** (1.923)
Spatial lag variable 0.431*

2
Adj. R 0.75 0.78 0.756
Signifance of:
sectoral effects 0.191 (0.981)
[~F(S-1, S(J-1)-k)]
regional effects
[~F(J-1, SJ-J-k)] 4.301 (0.000)
LM Test
2 32.201 (0.000)
[~ ℵ 1]
Hausman Test
2 52.222 (0.000)
[~ ℵ (K) ]
Instrumental set Full (b) Full (b´)
Sargan Test (c)
2 8.509 (0.394)
[~ℵ 1]
Moran's I test (d) 13.981 (0.000)
Nº of Observations 450 450 450 450 450
The results in column (2) confirm the expected sign of the coefficients, even if they are now less well
determined than before. The GLS estimates reported in column (3) present, as before, smaller coefficients
without any change in the magnitude of coefficient's rank. The LM-test clearly validates the GLS against
the OLS model (the computed LM=32.20 is greater than the critical value of 3.84). Again, the Hausman test
of the within against the GLS estimator gives a χ 2 25 statistic of 52.222 against a critical value (for
α = 0.10 ) of 44.3, suggesting dependence between the regressors and the individual effects.
2
The high R values, unusual for this type of analysis, might suggest two types of questions. The first
refers to the role that a good dataset may play in testing the geography models. As highlighted by Grilliches
(1995), data gathering should constitute a major concern for applied econometrics as tool building does.
Our dataset, for instance, covers all firms and workers and contains a straightforward measure for multi-
regional transport costs. The second might suggest problems of endogeneity. Indeed, we should not endorse
the fixed effects estimator unless regressors are only correlated with regional effects.
In looking at the way the explanatory variables were built, it is reasonable to fear a violation of the
assumption that regressors are uncorrelated with the idiosyncratic shocks. First of all, endogeneity issues
exist because of the agglomeration's mechanism present in the model. Indeed, the revenue depends itself on
the employment and on the number of firms; on the other hand, the input-output parameters are, by
definition, also a source of endogeneity. Another major source of endogeneity is given by measurement
errors. So far, we have admitted that the model is expressed as:
l sj* = η j + β ' x sj* + e sj (17)
where the ηj are the unobserved individual effects which may be correlated with the independent variable
of interest x sj* . However, the x sj* are not observed directly. Due to the approximation process
(linearization) that we made - i.e. for regional prices -, only their erroneous reflection, the x sj
x sj = x sj* + v sj (18)
is observed, where v sj i.i.d. is a measurement error with variance σ v2 . If the OLS is applied to the
observed variables, the resulting parameters will be biased for two distinct reasons: the correlation of the
x sj with the left-out individual effects and the negative correlation between the observed x sj and the new
composite disturbance term. It is clear that in panel data we can eliminate the first source of bias by going
within and by analyzing deviations around means. But it is also well-known that going within might
exaggerate the second source of bias and make things worse.
The TSLS estimator is run to eliminate the correlation between right-hand variables and the
disturbances. Candidate instrumental variables for the 1998 regression are the exogenous regressors for
1985 and the regional sectoral production for 1998. Candidate instrumental variables for the 1985
regression are the regional sectoral production and regional sectoral wages for 1985. For testing the
exogeneity of the instruments we use a Sargan test. The diagnostic appears to be optimistic. According to
the value of the Sargan test, we cannot reject the null hypothesis that the overidentifying restrictions are
correct for both years. The coefficients displayed in column (4) of tables 2 and 3 are all positive and better
determined than before.
A final econometric point concerns the spatial nature of the model and data. We easily observe some
spatial orders in Portuguese economic geography in maps 1 and 3. These empirical observations lead to two
types of questions. The first refers to geographic effects (demand linkages, transport costs...) that we believe
are captured, at least to some extent, by the model mechanisms. The second refers to econometric methods
to estimate geo-economic phenomena since data are not randomly distributed spatially.
In fact, the test of Moran's I, extended to residuals in TSLS regression of column (4) of tables 2 and 3,
as suggested by Anselin and Kelejian (1997), indicates the presence of spatial dependence. The spatial
autoregressive model is then carried out. In this model, spatial correlation of observations is handled by the
endogenous spatial lag variable (Wl sj ) included as a regressor. The weight matrix W contains information
about the relative spatial dependence between all districts. The elements wii on the diagonal are set to
zero, whereas the elements wij are set as wij = 1 if districts i and j have a common border and
wij = 0 otherwise17. The endogeneity resulting from the introduction of (Wl sj ) as a regressor is handled
through a set of instruments. We use the regressors for 1998 as well as the exogenous variables weighted by
the matrix W as a set of instruments. As underlined by Anselin (2000), under a reasonable set of
assumptions that are easily satisfied when the spatial weights are based on contiguity, the spatial TSLS
estimator achieves the consistency and asymptotic normality properties of the standard TSLS18.
The results displayed in column (5) of tables 2 and 3 give us the spatial TSLS estimates. The spatial lag
variable presents a positive and significant coefficient. This result emphasizes the existence of spatial spill-
over effects, implying that the regional district employment level is influenced by that of its neighboring
regions. Concerning the coefficients directly related to the model, we note that the estimated values of
β parameters are indeed positive but slightly smaller than in the preceding model.
Since the signs and the significance of the estimates were the main points for proving the theory, these
results imply that the model à la Cournot works in a straightforward way in different periods of time. Given
our estimation specification, this is equivalent to saying that the industrial sectoral and regional employment
differences in Portugal could be explained on the basis of the forces present in the model. Such a result not
only confirms the main findings of Combes and Lafourcade (2001) for a French data set but also reinforces
the robustness of the theory in the sense that we used a slightly different econometric approach and a non-
truncated data set.
If the interpretation of the signs and the significance of the coefficients are now clear, the interpretation
of the magnitude of these coefficients remains an important issue. The wide variation of the coefficients
might capture the relative importance of the transport costs across industries. A straightforward way to
check this prediction is to produce estimates of the transport costs, according to the theoretical model, and
to compare them with the share of the transport of freight on the production function of firms.
The sectoral transport cost is obtained from the multiplication of the estimated parameter vs by the
regional transport cost matrix t ij . The sectoral marginal cost is obtained by plotting the estimated
s λ s s
parameter, v s '' , in its linear form ( c sj = w ) s P (1 + ∑ v s '' ∑ φ s ' s X s '' s

j ) ). The relative share of
B s '' s'
17
In order to normalize the outside influence upon each disctrict, the weight matrix is standardized so that the elements of a row
add up to one wij

* = wij .
∑ wij
j
18
As highlighted by Anselin (2000), the form of the weighting matrix (based on economic and geography theory) is at the
discretion of the analyst. However, this assumption is testable against other alternatives. Conley (1999) describes alternate
methodology for efficient IV estimation with spatial dependence. While his estimates are less parametric, they require
measures of neighborliness that are symmetric (as well as the choice of a cut-of distance). For example, Lisbon must have as
great an impact in Leiria as Leiria has in Lisbon (which seems unappropriated for our study).
the transport cost on the marginal cost of each sector s, weighted by the sectoral regional production
n is y is v s t ij is displayed in columns (1) and (2) of table 4 for both years.
s
( T share = ∑∑ )
i j ∑ n k y k v t ij + c i
k
s s s s
Table 4 – Transport cost and industrial agglomeration
Industry Share of the transport cost in the Concentration index **

Marginal cost*
1985 1998 1985 1998
Extraction
3.1 2.8 0.038 0.041
Food and beverages
1.9 1.8 0.012 0.099
Textile
0.7 0.4 0.165 0.096
Clothing
0.5 0.3 0.067 0.051
Leather and footwear
0.6 0.5 0.171 0.088
Wooden and cork
2.3 2.0 0.109 0.339
Paper and cardboard
4.2 4.0 0.108 0.072
Printing and publishing
3.1 2.8 0.069 0.049
Petroleum refining
5.1 3.2 0.572 0.482
Chemical products
2.8 2.4 0.239 0.051
Rubber and plastic materials
4.2 3.1 0.046 0.019
Other non-metal products
2.6 2.1 0.078 0.029
Basic Metallurgy
2.3 2.0 0.032 0.079
Metal products
2.6 2.3 0.012 0.028
Machines and equipment
1.7 1.2 0.061 0.099
Office machinery, equipment for treatment of information
1.9 1.7 0.017 0.044
Electronic machinery
2.3 1.9 0.061 0.071
Communication
1.7 1.4 0.086 0.091
Medical-surgical, Optical
and Watch making
2.4 2.3 0.085 0.115
Car and long vehicle
0.7 0.3 0.074 0.017
Other transport material

0.6 0.3 0.073 0.013
Furniture
3.5 3.1 0.127 0.444
Recycling
2.9 2.6 0.051 0.037
Table 4 – Transport cost and industrial agglomeration
Industry Share of the transport cost in the Concentration index **

Marginal cost*
1985 1998 1985 1998
Electricity, gas, water
1.0 0.8 0.563 0.021
Building
1.2 0.7 0.01 0.020
Notes to table 4 : *The relative share of the transport cost on the marginal cost of each sector s, weighted by the sectoral regional production is given by
 
 nsi ysi v s t ij 
T s share =  ∑∑ s s
− s s .
 i j ∑ n k y k v t ij + c i 
 k 
2
18
** The extent of agglomeration for industry i in time t is measured by a normalized Hirschman-Herfindahl index of concentration  Eij (t ) E j (t )  where Eij is industry i
g i (t ) = ∑  ( − 
j =1  E i (t ) E n (t ) 
employment in district j, Ei is national employment in industry i, and Ej and En are total manufacturing employment in district j and nationally. The concentration index is a measure of the
extent of agglomeration, the sum over districts of squared deviations of each district's share in national own-industry employment from its share of total national manufacturing employment.
With perfect deconcentration, the index is zero. Its highest possible value approaches 2. Given that deviations are squared, the index tends to be dominated by the two or three largest
employer-districts.
These results, drawn from the theoretical specification for 1985 and 1998, are close to the real figures19.
s
That it is to say, the estimates of v approximate the actual segmentation in Portuguese markets measured
by the share of the transport freight on the production function of firms20.
Now that we have a straightforward interpretation of the magnitude of the coefficients, we turn to an
examination of the degree of industry agglomeration during this period to try to relate it to the extend of
transport changes. It is well known that changes in transport costs are an important force driving the
location choices of industries in the increasing returns to scale sectors. Moreover, changes in transport costs
are likely to affect industries disproportionately, according to the relative share of these costs in their cost
production functions. Accordingly, sectors with a large share on transport costs and subjected to increasing
returns to scale are expected to have turned more concentrated over time. We examine this question by
means of the calculation of a simple measure of geographic concentration (a normalized Hirschman-
Herfindahl index) for each sector for the period 1985-1998, displayed in in columns (3) and (4) of table 4.
All industries are agglomerated in the sense that they present a positive index. However, agglomeration
is higher in high-tech industries (e.g. Medical-surgical and optical, Electronic machinery, Equipment for
treatment of information) than in capital goods (e.g. Chemical products, Rubber and plastic materials, Paper
and cardboard). Over time it increases in high-tech industries and declines in some capital goods industries.
Admitting that high-tech industries have more significant economies of scale than capital intensive
industries (Black and Henderson, 1999), this result suggests that the decline in transport costs might have
amplified such agglomeration. On the contrary, the decline in transport costs and costs of upstream and
downstream linkages seem to have weakened the benefits of agglomeration in a few sectors where
employment has spread out.
What does the reduction of the magnitude of all estimated coefficients over time mean ? Firstly, it
means that transportation technology and transport policy have reduced the cost of transport across space.
This result is confirmed by the data. Secondly, in light of the mechanism described in the model, this
decline in transport costs seems to have triggered the sequential (re)location process observed in Portugal.
5.2 Simulating the model
In order to (re)confirm the last statement concerning the magnitude of the coefficients over time,
another sensitivity test must run. Using Eq.[9], we simulate trade flows (the average of all sectors) between
regions for 1985 and 1998 for a representative firm. All the right-hand variables are known, since estimated
parameters were plotted into this equation. Because exports are the result of the interaction among different
variables (unitary prices, marginal costs, transport costs and demand) we first describe their spatial
pattern21. Unitary prices and marginal costs, in turn, are highly dependent both on the number of firms and
transport costs. Hence, we plot in maps A.1 and A.2 the distribution of firms and the average transport costs
(Appendix 3). The distribution of firms basically follows the pattern of the employment density described in
the map 3. Large districts concentrate the highest values of the concentration of firms, but intermediate
districts and the inland band are also well-identified. The average transport costs shows Coimbra as the
center – measured in terms of transport costs – of the Portugal mainland. As expected, this variable presents
a precise pattern, increasing its value as we go further from Coimbra (the value increases more rapidly in
the direction of the inland band than northwards or southwards). The unitary prices and the marginal costs
are plotted in maps A.3 and A.4 (Appendix 3).
By defining central and peripheral locations with respect to Coimbra22, we deduce that prices are
relatively high in peripheral regions. Since prices decrease in the number of firms and increase in the
average transport costs, we conclude that transport cost effects are especially important in central places
like Coimbra and effects of competition are especially important in (relatively less) central regions like
19
The corresponding real values are approximated by the weight of transport costs in the final (unitary) price of the products, for a
given distance (INE, 1999).
20
The values of this structural parameter when comparing with those obtained by Combes and Lafourcade (2001) present some
discrepancies, that might be due to differences in industries classification.
21
For saving space, the description of these variables are provided only for 1998.
22
In our terminology, the centrality of regions is defined with respect to the ``center'' of Portugal and the size (large, intermediate
and small) of regions is defined with respect to their industrial concentration.
Braga and Porto. All peripheral regions present simultaneously high transport costs and low competition –
with the exception of Faro concerning the number of firms – which contributes to their relatively high
prices. Since the marginal costs also depend on competitive and transport cost effects, the similarity relative
to the price pattern was logically expected. However, it is worth noting that prices present higher absolute
values than marginal costs.
Since the location of expenditure is an important force driving location, especially in increasing returns
to scale sectors, we compute and plot regional demands (final and intermediate goods) in map A.5
(Appendix 3). Demand is strongly polarized in the sense that large districts present high values and the
inland band presents very low values.
Now, it is possible to simulate trade flows between regions. Porto is defined as a benchmark for regions
due to its low marginal and average transport costs. Maps 5 and 6 show the relative (to the average for the
Portuguese mainland) Porto's domestic exports destinations for all sectors for 1985 and 1998.
In 1985, the main destination of Porto exports were its neighbouring districts (Aveiro and Braga) and
Lisbon, Coimbra and Leiria. Even if Porto's main destination markets remained the same, the export pattern
suffered slightly modifications over time. In 1998, some intermediate industrial regions (e.g. Viseu) and
some small industrial regions (e.g. Guarda) became relatively more important markets for Porto23. This
simulation confirms that trade flows have increased over time as a result of a general but not uniform
reduction in transport costs. Such a result is in perfect harmony with the theory, since high transport costs
may constitute a sort of “protection” for local markets; an improvement in the connection between a large
and a small region makes it easier for producers in the small region to market their products to large region,
but may also expose the region to the competition of more “advanced” products from the larger region and
so reduce the regional monopoly power. Nevertheless, this increase in regional trade seems to benefit the
23
Given the transport costs and demand spatial patterns, such a result leads us towards an interpretation of the main export's
determinants in the spirit of the market potential function.
large and intermediate industrial districts, to the detriment of the small and peripheral districts (with the
exception of Guarda).
All previous results constitute evidence about the effects of the Portuguese transport policy. Contrary to
the authorities' expectations, the general decrease in transport costs has not reduced agglomeration, which
raises the question whether Portuguese authorities are (re)creating the Italian Mezzogiorno experience
reported by some authors24.
In order to answer this question, we must simulate (using the estimates of 1998) a main factor for the
attraction of firms: the variable profit for a representative firm for all regions. Such a simulation might
provide ceteris paribus insights about future locations, largely determined by the distribution of the variable
profit25. However, in light of the model, the profitability distribution is itself determined by opposite forces.
In order to better understand the direction and the strength of these forces, we first provide a description of
the main determinants of the variable profit – the average mark-up on all markets and the total quantity
produced per firm – and then we analyze how these forces interact between themselves.
The mark-up for a representative firm in each region, increasing on unitary prices and decreasing both
on marginal and transport costs, is shown in map A.6 (Appendix 3). The pattern of the regional mark-up
(high in central and peripheral zones) enables us to conclude that marginal and transport costs dominate the
price effects in central regions like Aveiro or Porto. Instead, in peripheral regions (e.g. Bragança price
effects dominate those effects embodied in marginal and transport costs. Intermediate central regions (e.g.,
Lisboa or Setubal) present a lower mark-up than central and peripheral zones, which means that the
competition effect they suffered are not compensated by their relatively low transport and marginal costs.
Turning now to the total quantity produced per firm in each region, plotted in map A.7 (Appendix 3), it
can be easily deduced that the highest values of this variable appear in central – and large – regions (e.g.
Porto, Braga and Aveiro). Since the total quantity produced per firm is increasing both on the marginal
profits (high in central regions) and local demand (high in large regions) and decreasing on local prices (low
in central regions), such a result is widely coherent. Lisbon's high value on the total quantity produced per
firm means that local demand effects are stronger than the price and marginal profit effects in determining
this variable. Peripheral regions present low values of this variable since their high local mark-up values are
dominated by high local prices and low local demands (with the exception of Faro).
Finally, we show in map 7 the variable profit distribution as the result of the average mark-up on all
markets and the total quantity produced per firm.
24
For instance, Faini (1983) had already pointed out that, due to endogenous and self-sustaining mechanisms, transport network
improvements have produced perverse impacts in the Southern Italy.
25
Another main factor for location, the regional fixed cost, is ignored in this simulation.
We ascertain that large industrial regions present relatively high values on the variable profit variable.
Such a result implies that production and mark-up effects are convergent in these regions26. But these
effects are clearly divergent in peripheral districts. Indeed, besides their high value on the mark-up variable,
peripheral districts present a relative low power of attraction due to the dominance of production effects.
Large and intermediates regions present a higher potential for location than peripheral regions. Since the
inland band will continue attracting fewer firms than other regions, an increase in the spatial concentration
should be expected in the future27.
However, the ceteris paribus hypothesis is widely disproved in the sense that Portuguese authorities
have recently (re)designed transport networks as a major goal of their political action. It implies that more
important reductions in transport costs will be obtained. Indeed, using data provided by the Portuguese
Ministry of Public Works about the transport road network design in 2010 and using the computational
procedure described in Appendix 1 (where unitary costs (time and distance) were lowered by 10% with
respect to 1998 values), important transport costs savings (42% on average) are expected.
26
Production effects dominate mark-up effects in Lisbon.
27
It is admitted that regional fixed costs are the same across all regions.
180 Transport Policies in Light of the New Economic Geography: The Portuguese Experience
Map 8 shows that intermediates and small industrial regions are the main beneficiaries of improvements
to the road network in terms of access to others markets. Indeed, the highest transport cost savings value is
obtained by Faro (an industrial intermediate region); Bragança (a small industrial region) presents in turn a
more important transport cost savings value than any of the large regions. It results clear that public
authorities continue their political aim of a “more harmonious development”.
According to the NEG message, if transport costs are lowered sufficiently then the spreading of
economic activities will be achieved by action of the centrifugal forces. It is then worthwhile to simulate the
employment distribution in the next 10 years28.
The simulation of the employment distribution in 2010 is then displayed in map 9. All exogenous
variables were recalculated accounting for the new regional transport costs matrix values. These exogenous
variables were plotted in the Eq.[16] and the values of the coefficients correspond to the estimates of 1998.
28
We assume that the number and the location of firms is fixed but the size of firms (or the industries) may vary.
This simulation shows an important spatial dispersion of economic activity. Indeed, Lisbon and
Porto may lose their attractiveness as industrial locations because they are subjected to strong dispersion
forces. Other large industrial districts (Aveiro, Braga, Leiria and Setubal) are likely to attract a large amount
of those workers located in Lisbon and Porto. Some intermediate industrial districts (namely, Faro) may
increase their relative industrial strength. Faro may attract workers previously located in large industrial
districts like Lisbon. The inland band may present a heterogeneous evolution: some districts (mainly,
Bragança) may attract an important amount of workers and others districts (Beja, Portalegre...) may
definitely lag behind the all country.
More spatial equity is then likely to be achieved if the planned investments are implemented.
That is to say, the Mezzogiorno development experience will probably not occur in a Portuguese version.
Moreover, a natural question arises: is it the best allocation for public funds? According to Martin (1999,
2000), alternative allocations (e.g. promoting workers’ qualification or/and innovative initiatives for firms)
imply a greater overall growth process. However, if spatial equity is a value beyond economics (e.g. for
political reasons) it may well be worth continuing on the route to more roads.
6. Conclusion
Currently, the empirical mainstream in the NEG has focused on descriptive analysis of the data.
As a result, the direct testing of spatial agglomeration models has just reached (to use Ottaviano’s
metaphor) its stage of infancy. One of the goals of our work was to contribute to bring it to a more mature
stage, proving the empirical relevance of a model that relies explicitly strategic interaction between agents,
input--output linkages and endogenous demand. In light of new data, strong evidence was found showing
that sectoral and regional employment differences in Portugal could be explained on the basis of the forces
present in the model.
Our second goal was to conduct policy analysis within a NEG setting. The Portuguese transport
policy – a major instrument of economic policy for national and European authorities – was shown not to be
producing the desired effects on spatial equity grounds. The common belief that a good endowment of
transport infrastructure is a prerequisite to attract firms should then be strongly mitigated. In fact, the effects
of the new infrastructure on the entire economy are complex. This complexity is essentially due to the
network nature of transport systems, the existence of critical levels in the transport costs as well as self-
reinforcing agglomeration effects (particularly important for increasing returns to scale sectors).
However, simulations of the model have proven right another major prediction of the NEG: if
transport costs are sufficiently lowered the spreading of industry occurs. This is also an important result for
policy-makers that should be obviously related with important questions (not answered in our work)
concerning efficiency and the best allocation of public funds.
References
Anselin, L., 2000, Spatial econometrics, in: B. Baltagi, ed., Theoretical econometrics (Blackwell) 310-330.
Anselin, L. and H. Kelejian, 1997, Testing for spatial error autocorrelation in the presence of endogenous
regressors, International Regional Science Review 20, 153-182.
Black, D. and V. Henderson, 1999, Spatial evolution of population and industry in the United States,
American Economic Review, May, 321-327.
Boldrin, M. and F. Canova, 2000, Inequality and convergence: Reconsidering European regional policies,
paper presented to the 31st Panel Meeting of Economic Policy, Lisbon, April 14-15.
Chandra, A. and E. Thompson, 2000, Does public infrastructure affect economic activity? Evidence from
the rural interstate highway system, Regional Science and Urban Economics 30, 457-490.
Combes, P-Ph., 1997, Industrial agglomeration under Cournot competition, Annales d’Economie et de
Statistique 45, 161-182.
Combes, P-Ph. and M. Lafourcade, 2001, Transportation cost decline and regional inequalities: Evidence
from France, CEPR Working Paper 2894.
Contley, T.G., 1999, GMM\ estimation with cross sectional dependence, Journal of Econometrics 92 , 1-45.
European Commission, 2000, The impact of structural policies on economic and social cohesion in the
European Union 1989-99 (EC, Bruxelles).
European Commission, 1999, Cohesion and transports (EC, Bruxelles).
ECMT, 1999, Investment in transport infrastructure, country studies (OECD, Paris).
Fernald, J., 1999, Roads to prosperity? Assessing the link between public capital and productivity,
American Economic Review 83 (3), 619-638.
Faini, R., 1983, Cumulative process of de-industrialization in an open region: The case of southern Italy,
1951-1973, Journal of Development Economics 12, 277-301.
Fujita, M., Krugman, P. and A. Venables, 1999, The spatial economy: Cities, regions and international trade
(MIT Press, Cambridge).
Gramlich, E., 1994, Infrastructure investment: A review essay, Journal of Economic Literature 32, 1176-
1196.
Grilliches, Z. and J. Haussam,1986, Errors in variables in panel data, Journal of Econometrics 31, 93-118.
Grilliches, Z., 1995, Introduction to the applications, in: L. Mátyás, and P. Sevestre, eds, The econometrics
of panel data (Kluwer Academic, London) 653-659.
Hanson, G., 2001, Scale economies and the geographic concentration of industry, Journal of Economic
Geography 1, 255-276.
Hanson, G.,1998, Market potential, increasing returns and geographic concentration, NBER Working
Paper, 6429.
Instituto Nacional Estatística,1999, Inquérito aos transports rodoviários e de mercadorias (INE, Lisboa).
Limão, N. and A. Venables, 2001, Infrastructure, geographical disadvantage, transport cost and trade,
World Bank Economic Review 15, 451-479.
Link, H., Dodgson, J., Maibach, M. and M. Herry, 1999, The costs of road infrastructure and congestion in
Europe (Physica-Verlag, Germany).
Krugman, P., 1993, On the number and location of cities, European Economic Review 37, Issue 2-3, 293-
298.
Krugman, P., 1991, Geography and trade (LUP, Leuven).
Martin, P., 2000, The role of public policy in the process of regional convergence, European Investment
Bank Papers.
Martin, P.,1999, Public policies, regional inequalities and growth, Journal of Public Economics 73 (1), 85-
105.
Portuguese Ministry of Public Works, 2000, Internal Reports.
Puga, D., 1999, The rise and fall of regional inequalities, European Economic Review 43, 303-334.
Appendix 1. The transport cost matrix.
The transport cost matrix among districts is the result of a complex calculation29. It can be seen as a
simultaneous exercise in cartography and operational research.
The cartographic part is conveyed in a digital map of the national road network – the so-called National
Road Plan – of continental Portugal. The map outlines the structure of the road network, specifying the
technical characteristics of each category of road. The main network on the map (the Principal Itineraries
(IP) with a length of 2,600 km) connects the major cities and ports and leads to the most important land
borders. This main network is completed by a secondary network that assures the links to the interior
regions and medium and small urban centers. This secondary network is composed of two kinds of roads:
the Complementary Itineraries (IC) with a length of 3,200 km and National Routes (EN) with a length of
5,500 km30. Some regional routes are also incorporated in the secondary network.
Composed of 2,200 individual links (each with an average length of 4.3 km), connected together by
nodes, the digital network is an excellent indicator of traffic conditions. In fact, each link is defined
according to the type, width and surface quality of road as well as the number of lanes that it has. Moreover,
each link has the same amount of traffic, with crossroads treated like nodes. All traffic in the districts are
concentrated in “centroids”, which are artificially connected to the nodes of the digital network through a
“fictional” link. This network covers more than 85 % of the total traffic among all small, medium and large
urban centers.
Operational research is the second part of this simultaneous exercise. It is conveyed in the calculation
of the shortest, fastest and cheapest itinerary between two geographical points31.
The shortest itineraries (the simplest calculation) were created by means of a network simulation. That
is, the shortest itinerary between two points in the network was given without any concern for the physical
characteristics or traffic volume of a given link. Differences over time reflect changes in the state of the
network caused by construction of new roads.
The fastest itineraries require a more complex calculation, since it is necessary to determine an average
speed for each link. The estimates of traffic volume in each link are based on traffic flows at different times
of the day and week. These detailed data account, for instance, for the percentage of long vehicles in traffic
and other sources of congestion. The fastest itinerary between two points is then determined as a result of
the traffic volume capacity and physical characteristics of the link32.
Finally, the lowest cost itinerary between two points is calculated as a minimum pathfinder problem
subject to a feasible set. That is, the cost minimization problem is constrained by the existing road network.
It is then necessary to determine a unitary cost both for distance and time. The unitary cost per km (per
hour) is defined as the charge paid per km (per hour) by a typical road transporter. The parameters for this
calculation were provided by the National Association of Transport Operators (ANTRAM, 1985, 1998).
According to ANTRAM, a representative vehicle (40 tons) with an occupation rate of 100\% faces several
expenditures: fuel (whose consumption varies with the type of road), highway tolls (for a certain type of
road), insurance, maintenance, pneumatics, repairs, lorry drivers’ wages, haulers’ charges (insurance, taxes
and security costs), as well as the interim payments for equipment depreciation and renewal. Using these
expenditures, it is easy to establish a unitary cost of time travel as well as a unitary cost of distance (which
varies across the different types of road). The vehicle's loading/ unloading – an invariant time in the
itinerary – is accounted by adding 30 minutes to any fastest itinerary between two points.
The next step is to multiply each individual link by its correspondent costs (time and distance). Finally,
the pathfinder program is run.
29
In this respect, the Portuguese Ministry of Public Works is gratefully acknowledged.
30
These figures are relative to 1998.
31
These calculations were made using EMME/2 software.
32
Congestion must be considered as a major determinant of transport costs. For instance, Helke Link et al. (1999) evaluated the
yearly congestion for Portugal in a total amount of more than 5 million hours in road network travel for heavy goods vehicles.
Put into figures, their estimates suggest that yearly freights in Portugal rise by 27 millions Euros (5.2 Euros per hour, at 1995
prices.
In light of this result, it can be ascertained that neglecting congestion issues underestimates transport costs. To our knowledge,
this is the first work aiming at analyzing such a crucial issue in a NEG setting.
Appendix 2. Linearization procedure
s
Let p denote the price would prevail if manufacturing goods were traded without costs. Hence,
according to Eq.[8] and Eq.[5], is given by
N s ( ws )λ
s
ps = Ps (19)
( N s − 1) B s
with B s = (λs )λ ∏ (φ s's )φ and where P s = ∏ ( p s' )φ

s s's s's
would be the price index of the
s' s'
manufacturing good if markets were perfectly integrated. Taking logs in [19] leads to
64 4 4 4 4 4 4 4 4M74 4 4 4 4 4 4 44 8 

N 1( w1 ) λ 
     log( 
 log(
p1)   φ .. .. φ .. ..φ   log( p )  
11 s1 S 1 1 ( N 1 − 1) B1 

 ..   ... .. .. .. .. ..     
     ..   .. 
 ..   .. .. .. .. .. ..   ..   .. 
        (20)
 log( p s )  =  φ 1s .. .. φ ss .. ..φ Ss  x  log( p1)  +  N 1( w1 ) λ .
      log( 
 ..   .. .. .. .. .. ..   ..   ( N 1 − 1) B1 
 ..   .. .. .. ..     
.. ..   ..   ..
   
     
 log(
 p S )   φ 11 .. .. φ s1 .. ..φ S1   log( p1)   N 1( w1 ) λ 
       log( 

 ( N 1 − 1) B1 
Admitting that the equilibrium regional price may be written as p sj = p s (1 + δ js ) – where the
δ s
j represents the relative gap between national and regional prices –, and taking into account [19] and
[20], it is now possible to derive a JxS system for the ( δ ( j ) j , s ). This linear system is given by
s
s 's
X
6 4 4 4 4 4 447j 4 4 4 4 4 4 48
∑ ∑ ∑
1 (21)
δj =
s
v s' bijs ' s n ks ' t ik .
p s ' ( N s ' − 1)
s' i k
where bijs 's are terms of the inverse of ( ( I JxS − M ).
Both sectoral prices and marginal costs for each region are now given by p sj = p s (1 + ∑vj
s'
X sj 's ) and
s λs
∑ v ∑φ
s
(w ) P
c sj = (1 + s '' s 's
X sj ''s ' ) , respectively. Finally, plotting these two last expressions in the
Bs
s '' s'
Eq.[9] leads to the Eq.[11].
ACESSIBILIDADE ECONÓMICA E BEM-ESTAR:
EVIDÊNCIA DA PENÍNSULA IBÉRICA
Armando José Garcia Pires*

ISEG/UTL
Dezembro 2001
Resumo:
Este trabalho apresenta uma metodologia para quantificar acessibilidade
económica no contexto de um modelo de geografia económica. Ao contrário das
medidas tradicionais de potencial de mercado desenvolvidas de uma forma “ad-hoc”
sem uma estrutura teórica-económica sólida por de trás, a presente metodologia tem
os seus alicerces num modelo que explicita a estrutura de mercado, agentes e
interacções estratégicas. A aplicação do modelo desenvolvido ao caso das regiões
NUTS 2 da Península Ibérica permite tirar algumas conclusões prévias. Em primeiro
lugar em termos de centralidade a economia espanhola tem uma grande vantagem
sobre a economia portuguesa. De facto Portugal aparece neste estudo quase como
uma economia periférica da espanhola, onde nem sequer a região de Lisboa (região
mais desenvolvida de Portugal) consegue ter no conjunto da Península Ibérica uma
importância relevante em termos de potencial económico. Comparativamente às
regiões espanholas, a região de Lisboa apenas tem uma performance marginalmente
superior à pior região da Espanha (Galiza). O pico de acessibilidade é atingido pela
região de Madrid sendo seguida de perto pelo País Vasco e pela Catalunha,
formando estas três regiões os principais centros económicos da Península Ibérica.
Testes de sensibilidade dos parâmetros confirmam a robustez dos resultados. Em
segundo lugar os resultados do índice de bem-estar regional confirmam os do índice
de acessibilidade económica, com uma importante excepção: a região de Lisboa. De
qualquer modo é aparente uma clara correlação entre o nível de potencial económico
e o nível de bem-estar de uma região. Em terceiro lugar um cenário de integração
completa entre a economia portuguesa e a economia espanhola não aparenta ser
benéfica para uma parte considerável das regiões. Tal como previsto pela literatura
na “nova” geografia económica as regiões mais beneficiadas são as mais avançadas,
evidenciando portanto efeitos de “lock-in” e de “causalidade cumulativa”.
Palavras Chave: Geografia Económica; Acessibilidade; Potencial de Mercado; Península Ibérica.
Classificação JEL: C68; F12; F15; R12; R13.
e-mail: armandopires@net.sapo.pt
1. Introdução
O conceito de Centro e Periferia tem sido amplamente discutido no contexto da União

Europeia (U.E.). Associam-se a estes conceitos em primeiro lugar a ideia de
desenvolvimento económico, e em segundo lugar o de bem-estar das populações (sendo
*
O autor agradece ao Professor Renato Flôres e à Professora Paula Fontoura pelos valiosos comentários e
interessantes discussões durante a preparação deste trabalho. Esta pesquisa foi suportada por uma bolsa da
Fundação para a Ciência e a Tecnologia (Praxis XXI/BM/17786/98). Quaisquer erros ou omissões são no
entanto da minha inteira responsabilidade.
190 Armando José Garcia Pires
este uma consequência lógica do primeiro). Nomeadamente no caso Europeu, os países da

Europa Central e do Norte são apresentados como desenvolvidos (a nível económico e
social) e aos países do Sul da Europa como ‘menos’ desenvolvidos, e entre estes claro
Portugal. Foi também a partir deste conceito de Centro-Periferia, e da possibilidade destas
diferenças se agravarem ao longo do tempo e em especial como consequência de processos
de Integração Regional (como é o caso da U.E.), que durante os anos oitenta a Comissão
Europeia achou necessário dar uma nova dimensão à ajuda regional Europeia. Em Portugal,
esta mudança no contexto económico e político fez também (re)nascer um certo receio,
fruto de uma vizinhança histórica por vezes ambígua, que a economia espanhola pudesse
absorver a economia portuguesa ou que pelo menos esta se tornasse periférica da primeira.
Deste modo Portugal passou a ver o ‘perigo’ da periferialidade a dois níveis: a nível
Europeu e a nível Ibérico.
A nível Europeu é já bastante conhecida a periferialidade da economia Portuguesa. Um
dos estudos de referência que confirma este facto foi feito por Keeble et al. (1988), estudo
este encomendado pela própria Comissão Europeia. No entanto a nível Ibérico toda esta
ideia parece-nos que ainda falta de comprovação empírica. O objectivo deste estudo é pois
analisar se que o que já se sabe a nível Europeu, e se teme a nível Ibérico, também se
confirma neste último nível. Ou seja, este estudo propõem-se analisar a posição da
economia portuguesa no contexto Ibérico nomeadamente se as regiões portuguesas
(periféricas a nível Europeu) também o são a nível Ibérico.
O estudo desta questão revela-se importante por dois motivos. Em primeiro lugar,
porque não basta suspeitar que a economia portuguesa é periférica da economia espanhola.
De facto é estritamente necessário confirmar (ou não) esta ideia, e principalmente aceder de
forma exacta a posição da economia portuguesa na Península Ibérica. Em segundo lugar
porque só com base num conhecimento profundo do ‘real estado’ da
centralidade/periferialidade a nível Ibérico é possível construir políticas capazes de
enfrentar os reais problemas do desenvolvimento económico de Portugal e mais uma vez da
sua posição na Europa e na Península Ibérica.
Com este intuito em mente, partimos do conceito de ‘potencial/acessibilidade
económico’1 (desenvolvido inicialmente por Harris, 1954, e utilizado também pelo estudo
acima referido de Keeble et al., 1988), para analisarmos a questão da Centralidade e
Periferialidade na Península Ibérica.
Indicadores de acessibilidade têm tido uma série de aplicações em pesquisa empírica em
várias áreas da economia. Nomeadamente, em economia dos transportes para avaliar
investimentos em infra-estruturas de transporte (e.g. Vickerman, 1989)2; em economia
regional como método de cálculo do potencial de mercado de regiões (e.g. Keeble et al.,
1988)3; em economia internacional para prever os efeitos espaciais da integração
económica (e.g.: Clark et al., 1969)4; e mais recentemente na literatura em geografia
1
Alguns autores preferem chamar a esta abordagem acessibilidade económica enquanto que outros potencial
económico. No âmbito deste estudo deve-se entender estes dois termos como equivalentes.
2
Por exemplo no caso Europeu estes indicadores foram usados para avaliar a importância económica das redes-
transeuropeias.
3
Foi a partir deste estudo de Keeble et al. (1988) que se institucionalizou no discurso da União Europeia a ideia
de Centro e Periferia na Europa. O estudo foi aplicado a todas as regiões NUTS 2 da Comunidade Europeia a
doze.
4
A ideia de base desta abordagem é a de que a integração económica reduz a distância económica e portanto
favorece as regiões que comparativamente ganham mais em temos de acessibilidades a mercados.
Acessibilidade Económica e Bem-estar: Evidência da Península Ibérica 191
económica estes tem sido embebidos em regressões econométricas para analisar os

determinantes de localização industrial e crescimento regional (e.g.: Hanson, 1998; e Pires,
1999)5.
A ideia central subjacente ao conceito de acessibilidade económica é a da que os
agentes económicos preferem escolher como localizações para o desenvolvimento das suas
actividades económicas as regiões com melhor acesso aos mercados (quer estes sejam
mercados de trabalho, consumidores, industriais, ou de conhecimento). Baseado nesta ideia
Harris (1954) desenvolveu nos anos cinquenta índices de potencial económico para analisar
e explicar padrões de localização da actividade económica nos EUA. A aplicação destes
índices à economia Americana demonstrou que as regiões mais industrializadas dos EUA
eram as regiões com melhor classificação nos índices de acessibilidade. Para Harris este
facto confirma que a produção e a população tendem a concentrar-se nas regiões mais
desenvolvidas, e que as localizações nestas regiões tem melhor acesso aos mercados que
localizações fora destes centros industriais. Estudos posteriores também vieram a confirmar
a existência desta correlação entre desenvolvimento económico regional e índices de
potencial económico. Desde então, este tipo de índices tem sido utilizados por muitos
economistas para descrever as vantagens de proximidade aos grandes mercados e para
estudar padrões de localização industrial e económica.
No entanto a análise de acessibilidade económica acarreta uma outra implicação óbvia:
a possibilidade de processos de causalidade cumulativa, no sentido que aglomeração
económica poder ser um processo que se auto-reforça a si próprio. As empresas querem se
localizar onde acessibilidade económica é superior, mas estes mercados também tenderão a
ser maiores nas localizações onde empresas já estão lá instaladas. Como tal e pelo menos
em teoria existe a possibilidade de processos de crescimento ou declínio regional que se
auto-reforçam.
Uma medida típica de acessibilidade económica em pesquisa empírica é o já referido
índice de acessibilidade económica de Harris (1954). Para uma dada região r o potencial
( Pr ) é definido como:
Pr = ∑ c h
s
Ms f crs (1)
onde Ms é a “massa” na região s , crs a distância económica entre a região r e s , e f ccrs h

uma função decrescente da distância6. Como com Pr se pretende medir o potencial de
mercado de uma região a variável Ms denota a dimensão de mercado na região s 7. A
distância económica crs pode representar vários custos relacionados com a distância
(transporte, tempo, tarifas, tarifas não-pautais e diferenças culturais).
No entanto, apesar dos índices de acessibilidade económica se terem revelado uma
ferramenta importante em pesquisa espacial aplicada, estes não podem ser justificados com
base em modelos teóricos económicos. Por outro lado também não é possível através destes
dizer quais serão as consequências no rendimento, emprego, ou até mesmo bem-estar
resultantes de uma alteração no potencial de mercado de uma localização (resultante por
5
Estes dois estudos (no caso de Hanson aplicado aos counties dos EUA e no de Pires às regiões NUTS 3 da
Espanha) confirmam empiricamente a importância dos rendimentos crescentes à escala, custos de transporte e
padrões de procura como determinantes da localização da actividade económica.
f (c rs ) = e − βcrs ou f (c rs ) = c rs
−δ são especificações típicas para a função de distância.
6
7
Usualmente mede-se a “massa” de uma região pelo seu produto regional.
exemplo de uma diminuição nos custos de transporte). Para além do mais o conceito
intuitivo acima referido de causalidade cumulativa também não têm expressão directa neste
tipo de índices ad-hoc. Como tal é desejável desenvolver índices de acessibilidade que
tenham um claro significado económico, e que sejam derivados explicitamente de modelos
teóricos de interacção económica espacial.
O objectivo deste trabalho é duplo. Em primeiro lugar demonstrar que é possível
desenvolver índices de acessibilidade económica no enquadramento de um modelo espacial
de concorrência imperfeita com rendimentos crescentes à escala e custos de transporte. Para
tal vamos seguir o trabalho de Bröcker (1998b,c) em equilíbrio geral espacial, e em
particular Bröcker (1998a) que analisa a questão da acessibilidade em equilíbrio geral. Em
segundo lugar estudar a questão da acessibilidade económica (e em consequência a
problemática dos centros e periferias) na Península Ibérica.
Para além desta secção este trabalho tem mais oito secções. Na secção seguinte é
apresentado o modelo de geografia económica. A secção 3 introduz a questão da
acessibilidade no contexto do modelo apresentado. É demonstrado que algumas das
relações económicas presentes no modelo (como os preços) estão relacionados com
indicadores de potencial de mercado e medidas de bem-estar teórico. Com base nestas
derivações a secção 4 discute o modelo em causa. Na secção 5 o método de calibração do
modelo é analisado, sendo esta posteriormente utilizada na secção 6 para descrever a
acessibilidade económica na Península Ibérica. Na secção seguinte com base nos resultados
da calibração são realizados algumas análises de sensibilidade e simulações para dados
regionais da Península Ibérica. O trabalho conclui com algumas considerações sobre a
abordagem aqui realizada.
2. Modelo de geografia económica
O modelo espacial utilizado neste estudo descende directamente dos modelos

desenvolvidos na última década pela chamada “Nova” Geografia Económica. Esta
literatura foi iniciada pelo artigo de Krugman de 1991 “Increasing returns and economic
geography”, e foi posteriormente desenvolvida por este e outros como Fujita et al. (1999).
O modelo que se vai apresentar considera um determinado número de firmas em cada
região, produzindo um cabaz de bens diversificados horizontalmente, usando como inputs
trabalho, capital, e um bem compósito intermédio. Existe neste cabaz um número potencial
elevado de tipos de bens produzíveis (apesar de em equilíbrio o mais provável ser que nem
todos os bens sejam produzidos). Firmas e consumidores percebem cada um dos tipos de
bens como substituto imperfeito de todos os outros tipos de bens. Cada firma em equilíbrio
produz apenas uma variedade, e estas entre si praticam concorrência monopolística. As
firmas maximizam os lucros tomando em conta a responsividade preço da variedade que
produzem. A entrada e saída de firmas é permitida, e o número de variedades existente no
mercado (e consequentemente o número de firmas) é determinado endogenamente pela
condição de lucro zero.
O bem intermédio é composto por todas as outras variedades existentes no mercado no
momento. Todas as variedades entram simetricamente na função de produção, o que
significa que não há preferência por variedades específicas. Seguindo Krugman (1993) a
produtividade marginal de uma variedade é decrescente, de tal modo que maior variedade
de bens diferenciados intermédios aumenta a produtividade.
O pagamento dos factores das firmas de propriedade regional vai directamente para os
consumidores dessa região, que gastam o seu orçamento em bens consumíveis. Os
consumidores no modelo também representam toda a procura final, incluindo consumo
público e investimento.
A dimensão espacial é introduzida neste modelo na forma tradicional, considerando que
tem que se incorrer em custos para fazer comércio. Estes custos de fazer comércio
aumentam com a distância entre os mercados consumidores e produtores. No entanto ao
contrário do usual nesta literatura, não são considerados custos de transporte na forma de
“iceberg”.
Dada esta estrutura, o modelo utilizado é um de “quase” equilíbrio geral no sentido em
que este é caracterizado pelos agentes afectados a cada região, tecnologias, preferências, e
dotações atribuídas a cada agente, pela delimitação de mercados e pela estrutura de
mercado.
2.1. Produção
A produção é modelada como um processo em dois estágios. No primeiro estágio as

firmas na região r produzem um bem homogéneo (bem primário), usando uma tecnologia
Cobb-Douglas e tendo como inputs, trabalho, capital e um bem compósito. Para uma firma
“price-taker” o custo mínimo por unidade de produto do bem homogéneo na região r é:
pr = µ r wrα zrβ qrγ , (2)

com α + β + γ = 1 , e onde wr denota a taxa salarial, zr a taxa de juro, e qr o preço do bem
compósito na região r . Os parâmetros α , β , e γ são as elasticidades parciais da produção.
Estes parâmetros igualam respectivamente a parte do trabalho, capital e bens intermédios
no valor da produção. Por sua vez o parâmetro µ r representa as diferenças regionais no
nível de produtividade. Dado um vector de inputs constante, o nível de produção é
proporcionalmente inverso a µ r .
No segundo estágio de produção, as firmas tomam o bem homogéneo como bem
intermédio para a produção de diferentes variedades de bens finais8. Para produzir x
unidades de uma variedade do bem diferenciado a firma necessita de f + cx unidades do
bem primário9. Esta é a conhecida função de custo linear, onde f denota os custos fixos, e
10
c o custo variável unitário (ambos expressos em termos de unidades do bem primário ).
Usando a relação de Amoroso-Robinson-Marshall, uma firma maximizadora do lucro
(tomando pr como adquirido) estabelecerá o preço dos bens finais tal que:
σ
πr = cpr (3)
σ −1
8
Esta estrutura de formalização da produção é conhecida por concorrência monopolística e deve-se a Dixit
Stiglitz (1977).
9
Ver Flôres (1997) para uma calibração dos parâmetros da função de custos. Os sectores considerados são:
têxteis e couros; madeiras; papel, produtos em papel e impressão; produtos minerais não metálicos; produtos
metálicos, máquinas, materiais e aparelhos eléctricos; produtos químicos; produtos alimentares, bebidas e
tabaco; material de transporte.
10
Como tal o bem primário é neste modelo o numerário.
onde σ é a elasticidade preço da procura (com σ > 1 ). Como com livre entrada e saída de
firmas os lucros em equilíbrio são zero. Portanto vamos ter:
f σ
c+ = c (4)
x σ −1
Resolvendo para x obtemos:
x=
f
c
a f
σ −1 (5)
Desta relação retira-se o número de variedades produzido na região r :

Xr
lr = (6)
σf
onde Xr denota a quantidade do bem primário produzido em r .

Uma unidade do bem compósito é constituída por variedades individuais de acordo com
um índice simétrico com elasticidade constante de substituição (índice CES). Expressando
este em forma dual vai significar que qs o preço por unidade do bem compósito na região s
é igual a:
1
L O 1−σ
qs = M∑ lr cπ rτ rs h
1−σ P
MM PP (7)
Nr Q
onde σ denota a elasticidade de substituição entre variedades11, e τ rs um factor mark-up
representando os custos de transferir os bens entre a localização r e s (com τ rs > 1).
2.2. Consumidores
Os consumidores em cada região r recebem como rendimento yr resultante do

empréstimo ás firmas regionais do trabalho e capital que possuem12. Todo o rendimento
obtido pelos consumidores é gasto na compra de bens de consumo. Os consumidores têm
preferências CES simétricas e homotéticas sobre as variedades com elasticidade de
substituição σ , a mesma que no agregado dos bens intermédios por parte das firmas.
Portanto com rendimento yr a utilidade máxima que um consumidor representativo pode
obter é:
y
ur = r (8)
qr
2.3. Custos do Comércio Inter-Regional
11
Se o número de variedades é muito grande, σ é também a elasticidade preço-procura para uma variedade
individual tal como na equação (3).
12
Na aplicação empírica deste modelo dado a ausência de dados, vamos ser obrigados a assumir que todo o
serviço de factores usado por firmas regionais vem de consumidores regionais. Na prática isto significa que
não são considerados fluxos inter-regionais de rendimentos de factores. Esta hipótese não é realista, pelo
menos em relação ao rendimento do capital, mas como não é conhecida a distribuição regional do rendimento
do capital não é possível ultrapassar esta inconveniência.
Como já foi referido os custos de fazer comércio pretendem significar não apenas
custos de transporte mas quaisquer custos que tornem mais onerosa a transacção dos bens
no espaço (custos de transporte, custos de comunicação, barreiras comerciais). Os custos de
transporte podem tanto depender da quantidade transportada como do valor dos bens. Neste
último caso estão alguns dos custos de transferência inter-regional relacionados
especialmente com os custos de informação, e seguros. Nas análises económicas
usualmente assume-se que os custos do comércio dependem apenas na quantidade dos bens
comercializados, no entanto neste modelo vamos optar por fazer os custos de transporte
depender no valor comercializado uma vez que simplifica a análise.
O modelo considera dois tipos de custos de comércio: custos relacionados com a
distância, e custos relacionados com os impedimentos ao comércio internacional. Se uma
região r pertence a um país k e a região s ao país l (com k ≠ l ) então o factor mark-up
representando os custos de transporte é:
c h
τ rs = exp ηgrs δ kl (9)
onde grs denota a distância geográfica entre duas regiões, η mede os custos por unidade de
distância em percentagem do valor do bem, e δ kl − 1 ≥ 0 o equivalente de custos de comércio
para todos os custos que resultam de um bem ter que ser exportado do país k para o país l .
Pretende-se capturar com este parâmetro todos os custos relacionados por exemplo com
tarifas, barreiras não pautais, e diferenças culturais13.
Os impedimentos ao comércio internacional podem ser expresso como
equivalentes de distância. Para um impedimento δ kl , o equivalente em distância é igual a
logcδ kl h / η . Podemos então escrever:
c h
τ rs = exp ηcrs (10)
com:
crs = grs +
c h
log δ kl
(11)
η
Nesta especificação dos custos de transferência inter-regional é também assumido que

os custos de transporte de bens que chegam à região s , são pagos a uma empresa
transportadora, que consome o bem compósito, não tendo qualquer lucro nesta actividade.
Ou seja, parte da despesa nos bens é usada para comprar o bem (quer este seja para
produção intermédia ou consumo final), e outra parte é para comprar bens que vão para
serviços de transporte.
2.4. Equilíbrio
Da especificação de tecnologias, preferências, e estrutura de mercado, é agora possível

derivar o equilíbrio para um sistema fechado de regiões. Considera-se que cada região está
dotada com um montante fixo de trabalho e capital, que pertence aos consumidores que
residem na região. Os consumidores gastam todo o seu rendimento dos factores,
13
De notar que δ kl = 1 para k = l , e δ > 1 para k ≠ l .
maximizando a sua utilidade. As firmas concorrem entre si sob concorrência monopolística

a la Chamberlain. O preço dos bens e factores, e o número de variedades na economia
ajustam até que os mercados atinjam o equilíbrio e os lucros sejam zero.
O equilíbrio do modelo consiste nas equações do preço do bem compósito (2) e (7), e
três condições de equilíbrio para o mercado de bens, mercado do trabalho, e mercado de
capitais. Denotando o valor do produto como Vr = Xr pr , e usando as equações (3) e (6), a
equação (7) pode ser escrita:
1
F I 1−σ FG IJ
1 σc 1−σ
qs = G ∑ φ Vr pr−σ τ 1rs−σ J com φ =
GH r JK H
f σ −1 K (12)
O equilíbrio do mercado de bens requer que para cada região r , o valor do produto Vr
seja igual ao valor da procura de bens nessa mesma região. Esta condição conduz-nos à
equação:
c
h−σ Vs
Vr prτ rs
Vr = ∑s ∑ Vt c ptτ ts h−σ (13)
Para chegar a este resultado de notar que a forma CES na equação (7) implica que a
procura para uma variedade da região r por unidade do bem compósito comprado em s
( drs ) seja:
F qs I σ
drs = GH π rτ rs JK (14)
Como tal a parte da região r no valor da procura em s é igual a:
lr drsπ r
=
c h −σ
Vr prτ rs
(15)
∑l d π t ts t ∑ Vt c ptτ ts h−σ
t t
Aqui é usado o facto de π r ser proporcional a pr de (3), e de que lrπ r ser proporcional a
Vr de (5). Por outro lado se o rendimento não é transferido entre regiões, então o valor da
despesa em bens na região s é igual a Vs , uma vez que as firmas gastam todo o seu
rendimento nos bens intermédios, custos de comércio e rendimento de factores. O
pagamento de custos de transporte, e o rendimento de factores vai todo por seu lado para a
compra de bens.
A procura de factores é obtida da equação (2) pelo Lema de Shephard. Logo as
condições para o equilíbrio do mercado de factores são:
αVr
Lr = (16)
wr
βVr
Kr = (17)
zr
As condições de equilíbrio do sistema resumem-se às equações (2), (12), (13), (16) e

(17). Estas equações determinam pr , qr , Vr , zr , e wr para todo o r , dado o stock de factores
Lr e Kr para todo o r .
3. Acessibilidade e Bem-Estar
Como já foi referido o índice ur definido na equação (8) é o índice de bem-estar

regional neste modelo. Variações relativas de ur são iguais às medidas Hicksianas de
variação relativa do bem-estar. Numa análise estática comparativa os rendimentos regionais
para o equilíbrio de referência e para o equilíbrio contra-factual são obtidos multiplicando
o preço dos factores pela dotação de factores. Sendo urR e urC respectivamente a utilidade no
equilíbrio de referência e no equilíbrio contra-factual, então o ganho de bem-estar no
equilíbrio de referência em comparação com o equilíbrio contra-factual é:
uC
VERr = rR −1 (18)
ur
onde VERr é a medida de Hicks de variação equivalente relativa (VER). A VER representa
a variação equivalente expressa como a parte no rendimento do equilíbrio contra-factual.
Similarmente o ganho de bem-estar no equilíbrio contra-factual em comparação com o
equilíbrio de referência é:
uR
VCRr = 1 − rC (19)
ur
onde VCRr é a medida de Hicks de variação compensada relativa (VCR). A VCR é a

variação compensada expressada como percentagem no rendimento do equilíbrio de
referência.
O índice de bem-estar depende portanto nas dotações, tecnologias, e localização. Para
isolar os efeitos da localização, pode-se questionar qual o retorno real máximo em factores
de produção que uma firma representativa pode pagar em diferentes localizações, desde que
a tecnologia desta firma seja a mesma independentemente da região onde esta está
localizada.
Definindo a tecnologia da firma pela sua função de custo (2) mas substituindo ur por
uma constante temos:
_
pr = µ wrα zrβ qrγ (20)
Pode-se pensar nesta tecnologia como uma função Cobb-Douglas de dois níveis. No
nível mais baixo os serviços de trabalho e capital são combinados com um serviço de factor
com preço mr . No nível mais elevado este serviço do factor juntamente com o bem
intermédio produz o bem final. Então equivalente à equação (20), vai-se ter para o nível
mais baixo:
1
e j
mr = wrα zrβ α + β (21)
Para o nível mais elevado pode-se escrever:

_
pr = µ m1r −γ qrγ (22)
Daqui pode-se obter o retorno real do factor (ou seja o índice de acessibilidade
económica deste estudo):
1
F p I 1−γ
Fr = r = G _ r J
m
qr G
H µ qr JK
(23)
Portanto, Fr é uma função crescente com pr qr , que por seu lado é uma combinação de
um potencial da procura e oferta:
1 1
pr
= Srσ −1 Drσ (24)
qr
com o índice potencial de oferta:

Sr = ∑ φV p σ exp a1 − σ fηc
s
−
s s sr (25)
e com o índice potencial de procura:

Dr = ∑ b V expc−σηc h
s
s s rs (26)
onde bs é um índice potencial inverso:

1
bs = (27)
∑t
Vt pt−σ exp −σηcts c h
Como facilmente se pode ver Sr e Dr são similares aos índices potencial de mercado
padrão. A modificação em Sr é que a oferta é ponderada por ps−σ , ou seja a oferta tem
menor ponderação quanto maior o preço. Similarmente a modificação em Dr é a de que a
procura é ponderada por uma função potencial inversa, de tal modo que a procura é menos
importante quanto maior o potencial de ofertas concorrentes14. Deste modo Fr (equação
(23)) define o indicador de acessibilidade a utilizar neste estudo.
Outro facto interessante deste modelo é que ps−σ e bs são os factores compensadores de
um modelo gravitacional com duas restrições descrevendo os fluxos de comércio no
modelo. Pelas equações (13), (27) e (10) o valor do comércio entre duas regiões é:
c
trs = Vr pr−σ exp −σηcrs Vs bs h (28)
De (13) e (27), tem-se que trs deve somar para Vr e Vs . Logo obtém-se as duas restrições:
Vs = ∑t
r
rs (29)
14
De notar que funções de distância diferentes aparecem em (25) e (26). No entanto nenhuma delas é a habitual
forma “iceberg”.
Vr = ∑t
s
rs (30)
4. Discussão do Modelo
Nas secções anteriores foi apresentado um indicador de acessibilidade semelhante à

formulação clássica em economia regional. No entanto este ao contrário das medidas
tradicionais de potencial de mercado, tem uma base teórica bem formada num modelo
espacial de concorrência imperfeita.
Para além desta vantagem teórica, este índice tem mais duas vantagens. Em primeiro
lugar a parametrização do modelo é menos arbitrária do que na abordagem clássica. De
facto os padrões de acessibilidade calculados pela forma tradicional podem modificar-se
bastante com alterações nos parâmetros da função de distância (ver (1))15. Como as
acessibilidades não são directamente observáveis não é possível estimar ou calibrar
directamente este parâmetro. No entanto na abordagem aqui apresentada como os
parâmetros em causa aparecem em equações descrevendo fluxos de comércio, é possível
verificar a calibração contra dados da economia real (como comércio inter-regional).
Por outro lado este modelo permite fazer análises de estática comparativa. Por exemplo
é possível avaliar os efeitos de maior integração económica na acessibilidade, rendimento
dos factores e bem-estar. A secção 2.4 descreve como avaliar alterações na remuneração
dos factores; enquanto que a secção anterior explica como medir as alterações de bem-
estar. No presente estudo apenas vamos avaliar alterações na acessibilidade e no bem-estar
consequência da evolução do nível de integração económica.
5. Calibração
O modelo é calibrado para um sistema de 20 regiões NUTS 2 que compõem a Península

Ibérica16. O quadro 1 apresenta uma lista das regiões NUTS 2 consideradas neste estudo.
Os dados usados para a calibração são emprego total por região, distâncias inter-regionais,
e PIB regional17. Todos os dados utilizados referem-se ao ano de 1994. Com a excepção
dos dados das distâncias inter-regionais, todos os restantes foram retirados da base REGIO
da EUROSTAT18. Os dados da distância inter-regional foram compilados do CD-ROM
“Route 66 Europe”. Este programa contém uma informação geográfica vasta sobre todo o
Continente Europeu, que inclui entre outras, informações sobre distâncias por estrada entre
cidades e localidades europeias. Como tal para medirmos a distância inter-regional
utilizamos a distância em quilómetros por estrada entre cada par de regiões NUTS 2. A
distância entre duas regiões é deste modo medida como distância mais curta entre as
cidades principais de cada uma das regiões NUTS 2 em causa. Considera-se a cidade
15
Bröcker (1989) discute esta questão.
16
Ver EUROSTAT (1996) para definição das regiões NUTS de Portugal e Espanha. As regiões insulares de
Portugal e Espanha não são consideradas para manter continuidade espacial na unidade geográfica de análise.
17
Bröcker (1998a) utiliza apenas PIB a nível do país, como tal µ r também é apenas calibrado a nível nacional. O
presente estudo calibra este parâmetro a nível regional.
18
Ver EUROSTAT (1996) para uma definição das variáveis utilizadas.
principal de uma região NUTS 2, as mesmas que a EUROSTAT enumera nas definições da
base REGIO, o que em geral também corresponde à cidade mais populosa dessa região19.
Quadro 1
Península Ibérica: Regiões NUTS 1 e NUTS 2
País NUTS 1 NUTS 2

Espanha Noroeste Galiza
Principado das Astúrias
Cantábria
Nordeste Pais Vasco
Navarra
La Rioja
Aragão
Comunidade de Madrid Comunidade de Madrid
Centro Castela e Leão
Castilla-la Mancha
Estremadura
Este Catalunha
Comunidade Valenciana
Sul Andaluzia
Região de Múrcia
Portugal Continente Norte
Centro
Lisboa e Vale do Tejo
Alentejo
Algarve
Nota: Fonte EUROSTAT (1996)
Para calibrar o modelo apresentado nas secções anteriores é necessário conhecer o valor
de alguns dos parâmetros do modelo. Estes são: a elasticidade de substituição entre
variedades na procura intermédia e final ( σ ); a percentagem dos custos da distância por
unidade de distância ( η ); as elasticidades de produção ( α , β e γ ); os impedimentos ao
comércio internacional ( δ kl ); e o inverso do nível de eficiência regional ( µ r ). A estratégia
por nós adoptada é similar à utilizada neste tipo de estudos empíricos: fixa-se o valor de
alguns parâmetros do modelo enquanto que os restantes obtêm-se por calibração. Deste
modo para os seis primeiros parâmetros acima referidos são fixados valores de acordo com
os que usualmente são apontados para estes em estudos econométricos, estatísticas da
indústria ou na literatura teórica e empírica. No caso de µ r este é calibrado para as vinte
regiões consideradas. A calibração basicamente fixa estes parâmetros juntamente com todas
as variáveis endógenas do modelo, que são o preço dos bens ( pr e qr ), o preço dos factores
( wr e zr ), os valores da produção regional ( Vr ), e comércio inter-regional ( trs ), para obter
um valor para os parâmetros calibrados. Posteriormente com o modelo completo pode-se
19
Exemplificando, considere-se as regiões NUTS 2 de Madrid e Lisboa e Vale do Tejo. A distância entre estas
duas regiões não é mais que a distância mais curta em quilómetros por estrada entre a cidade de Madrid e
Lisboa.
realizar todo o tipo de análises económicas, tal como também simular alterações de política
ou ambiente económico.
Começando pelo parâmetro σ e η é possível observar que na equação (28) o produto ση
representa o parâmetro de distância usualmente utilizado nos modelos gravitacionais com
uma função de distância exponencial. Estimativas deste parâmetro, estão pelo menos para o
comércio internacional, disponíveis na literatura (ver Bröcker e Rohweder, 1990). No
entanto a incidência geográfica de Bröcker e Rohweder refere-se apenas ao comércio inter-
regional e não internacional. Bröcker (1984) estima o parâmetro em causa para os fluxos de
comércio entre regiões de seis países europeus. Infelizmente os países Ibéricos não estão
incluídos neste estudo. Para além do mais este estudo refere-se apenas ao ano de 1970, os
resultados podem portanto estar desactualizados20. Por seu lado Pires (1999) no contexto de
uma “função potencial de mercado” semelhante à de Harris (1954) e de um sistema de
equações representantes do equilíbrio do modelo de geografia económica de Krugman
(1991) estima um impacto da distância entre 0.0002 e 0.0006 por quilómetro para a
economia espanhola21. Parece pois adequado aceitar uma estimativa para este parâmetro
que se situe neste intervalo. Neste estudo optamos por um valor para o parâmetro de
distância de ση =0.0004 por quilómetro.
Examinemos agora estes dois parâmetros em separado. Se consideramos apenas os
custos de transporte convencionais22, o valor para η seria segundo a literatura pequeno em
percentagem. De facto estudos empíricos nesta área reportam custos de transporte médios
na indústria transformadora que não excedem dois porcento do valor das vendas (Keeble,
Offord and Walker, 1988; Bröcker, 1998a, Davis, 1998). Mas isto implicaria um valor para
o parâmetro σ na ordem das centenas, o que é pouco plausível pois isto significaria
estarmos a assumir substituibilidade perfeita no comércio inter-regional. Portanto torna-se
necessário considerar também os custos de transporte não-convencionais. Se tomarmos
apenas os custos logísticos médios relacionados com a distância, estes são segundo Bröcker
(1998a,b) e Davis (1998) superiores a vinte por cento do valor das vendas. Mas para além
destes existem custos também relacionados com a distância que não são considerados nos
custos de logística como os custos de contactos “cara-a-cara”. Para além disso a distância
média no comércio inter-regional é muito menor que milhares de quilómetros. Portanto um
valor superior a vinte porcento para η é justificável. Seguimos então Bröker (1998a) ao
considerar um η igual 0.00004 por quilómetro. Consequentemente atribuímos um valor de
10 para o σ 23.
Como já foi referido, as elasticidades α , β e γ da função de produção, igualam
respectivamente as percentagens do trabalho, capital, e bens intermédios no valor do
produto. Os valores para estes parâmetros são retiradas das contas nacionais de Portugal e
20
McCallum (1995) também analisa através de um modelo gravitacional os fluxos internacionais e inter-regionais
de comércio, só que a incidência espacial deste estudo é a América do Norte (EUA e Canadá).
21
O estudo incide sobre os anos 1981-1982-1983; 1988-1989-1990; e 1993-1994-1995.
22
Os custos de transporte são normalmente divididos em duas categorias: custos de transporte convencionais; e
custos de transporte não-convencionais (Davis, 1998). Os primeiros referem-se aos custos associados com
seguros, fretes, e tarifas. Os segundos são outros custos (normalmente de mais difícil mensuração)
relacionados com barreiras não tarifárias e custos de informação.
23
Outros estudos de equilíbrio geral e parcial como o de Gasiorek, Smith, e Venables (1994); Bröcker (1998a,b);
e Bröcker e Jäger-Roschko (1996) também trabalham com uma estimativa de 10 para o parâmetro elasticidade
de substituição entre variedades.
Espanha, onde para ambos os países estas são aproximadamente 28 porcento, 12 porcento,
e 60 porcento respectivamente.
Tomando em consideração o valor para δ kl pode-se argumentar que este parâmetro em
princípio não deveria entrar na presente análise, uma vez que Portugal e Espanha
pertencem ambos a uma acordo de integração regional que prevê livre circulação de bens
(para além de pessoas e serviços). Como tal δ kl deveria ser igual a um. No entanto Bröcker
(1984) estima para os impedimentos aos fluxos inter-regionais de comércio entre países da
Comunidade Europeia, uma distância equivalente a 375 quilómetros. Dado o valor para o
custo da distância por unidade de distância ( η =0.00004 por quilómetro), este corresponde a
umδ kl de 1.19. Ou seja, uma tarifa equivalente de 19 porcento24. Como tal no presente
estudo consideramos que este apenas é igual à unidade quando duas regiões pertencem ao
mesmo país. Se duas regiões pertencem a diferentes países este parâmetro toma um valor
superior à unidade, nomeadamente e como referimos atrás uma tarifa equivalente de 19
porcento.
O parâmetro que resta ( µ r ) é calibrado, resolvendo um sistema de equações que
determina este parâmetro assim como todas as variáveis endógenas, dado o valor dos
parâmetros discutidos nos parágrafos anteriores. Este sistema de equações fixa µ r de modo
a que na solução de equilíbrio, o PIB regional de uma região iguale o PIB regional
observado para esta mesma região.
Para derivar este sistema de equações considere-se inicialmente que Vr é conhecido para
todo o r . Já foi referido que trs respeita as equações de um modelo gravitacional com duas
restrições (equação (13)) ou as equações (28), (29) e (30):
trs = ar Vr bs Vs frs (31)
Vs = ∑t
r
rs (32)
Vr = ∑t
s
rs (33)
com ar = pr−σ e frs = τ rs

−σ
. Para se proceder à calibração vai-se assumir que o preço do capital
( zr ) é constante em todas as regiões25. Considera-se para tal uma taxa de remuneração do
_
capital, z =0.0526. Deste modo para uma região r obtém-se da equação (2):
24
Pode-se pensar que este valor é demasiado elevado uma vez que os impedimentos ao comércio quase
desapareceram no mercado Ibérico. No entanto considerando por exemplo o Canadá e os EUA que também
pertencem a um acordo de integração regional (para além de culturas e língua muito semelhantes), McCallum
(1995) estima impedimentos ao comércio entre estes dois países superiores ao aqui considerados. Seguindo o
valor avançado no parágrafo anterior para δ kl , o comércio intra-regiões do mesmo país é 5.6 vezes superior
ao comércio intra-regiões trans-fronteiriço. Se considerarmos o estudo de McCallum, o comércio intra-regiões
do Canadá é 22 vezes superior ao comércio intra-regiões do Canadá e EUA. Tomando a elasticidade de
substituição por nós considerada ( σ =10) isto seria equivalente a uma tarifa equivalente de 36 porcento.
25
Esta hipótese pode ser justificada se assumirmos que existe perfeita mobilidade do capital inter-regional, o que
pode de facto ser irrealista e muito discutível. Mesmo supondo que tal é verdade, apenas é possível argumentar
em favor de uma taxa de retorno igual nos activos comercializáveis. A taxa real de retorno em capital instalado
pode desviar-se da taxa igualizada entre regiões, porque a instalação deste tipo de capital é muito onerosa e
1
F −γ I α
Gpq J
wr = G r r J (34)
GH _z β µ r JK
_
onde z denota o preço constante do capital. Pode-se pois concluir da equação (16):
Lw
Vr = r r (35)
α
Substituindo em (35) wr pelo seu valor em (34) obtêm-se:
1 γ
− 1
Vr = ϕ r Lr prα qr α com ϕ r = (36)
F _ β I1 α
αG z µr J
GH JK
Neste momento estamos em condições de calibrar µ r fazendo uso da informação
disponível quanto ao PIB regional ( Yr ). Como tal µ r é calibrado (ou melhor dizendo, ϕ r é
calibrado, que é função de µ r e outras constantes), de modo a que o PIB regional observado
coincida com PIB regional do modelo. Isto vai implicar:
b
Yr = α + β Vr g (37)
ou substituindo Vr pelo seu valor em (36):
1 γ
−
b g
Yr = α + β ϕ r Lr prα qr α (38)
Esta equação dá-nos ϕ r e também Vr :
Yr
Vr = (39)
α +β
Deste modo o processo de calibração por nós adoptado parte da constatação de

que Vr = ∑
trs , então das equações (31), (32) e (33) tem-se:
s
∑t s
rs = ar Vr ∑ cb V f h ⇔ V = a V ∑ c b V f h
s
s s rs r r r
s
s s rs (40)
rígida no tempo (ver Hayashi, 1982). No entanto também se pode defender que estes desvios podem
desaparecer no longo prazo.
26
Bröcker (1998a) para além desta hipótese faz uma hipótese adicional. Ele supõe que a taxa de participação
laboral é constante em todas as regiões. Assim em Bröcker a força de trabalho regional ( Lr ) é proporcional à
população ( B r ), o que significa que Lr = λBr , com taxa de participação λ . Neste trabalho optou-se por
seguir as condições do modelo e utilizar dados do emprego regional (em vez de dados para a população como
em Bröcker) para não ser necessário fazer hipóteses adicionais (ver equação (16)).
ou seja:
1
ar = (41)
∑c
s
bsVs frs h
Como ar = pr−σ , obtém-se pr :
1
F Iσ
pr = G ∑ bsVs frs J
GH s JK (42)
Depois de se achar Vr por (39), fazendo pr0 = qr0 = 1, é possível achar bs0 de (27), p1r de
(42) e q1s de (12)27. Tendo estes, faz-se iterações sucessivas até se obter convergência em pr
e qs 28. Para concluir o processo de calibração, substitui-se pr e qs finais em (38) para obter
µ r . Tendo este é então possível calcular o índice de acessibilidade a partir de (23), o índice
de bem-estar de (8), assim como simular alterações de política económica e depois calcular
as variações no bem-estar resultante desta a partir de (18).
27
O super-escrito refere-se ao número da iteração.
28
O processo de calibração utilizado neste estudo difere do de Bröcker (1998a). Bröcker utiliza métodos de
“iterative scaling” (ver Darroch and Ratcliff, 1972) para encontrar a solução do sistema de equações. Ver
Bröcker (1998a) para mais pormenores
6. Acessibilidade na Península ibérica
A figura 1 ilustra os resultados do indicador de acessibilidade definido na equação (23),

escalado para a unidade. As linhas unem localizações com o mesmo potencial de mercado,
enquanto que as setas sinalizam as regiões com maior acessibilidade económica.
Figura 1. Linhas de Acessibilidade na Península Ibérica
A medida de acessibilidade para as regiões NUTS 2 de Portugal e Espanha também é

apresentada no quadro 2. O indicador de acessibilidade varia entre um mínimo de 0.60 na
região do Algarve, e um máximo de 1.44 na região de Madrid. Como tal dada a
distribuição espacial dos mercados, uma firma representativa em Madrid pode pagar aos
factores de produção localizados nesta região, cerca de duas vezes e meia mais do que pode
pagar na região do Algarve, se a tecnologia usada nas duas localizações fosse semelhante.
Quadro 2
Acessibilidade e Bem-Estar
Regiões pr qr Fr ur
Galiza 0.96693 1.025633 0.855389 0.833021
Principado de Astúrias 0.967127 1.020983 1.045477 0.997377
Cantábria 0.9791 1.00766 1.112158 1.058286
Pais Vasco 0.990085 0.9971 1.224864 1.295616
Comunidade Foral de Navarra 0.975864 1.009597 1.15798 1.329812
La Rioja 0.987651 0.996715 1.118597 1.247242
Aragão 0.99387 0.989953 1.150229 1.249279
Comunidade de Madrid 1.028415 0.958217 1.305891 1.425764
Castela e Leão 1.009861 0.97505 1.081449 1.049546
Castilla-la Mancha 1.012443 0.972991 1.097576 0.929932
Estremadura 0.986254 1.000851 1.008457 0.781893
Catalunha 1.007979 0.981039 1.230091 1.374406
Comunidade Valenciana 1.000017 0.986062 1.099615 1.048418
Andaluzia 1.000464 0.99095 1.089567 0.801896
Múrcia 0.960992 1.02619 1.050313 0.924575
Norte 0.983178 1.006632 0.638891 0.672055
Centro (P) 0.983786 1.004836 0.615436 0.644344
Lisboa e Vale do Tejo 0.98764 1.003775 0.784866 0.973198
Alentejo 0.984499 1.003987 0.678741 0.627629
Algarve 0.950778 1.04178 0.654414 0.735711
Considere-se a Espanha e Portugal isoladamente. No primeiro caso a região com menor

acessibilidade é a região da Galiza, sendo Madrid a região com mais potencial de mercado.
No segundo caso a região com melhor acessibilidade é Lisboa enquanto que a de pior grau
da acessibilidade é o Algarve. De notar no entanto que Lisboa apesar de ser a região de
Portugal com melhor acessibilidade na Península Ibérica, esta apenas tem uma
acessibilidade marginalmente superior à da região menos central da Espanha (0.87 de
Lisboa contra uns 0.85 da Galiza).
Para além do máximo global em Madrid, de destacar outros máximos locais à volta das
regiões da Catalunha (1.34) e do País Vasco (1.32). A actividade económica na Península
Ibérica parece pois estar fortemente enviesada para três centros geográficos à volta de
Madrid, Catalunha e País Vasco.
O quadro 2 também apresenta os níveis regionais de utilidade tal como definido na
equação (8), escalada para uma média da unidade. A utilidade varia entre um máximo de
1.57 em Madrid a um mínimo de 0.55 no Alentejo. No caso da Espanha a região NUTS 2
com menor nível de utilidade é a região da Estremadura (0.73). Em Portugal a região com
maior utilidade é Lisboa (1.07). Como tal em termos de ranking das regiões, os resultados
do índice de bem-estar são muitos semelhantes aos do índice de acessibilidade. No entanto
existe uma importante excepção: a região de Lisboa. De facto enquanto que no ranking de
acessibilidade Lisboa ocupa o décimo quinto lugar, no ranking de bem-estar sobe para o
oitavo lugar. Mais à frente vamos discutir esta ocorrência.
7. Análise de sensibilidade
O método de análise empírica utilizado neste estudo, a calibração, é criticada por várias
razões. Em primeiro lugar a calibração é feita em relação a um ano base em que se supõem
que as condições de equilíbrio do modelo se verificam na economia, o que como é óbvio é
bastante irrealista pensar que tal ano exista. Por outro lado temos o problema da escolha das
equações do modelo para a calibração. Desde logo se coloca a questão de quais as equações
do modelo com mais significado económico, uma vez que para diferentes selecções de
equações a calibrar, se pode obter diferentes valores para os parâmetros. Além disso muitas
vezes tem que se escolher valores para certos parâmetros (como por exemplo economias de
escala), e estes devem ser obtidos de preferência através de estudos econométricos. Por
último e confrontando a econometria e a calibração (apesar de todas as críticas que se
podem apontar também à primeira), na econometria utiliza-se métodos estocásticos, o que
implica intervalos de confiança para os parâmetros calculados, tal não é o caso na
calibração. Deste modo na calibração não se pode ter certeza da precisão dos valores
obtidos.
Sendo assim os resultados da calibração de modelos devem sempre ser encarados com
muita precaução. De facto estes modelos apesar de permitem quantificar interacções entre
variáveis, e a resposta da economia a alterações de política; não permitem testar a teoria.
Dados estes argumentos é muito aconselhável levar a cargo testes de sensibilidade para se
ter uma ideia da robustez dos resultados.
A questão que se coloca neste caso é a que parâmetros testar a sensibilidade. Em
principio todos os parâmetros calibrados e para os quais foram atribuídos valores deviam
ser sujeitos a um teste de sensibilidade (i.e. os parâmetros σ, η, α, β, γ, δ kl e µ r ). No
entanto usualmente em modelos semelhantes aos por nós utilizado, i.e. modelos em
concorrência monopolista, o parâmetro mais sujeito a este tipo de análise é o da
elasticidade de substituição (ver por exemplo Smith e Venables, 1988; Flôres, 1997; e
Gasiorek et al., 1994). De facto como demonstrado em apêndice este tipo de modelos de
concorrência imperfeita são bastante sensíveis a diferentes valores do parâmetro
elasticidade de substituição. Como tal os resultados abaixo apresentados da análise de
sensibilidade só se reportam a este parâmetro29.
Com este objectivo em mente, o modelo foi re-calibrado para um parâmetro de
elasticidade de substituição de σ = 5 e σ =15. No primeiro caso o valor de σ corresponde a
um parâmetro de distância de ση = 0.00023, e no segundo caso um ση =0.00069. Deste
modo com σ =5 é considerada uma menor elasticidade de substituição entre variedades, e
um menor impacto da distância nas variáveis em causa (ver quadro 3). Já com σ =15
considera-se uma maior elasticidade de substituição entre variedades, e um maior impacto
da distância no potencial de mercado de uma região (ver quadro 4).
29
A análise de sensibilidade dos restantes parâmetros pode ser obtida através do autor.
Quadro 3
Acessibilidade e Bem-Estar ( σ =5)
Galiza 0.897456 1.038561 0.840016 0.818575
Cantábria 0.946382 0.993061 1.122204 1.06852
Pais Vasco 0.955444 0.98574 1.232086 1.304051
La Rioja 0.967164 0.972322 1.140265 1.272195
Aragão 0.978084 0.965186 1.173157 1.274985
Castela e Leão 0.985058 0.948396 1.105641 1.073695
Estremadura 0.937544 0.987161 1.016747 0.788807
Catalunha 0.966948 0.993578 1.207793 1.350334
Andaluzia 0.940395 0.992841 1.081411 0.796402
Múrcia 0.935337 1.006174 1.06522 0.938294
Norte 0.871808 1.044112 0.612522 0.644719
Centro (P) 0.87817 1.035288 0.59401 0.622291
Alentejo 0.871272 1.044369 0.648852 0.60037
Algarve 0.826636 1.103304 0.614467 0.691242
Quadro 4
Acessibilidade e Bem-Estar ( σ =15)
Galiza 0.973614 1.006991 0.868985 0.846287
Cantábria 0.976283 1.003291 1.114142 1.060192
Pais Vasco 0.990609 0.98861 1.232249 1.30342
La Rioja 0.97679 0.998274 1.114002 1.24213
Aragão 0.982454 0.989463 1.147831 1.24672
Castela e Leão 0.995426 0.975227 1.078469 1.046688
Estremadura 0.955669 1.018545 0.9884 0.766357
Catalunha 1.007669 0.955797 1.259354 1.407117
Andaluzia 0.994389 0.974271 1.105374 0.813551
Múrcia 0.960661 1.02049 1.053467 0.927377
Norte 0.976522 1.012464 0.633582 0.666485
Centro (P) 0.971855 1.015918 0.607171 0.635695
Alentejo 0.950752 1.036621 0.655707 0.606326
Algarve 0.912745 1.088136 0.624929 0.702578
Tal como ilustrado no quadro 3 (para σ =5) e quadro 4 (para σ =15) os resultados
obtidos anteriormente não se alteram consideravelmente. De facto é possível constatar que
não se verifica qualquer variação qualitativa nos resultados, a própria variação quantitativa
não é significativa.
Pode-se pois, concluir que no que concerne ao parâmetro elasticidade de substituição e
distância, os resultados da calibração não demonstraram ser muito sensíveis30.
8. Alterações de Política
A alteração de política considerada nesta secção supõe um mercado Ibérico totalmente

integrado. Na prática isto equivale a fixar o parâmetro de impedimentos ao comércio
internacional ( δ kl ) igual à unidade para todas as regiões NUTS 2 e repetir o processo de
calibração acima descrito. O quadro 5 apresenta os resultados desta simulação.
30
De referir que análises de sensibilidade aos restantes parâmetros também confirmam a robustez dos resultados
da secção anterior.
Quadro 5
Acessibilidade com Mercado Ibérico Integrado ( δ kl =1, ∀r )
Região pr qr Fr ur
Galiza 0.940432 1.020205 0.853028 0.828588
Cantábria 0.966559 1.086228 1.016717 0.971344
Pais Vasco 0.978418 0.918034 1.316397 1.392303
La Rioja 0.975754 1.152813 0.953192 1.066941
Aragão 0.984214 0.997903 1.131997 1.226205
Castela e Leão 0.992065 0.940104 1.078134 1.077036
Estremadura 0.946212 1.059129 0.944973 0.731047
Catalunha 1.002049 0.896896 1.339999 1.48743
Andaluzia 0.975363 0.898903 1.195334 0.874651
Múrcia 0.950343 1.049203 1.022066 0.894722
Norte 0.932433 0.940527 0.682622 0.711675
Centro (P) 0.932861 1.008873 0.611749 0.634971
Alentejo 0.916811 1.127263 0.603366 0.553074
Algarve 0.873442 1.129949 0.60221 0.671122
Tal como na secção anterior a alteração no valor de um parâmetro não causou alterações
significativas nos resultados da calibração base: em geral as regiões Espanholas são mais
centrais que as regiões portuguesas; a acessibilidade atinge um pico nas regiões de Madrid,
Catalunha, e País Vasco; o potencial de mercado atinge os valores mínimos nas regiões
portuguesas. Os resultados do índice de bem-estar anteriormente obtidos também são mais
uma vez confirmados.
Este exercício permite também avaliar a alteração no bem-estar das regiões espanholas
e portuguesas em consequência de uma integração total das duas economias. Neste sentido
foi calculado o indicador de bem-estar de Hicks de variação equivalente relativa, tal como
definido na equação (18)31. O quadro 6 apresenta os resultados.
31
O equilíbrio de referência usado na computação desta medida de bem-estar foi o calculado na secção 5. O
equilíbrio contra-factual considerado é o da presente secção.
Quadro 6
Variação Equivalente Relativa (VER)
Regiões VER
Galiza -0.00532
Principado de Astúrias -0.0108
Cantábria -0.08215
Pais Vasco 0.074626
Comunidade Foral de Navarra -0.05613
La Rioja -0.14456
Aragão -0.01847
Comunidade de Madrid 0.10182
Castela e Leão 0.026192
Castilla-la Mancha -0.04343
Estremadura -0.06503
Catalunha 0.082235
Comunidade Valenciana 0.05578
Andaluzia 0.090728
Múrcia -0.03229
Norte 0.058955
Centro (P) -0.01455
Lisboa e Vale do Tejo 0.099197
Alentejo -0.11879
Algarve -0.08779
Da análise do quadro conclui-se que a maior parte das regiões tem uma perda de bem-
estar com o cenário de integração completa. As regiões de Rioja na Espanha e Alentejo em
Portugal são as regiões com as maiores perdas relativas (14% em Rioja, e 12% no
Alentejo). Das vinte regiões NUTS 2 de Portugal e Espanha apenas oito regiões conseguem
obter ganhos de bem-estar. Curiosamente entre estas encontram-se as regiões mais
desenvolvidas e avançadas de ambos os países. No caso da Espanha, Madrid, País Vasco e
Catalunha; e no caso de Portugal, Lisboa e Norte32. Os ganhos nestas regiões variam entre
uns 10% em Madrid e uns 2.5% em Castela e Leão. Este resultado confirma as hipóteses da
“nova” geografia económica de efeitos de “lock-in” e casualidade cumulativa, em que as
regiões mais desenvolvidas tendem a construir novas vantagens com base nas antigas
32
De notar que a região Norte portuguesa não pode ser considerada no seu todo uma região desenvolvida. De
facto é possível dividir esta em uma região Litoral desenvolvida, com uma elevada densidade populacional, e
com um grande peso da industria transformadora no total da actividade económica desta região; e uma região
interior economicamente atrasada, com uma fraca densidade populacional, onde predomina principalmente a
actividade agrícola tradicional. Como tal a parte interior da região Norte é mais similar à região do Alentejo do
que à sua contra-parte Litoral.
alargando o fosso económico em relação às regiões mais atrasadas. Neste caso particular o
que provoca tais efeitos são uma diminuição dos custos de transação económicas.
9. Discussão
Este trabalho apresenta uma metodologia para quantificar acessibilidade económica no

contexto de um modelo espacial de concorrência monopolista. No entanto ao contrário das
medidas tradicionais de acessibilidade económica desenvolvidas sem uma base teórica bem
formada, a metodologia proposta tem como suporte um modelo teórico-económico que
explicita claramente estrutura de mercado, agentes económicos e interacções económicas.
Esta especificação tem no entanto outras vantagens. Em primeiro lugar é possível
confrontar os resultados do modelo com dados da economia real, como fluxos de comércio.
Na medida tradicional tal não era possível porque a única relação testada era a
acessibilidade, e esta não é observável. Em segundo lugar cenários de política alternativos
são simulados de uma forma natural no modelo resolvendo para o novo equilíbrio, sem ter
que desta forma recorrer a hipóteses adicionais como acontecia na abordagem tradicional.
A aplicação do modelo apresentado neste trabalho ao caso da Península Ibérica, permite
tirar algumas conclusões prévias e não definitivas. Em primeiro lugar em termos de
centralidade a economia espanhola parece levar uma grande vantagem sobre a economia
portuguesa. De facto Portugal aparece neste estudo quase como uma economia periférica da
espanhola, onde nem sequer a região da capital portuguesa (região mais desenvolvida de
Portugal) consegue ter no conjunto da Península Ibérica uma importância relevante em
termos de acessibilidade económica. Comparativamente às regiões espanholas, Lisboa
apenas tem uma performance económica superior à da pior região espanhola (Galiza).
O pico de acessibilidade é atingido pela região de Madrid sendo seguida de perto pelo
País Vasco e pela Catalunha, formando estas três os principais centros económicos da
Península Ibérica. Estes resultados não se mostraram sensíveis a testes de sensibilidade
efectuados a alguns dos parâmetros do modelo teórico.
Em segundo lugar, os resultados obtidos com os índices de bem-estar regional
confirmam os resultados de acessibilidade económica: as regiões com melhor classificação
no índice de bem-estar regional são também no seu geral as regiões com melhor
classificação no índice de acessibilidade económica. Como tal existe uma clara correlação
entre acessibilidade económica e bem-estar. No entanto existe uma excepção importante: a
posição de Lisboa no ranking de bem-estar altera-se significativamente relativamente à
posição ocupada por esta região no ranking de acessibilidade. De facto Lisboa revela-se
uma das regiões Ibéricas com maior bem-estar económico. Pode-se especular que isto se
deve ao facto de em Portugal o investimento proveniente de fundos públicos e Europeus
nas duas últimas décadas ter sido fortemente enviesado para a região de Lisboa. No entanto
não nos é possível afirmar se esta é a única causa desta circunstância. Um dado merece no
entanto ser mencionado: em meados dos anos oitenta a região de Lisboa tinha um PIB per
capita de 90% da media Europeia e uma década depois este valor tinha já ultrapassado a
média Europeia. Não deixa no entanto de ser estranho de que Lisboa apresente uma má
performance económica (tal como demonstrado na classificação modesta no índice de
acessibilidade económica) e isto não se reflicta nos índices de bem-estar.
No outro lado da moeda estão as regiões menos desenvolvidas de Portugal que também
em parte devido ao fraco investimento público direccionado para estas, continuaram entre
as menos desenvolvidas da Europa. Por exemplo o PIB per capita da região do Alentejo
que em meados da década de oitenta era cerca de 40% da média Europeia não sofreu
alterações de monta no período acima considerado. Se a dualidade entre o caso de Lisboa e
das restantes regiões portuguesas demonstra alguma coisa é a de que se o investimento
público tem pouca influência sobre a eficiência económica de uma região, este pode pelo
menos ter uma palavra a dizer em termos de equidade regional. Achamos que pesquisa
futura deve ser feita nesta direcção para clarificar esta observação. Testes de sensibilidade a
parâmetros do modelo tendem mais uma vez a confirmar estes resultados.
Em terceiro lugar foi também simulado um cenário de integração completa entre a
economia portuguesa e a economia espanhola. Este cenário não demonstrou ser benéfico
para a grande maioria das regiões Ibéricas. Como previsto pela literatura na “nova”
geografia económica, as regiões mais beneficiadas com este cenário seriam as mais
avançadas. Este facto demonstra portanto a presença de efeitos de “lock-in” e efeitos de
casualidade cumulativa: as regiões mais avançadas são as que estão mais preparadas para
enfrentar o futuro e dificilmente abandonarão a suas posições de liderança. Mais uma vez
surpreendentemente a região de Lisboa parece ser das que pode beneficiar mais num
cenário de uma região Ibérica totalmente integrada. Isto pode indicar que esta região sofre
os efeitos negativos de estar limitada ao pequeno mercado domestico português e que num
mercado maior sofreria de menores constrangimentos de crescimento.
10. Apêndice
Para discutir a questão da sensibilidade dos modelos de concorrência monopolista ao

parâmetro elasticidade de substituição vai-se optar por simplificar o modelo acima
apresentado de forma a este se parecer o mais possível com o modelo original de Dixit e
Stiglitz (1977). O modelo de Dixit e Stiglitz considera uma economia em que se produzem
dois bens, um bem homogéneo (numerário) e um bem compósito diferenciado, utilizando
apenas como único input trabalho. A remuneração deste factor é normalizada para a
unidade função da hipótese do bem numerário. Os consumidores tomam as suas decisões
de consumo em dois estágios. No primeiro estágio escolhem quanto do seu rendimento ( I )
vão gastar no bem compósito e no bem homogéneo de acordo com uma função de utilidade
Cobb-Douglas. No segundo estágio os consumidores escolhem quanto da despesa no bem
compósito vai ser distribuído por cada um dos bens diferenciados. Esta divisão é feita
seguindo uma função de utilidade CES:
σ
L n σ −1 O σ −1
V = M∑ xi σ P
MM PP (A1)
N i =1 Q
Como se pode constatar são apenas duas as diferenças entre o modelo por nós utilizado
e o modelo de Dixit e Stiglitz. Em primeiro lugar Dixit e Stiglitz consideram apenas um
factor de produção enquanto que nós dois factores de produção (trabalho e capital). Em
segundo lugar Dixit e Stiglitz abordam o problema do lado do consumo enquanto nós
tomamos o problema do lado da produção (uma vez que estes problemas são duais, é
simples ver a equivalência entre ambos). Tomando logaritmos de ambos os lados de (A1),
tem-se:
σ
ln V = ln x + ln n (A2)
σ −1
Fazendo I = pnx denotar a despesa total em bens diferenciados (estamos portanto no

equilíbrio simétrico em que todas as variedades entram de igual modo no consumo).
Eliminado x e resolvendo para (A2) vem:
I 1
ln V = ln + ln n (A3)
p σ −1
Como tal (A1) aumenta com n , e é convexa para σ inferior a 2. À função de utilidade
(A1) está associada um índice de preços semelhante a (7) no texto:
n
q1−σ = ∑p
j
1−σ
(A4)
A equação (A4) não é mais do que o índice de preços usado no modelo original de Dixit
e Stiglitz (1977). A única diferença em relação a (7) é que (A1) não está indexado
regionalmente, daí a não inclusão de custos de transporte. Invertendo este índice de preços
no equilíbrio simétrico e tomando logaritmos obtemos:
1
ln q = ln p − ln n (A5)
σ −1
Como facilmente se pode verificar, este índice também é decrescente em n . A partir de
(A4) pode-se também ver como as propriedades económicas do modelo se alteram
drasticamente com o valor do parâmetro elasticidade de substituição. Para ver isto com
mais clareza normalize o preço de cada variedade individual ( p ) para a unidade no
equilíbrio simétrico (i.e. todas as variedades entram de forma igual no índice de preços) e
resolva para o índice de preços assumindo diferentes valores para σ , nomeadamente 1.5, 5,
10, e 15 (ver figura 2).
1.2
1
0.8
0.6
q
0.4
0.2
0
1
13
19
25
31
37
43
49
55
61
67
73
79
85
91
97
103
109
n
s=15 s=10 s=5 s=1.5
Figura 2. Índice de Preços e Elasticidade de Substituição
Como foi dito anteriormente e agora visível na figura 2, o índice de preços cai com o
número de variedades (os consumidores apreciam a variedade) e tanto mais quanto menor
for a elasticidade de substituição entre variedades (mais diferenciados são os bens). Mas o
ponto a notar é que pequenas alterações no valor deste parâmetro provocam grandes
mudanças numa das relações mais importantes do modelo. De facto é a partir da equação
(A4) que muitas das conclusões da nova teoria do comércio são tiradas tal como o
conhecido “efeito índice de preços”. Segundo este os países com maior actividade
industrial terão um menor índice de preços comparativamente aos países menos
industrializados. Por este facto é que muitas vezes se refere a (A4) como um índice do
custo de vida num país (ou indirectamente como um índice de bem-estar). Também a partir
de (A4) podem ser associados o também conhecido “efeito mercado doméstico”: os países
com maior mercado doméstico detém uma proporção superior de indústria. Para isso basta
interpretar (A4) como o índice de preços de uma industria verticalmente relacionada por
relações de input-output (ver Venables, 1996). Deste modo os países com maior mercado
serão os países que também possuem uma maior proporção de indústria porque nestes se
pode produzir em condições mais vantajosas uma vez que tem acesso a produtos
intermédios a preços mais vantajosos. Daqui por outro lado também surgem os já referidos
no texto efeitos de aglomeração da literatura na ‘nova’ geografia económica. Pode-se pois,
perceber porque razão é que este parâmetro é o principal alvo dos testes de sensibilidade
quando se calibra um modelo em concorrência monopolística à la Dixit-Stiglitz.
11. BIBLIOGRAFIA
Bröcker, J. (1984), “How do International Trade Barriers Affect Interregional Trade”, in:
A. Anderson; W. Isard e T. Pu (eds.), Regional and Industrial Development
Theories: Models and Empirical Evidence, Amsterdam: North Holland.
Bröcker, J. (1989), “How to Eliminate Certain Defects of the Potential Formula”,
Environment and Planning, 21, pp. 817-830.
Bröcker, J. (1998a), “Measuring Accessibility in a General Equilibrium Framework”, in: A.
Reggiani (ed.), Accessibility, Trade and Locational Behavior, Aldershot: Ashgate.
Bröcker, J. (1998b), “How Would an EU-Membership of the Visegrád-Countries Affect
Europe’s Economic Geography”, The Annals of Regional Science, 32, pp. 91-114.
Bröcker, J. (1998c), “Operational Spatial Computable General Equilibrium Modeling”, The
Annals of Regional Science, 32, pp. 367-387.
Bröcker, J. e Jäger-Roschko, O. (1996), “Eastern Reforms, Trade and Spatial Change in the
EU”, Papers in Regional Science, 75, pp. 23-40.
Bröcker, J. e Rohweder, H. (1990), “Barriers to International Trade: Methods of
Measurement and Empirical Evidence”, The Annals of Regional Science, 4, pp.
289-305.
Clark, C.; Wilson, F. e Bradley, J. (1969), “Industrial Location and Economic Potential in
Western Europe”, Regional Studies, 3, pp. 197-212.
Darroch, J. e Ratcliff, D. (1972), “Generalized Iterative Scaling for Log-Linear Models”,
Annals of Mathematical Statistics, 43, pp. 1470-1480.
Davis, D. (1998), “The Home Market, Trade, and Industrial Structure”, The American
Economic Review, 88, pp. 1264-1276.
Dixit, A. e Stiglitz, J. (1977), “Monopolistic Competition and Optimum Product Diversity”,
American Economic Review, 67, pp. 297-308.
EUROSTAT (1996), “REGIO: Banque de Données Régionales, Description du Contenu”,
Luxembourg: Commission des Communautés Européennes.
Flôres, R. (1997), “The Gains from MERCOSUL: A General Equilibrium, Imperfect
Competition Evaluation”, Journal of Policy Modeling, 19, pp. 1-18.
Fujita, M; Krugman, P. e Venables, A. (1999), “The Spatial Economy: Cities, Regions, and
International Trade”, Cambridge: MIT Press.
Gasiorek, M.; Smith, A. e Venables, A. (1994), “Modeling the Effect of Central and East
European Trade on the European Community”, in: The Economic Interpenetration
Between the European Union and Eastern Europe, European Economy No. 6,
Report and Studies Part C, Chapter 3, Brussels: European Commission.
Hanson, G. (1998), “Market Potential, Increasing Returns, and Geographic Concentration”,
NBER Working Paper, No. 6429, February.
Harris, C. (1954), “The Market as a Factor in the Localization of Production”, Annals of
the Association of American Geographers, 44, pp.315-348.
Hayashi, F. (1982), “Tobin’s Marginal q and Average q: a Neoclassical Interpretation”,
Econometrica, 50, pp. 213-224.
Keeble, D.; Offord, J. e Walker, S. (1988), “Peripheral Regions in a Community of Twelve
Member States”, Luxembourg: Commission of the European Communities.
Krugman, P. (1991), “Increasing Returns and Economic Geography”, Journal of Political
Economy, 99, pp. 413-499.
Krugman, P. (1993), “On the Number and Location of Cities”, European Economic
Review, 37, pp. 293-298.
McCallum, J. (1995), “National Boundaries Matter: Canada-US Regional Trade Patterns”,
American Economic Review, 85, pp. 615-623.
Pires, A. J. G. (1999), “Distribuição Espacial da Actividade Económica na Espanha:
Estimação dos Parâmetros Estruturais de um Modelo de Geografia Económica”,
Tese de Mestrado, ISEG/UTL.
Smith, A. e Venables A. (1988), “Completing the Internal Market in the European
Community”, European Economic Review, 32, pp. 1501-1525.
Venables, A. (1996), “Equilibrium Locations of Vertically Linked Industries” International
Vickerman, R. (1989), “Measuring Changes in Regional Competitiveness: The Effects of
International Infrastructure Investments”, Annals of Regional Science, 23, pp.
275-286.
Sessão 3 - Capital Humano e Crescimento
Session 3 - Human Capital and Growth

EDUCATION AND EARNINGS IN PORTUGAL1
Pedro Telhado Pereira2

Universidade da Madeira, CEPR and IZA
Pedro Silva Martins3
University of Warwick
Abstract
This reports presents evidence on the returns to education in Portugal. We

find that, at 11%, Portugal exhibits a considerably large private economic
return to schooling. This has been increasing since the early 1980s up until
the mid-1990s but has decreased since then.
Although the average return to education is large, the marginal return differs
substantially as it generally increases with the schooling level. We find that
Engineering is the best rewarded degree, while Human and Social Sciences
(except Economics) reap the lowest returns. Some evidence suggests that
holders of technical degrees have out-performed holders of similar degrees
but of a more academic nature and that the private returns to some currently
offered types of vocational schooling are very low. As to differences between
gross and net returns to education, the latter are systematically lower. Wage
dispersion is higher for workers of high schooling attainment levels. No
evidence of signalling or selectivity in the returns to education is found.
Our policy suggestions focus on the role of the schools’ governance structure
upon students’ performance. While Portugal would certainly benefit from a
more qualified workforce, the evidence suggesting low quality or inefficiency
in the teaching/learning process should also be acknowledged. The efficiency
and performance of the system may benefit from a gradual shift in the role of
the government in education, from that of a provider and a producer to that
of a provider and a regulator.
1
This report has been prepared for the Conference of the Bank of Portugal on “Desenvolvimento Económico
Português no Espaço Europeu: Determinantes e Políticas”, held in 24-25 May 2002. The authors thank Mário
Centeno, Reamonn Lydon, Robin Naylor, Ian Walker, participants in the European Commission “Public
Funding and Private Returns to Education” research network and the Scientific Commission of the Conference
for helpful discussions. They also gratefully acknowledge logistical support from the Bank of Portugal. All
errors are of our own responsibility.
2
Departamento de Gestão e Economia, Universidade da Madeira, Campus Universitário da Penteada, 9000-390
Funchal, Portugal. Email: ppereira@uma.pt. Web: www.uma.pt/ppereira Fax: +351 291 705 040. Phone: +351
291 705 043.
3
Department of Economics, University of Warwick, Coventry CV4 7AL, UK. Email: p.martins@warwick.ac.uk.
Web: www.warwick.ac.uk/staff/P.Martins Fax: +44 24 76523032. Phone: +44 24 76528240.
220 Pedro Telhado Pereira, Pedro Silva Martins
Introduction and Executive Summary
The importance of education for both individuals’ prospects and economic growth is
widely recognised. Education and, more generally, all investments in human capital are
currently considered as one of the best tools to improve the long-run economic outlooks of
nations. Consequently, many countries have been making substantial efforts to improve the
quantity and quality of the skills acquired by their present and future workforce.
This paper, in which we present some of the evidence available on the economic returns
to education in Portugal, is a contribution to this process. We also examine some other
results in the economics of education literature. Our focus is on understanding the private
value of schooling on a worker’s earnings and addressing some possible policy-related
implications of these results. We believe this approach is of some relevance, given the
relatively little attention paid by economists to education matters so far in Portugal.
We begin by briefly presenting the methodology considered in the first part of our
analysis. This stems from the pioneering work of Becker (1964), where this author
developed the concept of human capital. Becker advanced the parallel between investments
in physical and human capital, given that in the latter case individuals also forego benefits
in the present in sake of an expected flow of benefits to be derived in the future. Mincer
(1974) also made an important contribution, as he developed the empirical dimension of
this approach.
Given that our results are of an eminently empirical nature, we start by presenting the
data sources used. In most cases, these were samples obtained from “Quadros de Pessoal”,
a large data set that includes information about employees and their firms. This source is
representative of workers in firms based in Portugal. It covers the period between 1982 and
1998, thus allowing for a comparison of some major variables across the eighties and
nineties in Portugal. We also draw on the European Community Household Panel, 1994
wave, a household survey covering employees, other workers and other individuals.
We then survey some results on the microeconomic return to education in Portugal and
compare it to those in other developed countries. We find that Portugal exhibits a
considerably large return, of about 11%, as the other countries’ average is about 8%.
(These figures should be interpreted as the average increase in earnings related to the
attainment of an extra year of schooling.) The difference between the Portuguese case and
that of other countries is probably related to the low levels of human capital of the
Portuguese workforce, making skills a particularly scarce and thus valuable good.
There has also been a slight increasing trend in these returns in the period covered.
This trend seems however to have stopped and possibly reversed in the last years of our
data, 1996-1998. In order to explain the values and trend obtained, we examine the wages
of different skill groups, defined in terms of the workers’ age and schooling, across the
period. We find that both the high returns and their increasing trend during most of the
period are likely to be largely influenced by the earnings of the more schooled and more
experienced workers. These are the groups whose relative wages increase the most.
Moreover, no evidence of substantial gender differences in returns to education is found.
Finally, we find that, although on average the return to education is in Portugal of about
11%, this figure is substantially different when the earnings of workers of different
schooling levels are compared. In particular, the average return to an extra year of
schooling for the transition between secondary and university education ranges from 12%
to 18% (depending on the year considered). However, when comparing secondary
Education and Earnings in Portugal 221
schooling and the third cycle of basic education (the 12th and 9th grades), the average figure
is only 7%.
There are also some differences in the returns to different degree contents, at both the
Bacharelato and Licenciatura levels, and for both men and women. Engineering degrees
are the best rewarded, while Human and Social Sciences (except for Economics) reap the
lowest returns. Some evidence also suggests that holders of technical degrees have out-
performed holders of similar degrees but of a more academic nature following the
discontinuation of the latter schooling level. Other evidence reported suggests the private
returns to some currently offered types of vocational schooling are very low.
As to differences between gross and net returns to education, we find that the latter are
systematically lower than the former, probably because of the progressivity of the tax
system.
In the third section, we examine some extensions to these findings. An interesting result
concerns the differences in returns to education for workers with similar characteristics. We
find that returns to education across workers who have high wages are particularly higher
than those returns across similar workers but who have low wages. This result means that
wage dispersion is higher for workers of high schooling attainment levels. It may also be
interpreted as evidence of a powerful interaction between workers’ characteristics and their
schooling, in the wage determination. Furthermore, it may raise questions about the impact
of schooling upon wage inequality, about the homogeneity of schooling quality, or about
the riskiness associated with the investment in education.
No evidence of signalling or selectivity in the returns to education is found. The first
case is examined by comparing returns to education of employees and self-employed
workers. Given that, for the latter group, education should not play a signalling role, the
signalling theory would suggest that their returns to education would be lower. However,
returns to education for the two groups turn out to be very similar. The matter of
endogeneity is more difficult to address and no conclusive evidence could be found here, as
this typically relies on experimental evidence difficult to obtain in the realm of the social
sciences.
Another perspective on the economic value of education involves looking at
macroeconomic returns to education. This approach, followed in the fourth section, is of
particular relevance, as it may account for externalities in education production, which are
typically disregarded in microeconomic studies. As, to our knowledge, there is no research
on this matter for Portugal, we simply summarise here some of the key points in the
literature.
One strand of the literature has aggregated the micro equations to the macro level and
has drawn on panel data from several countries. The first set of evidence has suggested that
changes in schooling did not translate into economic growth, unlike initial levels of
schooling. A more recent group of papers claims that these results are due to poor data and
finds that changes in levels of schooling impact upon changes in levels of income. Another
strand of the literature has explicitly examined social returns to education, focusing on
single countries. Here the results are still conflicting, as they are not robust to the
identification strategy adopted: some papers suggest social returns to education are
relevant, others suggest they are negligible. No consensus on this matter has yet emerged in
the literature. (See Temple, 2001, for a recent survey.)
The fifth section addresses the policy implications of our study. We would emphasise
this is a difficult task. This is due to the fact that establishing a link between research
results and policy initiatives is not always straightforward. However, we believe that the
implications of our results warrant some detailed consideration by policy-makers.
We start by reviewing some international evidence on the link between resources and
outcomes in education production. This evidence suggests there is, at most, a weak link
between the two variables, possibly because of the difficulty of controlling for the quality
of the investments. Other results indicate that an effective governance structure is much
more important for students’ performance than high levels of funding.
After presenting some suggestions for the reform of the Portuguese education system
put forward by other authors, we present our own. We argue that Portugal would certainly
benefit from a more qualified workforce, given that schooling is still particularly valuable
in the labour market. This result stems from the high returns to education uncovered in the
first part of this report. However, we are also concerned with the evidence suggesting low
levels in the teaching/learning process, particularly if one bears in mind the large amount
of resources spent on the education system. Therefore, a stronger focus on quality and/or
efficiency seems to be warranted.
In this respect, we fear that more resources will not necessarily translate into better
outcomes if some structures in the education system are left unchanged. In particular, we
believe the efficiency and performance of this system may benefit from a gradual shift in
the role of the government, from that of a provider and a producer to that of a provider and
a regulator. Simultaneously, other steps could be taken such as improving the
accountability of schools, and providing more information about different school paths and
their labour-market implications to prospective students
The last section summarises and concludes.
Section 1. Methodology
1.1 Theory
The approach followed in this paper stems from the work on human capital of authors such
as Gary Becker, Jacob Mincer and Theodore Schultz. They have established an insightful
parallel between investments in physical capital and those in human beings. In both cases,
current consumption is foregone with a view to higher consumption levels to be obtained in
the future.
According to this theoretical framework, agents are assumed to choose their schooling
attainment level in order to maximise their expected present value of the stream of future
incomes, up to their retirement age. The optimum level of schooling, s, is thus obtained by
equating marginal benefits and marginal costs, as below:
T−s
∑
ws − ws −1
t
= ws −1 + cs ,
(1 + rs )
t =1
where T is the retirement date, cs are costs of education, ws are wages to be obtained with s
years of schooling, and rs is the internal rate of return.
After some algebra and assumptions4 that allow one to derive an empirically testable
relationship, one obtains the following earnings equation:
2
log wi = Xi β + rsi + δ 1 xi + δ 2 xi + ui ,
4
See Harmon et al. (2001), Bjorklund and Kjellstrom (2000) and Asplund and Pereira (1999).
where wi represents earnings of individual i, si is this individual’s schooling attainment

level, xi is an experience measure (typically age minus schooling attainment years minus
age at which schooling started), Xi represents other variables assumed to affect earnings,
and ui is other variables not controlled for, which are assumed to be independent of both Xi
and si.
The coefficient of r can be understood as the private return to education as this measures by
how much do wages increase on account of an extra year of schooling. This coefficient will
be the variable of greatest interest in this paper.5
1.2 Data
In order to implement empirically this theoretical approach, we resort to “Quadros de

Pessoal”, a rich Portuguese data set. This stems from a compulsory survey conducted by
the now Ministry of Work and Solidarity, where all firms are required to provide
information about their workers. These data include variables such as monthly
compensation, highest schooling level attained, age, tenure and monthly hours worked.
In this paper, we draw on samples of about 50,000 workers per year. We cover the period
1982-1998, with the exception of the year of 1990. Graphs 1 and 2 present some summary
evidence on the schooling and real wages of these workers, respectively. The first graph
illustrates the point often made about the particularly low educational attainment of the
Portuguese workforce. In 1982, almost 60% of this group had not more than four years of
schooling. This share declines to about 35% in 1998 but is still the dominant schooling
category. One can also notice a substantial upward trend in the six years of schooling level,
particularly during the 1980’s.
Graph 1 - Schooling Attainment, Portuguese Workforce, 1982-1998
Source: "Quadros de Pessoal"
70.0
60.0
50.0
Percentage
40.0
30.0
20.0
10.0
0.0
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
Years
4 6 9 11+12 14+15 16+17
5
See Harmon and Walker (2001) for a detailed study of the UK drawing on this methodology.
Graph 2 - Real Wages, 1982-1998

Source: "Quadros de Pessoal"
900 150000
850
140000
800
750 130000
700
120000
650
110000
600
550 100000
500
90000
450
400 80000
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
Real Hourly Wages Real Monthly Wages
In any case, by the end of this period there are not more than 5% of workers with a
Licenciatura (the ‘16+17’ years of schooling group), not more than 3% of workers with a
Bacharelato (‘14+15’ group) and not more than 16% of workers with secondary schooling
(‘11+12’ group).This poor educational level was certainly heavily influenced by the
specific education policies adopted at least during the first half of the 20th century in
Portugal.6
Graph 2 depicts real wages (at 1998 prices) in our samples during the 1982-1998
period.7 This variable increases from about 600 to about 850 PTE per hour during this
period or from about 110.000 to 140.000 PTE per month (both are gross figures).
A second data source used is the European Community Household Panel. This data set
stems from a single questionnaire adopted across a large number of European Union
member-states. Moreover, this includes information on other workers than employees (e.g.,
the self-employed), on the unemployed and also on those individuals outside the labour
market. In this paper, we use information from the 1994 wave.
6
See Pereira and Martins (2001, pages 214-216) for a description of the Portuguese education system during the
1926-1995 period. Four different sub-periods are considered: 1926-1955 (a period of ‘regression’), 1956-1973
(‘growth’), 1974-1982 (‘rupture’) and 1983-1995 (‘consolidation’).
7
Real wages were computed considering the Consumer Prices Index.
Section 2. Micro Returns to Education
2.1 Linear Specification
In this subsection we present evidence on the financial returns to education in Portugal

during almost two decades. We aim to understand how much more do the more educated
earn when compared to the less educated. We therefore consider the theoretical models and
data sources described in the previous section.
Our results – see Graph 3 – suggest that returns to education in Portugal are, on
average, about 12%. Furthermore, they have apparently been increasing during the 1982-
1998 period considered. However, this trend may have been reversed in the mid-90’s, as
returns apparently peaked in 1995.
Graph 3 - Returns to Education, Linear Specification, 1982-1995
13.0%
12.5%
12.0%
11.5%
11.0%
10.5%
10.0%
9.5%
9.0%
8.5%
8.0%
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
Years
All Men Women
We have also considered any possible gender differences in returns to education, by

allowing returns to education across women to differ from those across men. We find little
differences between the two however.
2.2 Wages by Skill Groups
Here we examine the evolution of average wages of different skill groups in the
Portuguese labour market in the 1982-1998 period. Our goal is to explain the substantial
increase observed in returns to education during these years, at least until 1995.
Two different explanations are consistent with this pattern: (1) an increase of the wages
of the more educated workers and (2) a decrease of the wages of the less educated workers.
In order to disentangle these two views, we examine the evolution in time of the average
(real) wages of different skill groups of the work force.8
The skill groups considered are defined according to the age, schooling and gender of
the worker. There are two age groups: ‘young’ and ‘old’ workers (below or above 40 years
old, respectively); three education groups: low-, medium- and highly-educated (below or
equal to four years; six and nine years; more than nine years); and the two genders. Here
we present only the case of men, as that of women is largely the same.
There are two sets of results. Firstly, the wage rankings of the six skill groups are stable
and are robust to considering men or women or monthly or hourly wages (see Graph 4;
only the results for men are presented). The old and highly educated workers invariably
lead the rankings. This group is followed in a second group by the old and medium
educated and the young and highly educated. In a third group, we find the old and low
educated, the young and medium educated and finally the young and low educated.9
Secondly, old and highly educated workers are those who benefit from the larger
increases in real wages during the period covered (see Graph 5). These men see their real
monthly wages increase by more than 40%. They are then followed by the old and low
educated (35%), the young and highly educated (28%), the young and low educated (25%),
the old and medium educated (20%) and the young and medium educated (10%).
Graph 4 - Average Hourly Real Wages, Different Skill Groups, Men, 1982-1998
1998 Escudos
3000
2500
2000
1500
1000
500
0
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
Years
Young, Low Ed Old, Low Ed Young, Med Ed Old, Med Ed Young, High Ed Old, High Ed
8
This analysis should be of interest by itself, as it helps one understand how the growth of real wages witnessed
during the period covered was distributed across the Portuguese labour force.
9
However, one should mention that differences across groups are less clear for the case of women, whose wage
distributions are less dispersed.
Graph 5 - Average Hourly Real Wages, Different Skill Groups, Men, 1982-1998
Base Period: 1982
180%
170%
160%
150%
140%
130%
120%
110%
100%
90%
80%
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
Years
Young, Low Ed Old, Low Ed Young, Med Ed Old, Med Ed Young, High Ed Old, High Ed
Overall, these results suggest two tentative conclusions. Firstly, the increase in returns
to education that we document, particularly up to 1995, is likely to have been driven by the
higher wages received by individuals with high levels of education. An alternative
hypothesis, not supported in the data, would be that it happened due to lower wages of the
workers with lower levels of schooling.
Secondly, older workers in general and older and highly educated workers in particular
have witnessed the largest increases in wages across the six skills groups considered. This
suggests the existence of a cohort effect or interactions between schooling and experience
in the wage determination process. It also suggests that skilled workers of different
age/experience levels are not strongly substitutable.
2.3 Other Countries
Here we try to put in perspective the figures presented before on returns to education in
Portugal by comparing them with those of other developed countries. We draw on Harmon
et al (2001) and Pereira and Martins (2001a), which are based on a large research project
that attempted at computing returns to education in a large number of countries in a
comparable manner.
These results, which refer to the year of 1995 or the closest available, are summarised in
the Graph 6. One can easily see that Portugal, at 12,6%, exhibits the highest rate of return
to education, clearly above the average of about 8%. It is difficult to find out a pattern that
accounts for differences in returns across countries, as factors such as wage dispersion,
income per capita or average schooling attainment do not explain easily these results.10
Graph 6 - Returns to Education, Western Countries, 1995 (or closest year)
14%
12%
10%
8%
6%
4%
2%
10
See, for a contrasting view, Denny et al. (2002), which considers other factors such as direct foreign investment
and international trade. However, their results are still consistent with the view that the high returns to
education found in Portugal are largely influenced by the low schooling attainment of the Portuguese
workforce.
2.4 Dummies Specification
Graph 7 - Returns to Different School Degrees, 1982-1998
500%
400%
300%
200%
100%
0%
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
Years
1st Cycle 2nd Cycle 3rd Cycle Secondary "Bacharelato" "Licenciatura"
Up until now, the results presented have assumed a constant marginal return to
education across different schooling levels. Here we drop this assumption by allowing the
wage premium associated with moving from one schooling level to another to differ
between different schooling levels.
We consider six schooling levels: 1st Cycle, 2nd Cycle, 3rd Cycle, Secondary,
Bacharelato and Licenciatura. At a first level of analysis (Graph 7), we notice that, as
expected, a higher schooling level is associated with higher earnings. For instance, workers
with a Bacharelato earn about 300% more than workers with no schooling (the comparison
category implicit in this specific analysis).
A second result concerns the time pattern exhibited by each degree. Most degrees are
relatively stable, although with a slightly downward trend in some cases. An exception is
the Licenciatura, whose returns increase markedly, at least up until 1996.
We extend this analysis by examining the marginal returns implicit in the transition
between adjacent schooling levels. We therefore compare in turn the 2nd and 1st Cycles, the
3rd and 2nd Cycles, the Secondary and 3rd Cycle, the Bacharelato and Secondary levels, and
the Licenciatura and Secondary levels (see Graph 8). In this analysis we take into account
the number of years required to “upgrade” from one schooling level to the next.
We find that the highest marginal returns are consistently higher in the Bacharelato and
Licenciatura level. This means that the greatest wage benefits from schooling could be
found in the transition between secondary school and higher education. Additionally, the
lowest marginal returns are found in the secondary/3rd cycle transition.
As to any time trends found in the 1982-1998 period considered, we find that the
marginal returns at the transition between the 2nd and 1st cycles fall considerably. The
opposite occurs in the transition between the Licenciatura and secondary levels. This
suggests that the marginal premium from attending a Licenciatura has increased in Portugal
during this period.
Graph 8 - Average Marginal Returns to School Degrees , 1982, 1991 and 1998
20%
18%
16%
14%
12%
10%
8%
6%
4%
2%
0%
2nd/1st cycles 3rd/2nd cycles Sec/3rd cycle Bachar/Sec. Licenc/Sec.
School Degrees Compared
1982 1991 1998
There is some extra evidence on the returns to school-based training in Portugal – see
Saraiva (1999) and Hartog et al. (2000). These authors’ conclusions are that there are
apparently no positive returns to the apprenticeship system and to vocational training
schools. However, there are positive returns to vocational training in colleges and in a
working environment.
Other results disentangle the returns to the academic and the technical 3rd cycle degrees
(see Pereira and Martins (2001)). They suggest that, while in the beginning of the eighties
these were very similar, a gap between the two grows up since then, as the returns to the
academic version decline while those of the technical branch remain stable. This may be
related to the decision taken of dropping the latter type, following the 25th April revolution,
which probably decreased the supply of workers with such skills, thus increasing their price
in the market.
2.5 Returns to Higher Education Degrees
Here we present returns to specific higher education degree types. We are based on the
Quadros de Pessoal samples which, since 1995, include also such information. We find
that (see Table 1), at both the Bacharelato and Licenciatura levels, and for the entire set of
workers or for men and women separately, Engineering consistently pays the highest
wages, while Human and Social Sciences (except for Economics) pay the lowest wages.11
11
The exception is for men who hold a licenciatura, as Economics is associated with higher wages.
Table 1
Returns to Degree Type and Content, 1995
Degree Type Degree Content All Men Women

"Bacharelato" Management and Economics 270% 262% 267%
"Bacharelato" Engineering 368% 325% 408%
"Bacharelato" Human and Social Sciences 240% 224% 262%
"Bacharelato" Other Degrees 318% 327% 319%
"Licenciatura" Management and Economics 500% 471% 498%
"Licenciatura" Engineering 504% 450% 548%
"Licenciatura" Human and Social Sciences 431% 331% 492%
"Licenciatura" Other Degrees 438% 373% 492%
The above results should be taken with caution as the data do not include self-employed
and public servants. Moreover, we are not controlling for selectivity issues. Instead, we
assume that wage differences are only driven by the content of the degree and preclude any
personality differences that may influence simultaneously degree choice and intrinsic
productivity.12
2.6 ECHP Results
Some issues of interest cannot be addressed with the “Quadros de Pessoal” data set. We
therefore switched to the European Community Household Panel (ECHP) data set (1994
wave), which includes extra socio-economic evidence, based on a survey of all members of
different households.
Given that the ECHP is a representative sample of the population, we compute OLS
estimates comparable to those obtained from the “Quadros de Pessoal” data set. We find –
see Table 2 – that returns to education in the ECHP are generally higher to those of the
main data set. This suggests that the highly educated self-employed are earning on average
relatively more than the highly educated employees.
Table 2
Comparison of Results using QP and ECHP
All sample Men only Women only

QP ECHP QP ECHP QP ECHP
Education .0966 .1037 .0941 .1047 .0963 .1002
12
Another related topic of research amounts to comparing these returns to those expected by students. Future
work by the authors, based on a large survey of Portuguese undergraduate students, will try to do this.
Another issue of interest concerns any potential differences between net and gross
returns. So far, we have only considered the latter, as the “Quadros de Pessoal” data set
includes only information on gross earnings. However, given that we are now focusing on
private returns to education, the former type of returns is of great relevance. Using ECHP
data and, as dependent variable, the logarithm of net and gross wages, we find that net
returns are systematically lower than gross returns (see Table 3). This finding is probably
related to the progressivity of the tax system. The benefit of extra schooling is partially
eroded by the impact of higher marginal tax rates.
Table 3
Net and Gross Returns to Education
All sample Men only Women only

Net Gross Net Gross Net Gross
Education .0928 .1037 .0941 .1047 .0913 .1002
We finally study the difference between returns to education in the private and the
public sector. The results indicate a higher return to education for the individuals employed
in the private sector.13 However, our results suggest a wage premium of around 39% in the
public sector. The return to experience is also higher in the latter sector.
Table 4
Returns to Education in the Public and Private Sector
All sample Public Private

Education 8.2% 7.4% 8.9%
3. Micro Returns to Education: Extensions
3.1 Quantile Regression.
In the previous section, we focused on the average return to education, given that we
used the OLS method. Here we allow the return to education to vary depending on the
relative location of the worker across the wage distribution. This means that we may know
whether returns to education for workers that do better or worse are higher or lower than of
those of their counterparts.14
We focus our attention at the 10th, 50th and 90th percentiles of the wage distribution. The
results – presented in Graph 9 – suggest there are clear differences in returns to education
across the wage distribution. In particular, such returns are particularly higher at the top
13
These results contrast to those obtained by Portugal and Centeno (2001), who do not find significative
differences between the private and public sectors in the returns to human capital.
14
See Buchinsky (1994), Hartog et al. (2001), Machado and Mata (2001) and Pereira and Martins (2000) for
applications.
than at the tail of the wage distribution. Moreover, this difference has widened over the
period covered.
Graph 9 - Returns to Education, Quantile Regression, 1982, 86, 91 and 95
16%
14%
12%
10%
8%
6%
4%
2%
0%
1982 1986 1991 1995
Years
10th percentile 50th percentile 90th percentile
These results suggest that the dispersion of earnings increases with the workers’ level of
schooling. Moreover, such increase in dispersion has itself increased over the period
covered. This finding may account for some of the increase in wage inequality documented
for this period – see Cardoso (1998).
3.2 Education and Wage Risk
The dispersion of returns to education across the wage distribution can be seen as being
related to a measure of wage risk. To the extent that returns to education exhibit a great
deal of variability, the investment in education is also riskier. In this subsection, we
examine the relationship between the dispersion and the average return to education (as
obtained by standard OLS methods) across a number of developed countries – see Table 5
and Pereira and Martins (2002).
Table 5
Country Rankings, OLS and QR results
Country OLS Rank 1 Diff Rank 2

Portugal 12.6 1 8.9 16
Austria 9.7 2 5.6 15
Switzerland 9.5 3 1.9 5
Ireland 8.9 4 2.6 8
Finland 8.9 5 3.3 11
Spain 8.6 6 2.4 7
UK 8.6 7 4.8 14
Germany 8 8 0.3 2
France 7.6 9 3.4 10
Netherlands 7 10 3 9
Greece 6.5 11 -1.9 1
Denmark 6.6 12 0.8 4
Italy 6.4 13 0.4 3
US 6.3 14 4 13
Norway 6 15 2.1 6
Sweden 4.1 16 3.8 12
Average 7.8 2.8
St Deviation 2 2.5
These results indicate a clear negative correlation between the average return to
education and the spread of the returns to education – as obtained from quantile regression
methods. Countries that rank highly in average returns are generally at the bottom in
dispersion and vice-versa (for a more detailed discussion, see Pereira and Martins, 2002).
This establishes an interesting parallel with the theory of financial assets and its trade-off
between return and risk.15
Moreover, this evidence suggests that average returns to education may be misleading
as to the financial impact of schooling upon earnings. High average returns may therefore
15
Somewhat similar results, concerning public funding, are presented in Asplund and Pereira (1999).
be simply a compensation for high risk levels in the education investment. Taking this into
account, the Portuguese returns to education may not be as high as initially thought.
3.3 Signalling
A contrasting interpretation of the strong correlation between schooling and earnings

documented here and in many studies for other countries involves information
asymmetries. Under this view, workers acquire costly education in order to signal their
higher ability to employers – who would otherwise know little about their prospective
workers skills (see Spence (1973)). Education per se plays no role in enhancing a worker’s
productivity and is almost entirely wasteful from the public or social point of view.
One way to test this idea involves comparing the returns to education between
employees and the self-employed.16 The intuition behind this results is whereas the former
may benefit from education as a signal, the others will not, given that they are their own
employers and have no informational asymmetries problems to deal with. Our results –
presented in Table 6 –suggest that returns to education for the self-employed are at least as
high as those for employees, therefore there is no sign of signalling.17
Table 6
Returns to Education, Employees and Self-Employed
Employees Self-Employed
Education .099 .117
3.4 Selectivity
Returns to education for women are based on a possibly non-representative sample of

women – those that work. This may give rise to a biased estimation of the true return to
education. Here we address this problem by drawing on the Heckman two-stage method
and ECHP data.
We identify the equation system by considering the number of children and living status
(living with partner, with partner and children, with children and with parents) of the
women. Our implicit assumption is that these variables affect the likelihood that women
enter the labour market but do not influence their wage rates.
Without going into much detail, our results – not presented – suggest there are no
significant differences between the two returns. The associated lambda is not significant
and there is no difference between the OLS (.091) and the Heckit (.092) coefficient.18
16
See Skalli (2001) and Barceinas-Paredes and al. (2001) for a discussion of the topic using data from a large
number of European countries
17
Some practical problems arise when dealing with the sample of the self-employed. We had to delete
observations that reported losses (or zero profits) – 32% of the self-employed. We also attributed all profits to
the self-employed that reported such earnings, disregarding partnerships. Since some self-employed reported
their profits in classes, we followed Stewart’s (1983) method.
18
See Pereira and Martins (2001).
3.5 Endogeneity
A crucial issue in the returns to education literature concerns the possible endogeneity
of the schooling variable. An assumption maintained so far in this report is that schooling is
randomly attributed across workers – that is, it is independent of any characteristics that
may influence a worker’s earnings but which are not directly controlled for in the model
(see Card, 1999).
One way to deal with this matter involves considering instruments for education:
variables correlated with schooling but not (directly) with wages. Here we replicated with
1995 data the approach of Vieira (1999), who follows Harmon and Walker (1995). The
instrument here is changes in compulsory school attainment. It is argued that youngsters
whose year of birth determined a higher school leaving age were likely to have had their
schooling exogenously increased by one or more years. This is the source of the pseudo-
counterfactual experiment which allows for the identification of the system.
Using a sample of males in Quadros do Pessoal, we divided the individuals in four
cohorts affected by different levels of compulsory school. The first cohort, those born until
1955, were required to study at least three years. The second cohort, born between 1955
and 1960, should study for at least four years, if they were boys. The third cohort is made
of those born between 1960 and 1964, for whom the law of four years was extended and
had its level of compliance increased. The last group of individuals, those born after 1964,
should study at least six years.
Our results – see Table 7 – suggest that IV estimates give rise to lower returns than
those obtained by OLS. The magnitude of the difference we find is similar to the one found
by Vieira, which suggests these results are robust. However, these results are at odds with
Harmon and Walker (1995), as these authors find higher returns after instrumenting the
schooling variable.
Table 7
OLS and IV Estimates, Using Date of Birth
OLS IV - Compulsory
Education .080 .047
The results presented in Table 7 are also in contradiction with those obtained using
ECHP data and parental education as instruments – see Table 8. Here, IV estimates are
higher than OLS estimates. However, one may question whether these are appropriate
instruments for education, as parental education characteristics may influence wages both
directly and indirectly.
Table 8
OLS and IV Estimates, Using Parents Education
OLS IV – Parents Education

Education .097 .127
It should be stressed that the results obtained using changes in compulsory education as
the exogenous variation used to identify the return to education may refer only to the return
to the group affected by the treatment, which can be different from the rest of the
population. Taken at face value, these results would suggest that the financial return of
those workers who had their schooling increased due to these laws was lower than that of
their counterparts.
3.6 Meta-Analysis
We study in this subsection the sensitivity of the return to education to different

specifications of the Mincer equation. In Pereira and Martins (2001b) we show that, from
the theoretical perspective, if more covariates are used in this equation and these covariates
are choice variables that depend on education, then the return to education should decrease.
This result is obtained in a meta-analysis using results for Portugal from a large number
of studies – see Table 9 for a description of the characteristics of these studies. The return
to education decreases with all combinations of variables used. It can drop to half of its
original size, especially when the worker’s industry is controlled for.19
Table 9
Descriptive Statistics, Other Results
Variable Mean Std. Dev. Min Max

Rate 7.791 1.854 3 12.4
SE 0.004 0.012 0 0.096
Year 10.573 5.045 0 18
Public 0.049 0.216 0 1
Private 0.049 0.216 0 1
All 0.902 0.298 0 1
Net 0.061 0.24 0 1
Gross 0.939 0.24 0 1
QP 0.915 0.28 0 1
ECHP 0.085 0.28 0 1
Explan 8.213 12.487 0 72
Sector 0.165 0.372 0 1
Size 21.439 23.47 0.308 226.891
Men 0.341 0.476 0 1
Women 0.293 0.456 0 1
All 0.366 0.483 0 1
Hourly 0.634 0.483 0 1
Monthly 0.366 0.483 0 1
Age 0.28 0.451 0 1
Computed 0.671 0.471 0 1
TrueExp 0.049 0.216 0 1
Published 0.305 0.462 0 1
Unpubl 0.098 0.298 0 1
PuRE 0.598 0.492 0 1
OLS 0.939 0.24 0 1
IV 0.055 0.228 0 1
Heckit 0.006 0.078 0 1
19
The link between schooling attainment and sector choice is an aspect we are currently researching. This may
translate into over-education in the high wages industries.
These variables describe the characteristics of the several studies surveyed, all of those
focusing on returns to education in Portugal. Their meaning is the following: Rate –
Coefficient obtained, SE – Standard Error of the coefficient, Year – Year of the data, Public
– Public sector workers, Private – Private sector workers, All – All types of workers. Net –
Net wages, Gross – Gross wages, QP – “Quadros de Pessoal” data set, ECHP – European
Community Household Panel, Explan – Explanatory variables considered, except
education, experience and experience squared, Sector – Controls for industries, Size –
Controls for size of firms, Men – Only male workers, Women – Only women workers. All
– Both male and female workers, Hourly – Hourly wages as dependent variable, Monthly –
Monthly wages as dependent variable, Age – Control for age, Computed – Control for
Mincer experience, TrueExp – Control for true experience, Published – Published results.
PuRE – Estimate obtained for the “PuRE” project, OLS – OLS estimate, IV – Instrumental
variables estimate, Heckit – Heckman correction estimate.
The following variables are relevant in the meta-analysis study: (1) the year considered:
returns to education have been increasing moderately in Portugal – about 0.07% per year -
during the 19-year period covered, although at a progressively slower rate. (2) the
public/private distinction: education is less rewarded in public than in private jobs. (3) job-
related controls in Mincer equations substantially decreased returns (by 2.6%). (4) IV
estimates are substantially lower than their more common OLS counterparts. (5) the dataset
used (either QP or ECHP), since ECHP results in slightly higher estimates.
Section 4. Macro Returns to Education
As we have shown above the microeconometric evidence points to an increase in

individual’s income due to the investment in education. If we drop experience from the
Mincer equation we used in section 2, we have log linear functions for the individuals that
can be aggregated into a “Macro-Mincer” wage equation (Heckman and Kleenow (1997)),
ln Y = α + β S + ε, where Y is the geometric mean of wages and S is the average education.
Heckman and Kleenow (1997) compare the coefficient on education from cross-country
log GDP equations to the coefficient on education from micro Mincer models and they find
that macro and micro estimates yield similar effects of education on income. This means
that there seems to be no human capital externalities.
Krueger and Lindahl (1998 and 2000) question the results of the macro literature that
suggested that there is no link between the increases in education and growth. After
correcting for measurement errors they find that the effect of changes in educational
attainment on income growth in cross country data is at least as great as microeconometric
estimates of returns to years of schooling. They also find that the initial level of education
had little effect on growth for the average country.
De la Fuente and Domenech (2000) argue that the counterintuitive findings of no
influence of education on growth are due to deficiencies of human-capital data used in
empirical studies. They remove deficiencies from the data for a set of OECD countries and
show that, after this step, the results suggest that human capital is a crucial productive
input.
Cohen and Soto (2001) present a new set of data on human capital. This data set tries to
incorporate all the available information from a large number of sources: OECD databases
on education, national censuses, surveys published by UNESCO’s Statistical Yearbook and
censuses obtained directly from national statistical agencies’ web pages.
They start from a very simple neo-classical production function as in Mankiw et al

(1992) and a macro-Mincerian approach to human capital (as described above). The
estimated value for the return to education was 8.4%, which is “fairly much in line with the
average return obtained from micro data”.
The authors addressed the problem of endogeneity of education, and instrumented
schooling with the 1900 school enrolment level. The IV results are slightly higher when
compared with the OLS estimates. To test the robustness of the results they regressed the
growth rate of the income per capita on the increase in the number of years of schooling.
The result obtained was 8%, very similar to the one referred to above.
When the educational level was added as an explanatory variable in the growth equation
it appears to be not significantly different from zero. According to the authors, “this settles,
at least for these data [the data used in this authors’ paper], the long standing opposition
between the effects of levels and the effects of the increase of human capital on growth. We
find quite simply that levels are correlated to levels and growth rates to growth rates”.
Again in this paper we see that human capital seems to have social returns that are identical
to the private ones.
Section 5. Policy Implications
This section surveys some policy-oriented research on education, from the economics
perspective, developed elsewhere (in the US) and in Portugal. We then present some
suggestions as to policy implications that may be inferred from the results presented before
in this paper and link them with those put forward by other authors.
5.1 International Evidence
Hanushek (1994) studies the performance of US public schools, in terms of the

resources invested and the results of students in international comparative examinations.
This author notes the poor productivity in such relationship as “despite ever-rising school
budgets, student performance has stagnated.” Hanushek also underlines the point that
“although the case for investment in education is in large measure an economic one
economic considerations have been entirely absent from the development of educational
policies.” His results suggest that the improvement of schools today depends more on better
use of resources than on provision of added funds. The author calls for “more efficient use
of resources, greater performance incentives, and continuous learning and adaptation:
Rather than concentrating on spending more, schools must learn to consider trade-offs
among programs and operations and must evaluate performance and eliminate programs
that are not working.”20
Burtless (1996) presents a large body of evidence on the relationship between school
resources and school/students performance. This is partially motivated by a paradox in the
literature: on the one hand, extra school resources seem to play a negligible role in
improving student achievement while children are in school; on the other hand, students
who attend well-endowed schools grow up to enjoy better job market success than children
whose education takes place in schools where resources are limited. The results suggest
20
Woessmann (2001), one of the latest contributions to this debate, also fails to find a link between resources and
results across different countries and years.
that the payoff to money spent depends crucially on the way the money is spent: while
some schemes may indeed work, the simple injection of money into the system does not
offer guarantees of success
Another important contribution is that of Heckman (1999). This author puts forward a
human capital investment strategy, having the case of the USA in mind. Drawing on his
literature survey of a large number of studies, he makes the point that “Skill begets skill.
Early investment promotes later investment”, given the evidence that investments on
children are those that benefit from the highest payoffs. According to him, “American
society underinvests in the very young and overinvests in mature adults with low skills”.
This author also supports a stronger involvement of the private firms, as “private training
programs can train workers who are likely to benefit most and can tailor the training
programs to market needs”. Concerning the scope of schooling to erode wage inequality –
which increased considerably in the USA, particularly during the 1980’s – Heckman argues
that wage subsidies may be better tool to increase the incomes of older workers whose
skills may be obsolete. He finally suggests that reforms in the administrative structure of
education and the infusion of incentives and competition are more likely to be effective
than additional spending on schooling quality.
5.2 Portuguese Evidence
Clements (1999) studies the efficiency of education expenditures in Portugal. He

presents evidence of considerable inefficiency, which he obtains from contrasting the large
amount of resources spent and the poor performance of students. He suggests a number of
reforms. These involve: (1) setting targets for performance by school (e.g., success rates);
(2) the adoption of a goal-oriented management and incentive system (i.e., increasing the
accountability of school managers); (3) establishing minimum student/teacher ratios; and
(4) easing employment and work rules governing public school teachers.
Martins (2000) presents a succinct description of the performance of the Portuguese
education system, when compared to that of other developed countries. The panorama is
less than impressive, given the poor performance of the Portuguese students and the low
schooling attainment levels, even for the cohort in their early 20’s. A comparison of such
indicators that includes evidence from enlargement countries makes the Portuguese
situation particularly disappointing. The author also suggests some possible reforms in the
Portuguese education system, partially in line with those suggested by Clements (1999).
Martins (2001) presents some general principles that may guide further reform in the
Portuguese education system, notably accountability, transparency, and demand.
Associação Empresarial de Portugal (Portuguese Entrepreneurial Association, 2001)
considers that the State has been unable to provide the amount of human capital needed by
private firms and that this fact has significantly hindered the latter’s attempt to improve
productivity. They motivate this claim by noting the large percentage of workers with low
skills, even within the younger cohorts; the weak performance levels of Portuguese
students in international examinations; the lack of information about required skills in the
job market; and the low coherence of the training system. These authors also notice what
they claim is a large increase in expenditures (particularly in the kindergarten and higher
education levels) with little results in terms of performance.
In order to deal with this, these authors put forward a number of reforms (page 104): (1)
Making the school management system more accountable, namely by producing more
information about the performance of the system. (2) Improving the link between the
teaching and training sub-systems. (3) Adjusting the State’s interference in the education
system, so to become more of a regulator and less of a provider. (4) Increasing the quality
levels; increasing the supply of semi-highly skilled workers. (5) Boosting the importance of
the “Bacharelato” level and making it more linked with the private sectors. (6) Supporting
firms that invest in the human capital of their workers.
Neves and Rebelo (2001) also emphasise the poor performance of Portuguese students
in international tests. However, they note that more funds may be required, as expenditure
per student ratios in Portugal are still below those of the OECD average. Notwithstanding
this, they are careful to point out that increasing expenditures without changing structures
does not ensure that performance will pick up. As suggestions to improve the current
education system, they mention: (1) promoting school competition; (2) national
examinations that assess the performance of schools and students; and (3) hiring the best
university professors in the international market.
Another important recent contribution stems from a Portuguese newspaper (Público,
2001). These authors compiled rankings of the Portuguese secondary education schools,
obtained from their students’ national examinations results, which are required for
university admittance. This was made possible when the Ministry of Education decided to
disclose such information to outside groups.
Some interesting conclusions can be drawn: (1) The average performance of the schools
was low, particularly in core modules such as mathematics. (2) Private schools were over-
represented both in the top and in the bottom of the distribution of results. (3) There were
large average differences between schools located in the inland and along the coastline (the
latter delivering the higher results). (4) Some schools presented considerable differences
between the grades obtained by their student in the internal and external examinations. (5)
Some schools located in the same geographical areas exhibited large differences in results.
An important caveat to bear in mind when analysing these results concerns the lack of
controls for other factors that may also influence students’ performance and which may be
correlated to the variables of interest in this study, i.e. the school attended by each student.
For instance, the students’ parental background and family income are likely to differ
considerably across schools. This means that the different averages obtained cannot be
simply interpreted as the impact of each school on their students’ learning. However, we
believe the results described above are in any case very important in shedding some light
on the characteristics of the schooling system.
Finally, Neto (2001) provides an interesting account of the education policy followed in
Portugal in the 1995-1999 period, by describing the main policy reforms undertaken then
and discussing some related political economy factors. These include developing and
strengthening the kindergarten schooling level, and the adoption of a new procedure for the
allocation of financial resources to higher education institutions.
5.3 Suggestions for Reform
Which policy-oriented implications can be drawn from the empirical results outlined
here? High returns to education suggest there is still an important demand for skilled
workers, given that they benefit from a large wage premium when compared to those less
skilled. This should not come as a surprise, at least to the extent that the average schooling
attainment of the Portuguese labour force is very low by international standards.
Given that Portuguese workers do still benefit from large private returns to their
schooling, one may argue that education is still a valuable investment from the private
point of view. To the extent that the market for schooling performs poorly (because
education cannot act as collateral and/or because there are large positive externalities), the
involvement of the government in this sector, by providing schooling at subsidised prices,
is still warranted. However, the case for the public production of schooling goods is less
clear-cut.
We have also found some preliminary evidence suggesting that both the high returns to
education and the slight increasing time trend in such returns are determined at least partly
by the earnings of more experienced workers. This may mean that highly schooled workers
of different experience levels are far from being close substitutes. In particular, the
increasing demand for skilled workers may have benefited largely the more experienced
workers. Taken at face value, this finding may support those authors who have voiced
concerns about the quality of the teaching/learning of the new cohorts entering the labour
market.
Indeed, drawing on the impressive evidence about the performance of Portuguese
students in international comparative tests, we believe there should be a stronger focus on
the quality of the education provided in the system. This may involve an improvement in
the structure of incentives faced by teachers and school managers, namely by enhancing
their accountability towards parents, taxpayers and the Ministry of Education. However,
care should be taken to acknowledge the specific nature of the service being provided
(which is considerably different from those on offer by private firms). This means that
changes in the structure of incentives should be particularly well designed.21
Related guidelines involve improving the interactivity of the education system, whereby
more information is provided about the efficiency of different initiatives, so to strengthen
or discontinue any such programmes. This probably requires that the Portuguese education
system becomes less centralised and that the Ministry of Education progressively adopts
the role of a regulator and a provider rather than that of a producer.
Another finding with some policy relevance concerns the increasing returns to technical
schooling, after this degree was discontinued in the mid-1970’s. This illustrates the
potentially important interactions between education policy and economic outcomes. It
suggests that some care should be taken to study the implications of education policy upon
the labour market. More generally, a closer look at which specific skills are in high demand
should be relevant towards the allocation of public resources throughout the different
branches of the education system. In a similar way, the payoff to providing more
information about labour-market prospects of different schooling trajectories to prospective
students is also probably considerable.
21
Burgess et al (2001) discusses the implications of the introduction of the “Performance Related Pay” system in
the United Kingdom. They survey some evidence on this matter, mostly drawing on the US experience. Their
conclusions suggest that these schemes can result in “small positive gains in pupil attainment”. However, they
find that the British scheme is little else than a “general pay rise for eligible teachers”.
Section 6. Conclusions
This reports presents some of the evidence available on the economic returns to
education in Portugal. We also examine some other results in the economics of education
literature. Our focus is on understanding the private value of schooling on a worker’s
earnings and addressing some possible policy-related implications of these results.
We find that Portugal exhibits a considerably large microeconomic return to education,
of about 11%, where the other countries’ average is about 8%. The difference between the
Portuguese case and that of other countries is probably related to the low levels of human
capital of the Portuguese workforce, making skills a particularly scarce and thus valuable
good.
There has also been a slight increasing trend in these returns in the period covered.
This trend seems however to have stopped and possibly reversed in the last years of our
data, 1996-1998. We find that both the high returns and their increasing trend during most
of the period are likely to be largely influenced by the earnings of the more schooled and
more experienced workers. Moreover, no evidence of substantial gender differences in
returns to education is found.
Finally, we find that, although on average the return to education is 11%, this figure is
substantially different when the earnings of workers of different schooling levels are
compared. In particular, the average return to an extra year of schooling in the transition
between secondary and university education is between 12% and 18% (depending on the
year considered). The same figure, but when comparing secondary schooling and the third
cycle of basic education (the 12th and 9th grades) is only 7%.
There are also some differences as to the returns to different degree contents, at both
the Bacharelato and Licenciatura levels, and for both men and women. Engineering
degrees are the best rewarded, while Human and Social Sciences (except for Economics)
reap the lowest returns.
Some evidence also suggests that holders of technical degrees have out-performed
holders of similar degrees but of a more academic nature following the discontinuation of
the latter schooling level. Other evidence reported suggests the private returns to some
currently offered types of vocational schooling are very low.
As to differences between gross and net returns to education, we find that the latter are
systematically lower than the former, probably because of the progressivity of the tax
system.
In the third section, we examine some extensions to these findings. An interesting result
concerns the differences in returns to education for workers with similar characteristics. We
find that returns to education across workers who have high wages are particularly higher
than those returns across similar workers but who have low wages. This result means that
wage dispersion is higher for workers of high schooling attainment levels.
It may also be interpreted as evidence of a powerful interaction between workers’
characteristics and their schooling, in terms of the process of wage determination.
Furthermore, it may raise questions about the impact of schooling upon wage inequality,
about the homogeneity of schooling quality, or about the riskiness involved in the
investment in education.
No evidence of signalling or selectivity in the returns to education is found. The first
case is examined by comparing returns to education of employees and self-employed
workers. The matter of endogeneity is more difficult to address and no conclusive evidence
could be found here.
Another perspective on the economic value of education involves looking at
macroeconomic returns to education. This approach, followed in the fourth section, is of
particular relevance, as it may account for externalities in education production, which are
typically disregarded in microeconomic studies.
One strand of the literature has aggregated the micro equations to the macro level and
has drawn on panel data from several countries. A recent group of papers finds that
changes in levels of schooling impact upon changes in levels of income. Another strand of
the literature has explicitly examined social returns to education, focusing on single
countries. Here the results are conflicting as no consensus on this matter has yet emerged.
The fifth section addresses the policy implications of our study. Although this is a
difficult task, we believe that our results warrant suggestions of interest to policy-makers.
The international evidence on the link between resources and outcomes in education
suggests there is, at most, a weak link between the two variables. Other results indicate that
an effective governance structure is much more important for students’ performance than
high levels of funding.
Our suggestions for the reform of the Portuguese education system are that Portugal
would certainly benefit from a more qualified workforce, given that schooling is still
particularly valuable in the labour market. However, we are also concerned with the
evidence suggesting low levels in the teaching/learning process. A stronger focus on
quality and/or efficiency thus seems to be warranted.
We fear that more resources will not necessarily translate into better outcomes if some
structures in the education system are left unchanged. In particular, we believe the
efficiency and performance of this system may benefit from a gradual shift in the role of
the government, from that of a provider and a producer to that of a provider and a
regulator. Simultaneously, other steps could be taken such as improving the accountability
of schools, and providing more information about different school paths and their labour-
market implications to prospective students
References
Asplund, R. and P. Pereira (1999), Introduction, in: R. Asplund and P. Pereira, eds, Returns
to Human Capital in Europe – a Literature Review, ETLA, Helsinki.
Associação Empresarial de Portugal (2001) Uma Nova Ambição Para Portugal. Repensar o
Estado, Dinamizar a Iniciativa Privada, Mobilizar a Sociedade Civil. Associação
Empresarial de Portugal.
Barceinas-Paredes, F., J. Oliver-Afonso, J. L. Raymond-Bara, J. L. Roig-Sabaté and A.
Skalli, Does Education Improve Productivity or Earnings Only, in: R.
Asplund,eds, Education and Earnings – Further Evidence from Europe, ETLA,
Helsinki.
Becker, G. (1964). Human Capital: A Theoretical and Empirical Analysis, with Special
Reference to Education. New York: National Bureau of Economic Research
Bjorklund, A. and C. Kjellstrom (2000). Estimating the Return to Investments in
Education: How Useful is the Standard Mincer Equation? Economics of Education
Review, forthcoming.
Buchinsky, M. (1994), Changes in the US wage struture1963-1987: Application of
Quantile Regression, Econometrica, 62, 405-458.
Burgess, S., B. Croxson, P. Gregg, and C. Propper (2001) The Intricacies of the
Relationship Between Pay and Performance for Teachers: Do Teachers Respond
to Performance Related Pay Schemes? Centre for Market and Public Organisation
Working Paper Series 01/35. University of Bristol.
Burtless, G. (ed.) (1996) Does Money Matter? The Effect of School Resources on Student
Achievement and Adult Success. Washington, DC. The Brookings Institution.
Card, D. (1999), The Causal Impact of Education on Earnings, in: O. Ashenfelter and D.
Card, eds, Handbook of Labor Economics, North Holland, Amsterdam and New
York.
Cardoso, A. R. (1998) Earnings Inequality in Portugal: High and Rising? Review of
Income and Wealth, 44, 325-343.
Clements, B. (1999), The Efficiency of the Education Expenditure in Portugal,
International Monetary Fund, WP/99/179.
Cohen, D. and M. Soto (2001), Growth and Human Capital: Good Data, Good Results,
CEPR, Discussion Paper 3025.
De la Fuente, A. and R. Domenech (2000), Human Capital in Growth Regressions: How
Much Difference Does Data Quality Make, CEPR, Discussion Paper 2466.
Denny, K., C. Harmon, and R. Lydon (2002) Cross Country Evidence on the Returns to
Education: Patterns and Explanations, CEPR Discussion Paper, forthcoming.
Hanushek, E. (1994) Making Schools Work: Improving Performance and Controlling
Costs. Washington, DC. The Brookings Institution.
Harmon, C. and I. Walker (1995), Estimates of the Economic Return to Schooling for the
United Kingdom, American Economic Review, 85, 1278-1286
Harmon, C. and I. Walker (2001), The Returns to Education: A Review of Evidence, Issues
and Deficiencies in the Literature. Department for Education and Employment
Research Report 254.
Harmon C., I. Walker and N. W. Nielsen, (2001), Introduction, in: C. Harmon, I. Walker
and N. W. Nielsen, eds, Education and Earnings in Europe – a Cross Country
Analysis of Returns to Education, Edward Elgar, Cheltenham, UK.
Hartog, J., P. Pereira and J. Vieira (2000), Vocational Training and Earnings in Portugal,
Economia, 24, 35-52.
Hartog, J., P. Pereira and J. Vieira (2001), Changing Returns to Education in Portugal
during the 1980s and early 1990s: OLS and Quantile Regression Estimators,
Applied Economics, 33, 1021-1037.
Heckman, J. (1999) Policies to Foster Human Capital. NBER Working Paper 7288.
Heckman, J. and Kleenow (1997) Human Capital Policy, Mimeo, University of Chicago.
Krueger, A. and M. Lindahl (1998), Education for Growth in Sweden and the World,
Princeton University, Working Paper # 411.
Krueger, A. and M. Lindahl (2002), Education for Growth: Why and for Whom? Journal of
Economic Literature 39, 1101-1136.
Machado, J and J. Mata (2001). Earnings Functions in Portugal, 1982-1994: Evidence from
Quantile Regressions. Empirical Economics, 26, 115-134.
Mankin G., D. Romer and D. Weil (1992), A Contribution to the Empirics of Economic
Growthy, Quarterly Journal of Economics, May, 107, 402-437.
Martins, P. S. (2000) Avaliação e Reforma do Sistema Educativo Português. Diário
Económico, 28 November.
Martins, P. S. (2001) Reforma da Educação em Portugal: O Que Fazer? Diário Económico,
16 August.
Mincer, J. (1974). Schooling, Experience and Earnings. New York: National Bureau of
Economic Research.
Neto, D. (2001) Difícil é Sentá-los. A Educação de Marçal Grilo. Oficina do Livro.
Neves, J. C., Rebelo, S. (2001) O Desenvolvimento Económico em Portugal. Fórum de
Administradores de Empresas and Bertrand Editora.
Portugal, P. and M. Centeno (2001) Os salários da função pública. Boletim Económico,
Banco de Portugal (Setembro).
Pereira, P. T. and P. S. Martins (2000a), Does Education Reduce Wage Inequality?
Quantile Regressions Evidence From Sixteen Countries, IZA Discussion Paper
120.
Pereira, P. T. and P.S. Martins (2000b), Returns to Education and Wage Equations, IZA
Discussion Paper 298.
Pereira, P. T. and P. S. Martins (2001), Portugal, in: C. Harmon, I. Walker and N. W.
Nielsen, eds, Education and Earnings in Europe – a Cross Country Analysis of
Returns to Education, Edward Elgar, Cheltenham, UK
Pereira, P. T. and P. S. Martins (2002), Is There a Return-Risk Link in Education?
Economics Letters, forthcoming.
Público (2001) “A Lista”.
Saraiva, A. F. (1999), Incidência e Impacto Salarial da Formação Profissional em Portugal,
Economia, XXIII, 27-43.
Skalli, A. (2001), The Role of Schooling: Screening versus Human Capital, in: R.
Asplund,eds, Education and Earnings – Further Evidence from Europe, ETLA,
Helsinki.
Spence, (1973) “Job Market Signaling," Quarterly Journal of Economics, 87, 355-374.
Stewart, B. (1983), On the Least Squares Estimation when the Dependent Variable is
Grouped, Review of Economic Studies, 50, 737-753.
Temple, J. (2001) Growth Effects of Education and Social Capital in the OECD Countries,
University of Bristol, Working Paper 01/520.
Vieira, J. (1999), Returns to Education in Portugal, Labour Economics, 6, 535-541.
Wossmann, L. (2001) New Evidence on the Missing Resource-Performance Link in
Education. Kiel Institute of World Economics Working Paper 1051.
Table A1 - Mincer equations, Linear schooling, All workers
Dependent variable: Log hourly total wages
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
Education 10,5% 11,4% 11,5% 11,5% 11,2% 11,5% 11,6% 11,9% 11,7% 12,3% 12,5% 12,2% 11,9% 12,1%
(112.11)** (128.33)** (129.67)** (130.77)** (128.17)** (135.00)** (129.76)** (126.12)** (128.09)** (134.31)** (151.24)** (151.98)** (153.93)** (165.81)**
Experience 4,7% 5,0% 5,5% 5,7% 5,3% 4,9% 4,8% 4,8% 4,7% 4,5% 4,6% 4,6% 4,5% 4,2%
(56.34)** (59.45)** (65.02)** (66.80)** (65.40)** (62.95)** (58.74)** (56.87)** (55.83)** (55.30)** (58.99)** (59.65)** (61.71)** (59.79)**
Experience2 -5,5% -5,5% -6,2% -6,3% -5,8% -5,3% -5,1% -5,1% -4,9% -4,8% -4,9% -4,9% -4,8% -4,1%
(x100) (42.40)** (42.83)** (47.17)** (48.33)** (46.22)** (42.78)** (38.78)** (37.27)** (36.44)** (35.26)** (36.89)** (37.78)** (38.64)** (34.66)**
Constant 29,1 44,8 49,6 55,8 63,8 73,4 102,4 114,6 124,2 130,5 131,3 138,8 148,3 164,0
(263.84)** (289.60)** (297.37)** (309.66)** (328.77)** (348.27)** (358.15)** (351.60)** (360.34)** (372.48)** (396.16)** (405.50)** (427.45)** (455.43)**
Observation 33307 33681 33550 34007 34551 37230 38783 39268 38643 38753 42528 42809 46243 47782
s
R-squared 0,29 0,35 0,36 0,36 0,35 0,35 0,32 0,31 0,31 0,33 0,36 0,36 0,35 0,37
Absolute value of t-statistics in parentheses

* significant at 5% level; ** significant at 1% level
Table A2 - Mincer equations, Linear schooling, Men
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
Education 10,0% 11,0% 11,4% 11,5% 11,2% 11,5% 11,4% 12,0% 11,6% 12,2% 12,4% 12,1% 11,9% 12,2%
(91.85)** (103.78)** (105.95)** (107.07)** (104.28)** (109.62)** (102.34)** (102.95)** (100.40)** (105.31)** (117.01)** (116.54)** (119.10)** (128.53)**
Experience 4,9% 5,3% 5,8% 6,1% 5,8% 5,4% 5,3% 5,3% 5,2% 5,1% 5,2% 5,1% 5,1% 4,8%
(50.77)** (52.04)** (55.31)** (57.73)** (56.52)** (55.23)** (50.96)** (49.85)** (48.18)** (48.00)** (50.57)** (50.39)** (53.07)** (51.99)**
Experience2 -6,0% -6,1% -6,7% -6,9% -6,6% -6,0% -5,9% -5,9% -5,8% -5,7% -5,8% -5,7% -5,7% -5,1%
(x100) (40.36)** (39.95)** (42.15)** (43.87)** (42.28)** (40.18)** (37.05)** (35.63)** (34.48)** (33.47)** (34.48)** (34.64)** (35.71)** (33.46)**
Constant 31,8 47,9 51,6 57,2 66,0 76,5 109,1 119,4 130,6 137,2 138,5 147,6 153,8 169,0
(231.43)** (241.44)** (240.73)** (250.50)** (264.13)** (282.42)** (284.45)** (280.56)** (280.42)** (291.92)** (307.55)** (313.90)** (329.45)** (346.83)**
Observations 23336 23164 22609 22838 22895 24143 24625 24645 23980 23497 25705 25850 27059 28056
R-squared 0,29 0,34 0,36 0,37 0,36 0,36 0,33 0,32 0,32 0,34 0,36 0,36 0,36 0,38

Table A3 - Mincer equations, Linear schooling, Women
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
Education 11,0% 11,5% 10,8% 11,1% 10,8% 11,2% 11,5% 11,1% 11,4% 12,3% 12,5% 12,3% 11,6% 11,6%
(64.66)** (77.25)** (75.40)** (77.98)** (78.83)** (84.02)** (84.73)** (76.36)** (85.40)** (89.27)** (104.90)** (107.26)** (106.76)** (116.73)**
Experience 3,7% 3,9% 4,4% 4,5% 4,2% 3,9% 3,7% 3,6% 3,6% 3,5% 3,7% 3,8% 3,7% 3,1%
(24.49)** (27.84)** (32.64)** (34.07)** (33.60)** (32.04)** (29.24)** (27.83)** (29.22)** (28.77)** (32.84)** (34.56)** (35.41)** (33.64)**
Experience2 -4,3% -4,3% -5,1% -5,1% -4,7% -4,2% -3,7% -3,7% -3,6% -3,5% -3,9% -3,9% -4,0% -3,1%
(x100) (17.48)** (18.46)** (23.01)** (23.74)** (22.78)** (20.69)** (17.41)** (16.81)** (17.60)** (16.96)** (19.80)** (20.73)** (21.97)** (18.57)**
Constant 28,0 44,9 53,1 59,2 66,5 75,2 103,5 120,5 126,9 130,6 129,5 133,7 148,3 165,5
(146.78)** (177.44)** (192.04)** (201.18)** (218.15)** (229.02)** (241.39)** (234.04)** (254.79)** (254.28)** (278.64)** (287.09)** (304.63)** (334.72)**
Observations 9971 10517 10941 11169 11656 13087 14158 14623 14663 15256 16823 16959 19184 19726
R-squared 0,3 0,36 0,35 0,36 0,36 0,36 0,34 0,29 0,33 0,34 0,4 0,4 0,37 0,41

Table A4 - Mincer equations, Schooling in levels, All workers
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
1st Cycle 18,9% 18,3% 18,2% 19,8% 16,3% 8,4% 7,6% 6,0% 3,5% 14,5% 15,3% 15,7% 17,2% 14,3%
(22.28)** (20.07)** (19.29)** (20.08)** (16.65)** (8.88)** (7.03)** (5.26)** (3.08)** (10.08)** (10.30)** (10.77)** (11.32)** (9.35)**
2nd Cycle 61,6% 64,0% 63,6% 62,4% 56,2% 46,8% 40,6% 35,4% 33,6% 44,5% 44,9% 41,5% 43,2% 40,9%
(46.25)** (46.97)** (45.94)** (44.36)** (41.11)** (35.61)** (28.67)** (24.36)** (23.47)** (25.48)** (25.21)** (24.22)** (24.24)** (22.76)**
3rd Cycle 111,1% 121,9% 123,0% 124,3% 112,1% 107,3% 101,0% 99,6% 90,0% 106,1% 105,9% 102,4% 102,4% 92,7%
(72.77)** (76.54)** (75.39)** (73.72)** (68.48)** (66.88)** (57.28)** (54.27)** (50.56)** (48.80)** (47.86)** (47.99)** (46.53)** (42.79)**
Secondary 158,3% 174,0% 166,4% 172,6% 162,0% 154,7% 150,9% 143,8% 136,1% 157,3% 155,7% 149,7% 141,6% 138,9%
(65.49)** (74.36)** (74.04)** (75.47)** (73.34)** (73.88)** (67.56)** (62.89)** (61.92)** (60.97)** (60.50)** (60.81)** (57.18)** (55.99)**
"Bacharelato" 305,1% 301,1% 313,7% 329,7% 304,7% 293,5% 290,8% 303,9% 301,1% 296,7% 304,3% 288,1% 293,1% 268,0%
(43.86)** (50.86)** (48.21)** (48.39)** (46.94)** (48.12)** (44.88)** (47.43)** (48.18)** (53.56)** (59.60)** (62.30)** (64.32)** (63.92)**
"Licenciatura" 363,7% 394,8% 416,5% 400,3% 394,8% 403,3% 419,1% 449,0% 446,3% 483,0% 499,5% 473,2% 442,5% 449,0%
(64.36)** (74.78)** (77.13)** (77.25)** (76.88)** (78.88)** (80.75)** (80.88)** (81.46)** (81.33)** (92.87)** (93.90)** (93.20)** (97.07)**
Experience 5,1% 5,4% 6,0% 6,1% 5,9% 5,7% 5,4% 5,4% 5,3% 4,8% 5,0% 4,9% 4,8% 4,5%
(60.00)** (63.83)** (69.02)** (70.10)** (69.73)** (69.55)** (64.30)** (62.14)** (61.88)** (57.02)** (61.54)** (62.27)** (64.13)** (63.19)**
Experience2 -6,2% -6,4% -6,9% -7,0% -6,7% -6,4% -6,0% -6,0% -5,9% -5,3% -5,4% -5,4% -5,3% -4,7%
(x100) (46.71)** (47.86)** (51.82)** (52.38)** (51.31)** (50.57)** (45.63)** (43.95)** (43.74)** (37.76)** (40.18)** (41.25)** (41.78)** (38.85)**
Constant 34,2 53,5 59,4 66,6 76,9 93,5 134,1 154,2 169,2 176,7 177,8 187,5 196,0 223,5
(266.39)** (285.65)** (291.09)** (296.19)** (311.71)** (331.80)** (329.33)** (324.04)** (330.79)** (306.46)** (309.27)** (319.24)** (318.65)** (327.41)**
Observations 33307 33681 33550 34007 34551 37230 38783 39268 38643 38753 42528 42809 46243 47782
R-squared 0,3 0,36 0,37 0,37 0,36 0,37 0,34 0,33 0,34 0,33 0,36 0,36 0,35 0,38

Table A5 - Mincer equations, Schooling in levels, Men
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
1st Cycle 18,4% 16,5% 17,9% 18,2% 16,3% 7,8% 6,5% 8,1% 2,9% 16,8% 14,8% 14,9% 17,8% 16,4%
(18.32)** (14.99)** (15.36)** (14.92)** (13.58)** (6.64)** (4.89)** (5.60)** (2.00)* (9.33)** (7.92)** (7.96)** (9.04)** (8.11)**
2nd Cycle 57,9% 60,6% 63,6% 61,0% 57,8% 47,1% 40,1% 40,2% 35,1% 46,5% 44,9% 40,6% 45,1% 42,9%
(37.43)** (36.93)** (37.09)** (35.07)** (34.01)** (28.84)** (22.55)** (21.25)** (18.90)** (21.22)** (19.87)** (18.32)** (19.40)** (18.15)**
3rd Cycle 101,8% 114,3% 118,8% 119,7% 109,6% 105,2% 96,4% 104,2% 88,1% 107,7% 103,6% 98,0% 104,6% 97,0%
(57.84)** (59.94)** (59.13)** (57.57)** (54.33)** (52.95)** (44.15)** (43.97)** (38.56)** (39.53)** (37.20)** (35.90)** (36.52)** (33.79)**
Secondary 143,5% 163,0% 171,6% 172,9% 165,6% 158,8% 153,2% 155,5% 134,4% 163,5% 159,3% 153,2% 149,4% 151,4%
(50.50)** (56.59)** (58.42)** (59.01)** (58.29)** (58.22)** (52.65)** (50.95)** (46.62)** (49.37)** (47.97)** (47.27)** (45.32)** (45.02)**
"Bacharelato" 280,8% 271,7% 307,1% 309,6% 297,9% 274,3% 284,2% 299,1% 295,9% 334,9% 323,8% 293,9% 317,5% 293,5%
(37.31)** (42.50)** (41.49)** (42.19)** (40.53)** (40.88)** (37.01)** (38.64)** (38.45)** (42.48)** (43.79)** (45.35)** (48.36)** (48.47)**
"Licenciatura" 338,9% 370,2% 397,3% 393,8% 384,5% 387,4% 381,6% 433,3% 429,6% 459,0% 464,6% 439,2% 428,0% 449,6%
(56.02)** (62.98)** (64.34)** (64.57)** (64.40)** (65.65)** (64.13)** (65.61)** (64.57)** (64.32)** (71.45)** (70.72)** (71.09)** (74.47)**
Experience 5,3% 5,8% 6,3% 6,5% 6,3% 6,1% 6,0% 6,0% 5,9% 5,3% 5,5% 5,4% 5,4% 5,1%
(53.44)** (55.43)** (58.53)** (60.60)** (60.21)** (61.03)** (55.90)** (54.16)** (53.30)** (49.34)** (52.73)** (52.63)** (54.97)** (54.53)**
Experience2 -6,7% -6,9% -7,4% -7,7% -7,3% -7,1% -6,9% -6,9% -6,9% -6,1% -6,3% -6,3% -6,2% -5,6%
(x100) (43.41)** (43.73)** (45.69)** (47.26)** (46.42)** (46.73)** (42.76)** (40.75)** (40.33)** (35.24)** (37.02)** (37.37)** (37.93)** (36.44)**
Constant 36,9 57,2 61,6 68,6 78,8 97,0 141,2 156,9 176,3 181,0 185,6 197,7 200,1 226,2
(232.15)** (238.11)** (235.17)** (240.64)** (251.90)** (269.46)** (263.54)** (255.75)** (257.95)** (243.18)** (243.29)** (247.80)** (245.46)** (250.07)**
Observations 23336 23164 22609 22838 22895 24143 24625 24645 23980 23497 25705 25850 27059 28056
R-squared 0,3 0,35 0,37 0,38 0,37 0,38 0,34 0,34 0,34 0,34 0,37 0,36 0,36 0,39

Table A6 - Mincer equations, Schooling in levels, Women
1982 1985 1986 1987 1988 1989 1991 1992 1993 1994 1995 1996 1997 1998
1st Cycle 7,9% 10,4% 7,3% 12,0% 6,6% 1,7% 1,7% -7,4% -4,3% 5,5% 8,5% 10,2% 9,6% 4,2%
(5.72)** (7.32)** (5.24)** (8.10)** (4.54)** -1,23 -1,07 (4.79)** (2.84)** (2.63)** (4.02)** (5.10)** (4.65)** (2.10)*
2nd Cycle 49,2% 52,0% 44,2% 48,9% 39,9% 33,0% 31,3% 15,8% 19,1% 33,2% 35,7% 34,3% 31,1% 26,9%
(22.57)** (24.82)** (22.02)** (23.55)** (20.17)** (17.59)** (15.42)** (8.08)** (10.04)** (13.08)** (14.05)** (14.63)** (12.96)** (11.52)**
3rd Cycle 106,7% 113,2% 106,3% 112,5% 99,4% 94,6% 92,3% 73,8% 75,6% 91,4% 94,3% 95,4% 83,9% 72,1%
(41.74)** (45.75)** (44.04)** (44.71)** (41.10)** (40.92)** (36.17)** (29.67)** (31.43)** (28.79)** (29.95)** (32.35)** (28.41)** (25.91)**
Secondary 156,8% 166,4% 134,9% 150,4% 137,3% 135,1% 134,7% 108,5% 119,9% 136,8% 138,0% 134,4% 121,2% 111,1%
(40.58)** (47.97)** (44.24)** (46.84)** (44.79)** (47.19)** (43.75)** (36.20)** (41.35)** (37.06)** (38.47)** (40.53)** (36.79)** (35.29)**
"Bacharelato" 287,7% 296,7% 236,7% 266,2% 234,7% 248,3% 227,4% 245,9% 243,2% 240,1% 274,0% 272,5% 253,2% 224,5%
(22.14)** (25.56)** (21.62)** (19.14)** (21.60)** (22.74)** (23.91)** (26.61)** (28.42)** (34.36)** (42.94)** (46.36)** (45.13)** (44.81)**
"Licenciatura" 327,2% 346,8% 359,5% 326,3% 335,4% 345,5% 406,8% 384,0% 384,0% 457,9% 483,0% 466,9% 405,3% 387,0%
(29.91)** (37.50)** (39.83)** (40.65)** (41.19)** (42.85)** (48.85)** (46.38)** (49.66)** (51.15)** (61.39)** (65.03)** (62.80)** (65.56)**
Experience 4,3% 4,7% 5,1% 5,0% 4,8% 4,6% 4,3% 4,4% 4,2% 3,8% 4,1% 4,1% 4,0% 3,6%
(28.66)** (32.41)** (37.03)** (37.24)** (37.43)** (37.03)** (33.75)** (33.03)** (34.33)** (30.52)** (35.32)** (36.85)** (38.06)** (37.28)**
Experience2 -5,7% -5,7% -6,4% -6,1% -5,8% -5,5% -5,0% -5,2% -4,9% -4,2% -4,7% -4,6% -4,7% -4,0%
(x100) (22.88)** (24.11)** (28.43)** (27.74)** (27.56)** (26.87)** (23.04)** (23.47)** (23.99)** (19.46)** (23.04)** (23.84)** (25.48)** (23.05)**
Constant 35,4 56,2 66,9 73,4 85,6 100,6 142,2 176,5 184,7 190,3 185,6 190,5 208,6 243,9
(154.22)** (179.70)** (193.54)** (194.88)** (209.48)** (223.00)** (224.85)** (225.43)** (238.83)** (207.54)** (216.02)** (228.47)** (230.22)** (243.92)**
Observations 9971 10517 10941 11169 11656 13087 14158 14623 14663 15256 16823 16959 19184 19726
R-squared 0,33 0,39 0,38 0,38 0,38 0,39 0,36 0,33 0,37 0,35 0,4 0,41 0,38 0,42

EVALUATING EFFICIENCY IN THE PORTUGUESE HEALTH
AND EDUCATION SECTORS
Miguel St. Aubyn1
ISEG – UTL
mstaubyn@iseg.utl.pt
Abstract
Public expenditure in Portugal has been growing in recent years and approached
values close to the EU average. Social expenditure, namely with health and
education, increased in relative terms. Also, recent growth in public employment
resulted greatly from new hirings in these two sectors.
Taxation is the dominant source of financing education and health costs. Evaluating
efficiency in these sectors is of great importance for public policy – an inefficient
situation will imply the possibility of a better performance without increasing
allocated resources.
Increased health expenditure has been accompanied by a better health status of the
Portuguese. Nevertheless, efficiency frontier estimation results provide evidence in
favour of inefficiency. A more detailed analysis points towards a structure where
nurses and hospital beds are relatively scarce, resources are asymmetrically
distributed in geographical terms, and expenditure in pharmaceuticals and ancillary
diagnostic services is excessively high. Suggested policy implications include the
promotion of cost effectiveness in the public provision of health, a better co-
ordination and integration between hospitals and health centres, a review of
personnel management and payments policy, and induced changes in the
pharmaceutical market, in users behaviour and in formation policy.
In what concerns education, enrolment rates have progressed along with
expenditure growth, school expectancy in Portugal being close to EU average
values. Even if average expenditure per student is lower than the EU average,
results presented here point towards inefficiency in this sector. Graduation rates at
the end of secondary education are too low and Portuguese student performance is
disappointing at an international level. It is suggested that promoting efficiency in
this sector should include a careful examination and reassessment of teaching
standards in secondary schools, namely in key knowledge areas, a careful and open
assessment of school performance, a review of personnel management and
payments policy, a review of the breakdown of educational expenditures by resource
category and the development of a life-long education program.
JEL codes: I18, I21, I28, H51, H52.

Keywords: government expenditure and education, government expenditure and health,
government expenditure efficiency.
1
The author wishes to thank the referees, and Prof. Manuela Arcanjo, Prof. Gouveia Pinto and Dr. Manuel
Teixeira for their very useful suggestions. They are not in any way responsible for anything written in this
paper.
254 Miguel St. Aubyn
1. Introduction
Portuguese public expenditure weight on GDP has grown recently and attained levels
that are close to the European Union average. According to the European Commission
AMECO database, these figures were, 46.2 and 46.5 percent, respectively, in 2001.
Growth in public expenditure since 1995 resulted, in a functional classification, from
growth in the social functions. They grew 4.9 percent in real terms per year, the economic
functions having declined (-4.4 percent per year). Among the social functions, the more
important in 2001 were "education" and "health" (5.4 and 4.8 percent of GDP,
respectively). These were precisely the items that grew at a higher rate in the last 6 years
(5.5 and 4.8 percent, respectively)2.
Growth of expenditure on education and health has been accompanied by growth in
public employment in those sectors. From 1997 to 2000, the percentage of new subscribers
to the Caixa Geral de Aposentações related to these sectors varied between 44,1 and 53,6
percent3.
The quantities of public provision of health and education have a direct impact on
welfare and are important for the prospects of growth of the Portuguese economy4. The
efficiency in spending on health and education becomes a relevant topic because there is no
necessary connection between spending levels and provision levels. Increased spending
levels may lead to very little improvement in provision if inefficiency predominates. If
there are important inefficiencies, higher provision is best attained by eliminating them.
This paper is a contribution to the evaluation of efficiency in the health and education
sectors in Portugal. It is structured as follows. After this introduction, section 2
summarises some methods to be used in measuring efficiency. Section 3 reviews some
recent trends in the health sector, provides some new results concerning efficiency, and
establishes some policy implications for health. Section 4 is devoted to education and
similar in structure to section 3. Even if both sections have their own conclusions and are
independent from each other, the paper ends with some final remarks.
2. Evaluating efficiency - methodology

Results presented in this paper concerning efficiency evaluation in education and health
sectors are based in an estimation of efficiency frontiers. We follow a non-parametric
method known as "free disposal hull analysis" (FDH). Some results from an alternative
parametric method – corrected least squares (CLS) to health expenditures and outcomes are
also included. Also, previous results from other authors and to whom reference is made
here are based in similar methods. It is therefore important to have a grasp of this
methodology before proceeding5.
2
The author computed these figures from the Contas Gerais do Estado and from the last Orçamento Geral do
Estado. Nominal values were deflated using the GDP deflator.
3
See Caixa Geral de Aposentações (1998, 1999, 2000, 2001).
4
Most recent theoretical and empirical research on growth emphasise the importance of human capital. Temple
(2001) reviews this literature. Bassanini, Scarpetta and Hemmings (2001) provide empirical evidence that low
levels of human capital have hindered economic growth in Portugal in recent years.
5
The interested reader may refer to Fried, Lovell and Schmidt (1993), a book with several contributions on
efficiency frontier techniques and applications. Gupta and Verhoeven (2001) apply FDH analysis to education
Evaluating Efficiency in the Portuguese Health and Education Sectors 255
Suppose that under efficient conditions, health or education status of a population i,

measured by an indicator I i , the output, depends on health or education expenses per
habitant, the input, and on other variables (controls), C1i , C 2i ,..., C ni :
I i = F ( Di , C1i ,..., C ni ) .
If I i < F ( Di , C1i ,..., C ni ) , it is said that country i exhibits inefficiency. For observed
expense level and controls, the actual output is smaller than the best attainable one. FDH
and CLS are two different methods of estimating function F, the efficiency frontier.
In a simple example without controls, three different countries display the following
values for indicator I and expense level D:
Table 1
Fictitious values for countries A, B and C
Indicator Expenditure
Country A 65 800
Country B 75 1000
Country C 70 1300
Expenditure is lower in country A (800), and health or education level is also the lowest
(65). Country C exhibits the highest expenditure (1300), but it is country B that attains a
better level of education or health (75).
and health spending in Africa. Clements (1999) applies it to Portuguese education. Evans, Tandon, Murray and
Lauer (2000) include a discussion of different techniques and their application to health spending.
Graph 1
FDH frontier
B
75
C
70
A
65
D
800 1000 1300
Country C may be considered inefficient, in the sense that it performs worse than
country B. The latter achieves a better status with less expense. On the other hand, neither
country A nor country C shows as inefficient using the same criterion.
In FDH analysis, both countries A and C are supposed to be located on the efficiency
frontier. This frontier takes the following form in this example:
 65, 800 ≤ D < 1000

I = F ( D) = 
75, 1000 ≤ D ≤ 1300
This function is represented in graph 1.
It is possible to measure country C in two different ways:

i) Inefficiency may be measured as the vertical distance between point C and the
efficiency frontier. Here, one is evaluating the difference between the level of health or
education that could have been achieved if all expense was applied in an efficient way, and
the actual level of health or education. In this example, the efficiency loss equals 5 –
country C should, at least, achieve the same indicator level as country B, under efficient
conditions.
ii) If one computes the horizontal distance to the frontier, the efficiency loss is now 300,
in units of expense. It can be said that efficiency losses in country C are about 24 percent
(=300/1300) of total expense. To attain an indicator level of 70, it is necessary to spend no
more than 1000, as shown by country B.
FDH analysis is a non-parametric method, as frontier F is not previously specified.

Suppose now that, a priori, frontier F is considered to be linear and its parameters
estimated in a second step. One possibility is to adjust a least squares line to points A, B
and C. This line, depicted as a dotted line in graph 2, is not yet an efficient frontier, as it has
necessarily a point above it. Nevertheless, it is possible to shift the line upwards, adding to
it the symmetric of the smallest of the residuals. This method is known in the literature as
"corrected least squares", also represented in graph 2. In this case, one country only is on
the frontier (country B), and country A is now deemed inefficient. Vertical and horizontal
distances to the border can again be computed with the same economic interpretation.
Graph 2
CLS frontier
CLS frontier
B
75
C
70
A
65
D
800 1000 1300
As we do not have any particular reasonable assumption concerning the functional form of
the frontier function, we apply the FDH analysis to health and education data. We also
present some CLS results for health, for the sake of comparability.
3. Public expenditure with health

3. 1 Some recent trends in expenditure
Table 2
Public expenditure on health
1995 1996 1997 1998 1999 2000 2001
6
In 1000 euros 34806,1 37759,0 40188,1 43131,1 49236,3 52715,0 57942,0
% of total public expenditure 9,6% 9,6% 9,7% 9,7% 10,3% 10,4% 10,3%
% of GDP 4,3% 4,4% 4,3% 4,3% 4,6% 4,6% 4,8%
Real change
5,3% 2,6% 3,3% 10,2% 4,4% 5,4%
Source: CGEs, OGE 2002.
Graph 3
Source: OECD (2001a).
As mentioned before, public expenditure with health has clearly increased, both as a
percentage of GDP and as a percentage of total public expenditure. In 2001, health
expenditure accounted already for a bit more than a tenth of public expenditure, and this
6
Does not include regional and local government expenditure.
amounted to 5 percent of GDP. Considering the last 6 years, expenditure on health grew at
a real rate of 5,2 percent (table 2). Nevertheless, and despite its recent progress, Portuguese
public expenditure on health is below the EU average as a percentage of GDP, as shown in
graph 3.
Portuguese private expenditure on health is high in international terms. Namely, its
weight on GDP is above EU average (graph 47). Private expenditure on health in Portugal
amounted to approximately a third of total expenditures, compared to a figure close to a
quarter in the EU.
Graph 4
Due to the recent increase in both private and public health expenditure, total
expenditure on health as a percentage of GDP in Portugal was close to average values in
the EU (7.7 and 7.9 percent, respectively). These two time series are plotted in graph 5.
7
The drop in Portuguese data in 1989 is probably due to a break in the series.
Graph 5
Since Portuguese GDP per capita measured in purchasing power parities is clearly
below EU average, it results that total health expenditure in per capita terms and purchasing
power parities is also below EU average values. This expenditure is depicted in graph 6. In
less than 30 years, expenditure on health per capita grew from 27.1 percent of EU average
to 67.9 percent in 1998.
Graph 6
Source: OECD (2001a)
3.2 Some recent trends in health indicators

As shown above, Portuguese health expenditure, both private and public, has been
growing and its level is now closer to the EU average. This convergence trend in what
concerns costs is mirrored by convergence in results – in general terms, health indicators
show an approximation towards average EU values. Two indicators that display a similar
qualitative time trend are plotted below – infant mortality and potential years of life lost8
(graph 7 and 8, respectively). Portuguese, EU average and best performing country levels
are plotted in both figures.
8
The potential years of life lost potential years for all causes is the number of years of life lost when a person dies
"prematurely", before age 70.
Graph 7
In both cases there has been some convergence of Portuguese values towards European
values, in the sense that the gap between the series is much smaller in recent years.
However, the gap is still positive, meaning that Portugal is still one of the EU worst
performing countries.
Graph 8
3.3 Evaluating health expenditure efficiency

3.3.1 Previous results
The OECD health sector efficiency evaluation

The OECD 1997-98 report on Portugal includes a chapter where the health sector is
analysed in terms of size and performance, some cost pressures are identified and some
policy measures are recommended9.
In a graphical and relatively informal analysis, it is concluded that efficiency in this sector
is low when compared to other OECD countries. The study rests on the use of two health
output indicators, infant mortality and potential years of life lost, and one input variable,
total health expenditure as a percentage of GDP.
Results from this study deserve some criticism. Not all countries that spend the same
percentage of GDP on health spend the same amount of resources per habitant, as GDP per
head varies. Portugal is a relatively poor country within the OECD, so its expenditure per
habitant is lower. One could expect lower results even under efficient conditions10.
Moreover, it could be the case that GDP per head is also important in determining health
9
See OECD (1997).
10
In fact, in OECD (1998) all countries that spend about the same percentage of GDP on health have a higher
GDP per head. These countries are Australia, Austria, Finland, Iceland, Italy, Japan, New Zealand, Netherlands,
Norway and Spain.
status, i.e., GDP per head could be a control variable. Therefore, results from OECD (1998)
are not taken here as perfect evidence of Portuguese inefficiency in this sector.
The WHS evaluation of the comparative efficiency of national health systems in

producing health
The World Health Organisation (WHO) published in 2000 an extensive report on the
performance of health systems in 191 countries11. Portugal was well placed in the
efficiency ranking. All countries together, the Portuguese health system was the 13th more
efficient. Table 3 presents the ranking for OECD countries12.
Table 3
Ranking for OECD countries in WHO (2000)
WHO ranking
France 3
Italy 4
Spain 6
Japan 9
Greece 11
Portugal 13
Austria 15
Norway 18
Netherlands 19
Sweden 21
United Kingdom 24
Switzerland 26
Iceland 27
Belgium 28
Luxembourg 31
Ireland 32
Turkey 33
Canada 35
Australia 39
Germany 41
Finland 44
Mexico 64
Denmark 65
USA 72
New Zealand 80
Czech Republic 81
Poland 89
Hungary 105
Korea 107
11
See WHO (2000).
12
Note the probably unexpected US position. This is due to the very high US health expenditure per habitant. In
fact, this country expended approximately 41 percent more than Switzerland, the 2nd country in the
expenditure ranking. Moreover, health indicators in the US are below average (see table 4).
With these results, Portugal is the sixth more efficient country in producing health in the
OECD, which runs counter the perception conveyed by this same organisation in the report
mentioned before. To understand this result, it is important to have a grasp of the
methodology in which it is based.
Evans, Tandon, Murray and Lauer (2000) describe the methods and results used in
WHO (2000) in detail. The efficiency frontier was estimated by an extension of corrected
least squares to panel data, as data from 1993 to 1997 were considered when available. In
log form, it was considered that:
ln DALE = α1 + α 2 ln D + α 3 ln S + α 4 (ln S )
2
where DALE is the "disability-adjusted life expectancy", an output indicator, D is total

health expenditure per head in purchasing power parity, and S a control variable, the
average number of schooling in the population13. Countries were then ranked in accordance
with the vertical distance to the efficient DALE.
There are several differences from OECD (1998) that can well explain different results:
i) the sample is much larger, with 191 countries;

ii) resources were measured in per head terms, and not as a percentage of GDP;
iii) one control variable was considered.
iv) an efficient frontier was formally estimated and efficiency was effectively
measured.
The main criticisms one could point in what concerns the WHO study derive from the
high number of countries considered and from the no presentation of results with other
specifications for the efficiency frontier or with alternative ranking criteria.
Results could have been slightly different if sub-samples of more similar countries were
considered, as for instance, the OECD subset. It would be interesting to show the ranking
robustness to alternative ways of estimating the frontier. Moreover, and as explained in
section 2, efficiency can also be measured by the horizontal distance to the frontier.
Rankings with this other measure rarely coincide with vertical rankings.
13
"DALE is most easily understood as the expectation of life lived in equivalent full health." (in WHO (2000), p.
146). This reference includes a discussion on the advantages of using DALE and some technical information
on how it is estimated.
3.3.1 New results

Table 4
Basic data for efficiency in health evaluation
Health DALE Infant Medium Expen- DALE Inf. Educ.
expen- Mortality or higher diture ranking Mort. ranking
diture per education ranking ranking
head. ppp (%)
Australia 1601 73.2 5.3 57 16 2 12 20
Austria 1960 71.6 4.7 71 6 13 6 11
Belgium 1738 71.6 6 53 14 13 18 21
Canada 1836 72 5.5 76 10 9 14 7
Czech Republic 640 68 5.9 84 25 24 16 2
Denmark 1940 69.4 5.3 66 8 21 12 13
Finland 1539 70.5 3.9 67 17 17 3 12
France 2125 73.1 4.7 60 4 3 6 18
Germany 2365 70.4 4.8 81 3 18 8 4
Greece 964 72.5 6.4 44 23 7 21 23
Hungary 372 64.1 9.9 63 28 28 26 14
Iceland 1757 70.8 5.5 62 13 16 14 16
Ireland 1200 69.6 6.2 50 20 20 19 22
Italy 1824 72.7 6.2 38 11 6 20 24
Japan 1759 74.5 3.7 80 12 1 2 5
Korea 862 65 7.7 61 24 26 25 17
Luxembourg 1985 71.1 4.2 29 5 15 5 26
Mexico 421 65 16.4 21 26 26 28 27
Netherlands 1911 72 5 63 9 9 10 14
New Zealand 1393 69.2 6.8 60 18 23 23 18
Norway 1708 71.7 4.1 82 15 11 4 3
Poland 392 66.2 10.2 74 27 25 27 9
Portugal 1060 69.3 6.4 20 22 22 21 28
Spain 1211 72.8 5 30 19 5 10 25
Sweden 1943 73 3.6 74 7 4 1 10
Switzerland 2644 72.5 4.8 80 2 7 8 5
Turkey 231 62.9 39.5 17 29 29 29 29
United Kingdom 1193 71.7 5.9 76 21 11 16 7
USA 3724 70 7.2 86 1 19 24 1
Taking into account criticisms to the two studies on health production discussed
previously, the analysis presented here is restricted to OECD countries. Also, we
considered two alternative ways of uncovering the efficiency frontier (FDH and CLS) and
present rankings based both on vertical and horizontal distances. Moreover, two different
output indicators were considered (DALE and infant mortality). Health expenditure, the
input variable, is measured as total health expenditure per head in purchasing power
parities. In some specifications the percentage of the population without higher secondary
or tertiary education enters as a control variable. Data concerns 1997 and can be read from
WHO (2001), except for education, where the source was OECD (2001a).
Basic data for this study is reproduced in table 4.
Portugal is one of the countries that spend less on health per habitant (22nd place).
Also, it is one of the countries with worse results in health indicators (22nd place with
DALE, 21st place with infant mortality). As far as education is concerned, only Turkey has
a worse indicator than Portugal. This is a country were, in international terms, not much is
spent on health, and where health is relatively poor.
Clearly, a more detailed analysis is required to evaluate efficiency, and to see to what
point the same expense could buy better health, or the same health could be bought with
less spent resources. We present detailed results from FDH results for DALE, followed by
results for infant mortality. Results using a parametric approach (CLS) are only sketched in
the end of this section, details being available from the author on request.
Health expenditure and DALE

Graph 9 and table 5 summarise results from FDH analysis to the relationship between
health expenditure and DALE in 29 countries.
Graph 9
Countries are ordered by the percent value of expenditure inefficiency, a measure of the
horizontal distance towards the efficiency frontier in graph 9. The first eight countries
(Japan, Australia, Spain, Greece, Czech Republic, Poland, Hungary and Turkey) are
efficiency cases. There is no country in the sample that achieves the same level of DALE
attained by these countries with less expenditure.
Table 5
FDH analysis- DALE
Expen DALE Expen Efficient DALE Efficient
diture ranking diture losses expen losses DALE
ranking (%) diture
Japan 1 1 0.0% 1759 0 74.5
Australia 1 1 0.0% 1601 0 73.2
Spain 1 1 0.0% 1211 0 72.8
Greece 1 1 0.0% 964 0 72.5
Czech Republic 1 1 0.0% 640 0 68
Poland 1 1 0.0% 392 0 66.2
Hungary 1 1 0.0% 372 0 64.1
Turkey 1 1 0.0% 231 0 62.9
Italy 9 14 3.6% 1759 1.8 74.5
Canada 10 18 4.2% 1759 2.5 74.5
Norway 11 12 6.3% 1601 1.5 73.2
Mexico 12 10 6.9% 392 1.2 66.2
Belgium 13 13 7.9% 1601 1.6 73.2
Netherlands 14 18 8.0% 1759 2.5 74.5
Iceland 15 17 8.9% 1601 2.4 73.2
Portugal 16 23 9.1% 964 3.2 72.5
Denmark 17 28 9.3% 1759 5.1 74.5
Sweden 18 12 9.5% 1759 1.5 74.5
Austria 19 20 10.3% 1759 2.9 74.5
Luxembourg 20 24 11.4% 1759 3.4 74.5
New Zealand 21 25 13.1% 1211 3.6 72.8
France 22 11 17.2% 1759 1.4 74.5
United Kingdom 23 9 19.2% 964 0.8 72.5
Ireland 24 20 19.7% 964 2.9 72.5
Finland 25 16 21.3% 1211 2.3 72.8
Germany 26 26 25.6% 1759 4.1 74.5
Korea 27 22 25.8% 640 3 68
Switzerland 28 15 33.5% 1759 2 74.5
USA 29 27 52.8% 1759 4.5 74.5
Portugal comes in 16th place. 9.1% of the Portuguese expenditure is inefficient. Greece
achieves a DALE of 72.5 years, with an expense that is lower in 9.1 percent. Portugal only
attains a DALE of 69.3. Therefore, efficiency losses measured in DALE amount to 3.2
years (vertical distance, column 6). If the ranking was done based on the vertical distance,
Portugal would come 23rd, and therefore as one of the countries where more years of life
are lost.
Health expenditure and infant mortality

FDH analysis to the relationship between health expenditure and infant mortality is
synthesised in table 6. As before, countries were ranked by relative efficiency losses in
expenditure14.
Table 6
FDH – Infant mortality
Expen Infant Expen Efficient Inf. mortality Efficient

diture mortality diture expen losses inf.
ranking ranking losses diture mortality
(%)
Sweden 1 1 0.0% 1943 0 3.6
Japan 1 1 0.0% 1759 0 3.7
Finland 1 1 0.0% 1539 0 3.9
Spain 1 1 0.0% 1211 0 5
Czech Republic 1 1 0.0% 640 0 5.9
Hungary 1 1 0.0% 372 0 9.9
Austria 7 14 0.9% 1943 1.1 3.6
Luxembourg 8 13 2.1% 1943 0.6 3.6
Italy 9 26 3.6% 1759 2.5 3.7
Australia 10 19 3.9% 1539 1.4 3.9
Canada 11 22 4.2% 1759 1.8 3.7
Poland 12 8 5.1% 372 0.3 9.9
Netherlands 13 18 8.0% 1759 1.3 3.7
France 14 14 8.6% 1943 1.1 3.6
Denmark 15 20 9.3% 1759 1.6 3.7
Norway 16 8 9.9% 1539 0.2 3.9
Belgium 17 25 11.4% 1539 2.1 3.9
Iceland 18 20 12.4% 1539 1.6 3.9
New Zealand 19 22 13.1% 1211 1.8 5
Germany 20 16 17.8% 1943 1.2 3.6
Korea 21 22 25.8% 640 1.8 5.9
Switzerland 22 16 26.5% 1943 1.2 3.6
Greece 23 11 33.6% 640 0.5 5.9
Portugal 24 11 39.6% 640 0.5 5.9
United Kingdom 25 1 46.4% 640 0 5.9
Ireland 26 10 46.7% 640 0.3 5.9
United States 27 27 47.8% 1943 3.6 3.6
It is interesting to note that, among the six countries that define the efficiency frontier,
now with a different indicator, four of the previously DALE efficient countries are found
(Japan, Spain, the Czech Republic and Hungary).
Portugal comes now in the 24th place, with an inefficient estimated expenditure of 39.6
percent. Under efficient conditions, Portuguese infant mortality should have been 5.9 per
thousand, like in the Czech Republic, where expense is clearly lower, and not 6.4 per
thousand, as was observed. If countries are ranked by infant mortality losses (the vertical
distance), Portugal comes 11th, now in the first half of the table.
14
Turkey and Mexico were excluded here, as they were outliers, especially with corrected least squares analysis.
Summary of results
Table 7 summarises Portuguese results, also with corrected least squares (CLS).
Table 7
Results for Portugal – a synthesis
FDH-Inf. CLS – Inf.
FDH-DALE CLS-DALE
mort. mort.
Expenditure losses 9.1% 55.8% 39.6% 69.8%
DALE losses 3.2 3.9
Inf. mort. losses 0.5 2.1
Ranking by
16th (in 29) 14th (in 29) 24th (in 27) 15th (in 27)
expenditure
Ranking by DALE 23rd (in 29) 23rd (in 29)
Ranking by inf.
11th (in 27) 20th (in 27)
mortality
Portugal arises as an inefficient country in producing health, considering both indicators

and both methods. It is almost always placed in the bottom half of the table, considering
either vertical or horizontal distances. Estimated inefficient expenditure varies between 9.1
and 69.8 percent, depending on method and indicator used.
As total expenditure was used, inefficiency may arise either from public or from private
sector performances. They may also be explained by variables not considered here, like
different attitudes towards illness, or different levels of alcohol and drug abuse or AIDS
across countries. Nevertheless, the robustness and magnitude of results authorises the
serious consideration of inefficiency in the public sector, which is responsible by
approximately two thirds of total health expenditure. This is the reason why we turn now
into a more detailed quantitative analysis of public health expenditures.
3.4 A more detailed analysis

3.4.1 Resources in the Portuguese health sector – some international comparisons
The Eco-santé database (OECD (2001a)) allows an international comparison of

physical and human resources allocated to the health sector in 30 countries. A comparative
analysis of some indicators is done in this section, in order to characterise available
resources is the Portuguese health system. As shown below, the number of physicians in
Portugal is similar to the EU average, but there are relatively few general practitioners as
compared to specialists. They tend to concentrate in the Lisboa e Vale do Tejo region.
Nurses and hospital beds are relatively scarce.
Physicians
Graph 10
The number of physicians in Portugal per habitant increased through time in a

pattern very close to the EU average (3.2 physicians per 1000 habitants for both in 1999).
Nevertheless, specialists predominate in Portugal (2.2 per 1000 habitants in Portugal vs. 1.4
in the EU in 1999). On the other hand, the general practitioners are relatively scarce (0.9 in
the EU, 0.6 in Portugal). The time series for these indicators are depicted in graphs 10, 11
and 1215.
15
According to OECD (2001a), there was a break in the series in 1992. Before 1992 some specialists working in
hospitals were not registered as specialists in the Medical Order. Since 1992 they have been included.
Graph 11

Graph 12

National Health Service (or SNS – Serviço Nacional de Saúde) doctors tend to
concentrate in the Lisboa e Vale do Tejo region, as the majority of central hospitals is
located in Lisbon (see table 8).
Table 8
Geographical distribution of SNS doctors
Total % Per 1000 habitants
Norte 7 964 31.7% 2.54
Centro 4 909 19.5% 2.12
Lisboa e Vale do Tejo 10 567 42.0% 3.26
Alentejo 847 3.4% 1.90
Algarve 845 3.4% 2.42
Total 25 132 100.0% 2.65
Source: Departamento de Gestão Financeira (2001)
Nurses
Graph 13
Source: OECD (2001a)
Nurses are scarce in Portugal in European terms, as shown in graph 13. In the EU there
were 8.3 nurses per 1000 habitants in 1999. In Portugal, there were only 3.8. They were
more evenly distributed than doctors (see table 9).
Table 9
Geographical distribution of SNS nurses
Total % Per 1000 habitants
Norte 10 506 29.9% 3.35
Centro 8 223 23.4% 3.54
Lisboa e Vale do Tejo 13 073 37.2% 4.04
Alentejo 1 943 5.5% 4.37
Algarve 1 427 4.1% 4.09
Total 35 172 100.0% 3.71
Source: Departamento de Gestão Financeira (2001).
Hospital beds
Graph 14
There is a tendency in the EU and also in Portugal for a decreasing number of hospital
beds. Nevertheless, Portugal has had consistently much less beds than the EU average (see
graph 14). In 1998, there were 4 beds per thousand Portuguese, as compared to 6.9 per
thousand habitants in the EU.
3.4.2 National Health Service Costs

SNS is financed in about 90 percent by the state budget. Two cost items, labour costs
and services and suppliers amount to more than 80 percent of expenses, about 4 percent of
GDP (see table 10).
Table 10
SNS costs
1998 1999
nominal
% of total % of GDP % of total % of GDP
change
Purchases 18.3%
14.3% 0.66% 14.9% 0.73%
Services and suppliers 12.3%
38.7% 1.78% 38.2% 1.87%
Labour 13.2%
45.0% 2.06% 44.7% 2.19%
Other 33.3%
2.0% 0.09% 2.3% 0.11%
Total 14.0%
100.0% 4.58% 100.0% 4.90%
Labour costs
Additional remuneration is quite important in labour costs. These result both from nurse
and doctor labour costs (table 11).
Table 11
Labour costs breakdown, 1999
Wages, incl. holidays and Christmas 76.0%
Additional remuneration 14.9%
Other 9.1%
Total 100.0%
Table 12 compares mean incomes of nurses, doctors and of all workers in the health
sector to the mean income in the economy. Portuguese values are not much different from
those of other EU countries.
Table 12
Ratio of mean incomes in the health sector to mean incomes of all employees
1999 (or 1998)
Total health
Nurses Doctors
sector
Austria na na 1.00
Belgium 3.99 na na
Denmark 1.18 2.03 0.58
Finland 0.82 1.83 0.87
France 0.81 1.94 0.82
Greece na na 0.76
Ireland 1.15 na 0.97
Luxembourg na na 0.85
Netherlands na na 0.88
Portugal 0.99 1.57 0.85
United Kingdom 0.89 1.39 na
Source: Departamento de Gestão Financeira (2001); European Commission:

AMECO database; OECD (2001a).
Services and suppliers

This item includes essentially two expenditure items: ancillary diagnostic services and
expenditure with pharmaceuticals (table 13).
Table 13
SNS external services and suppliers costs– 1999
% of total % of GDP Annual nominal change
Ancillary diagnostic services 24.9% 0.4% 16.9%
Chemists 53.1% 0.9% 11.4%
Other 22.1% 0.4% -29.0%
Total 100.0% 1.7% 0.0%
Expenditure with ancillary diagnostic services, and more importantly, with

pharmaceuticals, attain very high levels in international terms. In 1998, Portugal was the
third EU country with higher public expenditure on pharmaceuticals per habitant and in
purchasing power parities (see graph 15). The Portuguese state spent more than 40 percent
in pharmaceuticals per habitant than the EU average16.
Graph 15
This high level of spending results from high state co-payments and from a very high
consumption of pharmaceuticals in the country. Considering all countries with available
data, Portugal came second in 1998 in what concerns sales of pharmaceuticals per habitant
in purchasing power parities (see graph 16).
16
This average excludes Austria, for which there were no data available on this matter in OECD (2001a).
Graph 16
3.5 Health: conclusions and policy implications
Health public spending has grown significantly in recent years. It is an important item
of the government budget and it amounts to nearly 5 percent of GDP. This is less than the
EU average, but it is composed by high private expenditure. Total health expenditure in
Portugal has been growing and is close to 8 percent of GDP. Increased resources allocated
to the health sector have been accompanied by an improvement of Portuguese health status.
Nevertheless, empirical evidence presented here favours the hypothesis that the sector
displays some important inefficiencies. The same expenditure level could buy better
health, or, alternatively, the same health status could be bought with less spent resources.
A more detailed analysis of the SNS showed that some resources are relatively scarce –
general practioners, nurses or hospital beds. Moreover, they tend to be asymmetricaly
distributed in geographical terms. This may well diminish efficiency in health care. On the
cost side, it partly explains costs with extra remuneration for doctors and nurses.
Also, it results from international comparisons that public spending in pharmaceuticals
and ancillary diagnostic services are excessively high in Portugal. Note that these
expenditure items amounted to 1.3 percent of GDP in 1999. This spending area has been
identified as one of the main sources of inefficiency in the Portuguese health sector in
different studies17.
We turn next into some policy implications of the inefficiency findings. Most of these
implications are shared by different authors on this subject, even if there is some
disagreement when it comes to implementation issues18.
Promotion of cost effectiveness in the public provision of health. This includes the
separation of provision from financing and regulation, more autonomous and flexible
management of institutions, the distribution of funds to hospitals based more on
performance and therapy-based indicators and less on past history, and to regions based on
health needs.
Better co-ordination and integration between hospitals and health centres. Better co-
ordination between different health institutions should ensure a better rate of utilisation of
installed capacity in the public sector.
Review of personnel management and payments policy. Staff remuneration, namely

doctors', could be reviewed in order to reward performance and to provide incentives for an
improved regional distribution.
Induce changes in the pharmaceutical market. Public policy should be designed in order to
induce more consumption of generic drugs. Competition in this market would be reinforced
by easing entry restrictions, by allowing the sale of some non-prescribed drugs in
supermarkets and other outlets, and by abolishing the fixed margin on the sale of
pharmaceuticals.
Induce changes in users behaviour. This includes incentives and educations of users in
order to behave economically (e. g. in accepting and asking for generic drugs or in not
overcrowding hospitals' urgencies).
Formation policy. Education policy should take into account the relative scarcity of some
health workers, namely nurses.
17
For example, in Pereira, Campos, Ramos, Simões and Reis (1999) or in Pinto and Oliveira (2001).
18
See OECD (1998), Pereira, Campos, Ramos, Simões e Reis (1999), Pinto e Oliveira (2001) and Conselho de
Reflexão Sobre a Saúde (1998). See also the interview of the minister of health to the newspaper "Vida
Económica" (22-28 de Fevereiro de 2002) for an exposé of recent governmental policy and goals.
4. Public expenditure with education

4.1 Education expenditure – some data
As documented before, public expenditure with education in Portugal has been growing
in the last decade19. Expenditure with education has become a more important item within
total public expenditure. According to educational data from the OECD and presented in
Table 14, expenditure with education in Portugal was 13.5 per cent of total public
expenditure in 1998 (12.5 percent in 1995). Portuguese behaviour in this matter was not
very different from OECD average behaviour. In fact, education expenditure was growing
in the OECD area at a comparable rate and its weight on public expenditure was not much
lower (12.9 percent).
Considering the structure by level, it is noticeable that about three quarters of public
education expenditure is devoted to non-tertiary education in Portugal. In the OECD area,
the weight of tertiary education is slightly higher (one third). The weight of public
education expenditure on GDP is very similar in Portugal and in the OECD area, 5.7 and
5.4 percent, respectively, in 1998.
Table 15 presents some numbers concerning educational expenditure per student in
purchasing power parity dollars, and therefore corrected for different price levels across
countries. Also, they include both public and private expenditure, and are based in full-time
equivalent students.
Portuguese expenditure per student is lower than the OECD average in real terms. The
difference is more striking for lower levels of education. Educational expenditure with a
pre-primary child is less than half of what is spent in other OECD countries. For a
secondary student, the difference is much smaller: in Portugal it is spent 87,6 percent of the
OECD average with this type of student. Some European Union countries spend less than
Portugal with secondary students, namely Greece, Spain and Ireland.
19
Note that private financing of the Portuguese educational system is of residual nature. According to Branco
(2000), this system works as "a quasi-monopoly of the public system, being financed almost exclusively by
taxation" (p. 212, in Portuguese).
Table 14
Total public expenditure on education - OECD countries
Direct public expenditure on educational institutions plus public subsidies to the private sector (including
subsidies for living costs, and other private entities) as a percentage of GDP and as a percentage of total public
expenditure, by level of education and year
Public expenditure on education as a percentage Public expenditure1 on education as a percentage

of total public expenditure of GDP
1998 1995 1998 1995
Primary, Tertiary All levels All levels Primary, Tertiary All levels All levels
secondary education of of secondary education of of
and post- education education and post- education education
secondary combined combined secondary combined combined
non- non-
tertiary tertiary
education education
Australia 10.2 3.6 13.9 13.4 3.5 1.2 4.8 5.0

Austria 7.8 3.2 12.2 12.0 4.0 1.6 6.3 6.5
Belgium 6.9 2.2 10.2 m 3.5 1.1 5.2 m
Canada 8.2 3.9 12.6 12.9 3.7 1.8 5.7 6.5
Czech Republic 6.3 1.8 9.3 8.7 2.9 0.8 4.3 4.9
Denmark 8.8 3.9 14.8 13.1 4.9 2.2 8.3 7.7
Finland 7.6 4.0 12.4 12.1 3.8 2.0 6.2 6.9
France 7.9 2.0 11.3 11.1 4.2 1.0 6.0 6.0
Germany 6.3 2.3 9.8 8.6 3.0 1.1 4.6 4.7
Greece 4.6 2.1 6.9 5.2 2.3 1.1 3.5 2.9
Hungary 7.8 2.4 12.4 12.2 2.9 0.9 4.6 5.0
Iceland 10.8 5.6 17.8 m 4.3 2.2 7.1 m
Ireland 9.9 3.5 13.5 13.0 3.3 1.1 4.5 5.1
Italy 7.1 1.6 10.0 8.7 3.5 0.8 4.9 4.6
Korea 12.7 1.8 16.5 m 3.1 0.4 4.1 m
Mexico 16.2 4.5 22.4 22.4 3.0 0.8 4.2 4.6
Netherlands 6.8 3.0 10.6 9.1 3.1 1.4 4.9 5.0
Norway 9.7 4.2 16.1 18.4 4.6 2.0 7.7 9.1
Poland 7.8 2.7 12.2 11.5 3.5 1.2 5.4 5.5
Portugal 10.2 2.4 13.5 12.5 4.3 1.0 5.7 5.4
Spain 8.1 2.2 11.1 10.6 3.3 0.9 4.5 4.7
Sweden 9.1 3.6 13.7 m 5.3 2.1 8.0 m
Switzerland 10.8 3.0 14.6 m 4.1 1.1 5.5 m
United Kingdom 8.3 2.6 11.9 11.2 3.4 1.1 4.9 5.2
Average 8.7 3.0 12.9 11.9 3.7 1.3 5.5 5.5
1. Public expenditure presented in this table includes public subsidies to households for living costs, which are
not spent on educational institutions.
2. Post-secondary non-tertiary is included in tertiary education and excluded from primary, secondary and post-
secondary non-tertiary education.
Table 15
Expenditure per student (1998)
Expenditure per student in US dollars converted using PPPs on public and
private institutions, by level of education, based on full-time equivalents
Pre-primary Primary All secondary

OECD countries
Australia m 3981 5830
Austria1 5029 6065 8163
Belgium2 2726 3743 5970
Belgium (Fl.) 2 2601 3799 6238
Canada 4535 m m
Czech Republic 2231 1645 3182
Denmark 5664 6713 7200
Finland 3665 4641 5111
France 3609 3752 6605
Germany 4648 3531 6209
Greece2 m 2368 3287
Hungary 2160 2028 2140
Ireland 2555 2745 3934
Italy1 4730 5653 6458
Japan 3123 5075 5890
Korea 1287 2838 3544
Mexico 865 863 1586
Netherlands 3630 3795 5304
Norway1 7924 5761 7343
Poland 2747 1496 1438
Portugal 1717 3121 4636
Spain 2586 3267 4274
Sweden 3210 5579 5648
Switzerland1 2593 6470 9348
United Kingdom2 4910 3329 5230
United States 6441 6043 7764
Average 3549 3932 5293
1. Public institutions only.

2. Public and government-dependent private institutions only.
Source: OECD (2001b).
If one examines the breakdown of educational expenditure by resource category (table

16), it is apparent that most of it is on compensation of staff, both for OECD countries in
general and for Portugal. In Portugal, the weight of compensation of staff is considerably
higher than average, especially in what concerns non-tertiary education – it amounts to
almost 90 percent of total expenditure.
Table 16
Educational expenditure by resource category (1998)
Distribution of total and current expenditure on educational institutions, by resource
category and level of education
Primary, secondary and

Tertiary education
post-secondary non-tertiary education
Percentage of total Percentage of current Percentage of total Percentage of current

expenditure expenditure expenditure expenditure
Compen- Compen-
Capital Other Capital Other
Current sation of Current sation of
current current
all staff all staff
Portugal 95 5 94 6 84 16 70 30
OECD
92 8 80 20 87 13 70 30
country mean
Table 17 provides an international comparison of statutory teachers' wages in the

OECD area. It is noticeable that Portuguese teachers receive a wage that is lower than the
OECD average, both at the start of a career and after 15 years of teaching. However,
teachers at the top of the scale are among the best paid in the OECD.
Table 17
Annual statutory teachers' salaries in public institutions at the lower secondary
level of education, in equivalent US dollars converted using PPPs (1998)
Starting Salary Salary at Ratio of Ratio of Ratio of Years from Percentage Salary
salary after 15 top of starting salary salary starting additional after 15
1
/minimum years' scale salary to after 15 after 15 to top bonus years'
training experience /minimum GDP years' years' salary experience
/minimum training per experience experience per
training capita to GDP per to starting teaching
capita salary hour
Australia 25,775 36,175 36,175 1.17 1.64 1.40 8 10 45

Austria 21,585 28,464 44,604 0.90 1.19 1.32 34 n 44
Belgium (Fl.) 19,472 27,932 34,262 0.80 1.15 1.43 27 n 40
Belgium (Fr.) 21,259 30,496 37,627 0.88 1.26 1.43 27 n 42
Czech Republic 7,027 9,342 12,477 0.50 0.67 1.33 32 20 13
Denmark 25,375 31,000 31,000 0.96 1.18 1.22 10 1 48
England 22,661 38,010 52,023 1.04 1.74 1.68 m m 48
Finland 20,660 27,942 29,127 0.93 1.25 1.35 20 20 58
France 22,579 29,615 42,697 1.01 1.32 1.31 34 12 47
Germany 32,769 38,640 43,156 1.43 1.68 1.18 28 n 53
Greece 19,871 24,337 29,165 1.35 1.65 1.22 33 n 39
Hungary 5,978 11,066 12,526 0.53 0.99 1.85 40 m 20
Ireland 23,303 36,151 40,708 0.99 1.53 1.55 22 n 49
Italy 21,108 25,773 31,546 0.95 1.16 1.22 35 m 42
Japan 21,899 41,201 52,867 0.91 1.72 1.88 31 31 m
Korea 24,150 39,921 66,269 1.62 2.67 1.65 41 6 80
Mexico 12,774 14,708 26,496 1.51 1.74 1.15 11 n 18
Netherlands 25,515 31,380 38,988 1.10 1.35 1.23 24 n 34
Norway 19,565 23,879 25,702 0.71 0.87 1.22 28 3 39
New Zealand 19,863 32,260 32,260 1.12 1.81 1.62 8 11 33
Portugal 16,429 26,288 47,975 1.07 1.71 1.60 29 20 42
Scotland 19,658 32,679 32,679 0.90 1.50 1.66 11 m 36
Spain 27,506 32,144 40,806 1.56 1.82 1.17 42 m 59
Sweden 18,389 23,896 m 0.84 1.09 1.30 m m m
Switzerland 38,143 51,361 59,133 1.44 1.93 1.35 23 n 60
Turkey m m m m m 1.17 20 m m
United States 24,624 32,713 43,458 0.78 1.03 1.33 30 22 34
Country mean 21,459 29,899 37,749 1.04 1.45 1.40 26 8 43
1. Percentage additional bonus is an average of two ratios: maximum bonus applicable to starting salary and
maximum bonus applicable to salary at top of scale.
4.2 Educational performance
When evaluating educational performance, we refer to two types of performance

indicators. Short of a better designation, we call them "quantitative" and "qualitative"
indicators.
Quantitative indicators, like enrolment rates, school expectancy and educational
attainment, assess the current situation and progress in schooling. Here, one is concerned
with the level of education attained by the population, and with the frequency of education
by people in age groups at the expected age of participation. It is important to stress a flux
versus stock interpretation. It is usual to use variables related to the educational attainment
of the population as proxy variables to "human capital", the existing stock of knowledge
and skills that exist in an economy. The educational system works mainly by a flux
mechanism akin to investment - as the participation rates of young people augment through
time, the existing human capital stock increases when these youth cohorts enter the labour
force. As it will be shown below, Portugal has a comparatively low human capital stock,
but an investment rate that has already converged to OECD average levels. With the
passing of time, and maintaining this quantitative performance, it is to be expected that
attainment levels will also converge to the more developed countries typical values.
Quantitative indicators of human capital stock and human capital formation have an
important drawback. When making international comparisons, there is the implicit
assumption that educational levels are equivalent across countries. More direct measures
are possible, by testing individuals for a number of abilities, like literacy or numeracy. This
is done both at the “stock” level, when adults are considered, and at the "formation" levels,
when student performance is assessed and compared internationally. These are the
indicators designated here by "qualitative". With this approach, and as documented below,
Portuguese performance becomes much lower than when quantitative indicators only are
considered.
Finally, and even if some characterisation is given on the stock side of the problem, it
should be emphasised that the performance and efficiency of the Portuguese educational
system will be evaluated considering the investment perspective. As in any other country,
most of the education expense is done with young people not yet in the labour force.
Table 18
Educational attainment of the population (1999)
Distribution of the population 25 to 64 years of age, by
highest level of education attained
Pre-primary Lower
and primary secondary Other
education education
Australia X 43 57
Austria1 X 26 74
Belgium 20 23 57
Canada 7 13 79
Czech Republic X 14 86
Denmark n 20 80
Finland X 28 72
France 20 18 62
Germany 2 17 81
Greece 41 9 50
Hungary 4 29 67
Iceland 2 35 63
Ireland1 23 26 51
Italy 25 32 44
Japan X 19 81
Korea 18 16 66
Luxembourg 24 14 62
Mexico 59 21 20
Netherlands 12 23 65
New Zealand X 26 74
Norway1 n 15 85
Poland 1 X 22 78
Portugal 67 12 21
Spain 42 23 35
Sweden 11 12 77
Switzerland X 18 82
Turkey 68 10 22
United Kingdom X 18 82
Country mean 16 20 64
X indicates that the data are included in the next column.

1. Year of reference 1998.
Table 18 clearly shows that the stock of human capital as measured by the educational
attainment of the population is much lower in Portugal than in most other OECD countries.
In 1999, only 21 percent of the population aged between 25 and 64 had attained a level
higher than lower secondary education, compared to a country mean of 64 percent. Portugal
was clearly behind other countries with similar GDPs per head, like Spain or Greece, and at
a level very similar to poorer countries, like Turkey and Mexico. As mentioned before,
empirical studies have shown that the lower stock of Portuguese human capital have
hindered economic growth in recent years20. Considering that the educational system is of
paramount importance to human capital formation, this becomes one of the sectors where
investment and efficiency will have more long-lasting effects.
Human capital formation measured in quantitative terms has clearly increased in
Portugal in recent decades. Graph 17 depicts enrolment rates for selected years between
1986 and 1998. These enrolment rates are defined as the percentage of the population with
a specific age that participates in the education system21.
Graph 17
Source: DAPP.
Progress in schooling among younger people is clearly noticeable as the curve shifts
upwards through time. In 1986, a minority of children attended pre-primary education.
Attainment between 6 and 12 years of age was universal, but there was a significant drop
after 12, with less than half of young people aged 16 years of age attending school. Since
then, significant progress has been made at both ends of the distribution: both pre-primary
and upper-secondary and tertiary education attendance became clearly more common.
20
See Bassanini, Scarpetta and Hemmings (2001).
21
Theoretical rates should not exceed 100 percent. Empirical rates may exceed 100 percent due to measurement
errors. According to DAPP, these measurement errors may result either from an underevaluation of the number
of individuals in the 1991 Census or from incorrect information transmitted at the school level. According to
the same source, schools may tend to inflate student numbers to justify tenure and teacher/student ratios and/or
more favourable budgets (DAPP, p. 12-13, in Portuguese).
In fact, Portuguese enrolment rates have converged to the OECD average in recent
years. From Table 19, one can notice that Portuguese values are not far from the OECD
country mean for all age groups considered. Consequently, a Portuguese child aged 5 is
expected to have almost 17 years of schooling under current conditions, the same as the
average child in OECD countries (see Table 20).
Table 19
Enrolment rates (1999)
Net enrolment rates by age group for full-time and part-time students in public and private
institutions
Students aged:
4 and under as a 5 to 14 as a 15 to 19 as a 20 to 29 as a
percentage of the percentage of the percentage of the percentage of the
population aged 3 to population aged 5 to population aged 15 population aged 20
4 14 to 19 to 29
Australia 33.8 97.7 80.3 27.3

Austria 56.2 98.7 76.7 18.2
Belgium 118.2 98.8 90.6 24.6
Canada 19.7 96.6 75.3 20.3
Czech Republic 66.9 99.3 74.8 12.9
Denmark 78.9 99.0 80.4 28.7
Finland 36.3 91.2 84.5 36.1
France 118.2 99.9 87.2 18.9
Germany 65.8 100.1 88.3 22.6
Greece 28.2 98.5 82.0 15.9
Hungary 78.6 99.8 78.1 17.2
Iceland 121.4 98.2 78.7 28.8
Ireland 27.8 99.9 79.8 15.0
Italy 98.0 99.2 70.7 16.9
Japan 76.3 101.2 m m
Korea 16.2 91.8 81.2 21.9
Luxembourg 57.9 95.3 73.8 4.7
Mexico 35.0 94.0 39.3 8.7
Netherlands 49.7 99.4 87.7 22.0
New Zealand 85.4 98.8 72.5 20.4
Norway 73.6 97.4 86.1 27.5
Poland 28.4 93.5 83.0 22.7
Portugal 61.9 105.6 76.3 18.8
Spain 97.0 104.8 76.3 23.7
Sweden 66.9 98.5 86.2 33.7
Switzerland 19.3 98.2 83.6 18.6
Turkey m 76.9 30.5 7.9
United Kingdom 77.4 99.0 72.5 23.6
United States 47.2 100.7 78.1 20.4
Country mean 62.1 97.7 76.9 20.7

Table 20
School expectancy (1999)
Expected years of schooling under current conditions in public
and private institutions, excluding education for children
under five years of age, all levels of education combined,
full-time and part-time
Sweden 20.3
Australia 19.9
United Kingdom 18.9
Belgium 18.5
Finland 18.3
Norway 17.9
Denmark 17.7
Iceland 17.7
Spain 17.3
Germany 17.2
United States 17.2
New Zealand 17.2
Netherlands 17.1
Portugal 16.8
Canada 16.5
France 16.5
Switzerland 16.3
Austria 16.0
Ireland 16.0
Poland 16.0
Hungary 16.0
Korea 15.8
Italy 15.8
Greece 15.6
Czech Republic 15.1
Mexico 12.4
Turkey 10.6
Average 16.7
Evidence from direct measures of human capital formation is not so favourable to

Portuguese performance. We refer here to two direct studies of student performance in an
international context with Portuguese participation: the Third International Mathematics
and Science Study (TIMMS), conducted in 1994-95, and the more recent PISA 2000
study22. In both cases students of about the same age from different countries were given
the same set of tests. The TIMM Study included Mathematics and Science tests, but not
22
More details on this studies can be found in TIMMS (1996), and OECD (2001b).
literacy ones. The PISA study included scientific, mathematical and reading literacy.
Results from both studies are summarised in Tables 21 and 22.
Table 21
TIMSS results (1995)
Mathematical literacy Scientific Average
literacy
Japan 605 571 588

Korea 607 565 586
Czech Republic 564 574 569
Austria 539 558 549
Hungary 537 554 546
Australia 530 545 538
Switzerland 545 522 534
Ireland 527 538 533
England 506 552 529
Belgium* 546 511 528
Sweden 519 535 527
Germany 509 534 522
France 538 498 518
Norway 503 527 515
Spain 487 517 502
Greece 484 497 491
Denmark 502 478 490
Portugal 454 480 467
Ranking Portugal 19/19 18/19 19/19
Average 526.39 531.03 528.71
*Simple average of Flemish and French speaking students

Source: TIMMS (1996).
Both studies included more countries than those presented in Tables 21 and 22. For the
sake of comparability, non-OECD countries are not considered here. Also, TIMMS results
are for 8th grade students, with an age close to those considered in the PISA study (15
years of age).23
Two main conclusions can be drawn from these results:
23
Note that Finland, Mexico, Poland and Sweden did not participate in the TIMM study. England, a participant in
TIMMS, is a part of the United Kingdom, present in PISA. As Belgium is concerned, the average of
performances of Flemish and French speaking students in TIMMS was computed by the author.
i) Portuguese performance is consistently low across categories (reading,

mathematics, and science) and in both studies. Portuguese students are the worse
or among the worst performers in every dimension one considers.
ii) Correlation between results is strong, in the sense that a country with a good
performance in reading is a good performer in mathematics or science, and that a
country that performed well in TIMMS also performed well in PISA. This last
correlation is illustrated in Graph 18. Note that the two best performers and the
two worst ones are the same, with Portugal included in the latter group.
Table 22
PISA results (2000)

Mathematical
Reading literacy Scientific literacy Average
literacy
Japan 522 550 557 543

Korea 525 552 547 541
Finland 546 538 536 540
Australia 528 528 533 530
United Kingdom 523 532 529 528
Ireland 527 513 503 514
Austria 507 519 515 514
Sweden 516 512 510 513
Belgium 507 496 520 507
France 505 500 517 507
Switzerland 494 496 529 506
Norway 505 500 499 502
Czech Republic 492 511 498 500
United States 504 499 493 499
Denmark 497 481 514 497
Hungary 480 496 488 488
Germany 484 487 490 487
Spain 493 491 476 487
Poland 479 483 470 477
Italy 487 478 457 474
Portugal 470 459 454 461
Greece 474 461 447 460
Mexico 422 422 387 410
Ranking Portugal 22/23 22/23 21/23 21/23
Average 499.51 500.17 498.70 499.46
Source: OECD (2001c).

Graph 18
4.3 Evaluating efficiency in education
The analysis so far has not been made in terms of a direct comparison between inputs
and outcomes in education. As with the health sector, some results based on efficiency
frontier methods are presented next. We first refer to some previous research on the subject
and then present some new results.
4.3.1 Previous results
Recent research by Clements (1999) applied FDH analysis to evaluate the efficiency of
educational expenditure in Portugal. In this study, two output indicators were considered:
(i) the ratio of secondary graduates to population at typical graduation age;

(ii) the scores on eight-grade exams in TIMMS (1996).
Different inputs were used, namely:
(i) Spending per student in purchasing power parities;

(ii) Educational expenditure to GDP, adjusted for population structure or for
student enrolment as a share of the population.
With different combinations of inputs and outputs, results were always in favour of an
inefficient performance of the Portuguese educational sector. In the OECD context,
Portugal was never in the efficient frontier. Most of the times it was quite far from it, being
placed in the second half of the ranking.
New results presented next may be interpreted as both an updating and a refinement of
Clements (1999) analysis. On one hand, we use more recent data, including the new PISA
(2001) study. On the other hand, we choose different input indicators. In fact, and as with
health, the consideration of expenses as a share of GDP may tend to bias results, as
Portugal is a poorer than average country. Also, it may be defended that spending per
student is not a very good input when the output is expressed in per population terms. In
our new results, spending was also expressed in population terms if the output is a
graduation rate. From a qualitative point of view, Clements (1999) results proved to be
robust to these changes – we also conclude clearly in favour of inefficiency.
4.3.2 New results
We apply here the techniques described in section 2 comparing inputs and outcomes in
education. In applying FDH analysis, two input measures were chosen: cumulative
expenditure per student and expenditure in primary and secondary education per habitant
aged between 5 and 19, in both cases in US dollars converted using purchasing power
parities. As the ultimate goal of education is successful learning, we have chosen two
different measures related to it: the PISA results described above, and the graduation rates
in upper secondary. The earlier is an international assessment, the latter being a national
standard of success. Raw data for OECD countries used in this analysis is presented in
Table 23.
The first evaluation was done comparing graduation rates and expenditure in primary
and secondary education per habitant aged between 5 and 19. Graph 19 depicts the inferred
efficiency frontier.
Graph 19
Portugal was the country where graduation rates were smaller. Two countries spending
less achieved considerably higher rates – the Czech Republic and Greece. Consequently,
Portugal was not in the efficient frontier. Numerical results for this analysis are presented in
Table 24.
With the Portuguese level of expenditure, a graduation rate of 82.71 should have been
achieved in efficient conditions. The vertical loss in graph 19 is therefore equal to 26.71,
the higher in the sample. Expenditure loss is estimated as 13.24 % of actual expenditure –
Portugal could have achieved this low graduation rate expending less (horizontal loss in the
graph). Portugal comes 7th in 11 countries when efficiency is measured by expenditure
loss.
Table 23
Raw data for efficiency analysis in education
Cumulative expenditure per Expenditure in primary PISA result, Upper Cumulative Expenditure PISA result, Graduation
student (US$ converted using and secondary education 2000 secondary expenditure, per habitant, ranking rates,
PPPs), 199824 per habitant aged 5-19 graduation ranking ranking ranking
(US$ converted using rates, 199825
PPPs), 1998
Australia 44 623 na 529.70 na 12 na 4 na
Austria 71 387 na 513.58 na 1 na 7 na
Belgium 46 338 49 200 507.49 84.2 10 4 9 4
Czech Republic 21 384 23 426 500.19 80.0 20 11 13 11
Denmark 65 794 na 497.45 na 3 na 15 na
Finland 45 363 na 540.12 na 11 na 3 na
France 50 481 48 596 507.46 86.6 9 5 10 3
Germany 41 978 48 329 486.97 93.3 14 6 17 2
Greece 27 356 28 425 460.41 82.7 19 10 22 6
Hungary 20 277 na 488.03 na 21 na 16 na
Ireland 31 015 na 514.32 na 17 na 6 na
Italy 60 824 na 474.14 na 6 na 20 na
Japan 53 255 43 634 543.08 96.0 8 7 1 1
Korea 30 844 na 541.24 na 18 na 2 na
Mexico 11 239 na 410.26 na 23 na 23 na
Norway 61 677 na 501.68 na 5 na 12 na
Poland 16 154 na 477.45 na 22 na 19 na
Portugal 36 521 na 460.96 56.0 16 9 21 11
Spain 36 699 37 715 486.60 67.1 15 8 18 10
Sweden 53 386 61 592 512.74 79.3 7 2 8 8
Switzerland 64 266 66 462 506.46 83.6 4 1 11 5
United Kingdom 42 793 na 528.22 na 13 na 5 na
United States 67 313 52 412 499.01 73.5 2 3 14 9
Source: OECD (2000, 2001b, 2001c).
24
Cumulative expenditure per student corresponds to average spending per student from the beginning of primary school up to the age of 15. See OECD (2001c).
25
Ratio of upper secondary graduates to total population at typical age of graduation.
Table 24
FDH analysis
Input: Educational expenditure per habitant aged 5-19
Outcome: Upper secondary graduation rates
Efficient
Expenditu Graduation Efficient
Expenditure expenditure Graduation
re loss rate loss graduation
loss (%) (1000 US$ rate loss
ranking ranking rate
ppp)
Japan 1 1 0.00% 43.63 0.00 95.99

Greece 1 1 0.00% 28.43 0.00 82.71
Czech Republic 1 1 0.00% 23.43 0.00 80.04
Germany 4 4 9.72% 43.63 2.71 95.99
France 5 5 10.21% 43.63 9.41 95.99
Belgium (fl.) 6 6 11.31% 43.63 11.77 95.99
Portugal 7 11 13.24% 28.43 26.71 82.71
United States 8 10 16.75% 43.63 22.47 95.99
Spain 9 8 24.63% 28.43 15.61 82.71
Sweden 10 9 29.16% 43.63 16.70 95.99
Switzerland 11 7 34.35% 43.63 12.43 95.99
Comparison of two other indicators allows having more countries in the sample, due to
more data availability. In graph 20 we plot the efficiency frontier derived from the
consideration of cumulative expenditure per student as an input and the average of PISA
results as an outcome.
Graph 20
Again, it is apparent that Portugal is far from the efficient position. Poland, Hungary,
the Czech Republic, Ireland and Korea spend less than Portugal per student and get better
results. It can be read from table 25 that expenditure loss (the vertical distance to the
frontier) amounts to 15.5 percent of actual Portuguese expense. With the actual expense per
student, a result 14.8 percent higher should be achieved in efficient conditions. Ranking
countries by expenditure loss, Portugal comes 11th in 23. It comes last when the ranking is
done by PISA results loss.
Table 25
FDH analysis
Input: cumulative expenditure per student
Outcome: Average of PISA results
PISA Efficient
Expenditure Expenditure PISA result Efficient
result loss expenditure
loss ranking loss (%) loss (%) PISA result
ranking (US$ ppp)
Japan 1 1 0.0% 53255 0.0% 543

Korea 1 1 0.0% 30844 0.0% 541
Czech 1 1 0.0% 21384 0.0% 500
Republic
Hungary 1 1 0.0% 20277 0.0% 488
Poland 1 1 0.0% 16154 0.0% 477
Mexico 1 1 0.0% 11239 0.0% 410
Sweden 7 12 0.2% 53255 5.6% 543
Ireland 8 10 0.6% 30844 5.0% 541
Italy 9 22 12.4% 53255 12.7% 543
Norway 10 16 13.7% 53255 7.6% 543
Portugal 11 23 15.5% 30844 14.8% 541
Spain 12 21 16.0% 30844 10.1% 541
Switzerland 13 15 17.1% 53255 6.7% 543
b 14 19 19.1% 53255 8.4% 543
Denmark
United States 15 18 20.9% 53255 8.1% 543
Greece 16 17 21.8% 21384 8.0% 500
Austria 17 11 25.4% 53255 5.4% 543
Germany 18 20 26.5% 30844 10.0% 541
United 19 9 27.9% 30844 2.4% 541
Kingdom
Australia 20 8 30.9% 30844 2.1% 541
Finland 21 7 32.0% 30844 0.2% 541
Belgium 22 13 33.4% 30844 6.2% 541
France 23 14 38.9% 30844 6.2% 541
4.4 Education: conclusions and policy implications
Results from the previous sections on education can be summarised as following:

i) Portugal is one of the OECD countries with less human capital, as measured by the
educational attainment of the population;
ii) There has been a very significant investment in education. This has allowed
enrolment rates and school expectancy to rise and to converge to more developed countries
levels.
iii) Investment in education has been done at a cost to public finance that is comparable
to other OECD countries, if one considers this cost in terms of its weight in total public
expenditure or in GDP. Moreover, educational expenditure per student is not very high in
Portugal in international terms. The weight of current expenditure, essentially staff

compensation, is higher than in other OECD countries.
iv) The Portuguese education system is not efficient if one takes into account the
relationship between the level of spending and different success measures. Portuguese
students perform badly in international exams, and upper secondary graduation rates are
low, due to high failure rates at the end of secondary school.
Considering this state of affairs, it seems sensible to consider the following avenues in
what concerns further research with strong policy implications26:
Careful examination and reassessment of teaching standards in secondary school,
namely in key knowledge areas, e.g. the languages, sciences and mathematics. International
student performance comparisons show that it is possible to better prepare students than it
is currently done in Portugal. In this respect, international assessment should be pursued
and deepened. The idea that course programmes, exams and classroom practices could be
improved deserves some attention. Namely, achievements in previous levels that are too
low in comparison to standards at the end of upper secondary may be one of the causes of
the high failure rate.
A careful and open assessment of school performance. School assessment should be
done carefully, independently and openly and not be confined to a simple comparison of
student average results in exams. Its main goal should be to identify best practices
susceptible of further dissemination. The adoption of best practices should be encouraged
by a well-designed system of incentives.
Review of personnel management and payments policy. Personnel management and pay
policy could be reviewed in order to introduce adequate incentives so that human resources
are allocated where they are most needed. Schools should not be either under or
overstaffed, and a correct composition of teaching staff should be achieved at school level.
Also, there seems to be scope to introduce some wage differentiation based on merit and
individual achievement, and less on length of service.
Review of the breakdown of educational expenditures by resource category. The fact
that in Portugal current expenditure is so much higher than capital expenditure deserves
some more research and points to the hypothesis of some reallocation of resources.
Develop a life-long education program. As education attainment of the population is
very low, convergence at a rewarding speed to more advanced international standards is
only possible with the re-qualification of the adult population.
4. Final remarks
This paper provided empirical evidence for the existence of important inefficiencies in
education and health spending in Portugal. Under the efficiency hypothesis, as the sector is
below the efficiency frontier, there is an apparent room for Pareto optimal improvements –
output can be expanded without increasing the cost. In practice, the implementation of
policies, either in education or health, such as the ones suggested in sections 3.5 and 4.4 or
any others, imply different and sometimes opposing choices that are ultimately the
26
Some of the ideas proposed here are adapted or inspired in Carneiro (2000) and Clements (1999).
responsibility of policy-makers. This research tries to inform these choices and point
towards some possible available options.
Some aspects, namely from a methodological point of view, deserve further research.
Namely, it would be important to further test robustness of results to more specifications,
different indicators or controls. Also, this author thinks this type of methods is rewarding
and could be applied to other public spending areas, as justice or local government. They
are also suitable to a more micro-oriented approach, and much could be learned from
applications of efficiency frontier methods to, say, Portuguese schools or hospitals.
References
Banco de Portugal (2001), Relatório do Conselho de Administração – 2000.

Branco, F. (2000), Estudo Prospectivo sobre O Financiamento da Educação em Portugal,
in Carneiro, Caraça and Pedro (2000), vol. 3.
Bassanini, A, S. Scarpetta and Ph. Hemmings (2001), Economic Growth: The Role of
Polcies and Insitututions. Panel Data Evidence from OECD Countries, Working
paper nº 283, OECD Economics Department.
Caixa Geral de Aposentações (1998), Relatório e Contas – 1997.
Caixa Geral de Aposentações (1999), Relatório e Contas - 1998.
Carneiro, R. (direction), J. Caraça and M. E. Pedro (co-coordinators) (2000), O Futuro da
Educação em Portugal, Tendências e Oportunidades – um estudo de reflexão
propectiva, 4 vols., Ministério da Educação, Lisboa.
Carneiro, R. (2000), 2020: 20 Anos para Vencer 20 Décadas de Atraso Educativo – Síntese
do Estudo, in Carneiro, Caraça and Pedro (2000), vol. 1.
Clements, Benedict (1999), The Efficiency of Education Expenditure in Portugal, IMF
working paper nº 99/179.
Conselho de Reflexão Sobre a Saúde (1998), Recomendações para uma reforma estrutural.
Departamento de Gestão Financeira (2001), Serviço Nacional de Saúde, Contas Globais de
1999, Instituto de Gestão Informática e Financeira da Saúde, Ministério da Saúde,
Lisboa.
DAPP – Departamento de Avaliação, Prospectiva e Planeamento, Séries Cronológicas
Básicas do Sistema Educativo, Ministério da Educação, available on the Internet:
www.dapp.min-edu.pt/d/estat/seri_cron/sc_98.html.
Evans, D., A. Tandon, C. Murray and J. Lauer (2000), The Comparative Efficiency of
National Health Systems in Producing Health: an Analysis of 191 Countries, GPE
Discussion Paper Series, nº 29, World Health Organization, Genebra.
Fried, Harold, C. Knox Lovell and Shelton Schmidt, editors (1993), The Measurement of
Productive Efficiency – Techniques and Applications, Oxford University Press,
Oxford.
Gupta, S. and M. Verhoeven (2001), "The efficiency og government expenditure –
Experiences from Africa", Jorunal of Policy Modeling, 23, 433-467.
Mossialos, E. and J. Le Grand (1999), Healh Care and Cost Containment in the European
Union, Ashgate, Aldershot, United Kingdom.
OECD (1998), Portugal, OECD Economic Surveys, OECD, Paris.
OECD (2000), Education at a Glance 2000, OECD database, CD-Rom, Paris.
OECD (2001a), Eco-Santé OCDE 2001 – Analyse Comparative de 30 Pays, CD-Rom,
Paris.
OECD (2001b), Education at a Glance 2001, OECD database, Paris, available on the
Internet: www.oecd.org.
OECD (2001c), Knowledge and Skills for Life – First Results from Pisa 2000, Paris.
Pereira, J.; A. C. de Campos, F. Ramos, J. Simões and V. Reis (1999), Health care reform
and cost containment in Portugal, in Mossialos and Le Grand (1999).
Pinto, C. G. and M. Oliveira (2001), The Portuguese Health Care System: Current
Organisation and Perspectives for Reform, Livro de Actas da IV Conferência
Sobre Economia Portuguesa, CISEP, ISEG, Lisboa.
Temple, J. (2000), Growth Effects of Education and Social Capital in the OECD Countries,
Working paper nº 263, OECD Economics Department.
TIMMS, International Study Center (1996), Highlights from TIMMS – Third International
Mathematics and Science Study. Availabe in the Internet:
www.csteep.bc.edu/timms.
World Health Organization (2000), The World Health Report 2000 – Health Systems:
Improving Performance, Genebra.
QUALIFICATION REQUIREMENTS AND EDUCATIONAL ATTAINMENT
IN PORTUGAL, 1985-1997
Maria Clementina Santos (∗)

M. Mendes de Oliveira (*)
Abstract
This paper uses data from staff logs (Quadros de Pessoal) to analyse the
evolution of over- and undereducation for the period 1985-1997 in Portugal.
In 1997, nearly 50 percent of workers have more education than their jobs
require. Furthermore, from 1985, a year prior to the accession to the
European Union, to 1997 it is found that the incidence of overeducation has
duplicated whilst that of undereducation had only a modest decrease.
Effects of over- and undereducation upon workers' earnings are also
evaluated. Results suggest that there are losses of productivity and earnings
due to the mismatch between educational attainment and qualifications
required in the labor market. Overeducated workers, holding a job for which
their qualifications exceed those that are standard among fellow co-workers,
would benefit from changing to a job where their levels of schooling were
fully utilised. On the other hand, undereducated workers (that is, those with
educational credentials below what is common in their occupation) would
stand to gain if they were to catch up with the level of schooling of fellow co-
workers. These findings justify some skepticism upon the ability of the
Portuguese economy, after access to European Community, in absorbing the
increasing number of more educated people supplied by the educational
system.
∗
Faculdade de Economia do Porto e CETE – Centro de Estudos de Economia Industrial, do Trabalho e da
Empresa.
304 Maria Clementina Santos, M. Mendes de Oliveira
1. Introduction
The existence of a mismatch in the labour market between skills required by available
jobs and school-provided workers’ qualifications has been an issue for decades. Concern
began earlier (e.g., Freeman (1980) or Thurow (1975)); but it was in the early 1980’s that
the concepts of overeducation and undereducation took root in the economics of education
literature, after seminal work by Rumberger (1981) and Duncan and Hoffman (1981).
Twenty years of research in the field have consistently established that overeducation and
undereducation are pervasive in the market; and, in what relates to earnings, that workers
whose educational qualifications exceed those required by the jobs they hold (overeducated
workers, for short) tend to earn higher returns to their years of schooling than co-workers
who are not overeducated, but lower returns than they would be able to get at a job where
their schooling credentials were fully utilized. Symmetrically, undereducated workers
receive lower returns than their co-workers with the required (and, thus, higher) level of
education, but higher returns than employees who have the same educational attainment
and work in jobs that require just their level of education.
These findings have been reported by a number of studies for economies on different
stages of development, with widely different labour-market institutional frameworks or
education systems, and employing strikingly different means of measuring overeducation
and undereducation. A representative survey across countries might include reference to
Duncan and Hoffman (1981), Rumberger (1987), Sicherman (1991), or Cohn and Khan
(1995) for the United States, Hartog and Oosterbeek (1988) or Groot (1993) for the
Netherlands, Groot and Maassen van den Brink (1997) for the United Kingdom, Alba-
Ramírez (1993) for Spain or Kiker et al. (1997) for Portugal. Some authors employed
workers’ self-evaluation of educational requirements to measure / undereducation, as is the
case with the seminal studies for the U.S. and the Netherlands, whereas others relied upon
independent evaluations of functional requirements and job-required qualifications
(Rumberger (1981), Burris (1983)), and others gave preference to data based criteria
(Verdugo and Verdugo (1989), Kiker et al. (1997)).
Spurred by accession to the European Union in 1986, the Portuguese economy
experienced a trend of strong increase of the educational level and vocational training of the
work force, at the same time that deep changes in the economic structure were called for. In
the presence of considerable rigidities of the labour market, adjustment was sluggish and
pockets of severe mismatch between educational qualifications and required job skills
developed. In this paper, we plan to address these issues, using Portuguese data for the
period 1985-1997. We will review the evolution of over- and undereducation measures over
the period and evaluate their effects upon workers’ earnings.
In Section 2, we discuss measurement issues and describe the data set. The incidence
and evolution of over- and undereducation in Portugal in recent years are presented in
Section 3 and their effects on earnings in Section 4. Concluding remarks are in Section 5.
2. Data and measurement
Data used in this study come from staff logs (Quadros de Pessoal) collected annually
by the Portuguese Ministry of Employment from all business firms with more than one
employee. Available is information on workers’ characteristics such as gender, age,
education, occupation, qualification level, years with the firm, hours worked and earnings,
Qualification Requirements and Educational Attainment in Portugal, 1985-1997 305
an on job attributes such as type of industry, geographic location and plant size. The
variables listed in Appendix Table A were constructed with this information. After
exclusion of observations with incomplete or inconsistent data and of a number of
categories of individuals for whom reported earnings may impart a bias upon correct
evaluation of labour income1, results to be reported below pertain to a random sample of
84873 workers surveyed in 1985 (25865 females and 59008 males), 78042 workers in 1991
(27183 females and 50859 males) and 83612 workers in 1997 (34671 females and 48941
males). Sample means are given in Appendix Table B.
In their pioneering study, Duncan and Hoffman (1981) employed the worker’s own
evaluation of educational standards necessary to perform his or her job to ascertain whether
there was overschooling, adequate schooling or underschooling. The measurement being
open to criticism on the grounds of subjectivity, biased recall and possibly conflicting
evaluation by workers holding identical jobs and schooling levels, subsequent research
attempted to improve upon the method. An alternative measure rested upon the evaluation
of required skills by independent job analysts and their correspondence to school-provided
abilities. Examples of this methodology can be found in the use of the general educational
development (GED) information contained in the Dictionary of Occupational Titles of the
U.S. Department of Labor (Rumberger, 1981) and in Dutch job qualification scales (Hartog
and Oosterbeek, 1988). In this setting, conversion of GED scores or qualification levels into
equivalent years of schooling has proved to be a difficult task.
Rather than relying upon the worker's own evaluation or upon exogenously designed
criteria, Verdugo and Verdugo (1989) proposed a different indicator of over-
education/undereducation based on the distribution of actual educational attainments in
each occupation defined at the 3-digit level. Their index classifies as undereducated a
worker whose schooling deviates (negatively) from the observed occupation average by
more than one standard deviation, and as overeducated those with (positive) deviation from
the mean in excess of one standard deviation.
In previous work (Kiker et al., 1997), we experimented with a variant of the procedure,
replacing Verdugo and Verdugo's mean-centred bracket with a different measure of central
location, the distribution mode. The mode is less sensitive to the presence of outliers in the
data and provides a more accurate measure of the extent of surplus or deficit schooling than
can be garnered from the mean.
Both arguments can best be illustrated by means of a simple numerical example.
Consider a group of 10 workers with the same 3-digit occupation code, of which 8 are
college graduates (16 years of schooling); the ninth is a high-school graduate (12 years) and
let x denote years of schooling of the tenth worker. According to the mode criterion, the
ninth worker will be classified as undereducated with 4 years of deficit schooling, no matter
what the education level of the last worker turns out to be. In contrast, the mean-centred
criterion will classify the ninth worker as undereducated or adequately educated depending
on whether the tenth worker has at least 8 (x ≥ 8) or less than 8 years (x < 8) of schooling.
Furthermore, if the ninth worker is deemed undereducated, the amount of the assigned
education deficit may vary from 0.1 years to around 2.3 years as the last worker's
attainment increases from 8 to 16 years of schooling. Particularly striking are the results
when x = 16, that is, when all but one worker are college graduates. The mode-based
1
Excluded were individuals who were simultaneously owners and executives, self-employed, part-time and
unpaid family workers, individuals under 14 years of age, farmers and farm labourers.
criterion plausibly takes 16 years to be the required schooling level and measures the extent
of the schooling deficit of the undereducated worker as 4 years, whilst the mean-centred
evaluation produces a 14.3-16.9 range for years of required schooling and a value of 2.3 for
the amount of undereducation.2
If one is to resort to observed data to establish incidence of overeducation/under-
education and measure the extent of surplus or deficit schooling, the arguments presented
above clearly suggest that a mode-based criterion will produce better results than the mean-
based alternative.
Furthermore, we aim in this paper at the evolution of over- and undereducation. In this
setting, methods of measurement that rest upon administrative sources, which are only
infrequently updated, fail to accommodate changes in technological conditions and
workplace organization. On the contrary, data based criteria, at least, have the potential for
self-correction. In fact, if a technological change takes place that has the effect of making
higher education the required level thereafter, firms will retrain the existing work force,
adjust hiring standards upwards and replace each vacant or newly created position with a
better-educated employee. Over time, this will lead to a change in the mean and modal
values of education and, thus, to a more correct evaluation of the extent of over- and under-
education.
In this paper, we classified as adequately educated workers with completed schooling
equal to the modal value of employees holding identical jobs, as defined by the
occupational codes recorded in the Quadros de Pessoal, at the 3-digit level in the 1985 and
1991 waves and at the comparable 4-digit level in 1997 data.3 Appendix Tables C and D
provide detailed information on the number of workers, mean and modal values of
schooling and percentage of workers deemed adequately educated (that is, with the modal
value of schooling) for each of the most representative occupational codes.4
If the quantification of over- and undereducation is still an unsolved matter in the
literature, the consensus appears to be wider on what the attempted target is: the possible
mismatch between the worker’s educational investment and the qualifications actually
required by the job he happens to hold. It can be argued however that, individual
heterogeneity being the norm, different educational attainments among workers with
similar jobs are more a reflection of different innate ability, productivity, social
background, quality of the schooling acquired or of other unobserved trait rather than
evidence of a mismatch between the individual’s and the job’s attributes. It might be the
case that workers with higher education levels than are average or standard among their co-
workers were less able, less productive or have lower-quality education than what is
common among workers with the number of years of schooling they possess and thus be
forced to take jobs further down the occupation ladder. Symmetrically, undereducated
2
Verdugo and Verdugo (1989) refrained from using a quantitative measure of over- and undereducation and
relied on their mean-centred bracket merely to define dummy variables as to whether a worker was over- or
underschooled.
3
A new National Classification of Occupations (Ministério do Emprego e da Segurança Social, 1994) was
adopted in 1994. This review of an earlier classification, dating from 1980 (Ministério do Trabalho, 1980), was
conducted on the basis of the International Standard Classification of Occupations (ISCO-88) published by the
International Labour Office. In both pre-1994 and post-1994 data we retained the level corresponding to the so-
called base groups of occupations.
4
Full results, relating to some 225 occupations in both the 1985 and 1991 waves and 350 in the 1997 wave, are
available from the first author on request.
workers might be viewed as overachievers who were able to get jobs normally slotted for
people with higher education credentials than they hold.
This view is hardly reconciled with overwhelming evidence to the effect that
overeducated workers are paid a premium over, and undereducated workers earn less, than
their (adequately schooled) co-workers. If they were equally productive as fellow co-
workers, despite different schooling levels, the persistence and generality of pay
differentials could hardly be explained.
2. Incidence of overeducation and undereducation
A rapid overview of the evolution of educational attainment in Portugal in recent years

can be garnered from Figure 1.
Figure
Percentage distribution of workers by years of schooling, 1985, 1991, 1997
1985
60%
1991
1997
50%
40%
30%
20%
10%
0%
< 4 yrs 4 yrs 6 yrs 9 yrs 11-12 yrs University
Years of schooling
In the Portuguese educational system, elementary schooling takes 4 years; as of 1985,

close to 57% of all workers in our sample had just 4 years of education, a figure that
declined markedly over the 13 years of the analysis, to less than 39% by 1997. A fringe of
workers however, mainly elder people, was illiterate or failed to complete elementary
school; they represented 9% of our sample in 1985 and 2.5% in 1997. Secondary school
currently requires 8 (formerly, 7) additional years, with different degrees being granted at
the end of the 6th, 9th and 12th years of schooling. Mandatory schooling has been set at 9
years for children born after 1980; as a result, the percentage of workers with 9 years of
education more than doubled from 1985 (7%) to 1997 (15%), when the effects of the legal
change took hold. Completion of secondary school climbed from 12% in 1985 to 15.5% in
1997. Just 1.9% of our workers in 1985, 2.3% in 1991 and 5.6% in 1997 held university
degrees (requiring from 14 to 18 years of education, depending on age cohort, type of
college degree and course of studies undertaken).
The incidence or overeducation and undereducation by gender for each of the years
1985, 1991 and 1997 is shown in Table 1.
Table 1
Percentage Distribution of Workers by Educational Status
Females Males
Worker’s status 1985 1991 1997 1985 1991 1997
Overeducated 16 24 34 19 25 34
Adequately educated 62 58 49 63 61 50
Undereducated 22 18 17 18 14 16
Total 100 100 100 100 100 100
Sample size 25865 27183 34671 59008 50859 48941
Notes: Workers with the modal number of years of schooling for their occupation are classified as adequately
educated. Overeducated are those who completed more years of schooling (less, for undereducated) than is the
mode among workers with the same base group occupation code.
Given the very low level of educational attainment in Portugal (an average close to 7
years in the 1997 sample), it is not surprising that overeducation is more prevalent than
undereducation. In the 1985-1997 period, overeducation is clearly on the rise, whilst
undereducation exhibits a slightly decreasing trend. Using a similar criterion, Cohn et al.
(2000) also report a greater incidence of overschooling than of underschooling for both the
United States and Hong Kong, but no discernible trend in either territory in regard to the
evolution of the two measures from 1986 to 1991. In a recent meta-analysis, Groot and
Maassen van den Brink (2000) report average rates of incidence of 21.5% for
overeducation and 13.9% for undereducation, in European countries covered in the 25
studies they surveyed.
Tables 2 (for female workers) and 3 (for males) contain information on the distribution
of over- and undereducation by years of potential experience, computed by subtracting
from the worker’s age at the time of the surveys the sum (number of years of schooling
required for degree held + 6). Particularly striking are the figures on the first line, which
refer to entrants in the labour market (5 years or less of potential experience). From them, it
can be garnered that in recent years the overwhelming majority (65% of female workers
and 73% of males) of entrants take jobs for which their educational qualifications are
higher than those common among co-workers. On the other hand, only 2% of new entrants
are found to possess lower educational level than is standard in their jobs. These findings
are in line with the increase in school enrolments observed in the last two decades, in the
wake of legislative action to raise mandatory schooling to 9 years.
Among more experienced workers, the percentages of incidence of over- and
undereducation take opposite directions, descending for the former and ascending for the
latter as new brackets come into light. Experience being correlated with age, it should be
expected that older cohorts would reveal higher prevalence of deficit schooling and that; as
they leave the market, undereducation becomes less and less significant, as in fact happens
with females from 1985 to 1997; for males, the picture is less clear. But it is somewhat
disturbing to find that the rate of incidence of overeducation increases from 1985 to 1997
for all the brackets of potential experience. The figures strongly suggest that younger and
more educated cohorts have been largely unable to find jobs in which their educational
attainments would be fully utilised, casting doubt about the ability of the changing
economic structure to take advantage of better endowments of human capital.
Table 2
Percentage Distribution of Workers by Years of Potential Experience, Females
Number of years of Overeducated workers Undereducated workers

potential experience 1985 1991 1997 1985 1991 1997
0-5 48 58 65 5 5 2
6-10 40 59 58 10 5 7
11-15 28 45 50 13 9 11
16-20 13 32 42 16 14 14
21-25 7 18 33 20 16 17
26-30 7 10 22 22 19 22
31-35 7 7 10 30 22 26
36-40 7 5 8 36 27 27
41 or more 4 4 6 52 42 33
Table 3
Percentage Distribution of Workers by Years of Potential Experience, Males

potential experience 1985 1991 1997 1985 1991 1997
0-5 71 76 73 4 4 2
6-10 54 63 65 6 7 6
11-15 39 51 52 9 9 10
16-20 23 38 45 12 10 12
21-25 16 29 35 14 11 15
26-30 12 19 25 15 13 18
31-35 10 13 15 18 15 23
36-40 8 11 11 23 16 26
41 or more 5 6 7 37 26 28
Tables 4 and 5 display information on the distribution by years of tenure with the same
employer. By construction, the first bracket (5 years of tenure or less) includes virtually all
of the new entrants in the labour market; it includes too workers with longer working lives
who held jobs with other firms before moving to their current (at the time of the surveys)
positions. Apparently, newcomers to the market accept positions for which they are
overqualified but gradually move into jobs where their education credentials are put to
better use. At the same time, extensive pockets of undereducation remain among
individuals with long tenures. However, the 1985 to 1997 trend is still one of rising extent
of overeducation, whereas the rate of adequate allocation is diminishing for most brackets.
Again, this lends support to the view that, despite the modernisation spurred by UE
accession in the last 15 years, the economy has been unable to absorb the higher-skilled
individuals that the educational system annually provides to the market. Lower productivity
growth and considerable social waste are to be expected.
Table 4
Percentage Distribution of Workers by Years of Tenure, Females

tenure 1985 1991 1997 1985 1991 1997
0-5 23 33 43 15 11 11
6-10 17 28 35 20 15 15
11-15 10 20 30 25 21 19
16-20 8 12 21 28 25 29
21-25 9 8 11 32 30 35
26-30 16 7 6 40 33 39
31-35 11 7 5 42 47 39
36-40 15 5 4 43 61 49
41 or more 17 10 3 46 50 61
Table 5
Percentage Distribution of Workers by Years of Tenure, Males

tenure 1985 1991 1997 1985 1991 1997
0-5 26 32 43 12 10 11
6-10 22 31 34 15 12 14
11-15 14 25 27 21 15 18
16-20 11 18 25 25 17 21
21-25 10 14 13 26 20 33
26-30 10 12 9 28 22 36
31-35 12 12 7 31 23 41
36-40 9 15 8 35 26 38
41 or more 12 13 9 33 35 43
4. Effects on earnings
The general form of the regressions reported below is
ln Y = α + S δ + H β + Z γ + u
where Y denotes labour earnings, α is a constant, S is a (row) vector of variables related to

the overeducation/undereducation issue, δ is a conformable (column) vector of coefficients,
H is a vector of human-capital and hours-worked variables (TENURE and its square,
EXPER and its square, LNHOURS), β is the vector of coefficients of the variables in H, Z
is a vector of dummy variables to control for plant size, industry and geographical location,
γ is the associated vector of coefficients, and u is a random term.
Following the usual pattern in the overeducation/undereducation literature, observed
years of completed schooling, EDUC for shortness, are decomposed in three parts:
EDUC = ADSCHOOL + OVERSCH − UNDERSCH .
After evaluation of the number of years of schooling required for adequate job performance
(ADSCHOOL), if the individual's observed educational attainment is different the excess is
assigned to OVERSCH or, alternatively, the deficit is attributed to UNDERSCH. Thus,
instead of a single explanatory variable for schooling, all three components belong in vector
S. Other than these, in S are also included interaction terms defined by the product of the
three main variables and those purporting to measure firm-specific and market-related
general human capital, TENURE and EXPER. The earnings equation can be rewritten as
ln Y = α + φa ADSCHOOL + φo OVERSCH + φu UNDERSCH +

+ δat ADTEN + δae ADEXP + δot OVTEN + δoe OVEXP + δut UNDTEN + δue UNDEXP + H β + Z γ
+ u,
as posited in earlier research (de Oliveira et al, 2000). Under the selected specification,
marginal returns to a year of required schooling are measured by the partial derivative
∂ ln Y
= φa + δat TENURE + δae EXPER.
∂ ADSCHOOL
Similarly, marginal returns to surplus schooling (for overeducated workers) and to deficit
schooling (for undereducated workers) are given by
∂ ln Y
= φo + δot TENURE + δoe EXPER
∂ OVERSCH
and
∂ ln Y
= φu + δut TENURE + δue EXPER,
∂ UNDERSCH
respectively. In this way, we allow for different profiles of the returns to overeducation and
undereducation in relation to tenure and experience.
Inclusion of the six interaction terms allows for substitutability or complementarity of
different forms of the human capital endowment (educational, general or firm-specific,
chiefly). Overeducated workers may substitute education for the lack of previous job
experience, accepting jobs requiring less education than they actually possess in order to
acquire the necessary experience and thus improve their chances of finding a better match.
Similarly, undereducated workers may substitute experience for the lack of education and
thus acquire a mix of human capital endowments adequate enough for effective job
performance. Supposedly, deficit schooling might be compensated by a larger stock of
firm-specific capital or of general market-market experience, whereas surplus schooling
might be viewed as a substitute for the lack of firm seniority or of work experience. It is
generally expected, and evidence has held that view, that signs are positive for φa and φo
and negative for φu, with φo < φa.
Table 6 conveys information on the estimates obtained with the sample of female
workers for each of the years 1985, 1991 and 1997; Table 7 refers to the sample of males.5
In both samples, disregarding possible interaction effects of tenure and experience,
estimates of the marginal returns to schooling adequate for job performance are 4.7%, 5.4%
and close to 7.4% for women in 1985, 1991 and 1997, respectively, and 5.1%, 7.5% and
8.6% for men. Computed instead at the mean values of tenure and experience (from
Appendix Table B), marginal returns climb from 6.1 to 7.0 and 9.1% for females and from
6.9 to 8.6 and 10.8 for males, for each of the years 1985, 1991 and 1997, respectively. All
four sets of estimates suggest an increase in the returns to education from the mid 1980's to
the late 1990's, a finding that may be surprising in light of the observed increase of the
average level of education of the Portuguese workforce. Nevertheless, our results are fully
corroborated by recent studies of the Portuguese labour market using different waves of the
Quadros de Pessoal data base: both Cardoso (1998) and Machado and Mata (2001) report
increasing returns, over the period 1983-1992 for the former and for the years 1982 and
6
1994 in the latter case.
Marginal returns to a year of overeducation fall in the range 3.7-5.0%, whilst those for
undereducation are between –3.0 and –7.1%. Groot and Maassen van den Brink (2000)
report an average (across studies) rate of return to required education of 7.6% (standard
5
Separate earnings regressions by gender are standard in the literature. The possible existence of gender
discrimination in the labour market would of itself recommend the approach. Evidence of possible
discrimination in the Portuguese labour market can be found in a number of studies. Kiker and Santos (1991)
report a gender differential of 19 to 22% with 1985 data. Using quantile regressions and the same data base
(the Quadros de Pessoal), Machado and Mata (2001) find that the distribution of women's wages is to the left
of men's by as much as 8.6% (in 1982) or 10.7% (in 1994) for the lowest decile, 13.2% (1982) or 14.3% (1994)
at the median and 19.9% (1982) or 17.2% (1994) at the 9th decile. Note, however, that our calculation of the
modal value of years of schooling for each occupation was made irrespective of gender, neither available
evidence nor logical reasoning having led us to believe that the adequate schooling for each occupation
depended upon the worker's being a woman or a man.
6
The increase in the "average" marginal returns, however, may be masking different evolution by schooling
level. See Machado and Mata (2001) for evidence regarding an increase in the dispersion of the rates of return
to education.
deviation: 2.4) for Europe, with rates of 2.1 (standard deviation: 4.6) for over- and –1.2
(standard deviation: 7.0) for undereducation7.
If differential effects of tenure and experience are allowed for, returns to education are
shown to improve more clearly with the former than with the latter, as the coefficient of
ADEXP fails to attain the 0.001 level of statistical significance in the regressions for female
workers.
Table 6
Selected Coefficient Estimates, Females
Variable Coefficient estimates

1985 1991 1997
ADSCHOOL 0.0474 0.0539 0.0743

OVERSCH 0.0374 0.0457 0.0417
UNDERSCH -0.0301 -0.0464 -0.0643
ADTEN 0.0012 0.0014 0.0019
ADEXP 0.0002 (*) 0.0002 (*) 0.0002 (*)
OVTEN 0.0013 0.0021 0.0015
OVEXP 0.0001 (*) -0.0005 -0.0008
UNDTEN -0.0006 -0.0005 -0.0005
UNDEXP -0.0000 (*) 0.0003 (*) 0.0004
TENURE 0.0130 0.0101 0.0080
TEN2 -0.0003 -0.0002 -0.0002
EXPER 0.0123 0.0133 0.0166
EXPER2 -0.0002 -0.0002 -0.0003
LNHOURS -0.1534 -0.1208 0.2863
R2 0.633 0.577 0.580

Sample size 25865 27183 34671
Notes: Dependent variable is ln Y. Besides those reported in the table, included as regressors were a
constant, 3 plant-size dummies, 4 region dummies and 9 industry dummies. All coefficients are
statistically significant at the 0.001 level or better, except for those noted (*).
Returns to schooling in excess of that required by the job are positive, but lower than
those obtained in a job for which educational attainment is fully adequate. Returns to
overeducation improve with the accumulation of firm-specific capital, as evinced from the
positive and statistically significant estimates for δot; the evidence is not as clear in what
concerns δoe. The results run counter the view of overeducated workers as being in a
transient stage towards a better match. Tenure, in particular, is fairly well rewarded, which
we take to be strongly suggestive of sizeable investments in firm-specific capital by both
employee and employer.
7
In fact, our numbers are not strictly comparable to those in Groot and Maassen van den Brink (2000). The rates
of return we chose to compare with theirs apply to individuals with 0 years of tenure and experience. Actual
returns, in our study, should include additional terms for tenure and experience, as can be garnered from the
expressions for the partial derivatives shown above.
Table 7
Selected Coefficient Estimates, Males
Variable Coefficient estimates

1985 1991 1997
ADSCHOOL 0.0508 0.0748 0.0857

OVERSCH 0.0388 0.0495 0.0344
UNDERSCH -0.0390 -0.0616 -0.0713
ADTEN 0.0011 0.0006 0.0019
ADEXP 0.0004 0.0003 0.0005
OVTEN 0.0010 0.0011 0.0018
OVEXP 0.0003 0.0001 (*) -0.0001 (*)
UNDTEN -0.0006 -0.0001 (*) -0.0002 (*)
UNDEXP 0.0000 (*) 0.0002 (*) -0.0001 (*)
TENURE 0.0163 0.0184 0.0179
TEN2 -0.0003 -0.0003 -0.0003
EXPER 0.0146 0.0178 0.0192
EXPER2 -0.0003 -0.0003 -0.0003
LNHOURS 0.2673 0.1727 0.7482
R2 0.554 0.492 0.547

Sample size 59008 50859 48941
Notes: Dependent variable is ln Y. Besides those reported in the table included as regressors were a
constant, 3 plant-size dummies, 4 region dummies and 9 industry dummies. All coefficients are statistically
significant at the 0.001 level or better, except for those noted (*).
Deficit schooling is penalized. The coefficient estimates of UNDERSCH and of the

interaction term UNDTEN consistently display a negative sign and are generally
significant. Marginal returns to the schooling investment of undereducated workers appear
to deteriorate with prolonged tenure with the same employer, a result that cannot be
reconciled with a view in which firm-experience compensates for the lack of adequate
education.
A possible explanation for the behaviour of the coefficients of the interaction terms lies
in technological change. If a process is taking place by which undereducated workers are
gradually being replaced by more qualified individuals, the obsolescence and devaluation
of the educational endowments of the former will show up in the form of negative signs for
δut. Employers will tend to shift their investments in training and firm-specific capital to
overeducated workers whose educational background provides a better match with the job
skills now rendered necessary by technological change and marginal returns to surplus
schooling would be positively correlated with tenure, leading to a positive sign for δot. The
results on Tables 6 and 7 are clearly supportive of this hypothesis.
The estimates of the coefficients in β and γ (the latter not shown, for the sake of brevity)
are standard. Tenure-earnings and experience-earnings profiles reveal the classic concave
shape.
It is also interesting to look at the 1985 to 1997 evolution of the estimates. As was
pointed out before, marginal returns to adequate schooling of both female and male workers
appear to have increased from 1985, prior to accession to the European Union, to the
1990’s. The premium for an extra year of unrequired schooling, though lower than one
could get at a job where that endowment were fully utilized, seems to be around 3 or 4%,
but the numbers suggest that it may be declining as the relative supply of better-educated
workers increases. The penalty associated with undereducation, on its turn, appears to have
gained in severity, rising from the 3 to 4% range of the mid 1980’s to 6 or 7% by 1997. The
coefficients estimated for ADTEN and OVTEN generally display an ascending trend,
which we take to reflect the increased importance attributed by employers to retaining and
rewarding better-educated workers.
5. Concluding remarks
In the presence of rigidities in the labour market, the possible existence of a severe
mismatch between skills required by available jobs and school-provided workers’
qualifications has recently drawn a considerable amount of attention. Undoubtedly, over-
and undereducation bring about significant costs for the individual, but social costs are also
of concern. In fact, previous evidence for other countries suggests that increased
educational qualifications may not automatically lead to an increase in skills and income,
giving rise to sizeable social waste.
In this paper, we addressed these issues, using Portuguese data for the period 1985-
1997. We should stress at the outset that use of the term overeducation in connection with a
country where average educational attainment is utterly poor, by comparison to trends
prevailing in Europe and in the United States, has the potential for being seriously
misleading. In fact, if we were to single out a most important result in our analyses it would
undoubtedly be the fact that returns to investment in formal education remain clearly
attractive, in spite of the increase in the average schooling of the Portuguese workforce over
the last decades. Returns appear to have increased from the mid 1980’s to the 1990’s and
compare favourably to what is standard in more advanced economies.
With that cautionary note in mind, we reviewed the evolution of over- and
undereducation measures over the period and found that there is strong evidence of a
persistent and growing disadjustment between school-provided skills and those required by
existing jobs in the Portuguese economy. Apparently, the rhythm at which the educational
system began to produce better-educated workers in the last 15 years outpaced
modernisation of the economic structure in the wake of the 1986 accession to the European
Union and sizeable pockets of overeducation (underutilization of skills) emerged.
Effects of over- and undereducation upon workers’ earnings were also evaluated.
Consistently with numerous studies in a variety of European and American countries, we
found adverse effects of misallocation of educational resources. Overeducated workers,
holding a job for which their qualifications exceed those that are standard among fellow co-
workers, would benefit from changing to a job where their levels of schooling were fully
utilised. On the other hand, undereducated workers (that is, those with educational
credentials below what is common in their occupation) would stand to gain if they were to
catch up with the level of schooling of fellow co-workers. Given that our figures for 1997
assert that one out of every two Portuguese workers is either over- or undereducated, by
comparison to people holding similar jobs, the productivity and income losses associated
with so large a mismatch may reach a significant extent.
It may be argued that the operation of technological progress, bringing about an
upgrading of required skills, would produce a picture similar to the one we observed. As
younger and more educated cohorts join the labour market and gradually permeate
occupations predominantly held by older and less educated people, our statistical
measurement tends to classify recent entrants as overeducated. In time, so the argument

runs, more educated workers become the majority in an occupation, overeducation declines
and eventually vanishes. And, indeed, our results are generally consistent with this
scenario. There is a downside to it, however. Our research suggests too that employers prize
overeducated workers above their (supposedly) adequately schooled counterparts, both in
current pay and in opportunities for future advance. In cruder words, a 30% figure for
overeducation may in fact be covering a 70% figure for undereducation. Either way, a
mismatch exists.
A more dynamic and modern economy would surely make room for a better use of the
educational skills of workers now deemed overeducated. But it is largely beyond the scope
of the paper the proposal of policy measures to overcome the shortcomings of labour
demand in Portugal. To the contrary, in what concerns the supply side our findings confirm,
if need be, the benefits of investment in education. Furthermore, they convey the advantage
of a massive effort to upgrade over a short span the market skills of a sizeable fraction of
Portuguese workers, both those now perceived to be undereducated and those who will
soon join their ranks as more ambitious educational achievements become the rule.
Perceived shortcomings of the preceding analysis strongly suggest that a panel study of
workers would be more adequate to settle some issues. Indeed, a panel study would better
control for unobserved individual heterogeneity and allow a firmer grasp on the effects of
technological progress and the increase in the relative supply of better-educated workers,
taking place simultaneously in the labour market. We plan to address these issues in future
work.
References
Alba-Ramírez, A. (1993). Mismatch in the Spanish labor market: Overeducation? Journal

of Human Resources 28, 259-278.
Burris, Val (1983). The social and political consequences of overeducation. American
Sociological Review 48, 454-467.
Cardoso, A.R. (1998). Earnings Inequality in Portugal: High and Rising? Review of Income
and Wealth 44, 325-343.
Cohn, E., Johnson, E. and Ng, Y.C. (2000). The incidence of overschooling and
underschooling and its effect on earnings in the United States and Hong Kong.
Research in Labor Economics 19, 29-61.
Cohn, E. and Khan, S.P. (1995). The wage effect of overschooling revisited. Labour
Economics 2, 67-76.
de Oliveira, M.M, Santos, M.C. and Kiker, B.F. (2000). The role of human capital and
technological change in overeducation. Economics of Education Review 19, 199-
206.
Duncan, G.J. and Hoffman, S. (1981). The incidence and wage effects of overeducation.
Economics of Education Review 1, 75-86.
Freeman, R.B. (1980). The facts about the declining economic value of college. Journal of
Human Resources 15, 124-142.
Groot, W. (1993). Overeducation and the returns to enterprise-related schooling. Economics
of Education Review 12, 299-309.
Groot, W. and Maassen van den Brink, H.M. (1997). Allocation and the returns to over-
education in the UK. Education Economics 5, 169-183.
Groot, W. and Maassen van den Brink, H.M. (2000). Overeducation in the labor market: a
meta-analysis. Economics of Education Review 19, 159-168.
Hartog, J. and Oosterbeek, H. (1988). Education, allocation and earnings in the
Netherlands: Overschooling? Economics of Education Review 7, 185-194.
Kiker, B.F. and Santos, M.C. (1991). Human Capital and Earnings in Portugal. Economics
of Education Review 10, 187-203.
Kiker, B.F., Santos, M.C. and de Oliveira, M.M. (1997). Overeducation and
undereducation: Evidence for Portugal. Economics of Education Review 16, 111-
125.
Machado, J.A.F. and Mata, J. (2001). Earning Functions in Portugal 1982-1994: Evidence
from Quantile Regressions. Empirical Economics 26, 115-134.
Ministério do Emprego e da Segurança Social (1994). Classificação Nacional das
Profissões, Versão 1994. Instituto do Emprego e Formação Profissional, Lisboa.
Ministério do Trabalho (1980). Classificação Nacional das Profissões, Versão 1980.
Secretaria de Estado do Emprego, Lisbon.
Rumberger, R.W. (1981). Overeducation in the USA Labor Market. Praeger, New York.
Rumberger, R.W. (1987). The impact of surplus schooling on productivity and earnings.
Journal of Human Resources 22, 24-50.
Sicherman, N. (1991). Overeducation in the labor market. Journal of Labor Economics 9,
101-122.
Thurow, L.C. (1975). Generating Inequality. Basic Books, New York.
Verdugo, R. and Verdugo, N.T. (1989). The impact of surplus schooling on earnings: Some
additional findings. Journal of Human Resources 24, 629-643.
Appendix Table A. Definition of Variables
Variable Description
ln Y Natural logarithm of total monthly earnings in escudos

EDUC Number of years of education completed
ADSCHOOL Number of years of adequate schooling
OVERSCH Number of years of overeducation
UNDERSCH Number of years of undereducation
TENURE Number of years of tenure in the current job
TENURE2 TENURE squared
EXPER Number of years of presumed work experience in firms other than the current one (age-
education-tenure -6)
EXPER2 EXPER squared
ADTEN Interaction term ADSCHOOL×TENURE
ADEXP Interaction term ADSCHOOL×EXPER
OVTEN Interaction term OVERSCH×TENURE
OVEXP Interaction term OVERSCH×EXPER
UNDTEN Interaction term UNDERSCH×TENURE
UNDEXP Interaction term UNDERSCH×EXPER
LNHOURS Logarithm of total monthly hours worked
PLANT4 Dummy variable, 1 if number of employees in the plant is ≤4
PLANT99 Dummy variable, 1 if number of employees in the plant is >4 and ≤99
PLANT499 Dummy variable, 1 if number of employees in the plant is >99 and ≤499
PLANTBIG Dummy variable, 1 if number of employees in the plant is ≥500
NORTH Dummy variable, 1 if job in the Northern region
CENTER Dummy variable, 1 if job in the Central region
LISBON Dummy variable, 1 if job in the Lisbon-and-Tagus-Valley region
ALENT Dummy variable, 1 if job in the Alentejo region
ALGAR Dummy variable, 1 if job in the Algarve region
TEXTILE Dummy variable, 1 if plant in the textile and garments industry
PRIMSECT Dummy variable, 1 if plant in the forestries, fisheries or mining industry
MANUF Dummy variable, 1 if plant in the manufacturing industry
UTIL Dummy variable, 1 if plant is a public utility
CONSTRU Dummy variable, 1 if plant in the construction and building industry
TRADE Dummy variable, 1 if plant in the wholesale or retail trade industry
RESTHOT Dummy variable, 1 if plant in the hotels and restaurants industry
TRANSP Dummy variable, 1 if plant in the transportation industry
FINANCE Dummy variable, 1 if plant in the finance, insurance and real estate industry
SERVICE Dummy variable, 1 if plant in the services industry
Appendix Table B. Sample Means of Variables
Variable Females Males

1985 1991 1997 1985 1991 1997
ln Y 10.181 11.025 11.440 10.449 11.323 11.694

ADSCHOOL 5.941 6.019 6.691 5.462 5.409 6.341
OVERSCH 0.604 0.867 1.230 0.807 1.044 1.239
UNDERSCH 0.796 0.724 0.675 0.699 0.582 0.684
TENURE 9.242 9.255 7.125 10.355 10.309 7.618
TENURE2 142.768 164.478 116.681 187.750 203.582 135.798
EXPER 12.805 14.165 13.984 16.382 17.009 16.434
EXPER2 257.158 301.371 301.162 384.762 405.934 394.734
ADTEN 54.485 58.346 49.021 59.792 58.040 51.397
ADEXP 68.885 78.126 83.567 82.168 86.060 95.367
OVTEN 4.413 5.148 4.937 6.295 8.315 5.388
OVEXP 4.369 7.676 11.434 7.918 11.015 13.220
UNDTEN 9.214 9.775 7.557 10.026 8.171 8.545
UNDEXP 12.442 11.668 11.104 12.962 10.936 12.637
LNHOURS 5.189 5.171 5.156 5.206 5.189 5.163
PLANT4 0.058 0.074 0.120 0.044 0.054 0.090
PLANT99 0.443 0.487 0.552 0.473 0.507 0.593
PLANT499 0.310 0.283 0.228 0.275 0.267 0.227
PLANTBIG 0.189 0.156 0.100 0.208 0.172 0.091
NORTH 0.422 0.351 0.408 0.352 0.348 0.392
CENTER 0.114 0.128 0.152 0.148 0.132 0.153
LISBON 0.418 0.477 0.372 0.441 0.469 0.385
ALENT 0.020 0.029 0.028 0.034 0.036 0.036
ALGAR 0.026 0.014 0.040 0.026 0.015 0.033
TEXTILE 0.314 0.301 0.220 0.082 0.093 0.060
PRIMSECT 0.002 0.001 0.003 0.012 0.010 0.015
MANUF 0.266 0.220 0.241 0.367 0.315 0.332
UTIL 0.006 0.005 0.001 0.021 0.018 0.001
CONSTRU 0.012 0.015 0.017 0.112 0.138 0.183
TRADE 0.162 0.157 0.191 0.158 0.152 0.179
RESTHOT 0.059 0.066 0.090 0.030 0.031 0.046
TRANSP 0.059 0.071 0.031 0.106 0.140 0.046
FINANCE 0.060 0.057 0.101 0.074 0.063 0.107
SERVICE 0.066 0.111 0.104 0.041 0.042 0.031
Sample size 25865 27183 34671 59008 50859 48941

Appendix Table C. Schooling Indicators for Selected Occupations,

1985 and 1991
1985 1991
Occup. No. of Mean Mode Ad. ed. Overed No. of Mean Mode Ad. ed. Overed
code obs. (yrs.) (yrs.) (%) (%) obs. (yrs.) (yrs.) (%) (%)
032 515 9,6 12 56,3 0,6 403 9,9 12 52,6 3,2
219 1364 11,2 12 37,6 29,0 1586 11,4 12 33,9 31,5
300 1309 9,3 12 43,5 4,4 1076 9,4 12 38,4 7,4
331 617 7,7 4 34,2 65,8 742 7,9 4 29,7 70,1
333 2641 9,6 12 41,5 2,2 1299 10,5 12 51,5 5,6
360 544 4,6 4 78,1 19,9 559 5,1 4 69,1 29,3
370 778 4,5 4 85,4 14,6 850 4,8 4 78,5 21,5
380 1028 7,7 4 29,0 70,8 957 8,5 12 31,5 0,6
391 2063 5,0 4 67,8 26,0 1897 5,3 4 61,2 35,3
393 7341 9,0 12 39,6 1,7 5870 9,1 12 40,8 1,8
399 635 5,5 4 49,8 45,0 606 5,3 4 56,9 37,5
432 1119 7,5 4 36,3 63,3 990 8,2 12 30,8 2,8
451 3860 5,5 4 63,7 34,9 3630 6,1 4 51,4 47,1
531 671 4,0 4 76,5 7,8 894 4,2 4 76,2 15,1
532 1745 4,3 4 77,4 14,9 1736 4,8 4 69,2 26,7
552 1653 3,8 4 63,0 10,6 1762 4,2 4 68,9 15,5
589 714 4,5 4 61,8 23,4 662 5,4 4 53,2 40,0
700 2774 5,2 4 67,6 25,0 3288 5,7 4 59,2 36,2
752 1602 3,9 4 74,7 7,5 1113 4,1 4 73,1 11,4
753 497 4,2 4 76,9 14,3 502 4,5 4 72,7 18,9
754 776 3,9 4 76,7 5,3 682 4,1 4 80,1 10,0
756 907 4,0 4 74,3 9,0 1047 4,4 4 72,1 19,5
759 784 4,0 4 73,3 11,5 712 4,3 4 70,7 19,5
776 519 4,0 4 81,7 7,9 405 4,4 4 77,8 17,3
795 3533 4,4 4 79,5 16,3 3880 4,8 4 64,9 32,6
802 1218 4,2 4 75,5 15,0 685 4,8 4 62,8 34,2
811 480 4,0 4 85,4 5,4 682 4,2 4 81,8 11,6
832 1685 5,1 4 68,0 27,7 1145 5,4 4 59,0 38,1
839 927 4,7 4 68,0 19,2 556 5,3 4 59,5 35,4
843 818 4,6 4 78,6 20,4 624 5,0 4 67,5 31,3
855 1262 6,1 4 53,7 44,1 1127 6,3 4 51,5 47,1
874 1563 4,5 4 74,4 18,2 1173 4,8 4 69,2 27,3
951 1770 3,6 4 70,1 4,8 1932 4,0 4 75,4 10,6
954 773 3,8 4 76,7 6,3 718 4,3 4 76,9 14,1
971 1194 4,2 4 68,8 15,7 854 4,6 4 68,7 23,7
974 447 3,7 4 69,6 7,4 547 4,1 4 74,2 12,8
976 624 3,8 4 68,6 9,5 430 4,5 4 69,1 20,7
985 3062 4,2 4 91,7 6,8 3468 4,5 4 83,2 15,7
999 7776 4,6 4 58,5 22,0 7661 5,5 4 53,9 35,7
Notes: See Ministério do Trabalho (1980) for description of the occupation codes; Appendix Table E provides the
Portuguese title for the selected occupations. The numbers under the heading "Ad. ed." refer to the percentage of
workers with the modal value of schooling; "Overed" refers to the percentage of workers with schooling above the
modal value.
Appendix Table D. Schooling Indicators for Selected Occupations, 1997
1997
Number of Mean Mode Ad. ed. Overed
Occupation code
observations (yrs.) (yrs.) (%) (%)
1231 550 12,2 16 33,3 0,0
3119 528 9,4 9 28,8 40,3
3415 2271 9,5 12 36,3 9,6
3431 1422 10,5 12 37,8 15,8
4122 7510 10,1 12 46,0 6,1
4131 1393 6,6 4 39,2 58,9
4132 892 6,8 4 38,1 59,5
4142 836 7,8 9 34,5 21,3
5122 1137 4,9 4 64,5 30,6
5123 2780 6,2 4 43,8 54,3
5220 5034 7,2 4 30,9 68,4
7122 2362 4,5 4 71,3 22,3
7123 944 5,1 4 64,7 30,7
7124 978 4,8 4 65,0 31,6
7137 583 6,1 4 40,0 58,5
7214 1388 5,3 4 55,0 43,3
7222 724 5,6 4 51,8 46,4
7223 672 5,7 4 50,3 48,1
7241 1206 7,0 6 34,3 37,7
7411 813 5,0 4 58,2 35,2
7412 706 5,0 4 58,9 36,7
7422 704 4,8 4 60,7 34,0
7436 4630 5,2 4 51,0 47,1
7442 2125 5,0 4 53,6 43,7
8261 525 4,4 4 74,1 18,3
8262 866 4,9 4 61,3 34,8
8264 892 4,9 4 61,7 34,9
8281 511 7,9 9 33,9 23,1
8282 746 7,7 6 35,7 45,8
8322 606 5,4 4 59,9 40,1
8324 1436 5,0 4 66,8 33,2
9132 2548 4,8 4 64,3 29,1
9151 1099 6,2 4 43,9 53,6
9152 1286 6,6 4 39,7 57,4
9312 1943 4,8 4 60,0 30,8
9322 3737 5,5 4 49,4 45,8
9333 525 5,9 4 43,6 51,1
Notes: See Ministério do Emprego e da Segurança Social (1994) for description of the occupation codes;
Appendix Table F provides the Portuguese title for the selected occupations. The numbers under the heading "Ad.
ed." refer to the percentage of workers with the modal value of schooling; "Overed" refers to the percentage of
workers with schooling above the modal value.
Appendix Table E. Portuguese Title of Selected Occupation Codes, 1985 and 1991
Occup. code Title
032 Desenhadores
219 Directores e quadros dirigentes não classificados em outra parte
300 Chefes de secções administrativas
331 Empregados de contabilidade, caixas e trabalhadores similares
333 Empregados de transacções financeiras e operações de seguros
360 Condutores de comboios, cobradores, revisores de bilhetes e trabalhadores similares -
transportes
370 Carteiros, estafetas e trabalhadores similares
380 Telefonistas, telegrafistas, operadores de rádio e trabalhadores similares - excepto
radiodifusão e radiotelevisão
391 Empregados de aprovisionamento e armazém
393 Empregados de serviços administrativos
399 Pessoal administrativo e trabalhadores similares não classificados em outra parte
432 Representantes comerciais e caixeiros-viajantes
451 Vendedores e caixeiros
531 Cozinheiros
532 Empregados de mesa e trabalhadores similares
552 Pessoal de limpeza e trabalhadores similares
589 Pessoal dos serviços de protecção e segurança não classificados em outra parte
700 Encarregados e trabalhadores similares
752 Fiandeiros, bobinadores de fios e trabalhadores similares
753 Afinadores e preparadores de teares
754 Tecelões e trabalhadores similares
756 Branqueadores, tintureiros e acabadores de produtos têxteis
759 Trabalhadores têxteis e trabalhadores similares não classificados em outra parte
776 Padeiros, pasteleiros, confeiteiros e preparadores de massas alimentícias
795 Costureiras e bordadoras
802 Trabalhadores do fabrico de calçado
811 Marceneiros e trabalhadores similares
832 Serralheiros mecânicos, operadores de máquinas-ferramentas e traçadores
839 Forjadores, serralheiros mecânicos, operadores de máquinas-ferramentas e trabalhadores
similares não classificados em outra parte
843 Mecânicos de veículos a motor
855 Electricistas
874 Serralheiros civis, montadores de estruturas metálicas e trabalhadores similares
951 Pedreiros, ladrilhadores e trabalhadores similares
954 Carpinteiros de construção civil, naval e outras e trabalhadores similares
971 Embaladores
974 Condutores de máquinas de escavação, terraplanagem, construção civil e trabalhadores
similares
976 Bagageiros e outros carregadores e descarregadores
985 Condutores de veículos a motor
999 Trabalhadores indiferenciados não classificados em outra parte
Source: Ministério do Trabalho (1980).

Appendix Table F. Portuguese Title of Selected Occupation Codes, 1997
Occup. code Title
1231 Directores de serviços administrativos e financeiros

3119 Técnicos de investigação física e química, do fabrico industrial e trabalhadores similares,
não classificados em outra parte
3415 Representantes comerciais e técnicos de vendas
3431 Profissionais de nível intermédio dos serviços administrativos
4122 Empregados administrativos dos serviços financeiros e trabalhadores similares
4131 Empregados de aprovisionamento e armazém
4132 Empregados do planeamento e apoio à produção
4142 Carteiros e trabalhadores similares
5122 Cozinheiros e trabalhadores similares
5123 Empregados de mesa e trabalhadores similares
5220 Vendedores e demonstradores
7122 Pedreiros e calceteiros
7123 Trabalhadores da construção civil e obras públicas - betão armado
7124 Carpinteiros
7137 Electricistas da construção civil e trabalhadores similares
7214 Montadores de estruturas metálicas e trabalhadores similares
7222 Serralheiros mecânicos e trabalhadores similares
7223 Afinadores - operadores de máquinas ferramentas
7241 Electromecânicos e electricistas
7411 Magarefes, cortadores de carnes e trabalhadores similares da preparação de carnes e
peixes
7412 Padeiros, pasteleiros e confeiteiros
7422 Marceneiros, carpinteiros e trabalhadores similares
7436 Costureiras, bordadores e trabalhadores similares
7442 Sapateiros, trabalhadores de calçado e do couro
8261 Operadores de máquinas de fiar, torcer e bobinar
8262 Afinadores, preparadores e operadores de teares (tecelões)
8264 Operadores de máquinas de tratamento de produtos têxteis
8281 Montadores de construções mecânicas
8282 Montadores de aparelhagem eléctrica e electrónica
8322 Condutores de veículos ligeiros
8324 Condutores de veículos pesados de mercadorias
9132 Pessoal de limpeza de escritórios, hotéis e trabalhadores similares
9151 Estafetas, distribuidores, bagageiros e trabalhadores similares
9152 Porteiros, guardas e trabalhadores similares
9312 Serventes da construção civil e obras públicas, porta miras e trabalhadores similares
9322 Trabalhadores não qualificados da indústria transformadora
9333 Carregadores e descarregadores de mercadorias
Source: Ministério do Emprego e da Segurança Social (1994).

Sessão 4 - Instituições, Reformas e Crescimento
Session 4 - Institutions, Reforms and Growth

“FIRMS, FINANCIAL MARKETS AND THE LAW:
INSTITUTIONS AND ECONOMIC GROWTH IN PORTUGAL”*
José Albuquerque Tavares

Faculdade de Economia
Universidade Nova de Lisboa
July 2002
Abstract: This paper presents a broad diagnostic of the current level of development
of Portuguese institutions focusing on the legal, corporate governance and financial
systems. We provide a broad review of the determinants of effective institutions and of
their impact on economic growth for each of the above institutional areas. The
comparative assessment suggests that Portuguese institutions are generally less
developed than their European Union and East Asian counterparts but, more
developed than Greek institutions and on the same level as Spanish institutions. Using
data for a broad cross-section of countries since 1960, we estimate the empirical
relationship between institutional indicators and long-term economic growth and
identify the areas where reform may have a statistically and economically significant
effect on growth. The results, together with the assessment of relative institutional
development, allow us to construct three new indicators of the potential of
institutional reform: the impact of reform on growth, the required reform effort and
the efficiency of reform indices. These indices measure, respectively, which reforms
have the most payoff in terms of growth, which are “less costly” to undertake and
which deliver the most growth per required effort, all relative to a reform that would
bring Portuguese institutions to EU average levels of institutional development. The
analysis strongly suggests that comprehensive institutional reform will translate into
higher rates of economic growth in Portugal. In addition, out of the ten most
promising reforms, six are in the legal area, irrespective of which of the three indices
is considered. In addition, reform innthe legal area is promising both at the aggregate
and the microeconomic level. In the financial sector aggregate indicators offer the
wider scope for productive reform, while in the corporate governance area indices at
the firm level hold the most promise. Portugal has recently experienced a slowdown
in economic growth resulting in slower convergence vis-à-vis richer countries in the
European Union. Our results support the view that a comprehensive and sustained
reform effort is necessary for the Portuguese economy to grow at higher rates,
allowing faster real convergence with the richest countries in the European Union.
* This paper was prepared for the conference “Desenvolvimento Económico Português no Espaço Europeu:
Determinantes e Políticas”, organized by the Banco de Portugal. Comments from an anonymous referee are
sincerely appreciated. The usual disclaimer applies. Contact address: Faculdade de Economia, Universidade
Nova de Lisboa, Campus de Campolide, 1099-032 Lisboa, Portugal. E-mail: jtavares@fe.unl.pt
328 José Albuquerque Tavares
1. Introduction
“We cannot see, feel, touch, or even measure institutions; they are constructs of the human
mind.”
Douglass North
(1990)
In the last couple of decades the importance of institutions for economic growth has
received increasing attention by academics and policy-makers. The cliché that “institutions
matter” is likely to be among the set of priors held by most economists. Agreeing what
institutions are, why and how they matter, and what they matter for raises considerably
more disagreement. In this paper we provide a characterization of institutions based on the
economic roles they perform. Our focus is on three different areas of institutional
development: the legal system, the governance of the firm and the financial system. The
internal organization of the firm, its access to financial resources and the legal framework
within which it operates all affect the growth rate of an economy. After reviewing the
arguments and evidence linking each of the three policy areas to economic growth, we
conduct econometric tests to assess their impact on economic growth, as well as the costs
and benefits of specific institutional reforms.
Our analysis uses a broad cross section of countries and a wide data set to derive policy
implications for the Portuguese case. We characterize Portuguese institutional development
in the legal, financial and firm governance areas and compare Portugal to Spain, Greece, its
European Union partners and East Asian competitors. The assessment of relative
institutional development in Portugal is then crossed with estimates of the impact of each
institution in economic growth in order to compute the benefits of bringing Portuguese
institutions up to par with institutions in the European Union. We calculate three new
indices of institutional reform: the impact of reform on growth, the reform effort required to
raise Portuguese institutions to average European quality levels and the efficiency of
reform- the impact on growth per “unit” of effort required. We then identify specific
institutional reforms that are likely to have a substantial impact on Portuguese growth with
relatively low effort.
2. What Are Institutions and What Are They For?
In the economic analysis of institutions, it is natural to start from the work of Douglass
North, who defines institutions as “the humanly devised constraints that (…) structure
incentives in human exchange, whether political, social or economic” (North (1990, p. 3)).
The idea of institutions as constraints that, nonetheless, facilitate exchange is important.
The seminal work in Coase (1960) had shown that in a zero transaction cost world with
costless bargaining and contracting, competitive markets would reach the solution
maximizing aggregate income. In such a world institutions would be entirely redundant.
However, as interactions become mostly one shot events (not repeated), transactions occur
over an extended period of time and involve many players, and information is incomplete
or asymmetric, exchange and cooperation become increasingly difficult.1 In other words,
1
See, for instance, North (1990, p. 12).
Firms, Financial Markets and the Law: Institutions and Economic Growth in Portugal 329
transaction costs arise in the presence of problems of search and information, bargaining
and contracting, or policing and enforcing of contracts. In a context of transaction costs,
institutions do matter.
But what do institutions do? What roles do they perform that affect economic exchange
and growth? In spite of the multiple - often concurring - roles ascribed to institutions, a
short list of their functions should include the following:
A. Creating and Enforcing Rules – Institutions restrict the set of choices available to
individuals and may improve social outcomes when individualistic behavior does not
automatically lead to the most efficient outcome. Examples include situations characterized
by multiple equilibria, public goods, externalities or prisoner’s dilemma type of incentives.
A rule is relevant when enforceable and institutions should include mechanisms to enforce
rules, as private individuals may have no incentive to do so. Creating rules, determining if
and when they are violated and what to do when they are is an important role of
institutions.
B. Aggregating Information and Preferences – When goods and services have complex
characteristics, the differences in individual preferences cannot be overcome by simple
market mechanisms. An example is the case of public goods and externalities, where
private and social benefits are not aligned and individuals may even lack incentives to
reveal their true preferences. In these cases institutions may be set up with the objective of
aggregating information on preferences and delivering a collective choice. Institutions such
as a national parliament and the general assembly of a firm perform such a role.
C. Sharing Risks and Reducing Uncertainty – Institutions, as collective endeavors, can
be valuable as risk-sharing mechanisms. In general terms, institutions provide a context to
everyday transactions that reduces uncertainty, augmenting the set of fruitful exchanges
available. Particular institutions often have the effect of reducing specific risks.
D. Optimizing Resource Utilization – By pooling individual resources and capabilities,
institutions can increase total output in ways that individuals acting in isolation cannot.
Insitutions that take advantage of economies of scale are an example of such a benefit.
E. Redistributing Resources – Institutions may be created to alter the distribution of
resources away from that which would result from unfettered market interactions. A
redistributive institution can arise if society believes that certain groups – e.g. retired people
– or certain conditions – e.g. unemployed – merit special consideration and the related
market outcome is unsatisfying.
Once institutions are defined by the above functions, one realizes how they can combine
formal and informal features. In their formal incarnation, institutions are a set of explicit,
detailed, and coded rules applied within predefined limits. On the other extreme, informal
institutions are no more than an evolving pattern of social conventions and behavior that are
followed in a given society.
It is important to distinguish organizations from institutions. Organizations have a
clearly defined purpose that forwards the interests of their limited set of members.
Institutions have broader objectives as intermediate bodies facilitating the attainment of
diverse individual and group objectives. As an example, a given corporation is an
organization, while corporate law is an institution whose existence furthers the interests of
corporations, governments and individuals.
As institutions evolve through time they tend to do so slowly and incrementally rather
than through discrete changes. The inertia involved in institutional change has been
expressed by terms such as “history matters” or “path dependence.” However, inertia can
be an integral part of the role that institutions perform; if institutions changed constantly in
response to temporary economic and social circumstances, they would be unable to perform
most of their functions adequately.
3. Institutions and Economic Growth
Institutions matter for economic growth. A first approach to understand it is to depart

from neoclassical growth accounting, which sees production as resulting from the
combination of different inputs – say land, labor and capital – to produce an output. If
transaction costs - the cost involved in combining inputs in the most efficient way allowed
for by the production possibilities - are added to the direct costs of the inputs, institutions
become important insofar as they affect transaction costs and closeness to the production
possibilities frontier.2
A second approach sees the role of institutions in economic growth through the
encouragement of specialization. As North (1990) puts it, “non-specialization is a form of
insurance when the costs and uncertainties of transactions are high. The greater the
specialization and the number and variability of attributes, the more weight has to be put on
reliable institutions (…)”. The importance of institutions is evident in Adam Smith´s
understanding of specialization of production as the key mechanism in economic
development. By broadening the set of possible exchanges, institutions increase the
effective size of the market and allow further specialization.
3.1. The Legal System and The Law
The relationship between the law, the judiciary and economic growth has been the
subject of political economy from early on. Adam Smith (1755, p. 322) highlighted that “a
tolerable administration of justice,” in conjunction with peace and low taxes, could bring a
state “to the highest degree of opulence.”3 In his studies on society and economics, Max
Weber also examined the relation between the law, the judiciary and economic
development. There are two broad reasons why the judicial system may affect economic
performance. First, it enforces property rights, in particular keeps the government in check
by avoiding abuses of governmental power. Second, the judiciary facilitates exchanges
between private parties.4
The protection of property rights is generally accepted as the minimum necessary level
of government intervention. Coase (1960) discusses the importance of the law for economic
transactions. Income maximization can be independent of the legal system if the price
system is not subject to transaction costs. In this case individuals buy and sell rights as
2
See North (1990, p.28). Wallis and North measure the importance of transaction costs by summing the value of
banking, insurance and commerce activities as well as the income of lawyers, accountants, managers, etc. They
found that about 45 percent of total U.S. output was devoted to these activities, up from 25 percent a century
earlier. In a certain sense, the rise in the importance of services reflects the efforts to reduce transaction costs.
3
In the tradition of Adam Smith, North (1981) refers the emergence of a system of enforceable property rights as
the important institutional prerequisite for economic growth.
4
A 17th century political philosopher presented this argument by stating that in general contracts “he that
performeth first has no assurance the other will perform after because the bonds of words are too weak to bridle
men’s ambitions, avarice, anger, and other passions without the fear of some coercive power” As quoted in
Messick (1999).
needed.5 Notice that a legal system (potentially any legal system) needs to be in place and
duly enforced so that rights can be exchanged just as any other goods in the market. In
reality there are important transaction costs and the specifics of the legal system affects the
economy. Coase (1988, p. 28) restates the primacy of the law: “economic policy involves
choice among alternative social institutions, and these are created by the law or are
dependent on it.”
In the absence of clear property and contract laws, contracts tend to be simple,
involve fewer parties and be closed in “spot” markets as a way to reduce the scope for
breach of contract. Williamson (1995) noted the prevalence of long-term contracts in
developed economies.6 It is true that the extra gains from repeated interactions can be
sufficient inducement to honor a contract, as the discounted present value of earnings to be
realized in the future may exceed the one-time gain from breaching the agreement. In other
cases the benefits of acquiring and maintaining a good reputation provide the necessary
incentive.7 However, as Posner (1998) ably puts, the key to prosperity is the legal
infrastructure, centered on the protection of property and contract rights:
“In its ideal form (an important qualification), the machinery consists of
competent, ethical, and well-paid professional judges who administer rules that are
well designed for the promotion of commercial activity. The judges are insulated
from interference by the legislative and executive branches of government. (…They
are advised by competent, ethical, and well-paid lawyers…) Their decrees are
dependably enforced by (…) police or other functionaries (again, competent,
ethical, and well paid). The judges are numerous enough to decide cases without
interminable delay, and they operate against a background of rules and practices,
such as accounting standards, bureaus of vital statistics, and public registries of
land titles and security interests, that enable them to resolve factual issues relating
to legal disputes with reasonable accuracy and at reasonable cost to the
disputants.” (Posner (1998, p. 2-3))
The enforcement of property rights is an important element of a legal system. With

diffuse and non-enforceable rights production tends to rely as little as possible on physical
capital and transactions are limited both in time and as to the number of actors.
Common Law versus Civil Law
Most of the actual legal systems in place in the world trace their origins to the British or
the French system, respectively the common law and the civil/Roman law traditions.8
5
The University of Chicago approach to law and economics, of which Coase is a major exponent, emphasizes
market and non-market behavior reactions to (sometimes implicit) prices influenced by the workings of the
law.
6
In a parallel argument, Williamson (1995) suggests that the quality of a judicial system is indirectly assessed by
the complexity of the economic transactions it is able to support.
7
Similar principles are behind mechanisms that have gained force in developed societies, such as consumer
testing laboratories and better business bureaus, which “sell” their reputation through seals of approval or
quality guarantees.
8
Two other important legal systems are the German and Scandinavian, also derived from Roman law. The civil
law traditions in Germany and Scandinavia differ from the French system: while the content of legal rules is
While the French system relies heavily on legal codes, written rules and professional
judges, the British system relies on broader principles, oral arguments, rule of precedent
and lay judges. Why these systems evolved in different directions is beyond the scope of
this paper. But it is clear that the different characteristics of the systems are coherent. For
instance, codification is a natural response mechanism that helps control judges appointed
by a central state.
Often, as is the case in former European colonies, legal systems are involuntarily
acquired but legal systems may have been the object of conscious choice in England and
France. Mahoney (2000) suggests that the English common law developed as landed
interests and merchants wanted protection for property and contract rights that were strong
enough in themselves to limit the crown’s ability to interfere in markets. On the other hand,
French civil law may have originated in the desire of the French revolution generation and
of Napoleon thereafter to avoid that judges easily change the thrust of government
policies.9 The ideologies underlying common law and civil law systems differ markedly
and civil law systems are more comfortable with centralized and activist governments.10
There are many reasons why the origin and characteristics of the law may matter.
The quality of the legal rules seems to vary systematically across legal traditions.11 The
process of litigation, prevalent in the common law system is an example of a procedure that
may lead to the survival of the more efficient rules, as argued in Posner (1972). In other
words, judge-made law may be more efficient than statutory (coded) law.12 A second
possibility is that, even if the average quality of the rules is similar, their application differs
so that common law allows for greater predictability, thus reducing uncertainty.13
According to Mahoney (2000) there are two features that raise the predictability of
outcomes in common law systems: the respect for the precedent and the important powers
of appellate courts.14 It is extremely difficult to make an argument for or against a particular
legal system, given their complexities and inter-relatedness with cultural and political
similar, the political association is not. The independence of judges is upheld by Germany’s constitution, which
does not allow their involuntary reassignment. Still, German judges, just like the French, lack the prestige of
their common law countries counterparts.
9
Glaeser and Shleifer (2001) review the historical origins of the two legal systems and model their functioning
and consequences. One key problem of legal systems is the enforcement of the law and the protection of law
enforcers vis-à-vis powerful local interests. Coercion and corruption by local interests must be kept in check in
order for private property to be protected. In this case it is more efficient to leave dispute resolution to well-
informed local decision makers, such as juries; if local interests are very strong it may make sense to use less
informed but more insulated judges nominated by a central authority. The authors above suggest that France
adopted professional judges controlled by the King instead of the independent juries in England because local
interests (landlords) were much more powerful in France and thus benefited from a central (and relatively weak)
authority as the legal arbiter.
10
See Merryman (1985).
11
For example, La Porta et al. (1997, 1998) provide evidence that common law countries have superior systems of
corporate law on average. At a more general level, common law countries seem to have more efficient
governments, as shown by La Porta et al. (1999) after controlling for other determinants such as income per
capita, religion and ethnolinguistic fractionalization.
12
Tullock (1997) has put forth a strong refutation of this notion, arguing instead for the benefits of codification.
13
The 1997 World Bank Development Report presentes results from a survey of 3,600 firms in 69 countries where
around 70 percent of the interviewed mentioned an unpredictable judiciary as a major problem for their business
operations. In this survey, public confidence in government institutions - including the judicial system - was
highly correlated with the level of private investment and economic performance in general.
14
Nominally, these features are not present in civil law, where the code itself takes precedence as binding law, but
courts in civil law countries do in practice take into account precedence and the decisions of higher courts. As
noted in Merryman (1985).
traditions. In general, both legal traditions perform well on the main issue of the protection
of fundamental property and contract rights so that there are prosperous countries espousing
both legal tradition.
In Table I and Figure I below we present summary statistics for legal performance
indicators in common and civil law countries. These general indicators measure the extent
of the rule of law, the efficiency of the judicial system the level of corruption, as well as the
risks of contract repudiation and of expropriation. We report the average for countries
organized within families of legal systems. A higher value for an indicator suggests better
performance. Countries with legal systems of English origin perform better than countries
with French legal origin in all indicators. While Scandinavian and German legal origins are
associated with the best performances overall but, as these legal systems were not as widely
disseminated through colonization, there is a bias toward the inclusion of developed
countries (that would perform better anyway).
Table I
Performance of the Legal System in Different Legal Traditions
Efficiency Risk of
Risk of
Rule of Law of Judicial Corruption Contract
Expropriation
System Repudiation
English 6.46 8.15 7.06 7.41 7.91

French 6.05 6.56 5.84 6.84 7.46
German 8.68 8.54 8.03 9.47 9.45
Scandinavian 10 10 10 9.44 9.66
Whole Sample 6.85 7.67 6.9 7.58 8.05
Note: From La Porta, Lopez-de-Silanes, Shleifer and Vishny, (1998). European Union excludes Luxembourg.
East Asian Tigers refers to Hong Kong, Singapore, Taiwan and South Korea. A higher value corresponds to
better performance.
Figure I
Measures of Performance in Different Legal Traditions
Risk of Expropriation
Risk of Contract Repudiation
Corruption
Efficiency of Judicial System
Rule of Law
0 2 4 6 8 10 12
English French German Scandinavian
Djankov et al. (2001) assesses how dispute resolution is handled in practice in 109
different countries. The authors concentrate on the procedure to evict a tenant for non-
payment of rent and the procedure to collect a bounced check. Both are common economic
transactions in which substantial fractions of the population are potentially involved so they
are very informative as to the workings of the legal system.15 The authors find that in civil
law countries dispute resolution is associated with heavier regulation, longer duration of
legal proceedings, poorer access to justice and lower judicial efficiency. This represents a
greater deviation from an informal neighbor resolution ideal, involving voluntary and lower
cost resolution of conflicts through a friendly third part.
Restraining the State
The state is an economic player that enjoys a special power. The ability to restrain that
power so that it does not harm economic well-being is one of the main tasks of a legal
system. Weingast (1993) has posed this issue in the clearest terms: ”a government strong
enough to protect property rights is also strong enough to confiscate the wealth of its
citizens.” The legal and political systems need to extract credible commitments from the
state so it won’t arbitrarily alter property and contract rights. One such mechanism of
commitment is the separation of powers among the legislative, executive and judiciary
branches of government, such as the “checks and balances” of the US political system.16
The formal separation between the judicial system and other branches of government is
greater in common law countries. Actually, the vision of the judicial system as a branch of
15
Moreover, they involve lower level courts, whose functioning is extremely relevant to the mass of a country’s
citizens and thus to economic development.
16
Persson et al. (1999) model how different regimes affect public finance outcomes. They show that judicial
independence can reinforce the separation of powers and make fiscal redistribution costlier so that separation of
powers becomes associated with lower taxation and redistribution.
government in itself is very present in the Anglo-Saxon political tradition. Typically, judges
are appointed to a post following successful careers as lawyers, positions are prestigious
and well compensated. The role of the judiciary in common law countries is also
highlighted by the mobility between this branch and the executive, as judges often pursue
political careers.17 In civil law countries, becoming a judge is a separate career, embarked
on immediately after leaving law school and passing general examinations. Prestige is
associated with promotions and transfer to more desirable locations, which often depend on
the favors of the executive.18
The actual administration and enforcement of rules does not fall on the courts and the
judicial system but on the state’s administrative agencies. The independence of these state
agencies has to be weighed against their potentially negative role in rent extraction.19 The
law often enhances the independence of the administrative bodies, when the statutes of the
civil service explicitly protect administrative personnel against termination by the executive
on the basis of policy disagreements.
Shaping the Corporation
As will be discussed below, a country´s legal origin correlates strongly with creditor
and minority shareholder rights that are in turn associated with financial market
development. However, corporate law remains an unlikely place to uncover systematic
differences between legal systems: relative to other areas of commercial law such as
contracts, corporate law has been relatively more based on codes from early on, and this is
true even in common law countries. 20
Levine and co-authors have recently presented evidence that countries with legal
systems that enforce contracts effectively display more developed financial
intermediaries. Instrumenting for financial intermediation with legal system characteristics
shows that the component of financial intermediary development influenced by the legal
and regulatory environment is positively associated with economic growth. The evidence
also suggests that a better legal environment results in capital markets that are broader and
more valued and that French civil law countries tend to have the least developed capital
markets.21
17
Messick (1999) points out the United States may differ from other countries as judges play a significant role in
policymaking and lawyers can often engineer significant policy changes through litigation.
18
Mahoney (2000) presents evidence that the executive in civil law countries takes advantage of this dependence
relationship.
19
Motivated by private gain, including future employment in the private sector. The theory of “capture” of
regulatory agencies by private interests was proposed by Stigler (1971). Olson (1982) elaborates on the
possibility of governments responding to the demands of rent-seeking groups by modifying property or contract
rights.
20
As pointed by Mahoney (2000).
21
See La Porta et al (1998).
Formal Laws and Informal Rules
An explicit legal framework may not be essential for growth. First, it may be too
expensive for developing countries (though inexpensive in relative terms). Second, the
effects of a given legal framework are only as good as the laws themselves.22 Lastly, there
are good informal substitutes for protection and enforcement of contractual rights.23
Contracts signed under the guidance of formal rules typically have multiple dimensions but
cannot foresee all possibilities. They are typically incomplete and delegate to a third party
such as a court the resolution of disputes arising during the life of the contract. Informal
customs, rules and mores can complement formal rules. However, economic development
is almost always accompanied by a move from informal unwritten rules to formal written
rules.24
Among the informal substitutes for formal legislation and enforcement are arbitration,
reputation, merger - which internalizes disputes between independent firms -, bilateral
monopoly - a substitute for employment contracts enforced by law -, strong-arm tactics -
used in illegal markets - or altruism - enabling family-owned firms to operate effectively
outside the legal framework.25 These informal substitutes can be extremely important, even
in developed countries. One should recall that property rights and contract enforcement
were present before a formal state and an effective government came into existence.
These substitutes for formal law may be expensive. Enforcement of rights may
depend on the threat and the occurrence of violence, with the associated costs; family
alliances become dysfunctional in a complex economy; and when reputation constitutes the
basis of exchange, new firms without family or ethnic connections find it difficult to enter
the contracting network.26 These costs are important even if not apparent: new innovative
firms remain outside of established alliances; simple, one-time transactions, common in
modern economies, become less favored relative to complex transactions that ensure
informal enforceability.
Containing Crime
Several reasons suggest why crime may harm economic growth.27 Crime and
specially generalized violence deplete the physical capital stock and discourage the
accumulation of any kind of capital, including human capital. Crime diverts public
resources from other, potentially more productive uses. Studies that have computed the
22
As Posner (1998) puts it, the “legal system does more than enforce contract and property rights; it may also
enforce bad laws that reduce economic efficiency”.
23
A common example is that of commercial law. Historically, it originated in the customs of merchants enforced
privately by arbitration; only later was it adopted by the courts and codified.
24
The division of labor between formal and informal mechanisms is affected by the technical complexity involved
as more traders move in and it becomes harder to check their reputation. However, economic development may
lower some of the costs involved in operating an informal enforcement mechanism. The use of faxes,
computers, and other technologies, for example, reduces the costs of compiling and disseminating information
and thus favors the reemergence of informal mechanisms.
25
See Posner (1998).
26
Greif (1996) compares "collectivist" and “individualist” societies, where the former rely on personal reputation
and trust in contracting and the latter on formal legal mechanisms and reputation. As this author mentions "most
collectivist societies are low-income economies and most individualist societies are high-income economies…".
27
See Davis and Trebilcock (1999).
direct and indirect costs of crime generally find them to be substantial. Bourguignon (1999)
estimates that the aggregate cost of crime in Latin America is 7.5 per cent of GDP. In
addition to direct measurable economic costs, crime and the fear of crime impact the quality
of life for given income per capita.
What is the role for legal institutions in reducing crime? In virtually all countries all
deliberate acts of violence or harm to property are subject to penal sanctions so that the
good functioning of the criminal justice system can deter crime. A recent study by
Fanjzylber et al. (1998) analyzes the relation between the amount of policing and the rates
of homicide and robbery for the period 1970–1994 based on information from the United
Nations World Crime Survey. Among their findings is that, while economic downturns and
income inequality raise violent crime rates, the strength of the police and of the judicial
system can be significant deterrents of violent crime.
The Characteristics of Legal Systems: Portugal in Context
Tables II through IV present indicators of legal performance for Portugal as compared

to other economies. We have chosen to compare Portugal with Spain and Greece, southern
European countries that also joined the European Union in the mid-1980´s and have
sustained parallel experiences of political authoritarianism and delayed institutional
development. In addition, we report average values for institutional indices in the European
Union and the four East Asian fast-growing economies - Hong Kong, South Korea,
Singapore and Taiwan. The European Union and East Asian high-growth economies are
relevant benchmarks as these are likely to be Portugal´s strongest competitors in world
markets. As far as general indicators of legal performance are concerned, Table II shows
that Portugal compares well with Spain and Greece, performing as well or better than both
countries in Rule of Law, Corruption, Risk of Contract Repudiation and Citizen’s Access to
Justice and better than Greece in Risk of Expropriation.28 However, Portugal compares
poorly with East Asian countries and is well below the average European Union levels of
performance in all indicators. The difference is most acute in the indices of Efficiency of
the Judicial System, Corruption and Contract Enforcement.
Table II
Performance of the Legal System
Efficiency Risk of Citizen's

Rule of Risk of Contract
of Judicial Corruption Contract Access to
Law Expropriation Enforcement
System Repudiation Justice
Portugal 8.68 5.5 7.38 8.9 8.57 7.5 4.54

Spain 7.8 6.25 7.38 9.52 8.4 5 6.23
Greece 6.18 7 7.27 7.12 6.62 5 5.81
European Union 8.97 8.63 8.74 9.44 9.07 7.86 7.25
East Asian Tigers 7.67 8.19 7.22 8.76 8.86 4.38 6.49
Note: The variables Efficiency of Judicial System up to Risk of Contract Repudiation are from La Porta, Lopez-
de-Silanes, Shleifer and Vishny, (1998). European Union excludes Luxembourg. East Asian Tigers refers to Hong
28
The source, unit and description of all variables are presented in the Data Appendix.
Kong, Singapore, Taiwan and South Korea. Data on Citizen’s Access to Justice and Enforcement is from Djankov
et al. (2001). Higher values correspond to better performance.
In Table III we turn to a more specific measure of legal performance, the duration of
legal procedures.29 The data is quite revealing: Portugal has the highest duration for legal
procedures by far, well above Spanish, Greek, European and East Asian levels. The time
for completion of a legal procedure is 1.5 times that of the EU for the eviction of a tenant
and almost twice the EU level for collection of a bounced check. The data reveals the origin
of the problem more finely: it is the time spent in court that is well above international
standards, with Portugal at about five times Spanish levels; the time spent on other phases
of the legal process - service of the process and enforcement – is actually at or below
international duration levels.
Table III
Duration of Legal Procedures for Tenant Eviction and Bounced Checks
TENANT EVICTION
Service of Process Trial Enforcement Total Duration
Portugal 20 280 30 330

Spain 60 55 68 183
Greece 32 35 180 247
European Union 18.0 143.4 91.1 252.5
East Asian Tigers 19.0 94.8 108.4 221.1
BOUNCED CHECK
Service of Process Trial Enforcement Total Duration
Portugal 20 280 120 420

Spain 49 69 29 147
Greece 180 45 90 315
European Union 28.1 118.0 81.3 227.3
East Asian Tigers 16.9 39.4 41.9 98.1
Note: Data is from Djankov et al. (2001). The table presents estimates of the expected duration of each type of
dispute, from the original filing of a complaint to the ultimate enforcement of judgment for two common disputes,
the eviction of a non-paying tenant and the resolution of a bounced check case. Duration of enforcement refers to
the time between notification and actual enforcement.
Table IV below presents additional indices related to different aspects of legal

procedures. A higher index value represents a move towards more formalism and legalism,
except in the case of Defendant Protection, Time Limits and Administrative Procedure.30
Notice the high index levels for Portugal, generally higher than the European Union and
29
Data is from Djankov et al. (2001).
30
All variables are described in the Data Appendix.
East Asian levels, denoting excessive formalism. In particular, note that the Dispute
Resolution Index - measuring the number of independent procedures necessary to resolve
an issue - is higher in Portugal than in the EU, both for tenant eviction and bounced check
procedures. The difference is even higher as Portugal is compared with the highly
competitive East Asian economies. The relatively high levels of Defendant Protection,
Mandatory Time Limits and the presence of Administrative Procedures in the Portuguese
case do not seem to reflect positively on the efficiency of legal procedures. These
characteristics of the system may instead be poor substitutes for the lack of legal efficiency.
3.2. Firms and Corporate Governance
The neoclassical approach to the firm views managers as deciding what quantities of
inputs and technologies maximize the value of output. When information is freely
available, management lacks a meaningful role. Coase (1937) first pointed out that, in a
setting without transaction costs, there would be no reason for the existence of firms. All
transactions would be undertaken in the market. However, even if “(…) production could
be carried out in a completely decentralized way by means of contracts between
individuals, the fact that it costs something to enter into these transactions means that firms
will emerge (…) whenever their costs [are] less than the costs of carrying out the
transactions through the market.” (Coase (1988, p. 7).31 These relative costs determine what
the firms buy and sell in the market and what they produce in-house. In sum, firms arise
when it is less costly to take transactions out of the market place and bring them into a
controlled environment, for reasons of monitoring, enforcing, etc. A series of different and
complex contracts can be substituted for one simple contract stipulating that the employee
works for a certain remuneration, under the directions of a manager and within prescribed
limits.32
There are alternative views of the firm. It can be seen as a unit that transforms
uncertainty: the employee trades his autonomy for a fixed or stable income while the
employer directs all resources, obtains an income which is uncertain and becomes the
residual claimant.33 A third view is that firms result from the need to organize an ever-
increasing specialization and division of labor. But the central issue, as Coase (1988) has
pointed out, remains one of relative costs and under what circumstances organization by the
firm is more efficient than the price system.
31
The firm is, in such a view, an environment within which the price mechanism takes the back seat, even if its
internal transactions need to remain reasonably linked to market prices through what the firm buys and sells.
The entrepreneur becomes the substitute of the price system in the management of firm resources.
32
More generally, a firm may imply control over another person’s property/capital as well as labor.
33
This is the view of Knight (1933). The system under which “the confident and venturesome assume the risk and
insure the doubtful and timid by guaranteeing to the latter a specified income (…)”.
Table IV
Legal Procedures for Tenant Eviction and Bounced Checks
TENANT EVICTION
Statutory Control of Other Dispute Administrative
Professionals Written vs. Legal Defendant Mandatory
Regulation of Superior Statutory Resolution Eviction
vs. laymen Oral Justification Protection Time Limits
Evidence Review Interventions Index Procedure
Portugal 1 0.75 1 0.38 1 0 5.13 1 1 1

Spain 0.67 0.88 1 0.63 0.67 0.67 5.5 1 1 1
Greece 1 1 1 0.5 0 0.67 4.17 . 1 0
European Union 0.6 0.8 0.7 0.3 0.7 0.2 3.9 0.5 0.3 0.1
East Asian Tigers 0.6 0.7 0.6 0.3 0.7 0.1 3.4 0.2 0.3 0
BOUNCED CHECK
Statutory Control of Other Dispute Administrative
Professionals Written vs. Legal Defendant Mandatory
Regulation of Superior Statutory Resolution Collection
vs. laymen Oral Justification Protection Time Limits
Evidence Review Interventions Index Procedure
Portugal 0.67 0.75 1 0.5 0.67 0 4.58 0.5 0.8 0

Spain 1 1 1 0.63 0.67 0.67 5.96 0.67 0.4 0
Greece 0.67 1 1 0.5 0 0.67 3.83 0.5 0 0
European Union 0.6 0.8 0.7 0.3 0.6 0.2 3.7 0.6 0.3 0.1
East Asian Tigers 0.3 0.6 0.3 0.3 0.4 0.2 2.5 0.5 0.3 0
Note: Source is Djankov et al. (2001). Defendant protections represent a broad concept of efficiency.
There are different stakeholders in a firm, from managers and employees to owners and
creditors. While shares give firm owners the fundamental right to vote for firm managers,
debt gives creditors the power to repossess collateral when the company fails. These rights
are the instruments through which owners and creditors control the actions of managers,
however indirect and diffuse that control might be. As put in La Porta et al. (1998)
“shareholders receive dividends because they can vote out the directors who do not pay
them, and creditors are paid because they have the power to repossess collateral.”
Shareholder Rights
In securities markets, investors pay large amounts of money for intangible and diffuse
rights, which depend on information that is imperfect namely on the behavior of managers
of which investors know little about. In unregulated markets this is a difficult act to pull.
Black (2000) addresses the issue of how to create strong securities markets and argues for
two essential prerequisites: quality information about the firm and confidence that
company’s insiders (managers and controlling shareholders) won’t cheat minority
shareholders out of the value of their investment.34 Institutions thus have an important role
in prodding managers to maximize firm value, instead of other objectives such as individual
prestige or size of the firm.
Bankruptcy codes, company laws and commercial codes, all affect the relation between
the firm managers and the the firm stakeholders, shareholders and creditors in particular.35
Through the internalization of monitoring costs, the concentration of ownership can
alleviate the agency problem – the difficulty of controlling the actions of the managers so
they reflect the interests of the owners.36 However, when certain shareholders have a
prevailing and controlling influence, conflicts of interest can arise between minority
shareholders and the controlling bloc, as the latter is tempted to extract private benefits. The
result may be a reduction in the incentives to invest by non-controlling shareholders and
debtors, decreasing the role of the stock market as a channel to mobilize savings. In other
words, the protection of minority owners enhances the liquidity of equity markets and
allows diversification but dampens the private incentives to monitor managers.
The main instrument of shareholders to exercise their control rights is the vote for
company directors or on specific company issues. This power to influence companies’
decisions through the vote is not equally distributed: there are non-voting shares and shares
that have diluted as well as increased voting powers. In some cases legislation enforces the
one-share/one vote rule for ordinary shares but this does not preclude ownership blocs and
ownership concentration since the interests of dominant shareholders and managers can
also be furthered by raising the cost of voting. The prohibition on mailing proxy votes,
mandatory share deposit prior to the general shareholder meeting, non-voting shares,
founders’ shares with extremely high voting rights, shares whose weight increases with
holding period, or even the option to first purchase of new stock issues are all mechanisms
34
Minority shareholders are generally defined as those holding less than 10 percent of the stock.
35
Company law regulates the relations between company shareholders, directors and other insiders, on the one
hand, and the corporation itself on the other. It also regulates the relations between the corporation and specific
outsiders such as creditors. Bankruptcy laws detail the procedures to be unravelled in the case of default and
firm failure. See La Porta et al. (1998) and Leahy et al. (2001) for discussions.
36
The controlling shareholders face strong incentives to monitor managers and maximize profits when they retain
substantial cash flow rights in addition to control. These incentives help restrain the diversion of corporate
resources and enhance the value of minority shares, even when minority shareholders are not directly involved.
to change the distribution of voting rights. There are mechanisms working in the opposite
direction: the possibility of cumulative voting (casting all votes for one candidate), of
proportional representation in the board, right to call a shareholder meeting, right to
challenge management decisions in court or to force the repurchase of stocks if
shareholders disagree with a fundamental decision.
Creditor Rights
Enforcing creditor rights is a way to elicit efficient decision-making by managers.

Creditors are an interested party since managers may be tempted to undertake projects that
are too risky. In other words, however high the expected return on the project, the
incentives of managers and creditors as residual claimants are not aligned. One possibility
to overcome this is the attribution of stronger creditor rights to secured lenders, which can
lower the cost of collateralized borrowing - the possible crowding-out of the remaining
(non-secured) borrowing and equity as lower payback priority discourages lenders.
The rights of creditors and bankruptcy provisions in general can change the incentives
of managers´ and lenders´. In reorganization procedures, an automatic stay on assets
prevents secured creditors from obtaining property rights over collateral, avoiding the
immediate liquidation of the firm and protecting the interests of managers, shareholders and
other creditors.37
The Structure of Ownership
The view of the corporation as an organization where ownership is fractured and diffuse
and management has substantial discretion - presented in Berle and Means (1932) - may not
correspond to reality. In a study of the ownership structure of large corporations in 27
developed countries, La Porta et al. (1998b) made a special effort to identify the ultimate
firm owners.38 Only in economies with very high levels of shareholder protection,
particularly in the richest common law countries such as the United States, do widely held
firms arise.39 Families, the State or other owners tend to control most of the financial
institutions. The power of the controlling shareholders is, by definition, much in excess of
what is justified by the capital invested and derives from direct participation in management
decisions, special voting rights, and pyramid ownership of different corporations.
The temptation to expropriate minority shareholders can be mitigated since controlling
shareholders also own rights to the cash flow. But the fact that controlling shareholders
generally do not support the protection of minority investors suggests that, for the former,
control rights are more valuable than other income. In countries where the protection of
minority investors is poor, the loss of control by controlling shareholders is extremely
costly so that controlling shareholders tend to bias voting rights in their own favor as well
as sell smaller fractions of the total shares. The result is that in countries that neglect the
rights of minority shareholder the voting premium - the relative price of shares with extra
37
Whereas in certain countries secured creditors are still paid first when creditors get paid, in others they take
second seat to other stakeholders, such as the government and employees.
38
So that when the shares in a company are held by another firm, the authors go back and investigate who owns
that firm, and so on.
39
Notice that the study of large firms in rich countries - such as in Berle and Means (1932) - biases the results
towards finding widely-held ownership.
voting rights versus ordinary shares – is higher. In sum, existing ownership structures are in
themselves a response to the domestic legal environments so that countries Italy and Israel
have a higher voting premium than the United States.40 The protection of minority investors
can be understood as a transfer from the controlling to the minority shareholders.
How can the power of controlling shareholders specifically be put in check? One
possibility is to improve the legal environment, mandate one-share one-vote rules and to
stress public disclosure of information. The recent trend to list non-US companies in United
States stock exchanges is part of such a move since companies can partially escape home
country institutions and enhance reputation by listing their shares at another country’s stock
exchange. However, investors realize that reputation alone is not enough and local
institutions and enforcement remain important. The need to attract capital may force firms
opt for regimes that are more protective of minorities.41 In this context, entrepreneurs may
be interested in minority shareholder protection if it increases access to equity and
facilitates firm creation but the voice of entrepreneurs is typically less vocal than the voice
of corporate families.
Firm Creation and Destruction
The ease with which firms are created and liquidated – and the issue of bankruptcy in
general – relates directly to the allocative and adaptive efficiency in an economy. The first
has to do with efficiency in a static context, a moment in time, and the latter with the
context of rules that affect how an economy evolves through time.42 Societies that allow
more trial and error can, over time, find better solutions to diverse problems. Adaptive
efficiency harks back to the concept of “creative destruction”, which Schumpeter (1934)
viewed as the basis for economic growth.
A recent paper by Djankov et al. (2001) analyzed entry regulations in several countries
and found official costs of entry to be very high in many countries. The official creation of
a firm tends to be a cumbersome, time-consuming, and expensive operation. The study
focuses on the number of entry procedures required and the time and dollar costs involved
in these procedures.43 A high cost of entry is associated with higher levels of corruption,
larger underground economies, larger governments and less democratic countries. As to
possible benefits, larger costs to entry are not found to be associated with a higher quality
of public or private goods - such as pollution control, health outcomes or market
competition -, dispelling the “benevolent ruler” explanation. The association between better
governmental and political institutions and lower entry costs suggests that what is at the
root of entry costs is increasing the benefits of regulators through bribes, nepotism, etc.44
40
The results are reported in La Porta et al. (1998).
41
In Italy, a country where investor protection is specially low, firms go public much less often than in the United
States, according to La Porta et al. (1998).
42
See North (1990, p. 80).
43
Requirements to initiate operation of a business vary substantially from 16 procedures, US$3,946 in fees and use
of 62 business days in Italy as compared with 2 business days, 2 procedures and US$280 in fees in Canada,
according to Djankov et al. (2001). The lowest official cost is less than 0.5 percent of per capita GDP in the
United States, while the sample average is about 50 percent of per capita GDP.
44
This is what Djankov et al. (2000) designate as the tollboth theory of regulation, suggesting the creation of
barriers to entry in order to collect financial paybacks.
The Financial System, Privatization and Corporate Governance
The nature of the financial system has an impact on how firms raise funds and how
firms are managed, what Stulz (2000) has dubbed the connection between the financing and
the governance problems. The financial and corporate governance systems are indeed
intimately associated. Academic economists tend to take it as given that savings are
naturally directed to the best uses and ignore the role of firms. In the traditional neoclassical
view of perfect capital markets and risk-neutral agents, the interest rate is a sufficient
instrument to channel savings to the most valuable investments. But, more important than
differences in the cost of capital – which is small across developed countries – the
differences in management incentives and monitoring lead to the pursuit of different
investment projects in different countries.
The “elimination of politically motivated resource allocation [in favor of market
allocation] has unquestionably been the principal benefit of privatization around the world”,
as argued in Shleifer (1998). Johnson and Shleifer (2001) pursues the issue further and
argues that, for privatization to be successful, a proper legal framework that includes
protection of minority investors needs to be in place. The context of privatization is also
important. According to La Porta et al. (1998), shares in Russian firms were basically
worthless after privatization - 100 times cheaper relative to American firms backed by
similar assets - due to the almost absolute lack of protection of the rights of minority
investors.
Corporate Governance: Portugal in Context
Let us now compare indicators of corporate governance in Portugal and other countries.
Table VII displays aggregate firm level indicators related to the rights of shareholders in the
governance process of the firm. These rights are naturally defined vis-à-vis the power of
managers and the rights of minority shareholders, uncovering the balance between the
different stakeholders. In broad terms, and for the issues where data is available, Portuguese
corporate law does not seem to differ dramatically from the European Union norm.
Portuguese corporate law does not follow a strict “one share - one vote” rule, does not
allow proxy voting by mail and does not require that a dividend be paid. As to minority
shareholders – those holding 10 percent or less of the company – the law in Portugal is not
specially protective as it does not allow cumulative voting for a director, proportional
representation of minority interests, nor a clear judicial venue to settle grievances or a sell
option if company policy diverges fundamentally from shareholders desires. The summary
indicator of Anti-Director rights shows that the situation in Portugal, while more favorable
to the rights of minority shareholders than in Greece or the European Union in general, it is
less favorable that that of Spain or East Asian countries.
The panorama above reflects the formal content of Corporate Law but legal text is not
the whole story. Life in the firms is affected by the actual distribution of ownership. Table
IX shows that the average ownership share of the three biggest shareholders - for the ten
largest private, non-financial firms – is similar in Portugal, Spain and East Asia, at around
50 percent. For Greece this average share ownership is larger, while for the European
Union it is much smaller. The difference between the 51 percent average share ownership
in Portugal and the 44 percent in the EU is more real than one may think: while 51 percent
generally implies majoritarian control, 44 percent often does not.
In addition to shareholders, creditors are important stakeholders in the firm. The role of
creditors is especially important in the Portuguese economy, where a large part of the firms´
financial funds arrives via the banking system, not the stock market.45 In the indicator
Creditor Rights, Table X shows Portugal lives a situation similar to that of Greece but
worse than that in Spain, the EU and East Asia. The selected summary indicator takes the
value 1 for Portugal and 3.25 for East Asia, where a higher index corresponds to better
protection of creditor rights. This is the result of several concurring facts: in Portugal there
is no automatic stay on assets, there are few restrictions on reorganizations when companies
are failing and management may stay during reorganizations. The rule that creditors are
those first paid when a company fails is the sole sign of strong creditor rights in Portugal.
The ease with which firms are created and closed can be gleamed from data in Table IX,
which presents a summary of the procedural mechanisms needed to open a business in
different economies. It presents the costs of following those procedures and actually
opening a firm - in dollar terms (share of average yearly income) and in time costs. As to
the firm destruction side, we present information on the frequency of bankruptcies. The
emerging picture is clear: in Portugal the law demands a greater number of procedures to
open a firm than in the EU or the East Asia. Only Greece imposes heavier bureaucratic
steps but the time cost to open a firm is still higher in Portugal. It takes twice as much time
to open a business in Portugal than in a European Union country. In terms of total cost it is
two and a half times more costly to open a business in Portugal than in a East Asian
economy and about fifty percent more costly than in a European Union economy.
The number of bankruptcies as a share of total existing businesses reveals that a
Portuguese firm is almost 30 times less likely to go bankrupt than a firm in the European
Union and about 20 times less than in an East Asian country. The likelihood of bankruptcy
is lower for Portugal and Spain than for any other EU country, and significantly so. There is
a strong suggestion of a link between the cost of opening and of closing a firm, with both
rising or falling together: countries where it is hard to open a firm it is also hard to close.46
Innovation suffers as existing firms outlive their competitive advantage live off their one
unfair advantage, the fact that they are already in existence.
In sum, the situation of corporate governance development in Portugal does not differ
much from the European Union’s as far as the letter of the law. However, the comparison
of the costs of opening and closing a business firm show that Portugal, Spain and Greece
are characterized by less dynamism than the European Union or East Asia.
45
As analysed further below.
46
We have computed the sample correlation between the time required to open a business and the likelihood of
bankruptcy and found it to be –0.51, suggesting a strong link between barriers to creating and closing a
business firm.
Table VIII
Shareholder Rights
Cumulative
Shares not Preemptive Share of
One Share One Proxy by Mail Voting Oppressed Anti Director Mandatory
Blocked Before Right to New Capital to Call
Vote Allowed Proportional Minority Rights Dividend
Meeting Issues ESM
Representation
Portugal 0 0 1 0 0 1 0.05 3 0
Spain 0 0 0 1 1 1 0.05 4 0
Greece 1 0 0 0 0 1 0.05 2 0.35
European Union 0.07 0.14 0.43 0.07 0.21 0.79 0.10 2.50 0.03
East Asian Tigers 0.50 0.25 0.50 0.25 1.00 0.50 0.07 3.50 0.00
English 0.17 0.39 1 0.28 0.94 0.44 0.09 4 0

French 0.29 0.05 0.57 0.29 0.29 0.62 0.15 2.33 0.11
German 0.33 0 0.17 0.33 0.5 0.33 0.05 2.33 0
Scandinavian 0 0.25 1 0 0 0.75 0.1 3 0
Whole Sample 0.22 0.18 0.71 0.27 0.53 0.53 0.11 3 0.05
Note: From La Porta, Lopez-de-Silanes, Shleifer and Vishny, (1998). European Union excludes Luxembourg. East Asian Tigers includes Hong Kong,
Singapore, Taiwan and South Korea.
Table IX
Ownership by Large Shareholders
Average 3 Largest Median Capitalization

Shareholders 3 Largest Shareholders 10 Largest Firms
Portugal 0.52 0.59 259

Spain 0.51 0.5 1,256
Greece 0.67 0.68 163
European Union 0.44 0.41 4216.21
East Asian Tigers 0.53 0.55 1473.55
English 0.43 0.42 6,586

French 0.54 0.55 1,844
German 0.34 0.33 8,057
Scandinavian 0.37 0.33 2,644
Whole Sample 0.46 0.45 4,521
Note: The first two columns are the average and median percentage of common shares owned by the three
largest shareholders in the ten largest non-financial, privately owned domestic firms (state is not a known
shareholder). The third column provides average market capitalization of the largest 10 firms. From La Porta,
Lopez-de-Silanes, Shleifer and Vishny (1998). The sample has a total of 49 countries. European Union excludes
Luxembourg and East Asian Tigers refers to Hong Kong, Singapore, Taiwan and South Korea.
Table X
Creditor Rights
Restrictions No Automatic Stay Secured Creditors Management Does Creditor Rights Legal Reserve
Reorganization on Assets First Paid not Stay in Required as
Reorganization Percent of Capital
Portugal 0 0 1 0 1 0.2
Spain 0 1 1 0 2 0.2
Greece 0 0 0 1 1 0.33
European Union 0.57 0.36 0.86 0.14 1.93 0.13
East Asian Tigers 0.5 1 1 0.75 3.25 0.375
English 0.72 0.72 0.89 0.78 3.11 0.01

French 0.42 0.26 0.65 0.26 1.58 0.21
German 0.33 0.67 1 0.33 2.33 0.41
Scandinavian 0.75 0.25 1 0 2 0.16
Whole Sample 0.55 0.49 0.81 0.45 2.3 0.15
Note: From La Porta, Lopez-de-Silanes, Shleifer and Vishny, (1998). European Union excludes Luxembourg. East Asian Tigers includes Hong Kong, Singapore,
Taiwan and South Korea.
Table XI
Opening and Closing a Business
Number of Safety & Dollar Dollar plus Percentage

Environment Taxes Labor Screening Time Cost
Procedures Health Cost Time Cost Bankruptcies
Portugal 12 0 0 2 2 8 76 0.1844 0.4884 0.08
Spain 11 0 0 4 2 5 82 0.173 0.501 0.02
Greece 15 0 0 4 2 9 36 0.586 0.73 na
European Union 9.00 0.00 0.07 2.07 1.43 5.43 36.57 0.16 0.31 1.96
East Asian Tigers 8.25 0.00 0.00 1.00 2.25 5.00 25.25 0.10 0.20 1.64
Note: From Djankov, La Porta, Lopez de Silanes and Shleifer (2001). European Union excludes Luxembourg. East Asian Tigers refers to Hong Kong, Singapore,
Taiwan and South Korea.
3.3. The Financial System, Banks and Stock Markets
Capital is a key factor of production and its ready availability to firms depends on the
functioning of the financial system. Gerschenkron (1962) has highlighted the relationship
between finance and development by contrasting the experiences of Britain and other
European countries.47 The Schumpeterian view of the process of development as one of
“creative destruction” implies economies where amassing resources for new ventures goes
hand in hand with the prompt liquidation of unprofitable firms.48 In the case of OECD
countries, Leahy et al. (2001) have assessed the relation between growth, innovation and
indicators of financial development and have shown that stock market capitalization and
private credit issued by deposit banks are both important factors. Jayaratne and Strahan
(1996) has shown that, when individual states in the US relaxed intrastate banking
restrictions, the quality of bank lending improved and economic growth accelerated. Thus,
both the size of the financial market and the level of internal competition have a significant
impact on growth.
The study of financial markets within the larger role of institutions in development is
fairly recent. One of the preeminent economic institutions, secure property and contract
rights, is seen as key for banks and financial institutions to work properly since weak
contract enforcement creates incentives for default by debtors and decreases willingness to
lend.49 Less investor protection substantially affects opportunities for external finance and
leads to smaller capital markets.50
What functions do financial markets and intermediaries perform? An efficient financial
system increases the amount of savings and improves their allocation. The financial sector
can raise saving rates and increase liquidity, a process named financial deepening. First,
financial intermediaries such as banks and securities markets pool resources that would
otherwise not be available to firms seeking a larger and more efficient scale of operation.
Financial intermediaries can also channel funds to projects with a longer horizon. A second
role is to overcome idiosyncratic risk involved in single-project investment, even when the
amount of financial resources available to the investor is small. Risk is priced, diversified
and transacted so that projects with high return and high-risk can be financed. A third role
is that of acquiring information and evaluating projects, a process too costly to be
undertaken at the individual level. Fourth, financial intermediaries monitor firm managers,
overcoming another problem of high cost of information acquisition.
47
Wheareas Great Britain relied on a market-based system, Germany, as a latecomer, relied on a bank-based
system.
48
See Schumpeter (1934).
49
As argued in Shleifer and Vishny (1997). Using a dataset on legal and institutional country characteristics and
the level of legal protection and enforcement of investor’s rights, La Porta et al. (1997) find wide cross-country
differences: common law countries protect investors better and French civil law countries fare the worst in
investor protection.
50
As shown by La Porta et al (1998).
Trust and Financial Development
The relationship between trust and development has been put forward in recent work
on social capital and institutions. Knack and Keefer (1997) argues that institutions may be a
substitute for trust: when the latter is low, institutions that provide formal mechanisms for
contract enforcement become more valuable. A financial contract is, by nature, trust
intensive: resources are transferred today in expectation of future paybacks between parties
with imperfect and asymmetric information. Financial contracts are intrinsically incomplete
and monitoring costs high so that opportunism can easily arise. Collateral requirements
become the standard response. The effectiveness of institutions, particularly as they enforce
contracts and the repossession of collateral in disputes, affects aggregate outcomes directly.
The size and nature of financial markets, how much is borrowed and lent ultimately
depends on the levels of trust or the effectiveness of institutions that substitute for trust.
Guiso, Sapienza and Zingales (2000) study the relationship between trust and financial
development by contrasting trust attitudes between northern and southern Italians. Even
after taking full account of regional differences in education and law enforcement, their
evidence supports the hypothesis that trust and financial development go hand in hand: in
regions with higher levels of trust households hold less cash, have higher stock investments,
use more checks, have more access to credit, and use informal markets less frequently.
Even southern migrants in the north display financial behavior that is closer to their region
of origin. Caldéron et al. (2001) examines aggregate data and - after controls for the size of
the economy, level of human capital, inflation and law enforcement - finds higher levels of
trust associated with deeper financial markets, lower interest rate margins and overhead
costs and more dynamic stock markets. Countries with lower trust and poorer institutions
tend to have smaller and less efficient financial markets.
Bank-Based versus Market Based Financial Systems
An important issue is the comparative advantage of banks versus stock markets. At

low levels of financial development, the type of financial system affects the type of
external finance. In countries where the legal system suggests a developed securities
market, firms grow at rates that require sustained long-term external finance. Securities
markets may be better suited for long-term financing, whereas the banking sector is closely
related to the availability of short-term financing. The type of system may, thereby, have
implications for which firms and which projects are financed.51
Authors that argue that banks can better finance the expansion of existing firms as
well as the establishment of new firms emphasize banks´ low cost of screening and
monitoring, as well as the longer-term relationships with the firm. Stiglitz (1985) has
argued that there is a public good problem in acquiring information from developed
financial markets. As markets reveal new information – aggregated in stock prices –quickly
to all investors there are little incentives for individuals to spend resources researching
firms. Bank-based systems mitigate this problem through the longer-term relationships and
the fact that some of the acquired information is not publicly revealed.
Moreover, shareholders can easily sell their stake in a corporation and increase it in
another so there is little incentive to exert influence over management. Such is not the case
51
As argued in Demirgüç-Kunt and Maksimovic (1996).
with banks. Since banks enter a long-term relationship with firms, they can credibly commit
to make new resources available as the firm’s expansion and its performance justifies.52
The opposing view stresses the benefits of financial systems based on the stock
market. The stock market is large in terms of resources mobilized and very liquid, obvious
advantages in financing firm creation and growth. New technologies have a better chance
of being objectively appraised in such large and liquid markets. Efficient prices in
secondary markets help investors identify good and bad investments through a mechanism
similar to Tobin’s q. In contrast with the views expressed above, individuals can reap
substantial profits from acquiring information so that the acquisition of information in stock
markets may be at reasonable levels. In addition, the relationship between stock
performance and managers’ compensation and the possibility of takeovers are important
mechanisms of corporate control.
By acquiring inside information, banks can extract informational rents from firms so
that the price of finance increases. The insider status of banks is also in the fact that bankers
often hold equity and can collude with managers and the directors against the interests of
minority shareholders. Moreover, banks are debt issuers with a conservative bias against
risky projects, thus stifling innovation and growth. The benefits of solid collateral can bias
banks in favor of continually financing established firms with low-return projects instead of
new and innovative firms.53 In sum, there are also good reasons to believe developed stock
markets further innovation and growth in ways that banks cannot.
The Financial Services View
Research on the relative merits of the two types of financial systems uses long-term
analyses and comparisons of Germany and Japan - as bank-based systems – versus the
United States and Great Britain - as market-based. There does not seem to be a solid
difference between bank-based and market-based financial systems in terms of firm growth,
whether new or existing.54 A third view of financial systems, - the financial services view –
suggests that firms participate in both markets and that each type can gain the most in
different markets: smaller and younger firms with banks, larger and more innovative firms
with stock markets. Banks and stock markets are complements that can both promote the
flow of external funds to solid and entrepreneurial firms.55 Beck and Levine (2001)
decomposes sectoral growth into growth in establishment size and firm entry and finds that
overall financial development explains cross-country variation in the number of
establishments.
52
See Beck and Levine (2000).
53
As argued in Rajan and Zingales (1999).
54
See evidence in Beck and Levine (2001).
55
Garcia and Liu (1999) find that the development of financial intermediaries such as banks actually supports the
increase in stock market capitalization, further validating the complementary view.
The Legal Determinants of Financial Systems
Financial systems and their institutional characteristics have deep roots in the history
and political culture of a country. The legal system in particular conditions the type of
financial system that evolves over time. Legal traditions put different emphasis on the
comparative rights of individual investors vis-à-vis the state, with consequences for
financial development.56 Beck et al. (2001) show that the level of overall financial
development is indeed affected by the legal tradition. Moreover, the level to which firms
are financially constrained is more stringent when the risk of expropriation is high, the legal
systems inefficient and corruption high.57 When a legal system that can protect external
investors is absent, financial intermediaries who have sufficient bargaining power to
enforce their rights privately come forward and extract rents, as argued in Modigliani and
Perotti (1999).
A related prediction is that overall financial development as determined by the legal
system increases opportunities for the entry of new firms and the expansion of existing
ones. Rajan and Zingales (1998) and Beck and Levine (2001) decompose industrial growth
into two components: existing firm growth and the emergence of new firms. Rajan and
Zingales (1998) shows that firms dependent on external funds grow faster in economies
where the legal system duly enforces the rights of outside investors.58 The legal
environment also explains variation in the growth in the number of establishments across
countries. In accordance with this view, Demirgüç-Kunt and Maksimovic (2000) use firm-
level data for 40 countries to analyze how external finance impacts economic growth by
asking two questions: does the financial system have an effect on growth independent of
the legal system; is the use of external financing different in market-based and bank-based
systems and do market-based and bank-based systems differ in the provision of long-term
and short-term funds? They do not find evidence that differences across financial systems
unrelated to the legal system (such as bank versus market based) affect access to external
finance. Use of external finance is positively related to both developed banking and market-
based systems.
A variation on the law and finance view above highlights the ability of legal traditions
to adapt and evolve with development.59 Differences in the realm of adaptability may
affect financial development as legal traditions that adapt quickly to close any gap between
the needs of the economy and the capabilities of the legal system are likely to foster
financial development. The common law system is seen as particularly dynamic as
individual judges respond on a case-by-case basis and contribute to setting new precedent
in an expeditious way. The civil law tradition tends to see the codified written law as
immutable and unchangeable. Practice does not entirely accord with this dichotomy and all
countries adapt to a greater or lesser extent to new economic and contractual realities.60
56
This legal-based view thus rejects the dichotomy between bank and market-based systems.
57
The effect of these factors also works indirectly through financial development, as reported in Love (2001).
58
In previous studies, Levine (1999) and Levine, Loayza and Beck (1999) find an association between countries
that offer better investor protections and more developed financial markets.
59
As in Beck et al. (2001).
60
According to Beck et al. (2001), the German legal tradition has explicitly rejected the French approach and
sought to create a dynamic legal code. The emphasis on the evolutionary capabilities of legal systems
approaches the German and the common law systems.
Bank Regulation and Supervision
An important issue is how banks are regulated and the consequences of the regulatory
framework. The impact of banking crises on economic growth, of which the recent East
Asian crisis is the preeminent example, have put banking regulation reform high on the
agenda. International financial institutions, including the Bank for International
Settlement’s through the Basel Committee on Bank Supervision have devised a list of “best
practices” for regulation and supervision of banks with the hope of promoting bank “safety
and soundness”.61 But there is no evidence that the current set of best practices is best or
that more checks are always better than less. Instead, views on supervision reflect broader
views on government and its role in the economy.
Two contrasting views of regulation and supervision frame our discussion, the
“helping-hand” and the “grabbing-hand” views.62 The helping-hand view sees
governments as regulators that correct market failures whereas the grabbing-hand view sees
the support of political constituencies as the motivation of regulation. Pigou’s (1938)
treatment of regulation has put forth the view of government as correcting monopoly
power, externalities and informational asymmetries and thus increasing general welfare.
Applied to the banking system, this helping hand view of government favors close and
stringent regulation of banking activities on issues such as capital standards, range of bank
activities, firm entry and deposit insurance. Gerschenkron (1962) focused on the financial
prerequisites to start economic growth and argued that, even if privately owned commercial
banks are the main vehicle to channel savings to industry in countries such as Germany, in
other countries such privately-owned intermediaries are not available and the government
has to step in.63 The helping-hand view takes as given that the government can help
overcome existing market failures.
If it is the grabbing-hand view that is correct, one expects countries with strong
supervisory frameworks, entry restrictions and government ownership of banks to present
lower levels of political accountability and economic performance with no improvement
better bank performance. Encouraging private sector control of banks and high levels of
competition would be the needed remedy.64 Barth et al. (2001a) have developed a database
61
See Barth et al. (2001a) for a discussion.
62
As in Shleifer and Vishny (1998). These views translate into alternative views as to the effect of private versus
government supervision. An example is the possibility that the high cost of overcoming informational barriers
limits the monitoring of complex financial conglomerates by the private sector so that banks incur risk levels
which are higher than optimal. In this case it is beneficial for governments to monitor banks more closely.
63
In the developmental view, ownership of banks enables the government to collect savings and direct them
toward strategic long term projects. The government overcomes institutional failures which pervade private
capital markets, and generates aggregate demand and other externalities that foster growth. “The scarcity of
capital in Russia was such that no banking system could conceivably succeed in attracting sufficient funds to
finance a large scale industrialization; the standards of honesty in business were so disastrously low, the general
distrust of the public so great, that no bank could have hoped to attract even such small capital funds as were
available, and no bank could have successfully engaged in long term credit policies in an economy where
fraudulent bankruptcy had been almost elevated to the rank of a general business practice” (Gerschenkron
(1962), p. 19).
64
La Porta et al. (2002) show that government ownership of banks is ubiquitous and quantitatively important,
though higher in low-income countries with less deveoped financial systems and interventionist political
cultures. For the average country, 59 and 42 percent of the equity of the 10 largest banks is owned by the
government in 1970 and 1995, respectively. Higher government ownership of banks in 1970 leads to slower
financial development and slower growth in income per capita and productivity. Corroborating this view,
on bank regulation and supervision for 107 countries.65 The best mechanisms to promote
bank performance and stability seem to be private ownership and control together with
information disclosure. Over reliance on direct government oversight of banks needs to be
put in check. Authors such as Shleifer and Vishny (1998) advance the hypothesis that
government intervention serves to support private constituencies through employment,
subsidies and other benefits. In the end, government failure is at least as important as
market failures and overbearing regulation and supervision do not seem beneficial.
Deposit Insurance and Financial System Risk
Many countries have adopted deposit insurance programs with the objective of
reducing the risk of bank runs and generally stabilizing the financial sector. Deposit
insurance can increase the size of the financial pool through an increase in public
confidence, benefiting economic growth in the process. On the other hand, if the
government absorbs too much of the system’s risk, a moral hazard problem emerges where
self-monitoring decreases and excessive risk is taken by financial intermediaries.
Whenever, in a crisis, depositors are to be bailed independently of whether explicit deposit
insurance is available or not, systemic problems may arise.66 The negative effects of deposit
insurance may be compounded if the regulatory framework is lax.
Demirgüç-Kunt and Detragiache (2000) have analyzed banking crises data for 61
countries between 1980 and 1997. Variations in coverage, funding and management of
deposit insurance schemes significantly affect the likelihood of banking crisis, especially
when the broad institutional framework is weak. But the same explicit deposit insurance
program will have different effects within a different institutional context. Cull et al. (2001)
examined the effect of deposit insurance on the size and volatility of the financial sector for
long horizons. They find that insurance leads to financial instability in lax regulatory
environments - proxied by indices of the rule of law- and to financial development and
growth in sound regulatory environments. Furthermore, the longer a deposit insurance
scheme is in place - regardless of its specifics - the less concentrated the banking sector is.
Financial Development, Financial Constraints and Capital Allocation
Capital is scarce and its good allocation is an important economic objective. Bringing
capital to sectors where its return is highest is one of the ways financial intermediaries
promote economic growth. Wurgler (1999) uses data on investment for 65 countries and 28
manufacturing industries for over 30 years to uncover that the size of the domestic stock
and credit markets relative to GDP is associated with a better allocation of capital.
Financially developed countries finance growing industries better, at the expense of
investment in declining industries. This may be due to better flow of information on
Sapienza (2002) finds evidence that Italian state-owned banks pursue political objectives in their lending
policies.
65
These authors assess how the development and fragility of the banking system is affected by the mixing of
banking and commerce, regulations on bank entry and capital adequacy, deposit insurance, independence and
power of the supervisory agencies, information disclosure and government ownership of banks, among others.
66
Authors such as Gropp and Vesala (2000) argue that implicit insurance as has been the case in Europe may
imply more potential for moral hazard than explicit systems. The trade-off is between a higher uncertainty of
being bailed out versus a more encompassing implicit coverage.
specific firms, lower incidence of state ownership and better investor rights that limit
investment in declining industries.67 Love (2001) measures financial constraints by the
sensitivity of investment to the availability of internal funds and finds stricter constraints to
be associated with lower financial development, after controlling for determinants of
financial constraints including firm size and the business cycle.
Laeven (2001) confirms the results by Love (2001) that financial development and
financial constraints change together but adds that the effect is different for small and large
firms: small firms become financially less constrained after financial liberalization whereas
there is no significant difference for large firms. The result is that the relative ease of access
to finance decreases with financial development. This is probably due to large firms’ better
(and unfair) access to capital even in systems that are not fully developed or due to larger
firms ability to overcome informational asymmetries. To an extent, size is information.
The allocation of capital also impacts growth through the creation of new firms. Rajan
and Zingales (1998) and Beck and Levine (2001) find that industries that depend on
external funds grow faster wherever the financial system has a high level of development,
in addition to better legal protection of outside investors.68 These authors also corroborate
the view that new firms are more easily established in countries with developed financial
markets.
Financial Systems: Portugal in Context
We now provide a brief description of the state of development of financial systems in

Portugal, as compared to other countries. We move from broad general characteristics of
financial markets to firm-specific financial indicators. Figures II through IV present the
evolution of indicators of financial sophistication in Portugal, Spain, Greece, the European
Union and the East Asian high growth economies. Figure II reveals that liquid liabilities as
a share of GDP have been traditionally high in Portugal since the 1960’s. While Greece still
displays lower levels of liquidity, Spain has caught up and overcome Portugal in the
1990’s. The East Asian economies have not only caught up but now clearly surpassed the
level of liquid liabilities for Portugal, while the average level for the European Union is
similar to Portugal’s. The depth of stock market finance is given in Figure III as Stock
Market Capitalization to GDP. Portugal experienced a steady increase in this ratio since the
1960’s with a spurt in the late 1980’s that corresponded to the revival of the stock market´s
role in financing private equity. In international terms, however, the level of Portuguese
stock market capitalization is only above that of Greece, below Spain’s and well below the
European Union and East Asia. The value of firm stock quoted is below 20 percent of GDP
in Portugal, compared with levels above 40 percent for the European Union. Figure IV
presents private bond market capitalization to GDP, where Portugal displays levels above
those of Spain and Greece but again clearly below the EU levels. The value of private
bonds as a share of Portuguese GDP was around 20 percent in 1996, compared with 30
67
This latter channel suggests that, in the absence of efficient monitoring by investors, firm executives freely
allocate cash flow and tend to overinvest in declining industries.
68
Rajan and Zingales (1998) use industry-level data and focus on firms which are highly dependent on external
funds. In countries where financial markets are highly developed these financially dependent firms grow faster
than firms which rely more on internal capital. Demirgüç-Kunt and Maksimovic (1996) draw their attention to
firms within better developed financial markets and show that they grow faster than they would have grown
without access to external finance. Both papers can be seen as supporting the view that external finance matters.
percent for the EU. In terms of financial sophistication, Portugal is above that of Greece,
compares well with Spain but poorly with the European Union as a whole, while the
contrast with the East Asian high growth economies is even starker. Over time, Portugal
saw a deterioration of its relative level of financial development, at least as far as stock
market capitalization and liquid liabilities as a whole are concerned.
In Table V we present several indicators characterizing the financial system as a whole.
The first four indicators measure the intensity of financial activity, the next four the
structure of financial activity, namely the relative reliance on banks or the stock market as
the main source of finance. The very last three indicators in Table V measure aggregate
financial market efficiency through the net interest margin, the share of private credit and
the total value traded in the stock market as a share of GDP. The intensity of financial
transactions is lower in Portugal than in Spain, the European Union and East Asia. Only
Greece displays lower levels of financial deepening than Portugal. As to the balance
between banks and the stock market, all indicators suggest that Portugal relies relatively
more on financing by banks. This bias is more pronounced than that of Greece, with a low
level of financial development. The high level of private credit to GDP in Portugal, 0.63 as
compared to 0.72 for Spain and 0.40 for Greece, goes hand in hand with a value of total
stock traded that is only 2 percent of GDP in Portugal and Greece and 6 percent in Spain.
The European Union and East Asia have substantially higher shares of stock value trade.69
The net interest rate margin, the value for Portugal at almost twice that for the East Asian
economies, points towards low efficiency levels of the Portuguese financial system.70
Table VI presents indicators on the bank sector and bank supervision in general.
Portugal has the highest level of bank sector concentration: the five largest banks control
almost 82 percent of Portuguese total bank assets while even in financially unsophisticated
Greece the corresponding figure is below 70 percent.71 In Spain, and in spite of the
movement towards the merger of the largest banks, the five largest banks control only about
50 percent of total sector assets. Compared with the EU as a whole, Portugal has more
government ownership in the banking sector and weaker international bank presence. Spain
has levels of foreign ownership similar to Portugal’s.
As far as bank supervision, frequency, quality and independence indicators do not make
Portugal stand out from the general practice in the European Union. The number of
supervisors and the number of onsite examinations in Portugal is on a par with Europe and
the likelihood of supervisors being employed by the bank industry after abandoning their
role as public regulators is not higher. As in all other EU economies for which information
is available, there is an explicit deposit insurance scheme in Portugal.
At the level of the firm, the indication is that Portugal has financial development
indicators that are below those of Greece, Spain, the EU and the East Asian averages.
Interestingly, the firm level data corroborates the picture of the Portuguese financial system
69
One should note that the levels of stock market capitalization are much higher in the United States and even in
the United Kingdom than in the EU in general.
70
It is unclear whether a lower net interest margin is always for the better. The experience of the East Asian
economies leading to the financial crisis of 1997 suggests that the net interest margin in these countries was at
less than efficient levels, resulting in excessive lending to the private sector.
71
There may be an argument for higher bank concentration ratios in small countries as a minimum efficient level
has to be attained. It is specially useful, in this regard, to compare Portugal with a similar-sized economy such
as Greece.
as strongly biased towards the use of bank finance and against own equity and stock market
issues.
Figure II
Liquid Liabilities to GDP
1.6
1.4
share of GDP
1.2
1
0.8
0.6
0.4
0.2
0
60
62
64
66
68
70
72
74
76
78
80
82
84
86
88
90
92
94
96
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
year
Portugal Spain Greece European Union East Asia
Figure III
Stock Market Capitalization to GDP
1.6
1.4
share of GDP
1.2
1
0.8
0.6
0.4
0.2
0
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
year

Figure IV
Private Bond Market Capitalization to GDP
0.35
0.3
share of GDP
0.25
0.2
0.15
0.1
0.05
0
1990 1991 1992 1993 1994 1995 1996 1997
year

Table V
Financial System
Net
Finance- Finance- Finance- Finance- Structure- Structure- Structure- Structure- Private Total Value
Interest
Activity Size Aggregate Dummy Activity Size Aggregate Dummy Credit Traded
Margin
Portugal 4.23 4.31 0.12 1.00 -4.26 -2.66 -1.49 0.00 0.04 0.63 0.02
Spain 5.71 4.43 0.49 1.00 -2.71 -1.55 -0.30 0.00 0.04 0.72 0.06
Greece 2.59 3.92 -0.46 0.00 -4.47 -1.62 -0.92 0.00 0.04 0.40 0.02
European Union 5.48 4.41 0.43 0.69 -2.65 -1.30 -0.13 0.46 0.03 0.71 0.10
Asian Tigers 7.36 4.88 1.11 1.00 -1.07 -0.32 1.00 1.00 0.02 1.04 0.41
Note: See the Data Appendix for sources and definitions.
Table VI
Bank Supervision
Percentage Onsite
Five-Bank Percentage of Bank Supervisors
Assets Percentage Examination Explicit Deposit
Concentration Top 10 Rated Supervisors per Employed by
Government Foreign Owned per Bank Last 5 Insurance
Ratio Internationally Institution Bank Industry
Owned Years
Portugal 81.70 20.80 11.70 100.00 2.40 2.50 2.00 1.00

Spain 49.00 0.00 11.00 100.00 0.60 2.00 2.00 1.00
Greece 70.00 13.00 5.00 50.00 1.50 2.00 1.00 1.00
European Union 59.19 10.81 16.29 66.15 0.78 2.43 2.07 1.00
Asian Tigers 31.25 24.23 25.00 100.00 8.07 3.00 2.00 0.50

We now turn to the financial life of firms, as captured by indicators of financial health
at the firm level. Table VII presents several measures of firm’s operating income
variability, the share of debt in equity and of net working capital in total assets. The picture
for Portuguese firms is revealing: operating income is more volatile than in other
economies and the levels of indebtedness – short or long term – are much higher, while the
share of net working capital in total assets is lower. In sum, Portuguese firms face more
volatile income streams, have less capital to work with and more of that capital is externally
lent so that Portuguese firms are very vulnerable to economic fluctuations. In spite of their
higher indebtedness to the bank sector, as compared with similar firms in other countries,
Portuguese firms are relatively capital starved: the 0.05 share of net working capital to
assets compares with 0.12 for Spain and 0.18 for Greece. The high short-term indebtedness
is only surpassed by the East Asian economies, which were to experience the financial
crisis of 1997, arguably due to their vulnerability to short-term capital flows.
Table VII
Finance and the Firm
Operating Total Debt to Long Term Debt Short Term Debt Net Working
Income Market Value of to Market Value to Market Value Capital to Total
Variability Equity of Equity of Equity Assets
Portugal 0.87 0.59 0.22 0.37 0.05

Spain 0.77 0.29 0.10 0.19 0.12
Greece 0.46 0.17 0.01 0.16 0.18
European Union 0.69 0.39 0.16 0.23 0.17
Asian Tigers 0.41 0.83 0.19 0.65 0.11
4. Institutions and Growth: A Quantitative Exploration
The most widely recognized framework to analyze economic growth draws on

neoclassical growth theory. The neoclassical growth model, first developed in Solow
(1956), associates the rate of economic growth with several factors, preeminently among
them the level of capital per worker in the economy. All else equal, a higher level of capital
per worker implies a lower growth rate as the marginal product of an extra unit of capital
decreases with aggregate capital intensity. In other words, as countries accumulate more
capital (relative to labor) the marginal product of the additional capital decreases, causing
growth to taper off.72 Countries that differ in fundamentals such as the production function
or the savings rate will also differ as to the level of capital per worker at which growth will
taper off, i.e. the steady-state level of income per capita. The higher the distance between
the a country´s current level of capital per worker relative to the steady-state, the higher that
country´s rate of economic growth. This is the so-called convergence effect. In empirical
studies the level of income per capita is used as a proxy for distance to steady-state to test
for the significance of the convergence effect. For this reason, a sensible examination of the
72
Barro and Sala-i-Martin (1995) provide an extensive survey of the basic model and extensions.
influence of an economic or social variable on economic growth should control for a

country’s initial level of income per capita.
In contrast with neoclassical growth theory, the theory of endogenous growth highlights
the importance of technological progress to economic growth and the different rates at
which countries may grow in steady-state. The focus is shifted towards identifying the root
causes of technological progress.73 In endogenous growth theory, the preeminent candidate
to explain technological progress is the scale of the market: larger scale makes innovation
more profitable and accelerates the rate of technological progress. For the purposes of the
current paper it would make sense to interpret institutions as one of the factors that can
affect the effective scale of the market and thus growth in steady state. As laws become
clearer and more efficient, as financial markets grow deeper and more sophisticated and
firms operate in a governance framework that fosters the efficient use of capital and labor,
the number of transactions and the effective scale of the market should increase.
In this paper, we choose to conduct our empirical exercise in the context of the
neoclassical growth theory, for two reasons. First, this is the simpler and more widely
spread framework, allowing a clearer evaluation of our results in the context of previous
results in the literature. Second, there is a gaping lack of empirical support for the role of
scale in economic growth so that the empirical relevance of endogenous growth theory is
still to be determined. The interpretation of the role of institutions in terms of scale is,
consequently, of lesser relevance at this point.
This section uses data on institutions, most presented in the above section, to investigate
which institutions matter for growth and how much they matter.74 Since institutions tend to
change slowly and what we care for is long-term growth, we use average yearly growth in
the 1960-1995 period as our dependent variable to be explained.75 As right hand-side
variables we always use the level of country GDP per capita in 1960 to assess the
convergence effect. In addition, and in succession, we examine the role of the various
institutional measures.
Our empirical specification is thus:
Growth GDP 1960-95 = ”0 + ”1 . GDP 1960 + ”2. Institution + ε
Several of the institutional measures used in this paper were available only for later sub-
periods, sometimes only for the 1990-1995 period. For reasons of lack of data availability
and because institutions tend to change slowly over time, we use the indicators for the
1990’s as proxies for a country’s level of institutional development in the whole 1960-1990
period. We can interpret our results as assessing the long-term relationship between
institutions (evolving slowly over time) and growth, which is volatile in the short-run but
less so in the long-run, when measured over several decades. Our exercise allows the
identification of the institutions that affect growth significantly, in statistical and in
economic terms. The results are grouped by area and category of the indicators and
presented in Tables XII through XIX below. We present the results for Ordinary Least
Squares estimates and present standard errors that are heteroskedastic-consistent.
73
Aghion and Howitt (1998) review the literature on endogenous growth.
74
For a more general introduction to the issues of the economic determinants of institutions and their
consequences for economic growth see Barro (1999) and Tavares and Wacziarg (2001).
75
Other definitions of the dependent variable were used without changing the qualitative results appreciably.
The objectives of the empirical exercise described above should be stated beforehand.
Our aim is not to determine causality. Given the broad scope of the current study as well as
the variety of institutions under scrutiny, we leave the test of specific causality relationships
to future studies. Instead our aim is to suggest patterns of correlation between indices of
institutional development and average long-term growth. The breadth and scope of the
paper also determined our choice not to present robustness results and not adding additional
control variables other than initial income per capita. The small sample size - due to limited
availability of data on institutions – and the difficulty in measuring some of the institutional
indices support our choice of a simple and unencumbered empirical specification that
captures broad patterns of correlation rather than assess causation. In fact, for the samples
involved, the convergence effect is the most significant single factor that robustly affects
economic growth.
We will interpret statistical significance of the coefficient on the institutional index as
pointing towards an association between institutional development and economic growth.
If, in a given policy area – legal, corporate governance and financial systems – we find that
regression coefficients tend to come out as statistically and economically significant, we
will conclude that reform in this area is likely to deliver important growth benefits.
The first tables present results for the legal system, from aggregate performance
indicators to specific characteristics of tenant eviction and bounced check collection
procedures. Table XII shows that all aggregate legal performance indicators are associated
with growth in the predicted way. Since higher levels denote better institutions the positive
regression coefficients suggest that improving legal institutions is associated with higher
growth. All the indicators with the exception of judicial efficiency have statistically
significant coefficients, and even the former is almost significant. Better rule of law, lower
corruption, lower risk of expropriation and of contract repudiation, and better citizen
access to justice all foster growth. Respect for contractual agreements – low expropriation
and repudiation risks – seem to have the strongest impact on growth, followed by a
predictable legal framework characterized by a strong rule of law and low corruption.
Tables XIII and XIV present the relationship between the procedural characteristics of
check collection and tenant eviction and economic growth. The first noticeable regularity
is the significance of virtually all procedural characteristics of check collection, the
exception being defendant protection and the written versus oral legal bias of the law. In
contrast, for the case of tenant eviction only statutory regulation of evidence and the
dispute resolution indices matter for economic growth. These results suggest that indeed
our exercise is not spurious: check collection - the procedure more directly connected with
economic transactions – is indeed the one that more directly correlates with rates of
growth. All these specific indices grow in the direction of less efficiency and efficacy so
that the negative sign suggests that lower efficiency of legal procedures leads to lower
average growth over time. Overall, legal indices are strongly related to economic growth
at the aggregate and the disaggregated levels, suggesting legal institutions are key to long-
term economic development.
Tables XIV and XV present regression results for corporate governance indices, the
first set related to the power relations between firm stakeholders and the second set related
to the opening and closing of firms. Our findings suggest that the first group of
governance indicators is generally not statistically significant as a determinant of long-
term growth. Only the indicators of the percentage of share ownership required to call a
general meeting, the average size of the three largest shareholders and the quality of
accounting standards seem to matter. Nevertheless, these three indicators matter in a way
that is compatible with the most common prior: higher minimum shares to call a general
meeting lower growth, as well as more concentrated ownership and worse accounting
standards.
In contrast, all measures indicating a barrier to opening a firm are associated with
lower growth. A larger set of required procedures, higher time, dollar or total cost all
associate with slower growth. The addition of one more procedure as a necessary
requirement to open a firm is associated with a 0.10 percent decrease in average yearly
growth. Thus, the apparently small difference between Portugal and the EU - three
required procedures – may have an economically important impact on growth. On the
other hand, the number of bankruptcies is totally unrelated with long-term growth in this
sample. It seems that it is the ease with which a firm is opened that most determines the
quality of the existing pool of firms and the rate of economic growth.
Tables XVII through XIX analyze the relation between financial indicators and growth,
assessing the importance of the depth of the system, the bank versus stock market bias and
indicators of firm financial health. We find that all indicators related with the depth of the
financial system affect growth positively. By contrast, only one of the indicators related to
the structure of the financial system – bank-based versus stock-market based – is
significantly related to economic growth. This is in line with previous results in the
literature, showing that it is the development of the financial system and not its structure
that matter for economic development. Other indicators of the general performance of the
financial system relate significantly with country growth and the sign of the estimated
coefficient is also as expected: higher net interest margins lower growth while more private
credit, higher value traded in the stock market and more non-bank credit all increase
growth.
Table XVII shows that the nature of bank supervision and regulation do not seem to be
important determinants of an economy’s growth after the convergence effect is taken into
account. The percentage of banks rated internationally and the existence of an explicit
deposit insurance scheme are the only indicators that are significantly associated with
growth. In contrast, bank concentration, government ownership, the intensity and
independence of supervision and the nature of funding for the deposit insurance scheme are
not important for long-term growth.
As to disaggregated financial indicators at the level of the firm we find that total debt
and short-term debt as a share of equity positively impact growth. None of the other factors
significantly affect growth, even if higher income variability and the availability of long-
term debt approach significance with the expected signs.
Table XIX on firm finance shows that only the level of short-term debt is significantly
related to economic growth: the higher the access to short-term financing through banks,
the higher the rate of growth. At the firm level, a lower variability of operating income,
higher working capital per assets and even more long-term financing do not seem to raise
the growth rate for an economy. The gist in terms of financial system is that financial
sophistication and depth matter while financial structure – stock markets versus banks - and
firm level indicators do not.
Table XII
Performance of the Legal System
Dependent Variable: Growth of Real per capita GDP 1960-1995
Rule of Law Corruption Risk of Contract Judicial Efficiency Citizen's Access to

Expropriation Repudiation Index Justice
(1) (2) (3) (4) (5) (6)
Initial Income -0.0006 -0.0004 -0.0006 -0.0005 -0.0001 -0.0002

-6.40 -3.97 -8.47 -7.97 -6.49 -2.60
Institutional Indicator 0.58 0.48 1.03 0.89 0.16 0.21

6.52 3.55 9.40 9.10 1.49 3.14
Number of Observations 48 48 48 48 53 70
R2 0.56 0.34 0.72 0.72 0.21 0.13
Note: The description of each variable and its source are explained in the Data Appendix. t –statistics are presented in the even rows using heteroskedastically-consistent
standard errors. The first row of the Table designates the institutional indicator used as independent variable.
Table XIII
Procedures for Check Collection
Index Index Log of Index Index Index Legal Index Index Dispute
Defendant Mandatory Duration Professionals Written-Oral Justification Statutory Control of Resolution
Protection Time Limits -Laymen Regulation of Superior Index
Evidence Review
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Initial Income -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001
-4.35 -5.35 -4.44 -5.37 -4.91 -5.07 -6.17 -4.22 -5.54
Institutional Indicator -1.07 -1.41 -0.38 -1.29 -1.15 -0.93 -2.12 -1.04 -0.27
-0.98 -2.57 -1.73 -2.13 -1.18 -1.70 -2.44 -1.79 -2.02
Number of Observations 80 80 80 80 80 80 80 80 80
R2 0.08 0.12 0.09 0.12 0.08 0.10 0.11 0.09 0.11
Note: The description of each variable and its source are explained in the Data Appendix. t –statistics are presented in the even rows using heteroskedastically-
consistent standard errors. The first row of the Table designates the institutional indicator used as independent variable.
Table XIV
Procedures for Tenant Eviction
Index Index Log of Index Index Index Legal Index Index Dispute
Defendant Mandatory Duration Professionals Written-Oral Justification Statutory Control of Resolution
Protection Time Limits -Laymen Regulation of Superior Index
Evidence Review
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Initial Income -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001
-3.57 -4.76 -4.08 -4.71 -4.60 -4.76 -6.21 -4.22 -5.28
Institutional Indicator -1.15 -1.02 0.18 -0.32 0.24 -0.10 -2.11 -0.62 -0.23
-1.22 -1.66 0.77 -0.63 0.24 -0.16 -2.45 -1.15 -1.83
Number of Observations 80 80 80 80 80 80 80 80 80
R2 0.08 0.09 0.07 0.07 0.06 0.06 0.11 0.07 0.08
Table XV
Firm Governance Indicators
Market
One Share Percentage Anti Director Mean 3 Largest Accounting
Creditor Rights Capitalization 10
One Vote General Meeting Rights Shareholders Standards
Largest
(1) (2) (3) (4) (5) (6) (7)
Initial Income -0.0001 -0.0002 -0.0002 -0.0001 -0.0003 -0.0003 -0.0002

-1.93 -2.82 -2.49 -2.03 -3.41 -3.31 -2.25
Institutional Indicator 0.57 -6.18 0.10 0.17 -4.57 0.05 -0.02

0.94 -2.29 0.68 1.12 -2.82 2.84 -0.08
Number of Observations 48 47 48 47 44 40 46
R2 0.11 0.18 0.09 0.11 0.27 0.31 0.10
Table XVI
Opening and Closing a Firm
Number of Time Cost Cost and Percentage

Procedures Time Bankruptcies
(1) (2) (3) (4) (5)

(6)
Initial Income -0.0001 -0.0001 0.0000 -0.0001 -0.0002
-1.15 -1.21 -0.26 -1.36 -1.78
Institutional Indicator -0.10 -0.02 -0.01 -0.61 0.66

-1.97 -2.04 -6.13 -1.95 0.06
Number of Observations 68 68 68 67 33
R2 0.06 0.07 0.03 0.09 0.12
Note: The description of each variable and its source are explained in the Data Appendix. t –statistics are presented in the even rows using
heteroskedastically-consistent standard errors. The first row of the Table designates the institutional indicator used as independent variable.
Table XVII
Financial System Indicators
Finance- Finance- Finance- Finance- Structure- Structure- Structure- Structure Net Private Total Non-
Activity Size Aggregate Dummy Activity Size Aggregate -Dummy Interest Credit Value Bank
Margin Traded Credit
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Initial Income -0.0003 -0.0002 -0.0003 -0.0002 -0.0002 -0.0001 -0.0001 -0.0001 -0.0001 -0.0002 -0.0002 -0.003
-3.39 -2.68 -3.13 -2.89 -2.14 -1.11 -1.50 -1.26 -1.93 -3.35 -2.35 -2.63
Institutional Indicator 0.36 1.11 0.95 1.71 0.35 0.02 0.31 0.35 -33.03 2.57 3.17 4.96
4.24 2.91 3.68 3.95 2.47 0.06 1.15 0.68 -2.72 3.36 1.75 2.74
Number of Observations 41 41 41 41 41 41 41 41 54 60 50 42
R2 0.43 0.36 0.42 0.39 0.20 0.04 0.09 0.05 0.18 0.22 0.16 0.19
Table XVIII
Bank Regulation
5 Bank Percentage Percentage Percentage of Bank Onsite Supervisors Explicit

Concentration Assets Foreign Top 10 Rated Supervisors Examination Employed by Deposit
Ratio Government Owned Internationally per per Bank Last Bank Industry Insurance
Owned Institution 5 Years
(1) (2) (3) (4) (5) (6) (7) (8)
Initial Income -0.0004 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.00004
-0.58 -5.32 -5.06 -6.79 -4.18 -5.55 -4.12 -2.01
Institutional Indicator -0.01 -0.01 -0.005 0.02 0.10 -0.02 -0.09 0.87
-1.01 -1.23 -0.25 4.20 1.30 -0.39 -0.30 2.75
Number of Observations 65 65 59 59 63 61 70 121

R2 0.02 0.09 0.08 0.33 0.10 0.08 0.07 0.06
Table XIX
Firm Finance Indicators
Operating Income Total Debt to Market Long Term Debt to Short Term Debt to Net Working Capital
Variability Value of Equity Market Value of Market Value of to Total Assets
Equity Equity
(1) (2) (3) (4) (5)
Initial Income -0.0002 -0.0002 -0.0003 -0.0002 -0.0002

-4.07 -2.92 -3.87 -2.54 -3.31
Institutional Indicator -0.85 1.24 3.21 1.67 0.45

-1.43 3.20 1.41 3.56 0.13
Number of Observations 43 44 44 44 44
R2 0.26 0.29 0.23 0.30 0.19
A summary of our empirical results could run as follows. Institutions matter for growth
and their impact is quantitatively important. As far as the legal system is concerned, both
the general performance indicators and very specific procedural characteristics on check
collection are important. Aggregate indicators of corporate governance do not seem to
relate strongly to growth but faster and cheaper procedures to open a business firm do.
Finally, as far as the financial system is concerned, the existence of wide access to capital
is key, whether that is achieved through the banking system or the stock market is
irrelevant. Most of the disaggregated indicators related to the financial system are not
important.
5. Strategies for Institutional Reform in Portugal
“It must be remembered that there is nothing more difficult to plan, more doubtful of
success, nor more dangerous to manage, than the creation of a new system. For the
initiator has the enmity of all who would profit by the preservation of the old institutions
and merely lukewarm defenders in those who would gain by the new ones.”
Machiavelli
This paper now addresses the different options for institutional reform in Portugal. The
objective is to evaluate the benefits of each possible reform in light of the empirical results
from the previous section and the state of development of Portuguese institutions as
compared to EU institutions. The methodology we adopt is simple and suggestive. For each
institutional indicator – whether in the legal, governance or finance areas - we construct
three reform related indices: the Impact on Growth, the Required Reform Effort and the
Efficiency of the Reform Effort indices. The Impact on Growth index is measured as the
product of the regression coefficient by the difference between the institutional index i in
Portugal and in the EU as a whole:
Impact on Growth i =
[Institution i, Portugal – Institution i , EU Average] * Regression Coefficient i
This index measures the yearly increase in per capita growth – for the period 1960-1995
- that would likely result from an institutional reform that elevates Portugal to the European
Union level in the specific institutional area assessed. Evidently, a higher Impact on Growth
index indicates a more promising reform and so suggests these reforms should be
undertaken.
A shortcoming of the Impact on Growth index is that it abstracts completely from the
“cost of reform”, i.e., it ignores whether bringing Portugal to the EU level is more or less
hard for each specific issue considered. We thus find useful to compute a second index, the
Required Reform Effort index. It tries to compute the “cost of reform” by using the relative
position of Portugal and the EU for each institution divided by the Portuguese level. It is
based on the values of each institutional index i for Portugal and for the European Union
average:76
76
It is defined for the cases where Portuguese institutional development lags behind European levels and
computed always so that a higher index represents a higher level of institutional development.
Required Reform Effort i =

Absolute [Institution i , EU Average - Institution i , Portugal ] /
Absolute [Institution i , Portugal]
Thus, this index measures the absolute institutional change that Portugal needs to
undertake to achieve the EU average level, relative to the current absolute Portuguese
index. It is always positive and is measured as the percentage change required.77 It is a
measure of the “cost of reform” and a higher value indicates a higher required effort to
reform the specific institution. In contrast with the Impact on Growth index, the Required
Reform Effort index now totally ignores the impact on growth of the specific reform.
We realize that several different factors affect the cost of reform. These factors differ
across specific reforms, they are subject to discussion and, last but not least, extremely
difficult to evaluate quantitatively. The objective aof the index above is thus to provide an
expedite assessment of the necessary reform effort and does not intend to close the issue of
evaluating resistance to specific reforms.
Our third and last index is the Efficiency of Reform index. It combines the previous two
indices by diving the impact on growth by the required reform effort, providing a measure
of the yearly growth increase per unit of reform effort. It is computed for a move of
Portugal to the EU average level for an institution i:
Efficiency of Reform i =
Impact on Growth i / Required Reform Effort i
A higher index on Efficiency of Reform denotes a higher percent increase in economic

growth per unit of reform effort, in simplistic terms a higher “bang for the reform buck”.
The value of 1 denotes an increase of 1 percent in yearly economic growth for each reform
effort of 100 percent, that is, an effort that overcomes a Portuguese institutional index that
is half the EU average index.
Tables XX, XXI and XXII present an evaluation of the reform potential for different
legal, corporate governance and financial reform indicators. The first two columns present
the value of the index for Portugal and for the European Union average, the third column
the difference between the previous two and the fourth the regression coefficient from
Tables in the previous section. The Impact on Growth, the Required Reform Effort and the
Efficiency of the Reform Effort indices are presented in the last three columns of each
table. The indices for which the regression coefficients are statistically significant are noted
with an asterisk and the whole row of data noted in bold. As can be easily understood,
different reform issues correspond to widely different impacts on yearly growth, different
reform efforts and, importantly, very different reform efficiency.
77
In the few cases where the institutional index takes the value 0 for Portugal the denominator becomes the EU
average rather than the Portuguese level so that an index of 1 results, The interpretation remains clear: the
“maximum” reform effort is required for Portugal to achieve EU levels in this issue.
Table XX
Law and Reform
Impact on Growth, Required Reform Effort and Efficiency of Reform
(1) (2) (3) (4) (5) (6) (7)
Index (2)-(1) (4)*(3) (3)/(1) (5)/(6)

European Impact on Required Efficiency
Portugal Difference Coefficient
Union Growth Effort of Reform
Rule of Law** 8.68 8.97 0.29 0.58 0.17 0.03 5.03

Corruption** 7.38 8.65 1.27 0.48 0.61 0.17 3.54
Risk of Expropriation** 8.90 9.42 0.52 1.03 0.53 0.06 9.17
Contract Repudiation** 8.57 9.06 0.49 0.89 0.43 0.06 7.63
Judicial Efficiency Index 5.50 8.56 3.06 0.16 0.49 0.56 0.88
Citizen's Access to Justice** 7.50 7.88 0.38 0.21 0.08 0.05 1.58
Index Defendant Protection 0.20 0.47 0.27 -1.07 -0.29 1.36 -0.21
Index Mandatory Time Limits** 0.80 0.27 -0.53 -1.41 0.75 0.66 1.13
Log of Duration** 5.80 5.30 -0.50 -0.38 0.19 0.09 2.20
Index Professionals-Laymen** 1.00 0.64 -0.36 -1.29 0.46 0.36 1.29
Index Written-Oral 0.75 0.78 0.03 -1.15 -0.04 0.04 -0.86
Index Legal Justification** 1.00 0.69 -0.31 -1.93 0.60 0.31 1.93
Index Statutory Regulation of 0.38 0.25 -0.13 -2.12 0.27 0.33 0.81
Evidence**
Index Control of Superior 1.00 0.72 -0.28 -1.04 0.30 0.28 1.04
Review**
Dispute Resolution Index** 5.13 3.89 -1.24 -0.27 0.33 0.24 1.39
Index Defendant Protection 0.50 0.62 0.12 -1.15 -0.14 0.24 -0.58
Index Mandatory Time Limits 0.80 0.29 -0.51 -1.02 0.52 0.64 0.82
Log of Duration 6.04 5.21 -0.83 0.18 -0.15 0.14 -1.09
Index Professionals-Laymen 0.67 0.57 -0.10 -0.32 0.03 0.15 0.21
Index Written-Oral 0.75 0.76 0.01 0.24 0.00 0.01 0.18
Index Legal Justification 1.00 0.67 -0.33 -0.10 0.03 0.33 0.10
Index Statutory Regulation of 0.50 0.27 -0.23 -2.11 0.48 0.46 1.06
Evidence**
Index Control of Superior Review 0.67 0.62 -0.05 -0.62 0.03 0.07 0.42
Dispute Resolution Index** 4.58 3.72 -0.87 -0.23 0.20 0.19 1.05
Table XXI
Corporate Governance and Reform
(1) (2) (3) (4) (5) (6) (7)
Index (2)-(1) (4)*(3) (3)/(1) (5)/(6)

European Impact on Required Efficiency
Portugal Difference Coefficient
Union Growth Effort of Reform
One Share One Vote 0.00 0.07 0.07 0.57 0.04 1.00 0.04
Percentage General Meeting** 0.05 0.10 0.05 -6.18 -0.29 0.93 -0.31
Anti Director Rights 3.00 2.50 -0.50 0.10 -0.05 0.17 -0.30
Creditor Rights 1.00 1.93 0.93 0.17 0.16 0.93 0.17
Mean 3 Largest 0.52 0.45 -0.07 -4.57 0.32 0.14 2.38
Shareholders**
Accounting Standards** 36.00 63.62 27.62 0.05 1.38 0.77 1.80
Market Capitalization 10 0.03 0.29 0.26 -0.02 -0.01 8.51 0.00
Largest**
Number of Procedures** 12.00 9.00 -3.00 -0.10 0.30 0.25 1.20
Time** 76.00 36.57 -39.43 -0.02 0.79 0.52 1.52
Cost** 0.18 0.16 -0.02 -0.01 0.00 0.10 0.00
Cost and Time** 0.49 0.31 -0.18 -0.61 0.11 0.37 0.30
Percentage Bankruptcies 0.00 0.02 0.02 0.66 0.01 1.00 0.01
Table XXII
Finance and Reform
(1) (2) (3) (4) (5) (6) (7)

Index (2)-(1) (4)*(3) (3)/(1) (5)/(6)
Portugal European Difference Coefficient Impact on Required Efficiency of
Union Growth Effort Reform
Finance-Activity** 4.23 5.48 1.25 0.36 0.45 0.29 1.52

Finance-Size** 4.31 4.41 0.10 1.11 0.11 0.02 4.78
Finance-Aggregate** 0.12 0.43 0.31 0.95 0.30 2.59 0.11
Finance-Dummy** 1.00 0.69 -0.31 1.71 -0.53 0.31 -1.71
Structure-Activity** -4.26 -2.65 1.61 0.35 0.56 -0.38 -1.49
Structure-Size -2.66 -1.30 1.36 0.02 0.03 -0.51 -0.05
Structure-Aggregate -1.49 -0.13 1.36 0.31 0.42 -0.91 -0.46
Structure-Dummy 0.00 0.46 0.46 0.35 0.16 1.00 0.16
Net Interest Margin** 0.04 0.03 -0.01 -33.03 0.38 0.29 1.32
Private Credit** 0.63 0.71 0.08 2.57 0.20 0.12 1.62
Total Value Traded** 0.02 0.10 0.08 3.17 0.25 3.96 0.06
Operating Income Variability 0.87 0.69 -0.18 -0.85 0.15 0.21 0.74
Total Debt to Market Value of Equity** 0.59 0.39 -0.21 1.24 -0.25 0.35 -0.73
Long Term Debt to Market Value of Equity 0.22 0.16 -0.06 3.21 -0.19 0.27 -0.71
Short Term Debt to Market Value of Equity** 0.37 0.23 -0.15 1.67 -0.24 0.39 -0.62
Net Working Capital to Total Assets** 0.05 0.17 0.12 0.45 0.05 2.37 0.02
5 Bank Concentration Ratio 81.70 59.19 -22.51 0.45 -10.13 0.28 -36.77
Percentage Assets Government Owned 20.80 10.81 -9.99 -0.01 0.10 0.48 0.21
Percentage Foreign Owned 11.70 16.29 4.59 -0.01 -0.05 0.39 -0.12
Percentage of Top 10 Rated Internationally** 100.00 66.15 -33.85 -0.01 0.17 0.34 0.50
Bank Supervisors per Institution 2.40 0.78 -1.62 0.02 -0.03 0.67 -0.05
Onsite Examination per Bank Last 5 Years 2.50 2.43 -0.07 0.10 -0.01 0.03 -0.25
Supervisors Employed by Bank Industry 2.00 2.07 0.07 -0.02 0.00 0.04 -0.04
Explicit Deposit Insurance** 1.00 1.00 0.00 0.87 0.00 0.00 0.00
Figures V, VI and VII present the ten “most promising” reform areas, as suggested by
each of the three criteria computed above. In other words, we present in succession the ten
issues for which raising the Portuguese level to European Union levels raises growth levels
the most, the ten for which the required effort is the lowest and the ten for which the
efficiency of reform is highest. As an illustration, note that there may be four reasons why a
specific institutional reform is not recommended for Portugal on the basis of the Efficiency
of the Reform Effort. First, if Portugal is at a level of institutional development in a specific
area that is above the average level for the EU - that is, Portugal has better institutions than
the EU in that area - we automatically consider it a non-prioritary area of reform. Second,
any institutional indicator that was not associated with economic growth in a statistically
significant way is not a candidate for recommended reform. Third, when there is a
statistically significant impact of reform on growth, the economic impact may be too small
to make the reform effort promising. Finally, even when a given institution is significantly
related to growth – statistically and economically – if Portugal is very far from the EU
average level, this reform area is considered to require too much of a “reforming effort”.78
In Figure V we find that the two most growth enhancing reform issues are in the
corporate governance area. However, six out of the nine most promising reforms in terms
of induced growth relate to the legal system. Again, as far as the required reform effort is
concerned, five out of the six reform items requiring the least effort from Portuguese
institutions are in the legal area. The indices computing the growth impact and the required
reform effort show that two issues each of the finance and corporate governance areas are
within the ten most promising reforms. Interestingly, the finance and corporate governance
reform items that have the most growth impact and the least required effort are aggregate
and disaggregated level measures respectively. In contrast, in the legal area both aggregate
and disaggregated institutional indicators are classified within the ten most promising.
Figure VII presents the efficiency of the Efficiency of the Reform Effort index. The
three most efficient reform indices relate to legal issues, namely the risk of expropriation,
the risk of contract repudiation and the rule of law indices. All are aggregate measures of
the efficacy of the legal system. Three other legal indices are among the ten most efficient
reforms to undertake: decreasing the level of corruption to European Union levels is the
fifth most efficient; improving the Index of Legal Justification and the Logarithm of
duration of the check collection procedure are the seventh and eighth most efficient.
Increasing the depth of the financial system and decreasing the ownership concentration are
also efficient reforms. A gauge of the actual impact of reform is revealing: the five most
efficient reforms deliver from 0.3 to 0.9 percent extra growth per year for each 10 percent
increase in the institution index towards the European Union level. For a relatively small
effort in reform, Portugal can obtain important gains in terms of growth. These results give
us ample reason to believe that institutional reform is an important instrument to foster
Portuguese economic growth.
78
Note that the reform may be excluded from the most promising as to Impact of Growth if reasons one through
three are present and excluded from the Required Reform Effort if reasons one, two and four are present.
Looking at the results more finely, one concludes that the legal system is the most
promising area for reform, both in its general workings – measured by aggregate indices –
or in the specific procedures. Reform in the corporate governance and financial areas is also
important but while overall change in the workings of the financial system is most
promising, it is the specific issues related to firm governance reform that come out as the
most promising.
F ig u r e V
G ro w th Im p a ct o f R efo rm
F in a n c e -A c tiv ity
In d e x P r o f e s s io n a ls - L a y m e n
In d e x S t a tu to r y R e g u la tio n o f
E v id e n c e E v ic tio n
R is k o f E x p ro p ria tio n
S tru c tu re -A c tiv ity
In d e x L e g a l J u s tif ic a tio n
C o rru p tio n
In d e x M a n d a to r y T im e L im its
T im e
A c c o u n tin g S ta n d a rd s
0 .0 0 0 .2 0 0 .4 0 0 .6 0 0 .8 0 1 .0 0 1 .2 0 1 .4 0 1 .6 0
P e rc e n t G ro w th p e r y e a r
L aw G o v ern a n ce F in a n c e
F ig u r e V I
R e q u ir e d R e f o r m E f fo r t
C o rru p tio n
M e a n 3 L a r g e s t S h a r e h o ld e rs
P riv a te C re d it
C ost
L o g o f D u ra tio n
C o n tra c t R e p u d ia tio n
C i t i z e n 's A c c e s s t o J u s t i c e
R u le o f L a w
F in a n c e -S iz e
0 .0 0 0 .0 2 0 .0 4 0 .0 6 0 .0 8 0 .1 0 0 .1 2 0 .1 4 0 .1 6 0 .1 8 0 .2 0
P ercen t
Law G o v e rn a n c e F in a n c e
F ig u r e V I I
E f f ic ie n c y o f R e f o r m
P riv a te C re d it
A c c o u n tin g S ta n d a rd s
In d e x L e g a l J u s tific a tio n
L o g o f D u ra tio n
M e a n 3 L a r g e s t S h a r e h o ld e rs
C o rru p tio n
F in a n c e -S iz e
R u le o f L a w
C o n tra c t R e p u d ia tio n
0 .0 0 1 .0 0 2 .0 0 3 .0 0 4 .0 0 5 .0 0 6 .0 0 7 .0 0 8 .0 0 9 .0 0 1 0 .0 0
P ercen t G ro w th p er E ffo rt
Law G o v ern a n c e F in a n c e
6. Conclusions
This paper evaluates Portuguese institutional development in light of the experience of

Spain, Greece, the European Union and the high-growth East Asian economies. Three areas
are scrutinized, the legal, corporate governance and financial systems. We assess the
benefits of reforming Portuguese institutions by elevating them to EU levels, along three
dimensions: the impact on growth, the required reform effort and the efficiency of the
reform effort. Our analysis shows that Portugal is generally below European Union and
East Asian levels of institutional development in all areas – legal, governance or financial.
While on some issues Portuguese institutions are more developed than their Spanish
counterparts, there are as much instances where the reverse is true. As compared to Greece,
Portugal benefits from more developed institutions.
As to institutional reform, we find that the legal area is the most promising, as
improvement in different legal issues delivers a large increase in growth per unit of effort.
This is true for very aggregate as well as disaggregated indices of legal development. More
than half of the ten institutional indices that best foster growth and that require the least
reform effort are legal indices. It is true that indices from the corporate governance and
financial areas are also present among the most effective and least costly reforms to
undertake. Our conclusion is that comprehensive reform encompassing all three areas is
necessary. The high growth impact per reform effort required suggests that institutional
reform is sufficient to bring Portuguese economic growth to substantially higher levels.
Portugal has recently experienced a slowdown in economic growth that hinders the
convergence process vis-à-vis richer countries in the European Union. The fact that
institutions change very infrequently and very little strongly suggests that the low level of
development of Portuguese institutions severely constrains the growth rate of the economy.
A comprehensive and sustained reform effort may be the missing condition for Portuguese
growth to resume at higher rates and decisively reduce the income gap with other European
Union countries.
References
Aghion, P., and Howitt, P. (1998), “Endogenous Growth Theory”, Cambridge, MIT Press.
Barro, R. (1999), “The Determinants of Democracy”, The Journal of Political Economy,
107, S6, pp. S158 - S229..
Barro, R., and Sala-i-Martin, X. (1995), “Economic Growth”, MIT Press.
Barth, James, Gerard Caprio and Ross Levine (2001a), “Bank Regulation and Supervision:
What Works Best?”, Mimeo, World Bank.
Barth, James, Gerard Caprio and Ross Levine (2001b), “The Regulation and Supervision of
Banks Around the World – A New Database”, Mimeo, World Bank.
Barth, James, Gerard Caprio and Ross Levine (2001c), “Banking Systems Around the
Globe: Do Regulation and Ownership Affect Performance and Stability?,” Mimeo,
World Bank.
Barth, James, Gerard Caprio and Ross Levine (2001), “Financial Regulation and
Performance: Cross-Country Evidence”, Mimeo, World Bank Working Paper No.
2037.
Beck, Thorsten, Mattias Lundberg and Giovanni Majnoni (2001), “Financial Intermediary
Development and Growth Volatility: Do Intermediaries Dampen or Magnify
Shocks?” Mimeo, World Bank
Beck, Thorsten and Ross Levine (2001), “New Firm Formation and Industry Growth: Does
Having a Market- or Bank-Based System Matter?” World Bank Working Paper
Beck, Thorsten, Asli Demirgüç-Kunt and Ross Levine (2001), “Law, Politics, and
Finance”, World Bank Working Paper.
Berle, Adolf and Gardiner Means (1932), “The Modern Corporation and Private Property”,
New York: Macmillan.
Bhagat, Sanjai and Bernard Black (1999), “The Uncertain Relationship Between Board
Composition and Firm Performance”, Stanford Law School, John M. Olin
Program in Law and Economics Working Paper 175
Black, Bernard (2000), “The Core Institutions that Support Strong Securities Markets”,
Stanford Law School, John M. Olin Program in Law and Economics Working
Paper No. 200, Forthcoming, 55 Business Lawyer
Bossone, Biagio (2001), “The Role of Trust in Financial Sector Development”, Mimeo,
World Bank.
Bourguignon, François (1999), “Crime, Violence and Inequitable Development”, World
Bank Annual Conference on Development Economics, Washington.
Buscaglia, Edgardo, “Law and Economics of Development” Article 0580, Encyclopedia of
the Law
Calderón, César, Alberto Chong and Arturo Galindo (2001), “Structure and Development
of Financial Institutions and Links with Trust: Cross-Country Evidence”, Inter-
American Development Bank, Research Department Working Paper #444
Stijn Claessens, Simeon Djankov, Tatiana Nenova (2001), “Corporate Risk around the
World”, World Bank Working Paper
Coase, Ronald (1937), “The Nature of the Firm", Economica, November 1937, 4. Reprinted
in Coase, Ronald, (1988) “The Firm, the Market and the Law”, University of
Chicago Press, Chicago.
Coase, Ronald (1960), “The Problem of Social Cost", The Journal of Law and Economics,
Oct. 1960, 3, 1-44. Reprinted in Coase, Ronald, (1988) “The Firm, the Market and
the Law”, University of Chicago Press, Chicago.
Coase, Ronald, (1988) “The Firm, the Market and the Law”, University of Chicago Press,
Chicago.
Coffee, Kack (1999), “Privatization and Corporate Governance: The Lessons from
Securities Market Failure”, Working Paper No. 158, The Center for Law and
Economic Studies, Columbia Law School
Cull, Robert, Lemma Senbet, and Marco Sorge (2001) “Deposit Insurance and Financial
Development”, Working Paper 2682, World Bank.
Davis, Kevin and Michael J. Trebilcock, (1999), “What Role Do Legal Institutions Play in
Development?”, Prepared for the International Monetary Fund’s Conference on
Second Generation Reforms, November 8–9, 1999, Mimeo, Faculty of Law,
University of Toronto,
Demirgüç-Kunt, Asli and Detragiache, Enrica. (2000), “The Determinants of Banking
Crises: Evidence from Industrial and Developing Countries,” Policy Research
Working Paper 1828, World Bank.
Demirgüç-Kunt, Asli and Maksimovic, Vojislav (1996), "Financial Constraints, Uses of
Funds, and Firm Growth: An International Comparison," Mimeo, World Bank.
Demirgüç-Kunt, Asli and Maksimovic, Vojislav (2000), “Funding Growth in Bank-Based
and Market-Based Financial Systems: Evidence from Firm Level Data”, Working
Paper 2432, World Bank.
Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer,
(2000), “The Regulation of Entry,” NBER Working Paper #7892.
Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer,
(2001), “Legal Structure and Judicial Efficiency: the Lex Mundi Project”, World
Bank Working Paper.
David Dollar and Aart Kraay (2000), “Property Rights, Political Rights, and the
Development of Poor Countries in the Post-Colonial Period”, Development
Research Group, World Bank
Fajnzylber, Pablo, Daniel Lederman and Norman Loayza (1998), “Determinants of Crime
Rates in Latin America and the World, An Empirical Assessment”, Mimeo, World
Bank.
Foss, Nicolai, Henrik Lando and Steen Thomsen (1999), “The Theory of the Firm”,
Encyclopedia of Law
Garcia, Valeriano and Lin Liu (1999), “Macroeconomics Dteterminants of Stock Market
Development”, Journal of Applied Economics, Vol. II, No. 1 (May 1999), 29-59
Gerschenkron, Alexander (1962), “Economic Backwardness in Historical Perspective,”
Belknap Press of Harvard University Press, Cambridge, Massachusetts.
Gilson, Ronald (2000), “Globalizing Corporate Governance: Convergence of Form or
Function”, Stanford Law School, John M. Olin Program in Law and Economics
Working Paper No. 192
Glaeser, E., and Shleifer, A. (2001), “Legal Origins”, Mimeo, Third Draft, Department of
Economics, Harvard University.
Glaeser, E., and Shleifer, A. (2000), “On the Design of a Legal System”, Department of
Economics, Harvard University, Preliminary Draft for presentation at U. of C.,
May 2000
Greif, Aver (1996), “Contracting, Enforcement and Efficiency: Economics Beyond the
Law” in Michael Bruon and Boris Pleskovic (eds.) Annual World Bank
Conference on Development Economics 1996, World Bank.
Gropp, Reint and Jukka Vesala (2000), “Deposit Insurance and Moral Hazard: Does the
Counterfactual Matter?”, European Central Bank.
Guiso, Luigi, Paola Sapienza and Luigi Zingales (2000), “The Role of Social Capital in
Financial Development”, NBER Working Paper No. W7563.
Jayaratne, Jith and Strahan, Philip (1996), "The Finance-Growth nexus: Evidence from
Bank Branch Deregulation," Quarterly Journal of Economics, 111(3), pp. 639-670.
Johnson, Simon and Andrei Shleifer (2001), “Privatization and Corporate Governance”,
Mimeo, Harvard University, Prepared for the 12 th Annual East Asian Seminar on
Economics, June 28-30, 2001.
Klapper, Leora (2001), “Bankruptcy Around the World: Explanations of its Relative Use”,
Mimeo, World Bank
Knack, Stephen. and Keefer, Phillip (1997), "Does Social Capital Have an Economic
Payoff?: a Cross-country Investigation", The Quarterly Journal of Economics,
112(4): 1251.
Knight, Frank (1933), “Risk, Uncertainty and Profit”, London, London School of
Economics and Political Science.
Laeven, Luc (2000), “Financial Liberalization and Financing Constraints: Evidence from
Panel Data on Emerging Economies”, Working Paper 2467, World Bank.
Lora, Eduardo, Patricia Cortés and Ana María Herrera (2001), “Los obstáculos al desarrollo
empresarial y el tamaño de las firmas en América Latina”, Banco Interamericano
de Desarrollo Departamento de Investigación Research Department Working
Paper #447
La Porta, Rafael, Florencio Lopez-de-Silanes and Andrei Shleifer (2002), “Government
Ownership of Banks”, forthcoming, Journal of Finance, 2002.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny,
(1999), “The Quality of Government,” Journal of Law, Economics, and
Organization 15, 222-279.
La Porta, Rafael, Florencio Lopez-de-Silanes and Andrei Shleifer (1998b), “Corporate
Ownership Around the World”, Mimeo, Harvard University.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny,
(1998), “Law and Finance,” Journal of Political Economy 106, 1113-1155.
La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert W. Vishny,
(1997), “Legal Determinants of External Finance,” Journal of Finance 52, 1131-
1150.
Leahy, Michael, Sebastian Schich, Gert Wehinger, Florian Pelgrin and Thorsteinn
Thorgeirsson (2001), “Contributions of Financial Systems to Growth in OECD
Countries”, OECD Economics Department Working Papers No. 280.
Levine, Ross and Sara Zervos, (1998) “Stock Markets, Banks and Economic Growth.”
American Economic Review, 88 (June): 537-58.
Levine, Ross (1999), “Law, Finance, and Economic Growth”, Journal of Financial
Intermediation 8, 8-35.
Levine, Ross, Norman Loayza, and Thorsten Beck (1999), “Financial Intermediation and
Growth: Causality and Causes”, Journal of Monetary Economics, v46, n1: 31-77.
Love, Inessa (2001), “Financial Development and Financing Constraints: International

Evidence from the Structural Investment Model”, Mimeo, World Bank
Mahoney, Paul (2000), “The Common Law and Economic Growth: Hayek Might be
Right”, University of Virginia School of Law, Legal Studies Working Papers
Series, Working Paper 00-8.
Messick, Richard (1999), “Judicial Reform and Economic Development: A Survey of the
Issues, The World Bank Research Observer, vol. 14, no. 1, p. 117–36
Merryman, John Henry (1985), “The Civil Law Tradition: An Introduction to the Legal
Systems of Western Europe and Latin America”, Stanford, California, Stanford
University Press.
Merryman, John H. (1977), “Comparative Law and Social Change: On the Origins, Style,
Decline, and Revival of the Law and Development Western Europe and Latin
America. Stanford, CA, Stanford University Press.
North, Douglass (1981) “Structure and Change in Economic History”, New York, Norton.
North, Douglass (1990) “Institutions, Institutional Change and Economic Performance”,
Cambridge,U.K.: Cambridge University Press.
OECD Secretariat (2000), “Corporate Governance in OECD Member Countries: Recent
Developments and Trends”, OECD Paris.
Olson, Mancur (1982), “The Rise and Decline of Nations: Economic Growth, Stagflation,
and Social Rigidities”, New Haven, Yale University Press.
Persson, Torsten, Gerard Roland and Guido Tabellini (1999), “Comparative Politics and
Public Finance”, Mimeo, IGIER, Universitá Bocconi, Milan.
Pigou, Arthur C., The Economics of Welfare, 4th ed., (London: Macmillan and Co., 1938.)
Pistor, Katharina, Yoram Keinan, Jan Kleinheisterkamp and Mark West (2000), “The
Evolution of Corporate Law”, Mimeo, Kennedy School of Government, Harvard
University.
Pagano, Marco and Paolo Volpin (2001), “The Political Economy of Corporate
Governance”, Mimeo, London Business School.
Posner, Richard (1998), “Creating a Legal Framework for EconomicDevelopment”, The
World Bank Research Observer, vol. 13, no. 1, pp. 1–11.
Posner, Richard (1972), “Economic Analysis of Law”, Boston, Massachusetts, Little,
Brown and Company.
Rajan, Raghuram G. and Zingales, Luigi (1998), "Financial Dependence and Growth,"
American Economic Review, vol 88, pp 559-586
Rajan, Raghuram G. and Zingales, Luigi (1999), “Financial Systems, Industrial Structure,
and Growth,” Mimeo, University of Chicago.
Sapienza, Paola (2002), “The Effects of Banking Mergers on Loan Contracts”, Journal of
Finance, 57: 329-368..
Schumpeter, Joseph (1934) “The Theory of Economic Development,” Cambridge,
Massachusetts, Harvard University Press.
Shleifer, A. and Vishny, R. (1997), “A Survey of Corporate Governance”. Journal of
Finance, June, 1997.
Shleifer, Andrei and Robert Vishny (1998), “The Grabbing Hand, Harvard University
Press.
Shleifer, Andrei (1998), “State versus Private Ownership”, Journal of Economic
Perspectives, Fall.
Smith, Adam. (1755) “Unpublished manuscript quoted in Dugald Stewart, “Account of the
Life and Writings of Adam Smith.” In Adam Smith, Essays on Philosophical
Subjects. Edited by W. P. D. Wightman and J. C. Bryce. Oxford: Clarendon Press,
1980.
Solow (1956),"A Contribution to the Theory of Economic Growth", Quarterly Journal of
Economics
Stigler, George J., “The Theory of Economic Regulation,” Bell Journal of Economics and
Management Science, II (1971), 3-21.
Stulz, René (2000), “Does Financial Structure Matter for Economic Growth? A Corporate
Finance Perspective”, Mimeo, World Bank.
Tavares, J., and Wacziarg, Romain (2001), “How Democracy Affects Growth”, European
Tullock, Gordon (1997), “The Case Against the Common Law”, Durham, North Carolina,
Carolina Academic Press.
Weingast, Barry (1993), “Constitutions as Governance Structures: The Political
Foundations of Secure Markets”, Journal of Institutional and Theoretical
Economics 149, 286-311.
Williamson, Oliver (1995), “The Institutions and Governance of Economic Development
and Reform.” In Michael Bruno and Boris Pleskovic, eds., Proceedings of the
Annual World Bank Conference on Development Economics 1994. Washington,
D.C.: World Bank.
Wurgler, Jeffrey (1999), “Financial Markets and The Allocation Of Capital”, Yale ICF
Working Paper No. 99-08, Yale School of Management.
Data Appendix
Rule of Law
Source: Computed from Data in La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W.
Vishny, (1998), “Law and Finance”
Description: Assessment of the law and order tradition in the country produced by the country-risk rating agency
International Country Risk (ICR). Average of the months of April and October of the monthly index between 1982
and 1995. Scale from 0 to 10, with lower scores for less tradition for law and order. The scale was modified from
its original range of 0 to 6.
Corruption
Description: ICR’s assessment of corruption in government. Lower scores indicate “that high government
officials are likely to demand special payments” and “illegal payments are generally accepted throughout lower
levels of government” in the form of “bribes connected with import and export licences, exchange controls, tax
assessment, policy protection, or loans.” Average of the months of April and October of the monthly index
between 1982 and 1995. Scale form 0 to 10, with lower scores for higher levels of corruption (we changed the
scale form the original range going from 0 to 6).
Risk of Expropriation
Description: ICR’s assessment of the risk of “outright confiscation” or “forced nationalization”. Average of the
months of April and October of the monthly index between 1982 and 1995. Scale from 0 to 10, with lower scores
for higher risks.
Contract Repudiation
Description: ICR’s assessment of the “risk of a modification in a contract taking the form of a repudiation,
postponement, or scaling down” due to “budget cutbacks, indigenisation pressure, a change in government, or a
change in government economic and social priorities.” Average of the months of April and October of the monthly
index between 1982 and 1995. Scale from 0 to 10, with lower scores for higher risks.
Enforceability of Contracts
Source: Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer, (2001), “Legal
Structure and Judicial Efficiency: the Lex Mundi Project”
Description: Assessment of the “efficiency and integrity of the legal environment as it affects business,
particularly foreign firms” produced by the country risk rating agency International Country Risk (ICR). It may be
“taken to represent investors’ assessment of conditions in the country in question.” Average between 1980 and
1983. Scale from 0 to 10, with lower scores representing lower efficiency levels.
Efficiency of the Judicial System

Description: Assessment of the “efficiency and integrity of the legal environment as it affects business,
particularly foreign firms” produced by the country risk rating agency International Country Risk (ICR). It may be
“taken to represent investors’ assessment of conditions in the country in question.” Average between 1980 and
1983. Scale from 0 to 10, with lower scores representing lower efficiency levels.
Citizen's Access to Justice

Description: Assessment of the extent to which citizens are “equal under the law, do they have access to an
independent, non-discriminatory judiciary, and are they respected by the security forces”. Scale from 0 to 10. The
higher the rating, the greater the degree of equality under the law.
Index Defendant Protection

Description: Measures the protection granted by the procedural law to the tenant and his/her family, by reducing
the disparity in means and legal representation between landlord and tenant. The index ranges from 0 to 1, where
higher values mean a higher level of defendant protection, while 0 means a lower level. For the Eviction case the
index is formed by the normalized sum of the following variables : (i) mandatory legal aid by law or by court or
administrative order, (ii) attorney fees are fixed or limited by statute, court or administrative regulation, (iii) judge
has the independent legal obligation to investigate facts, (iv) tenant’s economic situation is considered at
judgment, and (v) tenant’s economic situation considered at enforcement. For the Check case the index is formed
by the normalized sum of the following variables: (i) mandatory legal aid by law or by court or administrative
order, (ii) attorney fees are fixed or limited by statute, court or administrative regulation, (iii) judge has the
independent legal obligation to investigate facts, (iv) attachment of defendant’s property only after judgment, (v)
transfer of defendant’s property only through public auction, and (vi) mandatory exclusion of defendant’s essential
survival assets.
Index Mandatory Time Limits

Description: The index measures the presence of mandatory time limits in the procedure. The index is calculated
as the average of the following variables: (i) term for admission, (ii) term to present evidence, (iii) term to present
defence, (iv) term for judgment, (v) term for compliance, (vi) term for notification of judgment. The index ranges
from 0 to 1, where higher values mean more mandatory deadlines. Available for two legal cases: tenant eviction
and collection of a bounced check.
Administrative Procedure
Description: The variable equals one if there is a generally available procedure for eviction or check collection
before an administrative officer, which may be used as a substitute to the judicial procedure, and which does not
imply any judicial involvement (such as issuance of warrants) or the participation of a housing or debt-collection
tribunal. The variable equals zero otherwise.
Index Professionals-Laymen
Description: The index measures whether the resolution of the case provided would rely mostly in the
intervention of professional judges and attorneys, as opposed to the intervention of other types of adjudicators and
lay people. The index is formed by the normalized sum of the following variables: (i) general jurisdiction court,
(ii) professional vs. non-professional judge, and (iii) legal representation is mandatory. The index ranges from 0 to
1, where higher values mean a higher intervention of professionals. Available for two legal cases: tenant eviction
and collection of a bounced check.
Index Written-Oral
Description: The index measures the written or oral nature of the actions involved in the procedure, from the
filing of the complaint, until the actual enforcement. The index is calculated as the number of stages carried out
mostly in a written form over the total number of applicable stages, and it ranges from 0 to 1, where higher values
mean higher prevalence of written elements. Available for two legal cases: tenant eviction and collection of a
bounced check.
Index Legal Justification

Description: The index measures the level of legal justification required in the process. The index is formed by
the normalized sum of the following variables : (i) complaint must be legally justified, (ii) judgment must be
legally justified, and (iii) judgment must be on law (not on equity). The index ranges from 0 to 1, where higher
values mean a higher use of legal language or justification. Available for two legal cases: tenant eviction and
collection of a bounced check.
Index Statutory Regulation of Evidence

Description: The index measures the level of statutory control or intervention of the administration, admissibility,
evaluation and recording of evidence. The index is formed by the normalized sum of the following variables : (i)
judge can not introduce evidence, (ii) judge can not reject irrelevant evidence, (iii) out-of-court statements are
inadmissible, (iv) mandatory pre-qualification of questions, (v) oral interrogation only by judge, (VI) only original
documents and certified copies are admissible, (vii) authenticity and weight of evidence defined by law, and (viii)
mandatory recording of evidence. The index ranges from 0 to 1, where higher values mean a higher statutory
control or intervention. Available for two legal cases: tenant eviction and collection of a bounced check.
Index Control of Superior Review

Description: The index measures the level of control or intervention of the appellate court’s review of the first-
instance judgment. The index is formed by the normalized sum of the following variables : (i) enforcement of
judgment is automatically suspended until resolution of appeal, (ii) comprehensive review in appeal, and (iii)
interlocutory appeals are allowed. The index ranges from 0 to 1, where higher values mean a higher control or
intervention. Available for two legal cases: tenant eviction and collection of a bounced check.
Dispute Resolution Index

Description: An independent procedural action is defined as a step of the procedure, mandated by law or court
regulation, that demands interaction between the parties or between them and the judge or court officer (e.g., filing
a motion, attending a hearing, mailing a letter, or seizing some goods). We also count as an independent
procedural action every judicial or administrative writ or resolution (e.g., issuing judgment or entering a writ of
execution) which is legally required to advance the proceedings until the enforcement of judgment. Actions are
always assumed to be simultaneous if possible, so procedural events that may be fulfilled in the same day and
place are only counted as one action. To form the index, we: (1) add the minimum number of independent
procedural actions required to complete all the stages of the process (from filing of lawsuit to enforcement of
judgment); and (2) normalize this number to fall between zero and one using the minimum and the maximum
number of independent procedural actions across the countries in the sample. The index takes a value of zero for
the country with the minimum number of independent procedural actions, and a value of one for the country with
the maximum number of independent procedural actions. Available for two legal cases: tenant eviction and
Duration
Description: The variable measures the total average duration in calendar days of the procedure under the factual
and procedural assumptions provided. It results form the sum of: (i) duration until completion of service of
process, (ii) duration of trial, and (iii) duration of enforcement. Available for two legal cases: tenant eviction and
One Share One Vote

Source: La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny, (1998), “Law and
Finance”
Description: Equals one if the Company Law or Commercial Code of the country requires that ordinary shares
carry one vote per share, and zero otherwise. Equivalently, this variable equals one when the law prohibits the
existence of both multiple-voting and non-voting ordinary shares and does not allow firms to set a maximum
number of votes per shareholder irrespective of the number of shares she owns, and zero otherwise.
Proxy by Mail
Finance”
Description: Equals one if the Company Law or Commercial Code allows shareholders to mail their proxy vote to
the firm, and zero otherwise.
Shares Not Blocked Before Meeting

Finance”
Description: Equals one if the Company Law or Commercial Code does not allow firms to require that
shareholders deposit their shares prior to a General Shareholders Meeting thus preventing them from selling those
shares for a number of days, and zero otherwise.
Cumulative Voting or Proportional Representation

Finance”
Description: Equals one if the Company Law or Commercial Code allows shareholders to cast all of their votes
for one candidate standing for election to the board of directors (cumulative voting) or if the Company Law or
Commercial Code allows a mechanism of proportional representation in the board by which minority interests
may name a proportional number of directors to the board, and zero otherwise.
Oppressed Minorities Mechanism

Finance”
Description: Equals one if the Company Law or Commercial Code grants minority shareholders either a judicial
venue to challenge the decisions of management or of the assembly or the right to step out of the company by
requiring the company to purchase their shares when they object to certain fundamental changes, such as mergers,
assets dispositions and changes in the articles of incorporation. The variable equals zero otherwise. Minority
shareholders are defined as those shareholders who own 10 percent of share capital or less.
Preemptive Rights to New Issues

Finance”
Description: Equals one when the Company Law or Commercial Code grants shareholders the first opportunity to
buy new issues of stock and this right can only be waved by a shareholders’ vote, and zero otherwise.
Percentage General Meeting

Finance”
Description: It is the minimum percentage of ownership of share capital that entitles a shareholder to call for an
Extraordinary Shareholders’ Meeting. It ranges from one to 33 percent.
Mandatory Dividend
Finance”
Description: Equals the percentage of net income that the Company Law or Commercial Code requires firms to
distribute as dividends among ordinary stockholders. It takes a value of zero for countries without such restriction.
Anti Director Rights

Finance”
Description: An index aggregating the shareholder rights which we labelled as “anti-director rights.” The index is
formed by adding 1 when: (1) the country allows shareholders to mail their proxy vote to the firm; (2)
shareholders are not required to deposit their shares prior to the General Shareholders’ Meeting; (3) cumulative
voting or proportional representation of minorities in the board of directors is allowed; (4) an oppressed minorities
mechanism is in place; (5) the minimum percentage of share capital that entitles a shareholder to call for an
Extraordinary Shareholders’ Meeting is less than or equal to 10 percent (the sample median); or (6) shareholders
have pre-emptive rights that can only be waved by a shareholders’ vote. The index ranges from 0 to 6.
Restrictions when going into reorganization.

Finance”
Description: Equals one if the reorganization procedure imposes restrictions, such as creditors’ consent, to file for
reorganization. It equals zero if there are no such restrictions. Bankruptcy and
No automatic stay on secured assets

Finance”
Description: Equals one if the reorganization procedure does not impose an automatic stay on the assets of the
firm upon filing the reorganization petition. Automatic stay prevents secured creditors to gain possession of their
security. It equals zero if such restriction does exist in the law.
Secured creditors first

Finance”
Description: Equals one if secured creditors are ranked first in the distribution of the proceeds that result from the
disposition of the assets of a bankrupt firm. Equals zero if non-secured creditors, such as the Government and
workers, are given absolute priority.
Management Does Not Stay

Finance”
Description: Equals one when an official appointed by the court, or by the creditors, is responsible for the
operation of the business during reorganization. Equivalently, this variable equals one if the debtor does not keep
the administration of its property pending the resolution of the reorganization process, and zero otherwise.
Legal Reserve
Finance”
Description: It is the minimum percentage of total share capital mandated by Corporate Law to avoid the
dissolution of an existing firm. It takes a value of zero for countries without such restriction.
Creditor Rights
Finance”
Description: An index aggregating different creditor rights. The index is formed by adding 1 when: (1) the
country imposes restrictions, such as creditors’ consent or minimum dividends to file for reorganization; (2)
secured creditors are able to gain possession of their security once the reorganization petition has been approved
(no automatic stay); (3) secured creditors are ranked first in the distribution of the proceeds that result from the
disposition of the assets of a bankrupt firm; and (4) the debtor does not retain the administration of its property
pending the resolution of the reorganization. The index ranges from 0 to 4.
Mean 3 Largest Shareholders

Finance”
Description: The average percentage of common shares owned by the three largest shareholders in the ten largest
non-financial, privately owned domestic firms in a given country. A firm is considered privately owned if the State
is not a known shareholder in it.
Accounting Standards
Finance”
Description: Index created by examining and rating companies’ 1990 annual reports on their inclusion or
omission of 90 items. These items fall into 7 categories (general information, income statements, balance sheets,
funds flow statement, accounting standards, stock data and special items). A minimum of 3 companies in each
country was studied. The companies represent a cross-section of various industry groups where industrial
companies numbered 70 percent while financial companies represented the remaining 30 percent.
Market Capitalization 10 Largest

Source: Computed from La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny,
(1998), “Law and Finance”
Description: Capitalization of the ten largest non-financial, privately owned domestic firms in a given country. A
firm is considered privately owned if the State is not a known shareholder in it.
Number of Procedures
Source: Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer, (2000), “The
Regulation of Entry”
Description: The number of different procedures that a start-up has to comply with in order to obtain a legal
status, i.e. to start operating as a legal entity.
Time
Description: The time it takes to obtain legal status to operate a firm, in business days. A week has five e business
days and a month has twenty-two.
Cost
Description: The cost of obtaining legal status to operate a firm as a share of per capita GDP in 1999. It includes
all identifiable official expenses (fees, costs of procedures and forms, photocopies, fiscal stamps, legal and notary
charges, etc). The company is assumed to have a start-up capital of ten times per capita GDP in 1999.
Cost and Time

Description: The cost of obtaining legal status to operate a firm as a share of per capita GDP in 1999. It includes
all identifiable official expenses (fees, costs of procedures and forms, photocopies, fiscal stamps, legal and notary
charges, etc) as well as the monetised value of the entrepreneur’s time. The time of the entrepreneur is valued as
the product of Time and per capita GDP in 1999 expressed in per business day terms. The company is assumed to
have a start-up capital of ten times the GDP per capita level in 1999.
Percentage Bankruptcies
Source: Klapper, Leora (2001), “Bankruptcy Around the World: Explanations of its Relative Use”
Description: Total number of bankruptcies as a percentage of total number of firms.
Finance-Activity
Source: Thorsten Beck and Ross Levine (2001), “New Firm Formation and Industry Growth: Does Having a
Market- or Bank-Based System Matter?”
Description: Logarithm of (Total value traded as share of GDP and claims on private sector by financial
institutions as share of GDP)
Finance-Size
Description: Logarithm of (Market capitalization and claims on private sector by financial institutions as share of
GDP)
Finance-Aggregate
Description: First principal component of Finance-Activity and Finance-Size
Finance-Dummy
Description: Dummy variable that takes the value 0 if total value traded as share of GDP and claims on private
sector by financial intermediaries as share of GDP are less than the respective sample mean, 1 otherwise
Structure-Activity
Description: Logarithm of (Total value traded divided by claims on private sector by commercials banks)
Structure-Size
Description: Logarithm of (Market capitalization divided by claims on private sector by commercials bank)
Structure-Aggregate
Description: First principal components of Structure-Activity and Structure-Size
Structure-Dummy
Description: Dummy variable that takes the value 1 if Structure-Aggregate is above the median, 0 otherwise
Net Interest Margin

Source: Barth, James, Gerard Caprio and Ross Levine (2001c), “Banking Systems Around the Globe: Do
Regulation and Ownership Affect Performance and Stability?”
Private Credit
Total Value Traded

Non-Bank Credit
Operating Income Variability

Source: Stijn Claessens, Simeon Djankov, Tatiana Nenova (2001), “Corporate Risk around the World”
Description: Cash flow risk: Operating income variability
Total Debt to Market Value of Equity

Description: Financial leverage: Total debt to equity
Long Term Debt to Market Value of Equity

Description: Financial leverage: Long-term debt to equity. Short Term Debt to Market Value of Equity computed
from data in Stijn Claessens, Simeon Djankov, Tatiana Nenova (2001), “Corporate Risk around the World”.
Net Working Capital to Total Assets

Description:
Five Bank Concentration Ratio

Source: Barth, James, Gerard Caprio and Ross Levine (2001b), “The Regulation and Supervision of Banks
Around the World – A New Database”
Description: The degree of concentration of deposits in the 5 largest banks.
Legal Origin
Source: La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny, (1999), “The
Quality of Government”
Description: Identifies the legal origin of the company law or commercial code of each country. Equal 1 of the
origin is English common law, two if the origin is the French commercial code, three if the origin is the German
commercial code, four is the origin is Scandinavian civil law, and five if the origin is Socialist civil law.
Percentage Assets Government Owned

Description: Government Owned Banks.
Percentage Foreign Owned

Description: Foreign Bank Ownership.
Percentage of Top 10 Rated Internationally

Description: Percent of 10 Biggest Banks Rated by International Rating Agencies.
Bank Supervisors per Institution

Description: Supervisors per Bank
Onsite Examination per Bank Last 5 Years

Description: Onsite Examination Frequency
Supervisors Employed by Bank Industry

Description: Likelihood Supervisor Moves into Banking
Explicit Deposit Insurance

Description: Takes value 1 if there is an explicit deposit insurance scheme.
LABOR MARKETS IN PORTUGAL: RECENT PERFORMANCE
AND CHALLENGES FOR DEVELOPMENT IN THE EUROPEAN CONTEXT1
Daniel Traça
INSEAD
1
Prepared for the conference on ‘Desenvolvimento Económico Português no Espaço Europeu: Determinantes e
Politicas’ organized by Banco de Portugal. For comments, please email: daniel.traca@insead.edu . This paper
has benefited dearly from comments by Mario Centeno and Ana Balcao Reis, as well as from the referees from
Banco de Portugal.
396 Daniel Traça
I. Introduction
The last quarter-century has been a period of heightened challenges for labor markets in
industrialized countries, namely for Portugal. The next quarter-century is likely to be even
more testing. The widely acknowledged role of labor market rigidities in hindering the
competitiveness of the EU area highlights the relevance of labor market performance for
the challenge of economic development.
The main goal of this study is to bring to light the shocks that will affect labor markets
in the EU area, in the medium term, focusing on the case on Portugal. Drawing on labor
market developments and performance of the last 30 years, we uncover the main drivers of
change in the near future, and assess potential outcomes. We conclude by suggesting
directions for reform, aiming at fostering competitiveness.
International background over the last 30 years
The difficulties of labor markets in the OECD over the last 30 years are well known.
After a period of striking rise in the standard of living of European workers in the 1950’s
and 1960’s, with low unemployment and high wage growth, the performance of labor
markets from the 1970’s to the mid 1990’s was lackluster. Slow wage growth and widening
wage differentials in the flexible labor market of the United States were matched in rigid
Europe by a dramatic and prolonged rise in unemployment rates, striking mainly the young
and unskilled.
Recently, the second half of the 1990’s saw sharp declines in unemployment rates in
many European countries, namely: the Netherlands, the UK and Ireland. Instrumental to
these successes, particularly in the Netherlands, has been the liberalization of labor market
institutions, highlighted in the recommendations of the OECD job strategy of 1994. Non-
reformers like France, Germany and Italy have lagged behind. Meanwhile, unemployment
in the United States fell to its lowest level, fuelled by the longest expansion in that
country’s history.
Comparative features and performance of labor markets Portugal
Against this international background, the performance of Portuguese labor markets had
some unique features. While the unemployment rate has remained low, closer to the
flexible US benchmark than to the high unemployment in the rigid EU area, the proportion
of long-term unemployment has been extremely high, mimicking the features of other EU
member countries. Job creation in Portugal has been higher than the European average, but
much lower than US levels.
A striking feature of job creation in Portugal has been the ability to create low skilled
jobs during the 1990’s, in contrast to the dwindling number of opportunities for unskilled
workers in Europe. On the other hand, sagging job creation for the highly skilled in
Portugal contrasts with the increasing skill intensity of employment in the rest of the EU,
and is a reason for concern. Another remarkable feature of job creation in Portugal has been
the steep increase in the share of employment in the government sector.
Driven by economic convergence, productivity growth has remained strong, but highly
volatile, and has been matched closely by compensation. Remarkably, the share of labor is
much lower in Portugal than for the US or the EU average.
Labor Markets in Portugal: Recent Performance and Challenges for Development in the European Context 397
The low unemployment rate of the Portuguese labor market is predicated in the
flexibility of its wage-setting mechanisms. A relatively competitive and decentralized
system of wage bargaining, where Trade Unions underbid each other to sign an agreement
with the employer, ensures that wages remain in line with productivity, for the different
industries and firms.
Other features of Portuguese labor markets are much less proficient. Taxes and benefit
systems are near the European average, but considered too high. Employment Protection
Legislation for regular workers is the most restrictive in the OECD, leading to a large and
growing share of fixed-term contracts in employment, despite very severe institutional
limitations on temporary work. Finally, active labor market policies in Portugal are very
inefficient, with the role of the public employment service in job placement well below
other EU countries and a perverse trend in public sponsored vocational training, whereby
the majority of investment has been allocated to employed workers.
Upcoming Trends
In this context, some new striking challenges loom for labor markets in Portugal, and
more generally, in Europe. We identify three main drivers of economic change: the
expansion of trade and capital flows, at a global and European scale; the increased pace of
technological change and institutional changes in capital markets; and the implementation
of the European Monetary Union (EMU), along with the fiscal constraints of the stability
pact.
The repercussion of these dramatic structural changes for labor markets can be
summarized in two main trends: a shift in the relative demand toward skilled worker,
continuing a trend of the last twenty years, and an increase in labor demand volatility.
Responses to the increase in demand for skills must encompass an effective education
policy. Because others in this conference will address the challenges for the education
system, they are beyond the scope of this study. Hence our main object of analysis is the
rise in labor demand volatility, and the policy responses it entails.
Policy Responses to Labor Demand Volatility
Responses to enhanced labor demand volatility are determined by the adaptability and
flexibility of labor market institutions and the workforce. Adaptability captures the ability of
workers and their employers to adjust to a shock by moving to another region (in the case
of a country-specific shock) or another industry (for an industry-specific shock). Flexibility
depicts the ability of wages and prices to adjust to the shock, bringing the real wage in line
with marginal productivity in the different industries and/or countries.
In an adaptable environment, workers move from countries, regions and industries
where shocks reduce productivity, to those where productivity is higher, enhancing the
standard of living. Flexibility is also important, for it ensures that wages adjust to maintain
full-employment. Without adaptability, flexibility constitutes only a second-best outcome.
The worst-case scenario is a rigid labor market, with low adaptability and flexibility, where
an increase in labor demand volatility is likely to increase unemployment and hinder wage
growth.
398 Daniel Traça
While the determinants of flexibility are well recognized, involving a decentralized

wage bargaining system, with scope for variability across firms, industries and regions, the
sources of adaptability have been less studied. We speculate that several factors are key. On
the one hand, education, by providing general human capital transferable across firms and
industries and increasing the ability to track down opportunities, raises the ability of
workers to move to industries and firms fast, as the occasion arises. On the other hand, two
aspects of labor market institutions are key to enhance adaptability: a liberal employment
protection legislation reduces barriers to entry, and reduces the cost of moving to expanding
industries; while an efficient active labor market policy ensures that search costs are low,
reducing the duration of increasingly frequent unemployment spells and ensuring
continuous access to training opportunities.
Implications for Portuguese labor markets and directions for reform
Given these success factors, the features of Portuguese labor markets mentioned above
imply a moderate success. On the one hand, the flexibility of wage formation systems in
Portugal ensures that the country will be able to face heightened labor demand volatility
without the fear of unemployment. In this sense, the consequences are likely to be more
serious in the rest of Europe, where wage-setting mechanisms are more centralized and less
sensitive to unemployment.
In contrast, performance in terms of adaptability is likely to be sub-standard. Indeed the
three less efficient aspects of Portuguese labor markets, as recognized by the OECD in
1998, correspond exactly to the main elements of adaptability; namely: the low skill
attainment of the labor force, the extreme strictness of employment protection legislation,
and the inefficacy of active labor market policies. Hence, while unemployment and severe
stress are not at risk, reform of labor markets to improve adaptability are crucial to ensure
that Portuguese labor markets will obtain maximum benefit for the changes that are likely
to occur in the next 30 years.
We conclude that reform of labor market institutions in Portugal is key to ensure
improved competitiveness and a successful response to the challenges ahead. The main
directions for reform include: (a) the relaxation of employment protection legislation, and
(b) an increased experimentation and monitoring of active labor market policies, in a
contestable framework and in interaction with the private sector.
Remainder of this study
Section II addresses the performance and institutional developments of labor markets

over the last 30 years in Portugal, vis-à-vis the international developments in the US and
the EU. In Part 1, we present traditional labor markets indicators of unemployment, job
creation, and compensation. Part 2 looks at the implications of the rigidities that shaped
European labor markets, and the recent reforms. Finally, Part 3 uncovers the features of
labor market institutions in Portugal, in comparison with its European partners. Section III
looks at the future, in the medium term, in Portugal and Europe. In Parts 1 and 2, we
anticipate the key drivers of economic change, and address their implications for labor
markets. Part 3 looks at the appropriate policy responses to these shocks. Finally, Part 4
evaluates the readiness of Portuguese labor market institutions, and suggests avenues for
reform. Sections IV concludes.
II. Performance over the last 30 years
This section addresses the performance of labor markets in Portugal over the last
quarter-century, against the background of international developments in the United States
(US) and the European Union (EU). The comparison with US data is important as
benchmark, due to its well-recognized flexibility. The evaluation against EU data serves a
double purpose: first, the EU data can be used as a benchmark for a rigid labor market,
given the well-known constraints to market forces in the labor markets of Continental
Europe; second, it allows us to assess the features of the labor market in the EU and in the
EURO area, within which the Portuguese economy is to become increasingly integrated.
In this context, all throughout this paper, we report the simple average for the available
EU member countries for the various variables, instead of the aggregate (weighted average)
EU value. Since labor market policies are defined at the level of each member country, the
simple average is a better measure of the implications of the different policies followed in
the different countries. (The weighted average would give excessive weight to the
extremely rigid policies of Germany, France and Italy.)
Part 1 depicts several indicators of labor market performance, for Portugal, the US and
the EU. We focus on the data on unemployment, job creation, and wages and productivity.
Part 2 addresses the labor market rigidities that characterize the labor markets of Europe,
their consequences and the evidence on reform. Finally, part 3 looks at the features of
Portuguese labor markets, assessing its rigidities vis-à-vis the international benchmarks.
1. Labor market performance: Portugal, United States and the EU
i. Unemployment
Unemployment rates
We start by looking at the evolution of unemployment. Figure 1 captures the evolution

of unemployment rates in the European Union, the United States and Portugal, starting in
1970. Figure 2 presents estimates of the Natural Rate of Unemployment ( NAWRU - Non-
Accelerating Wage Rate of Unemployment) from the OECD, for various years.
400 Daniel Traça
Figure 1 - Unemployment rates in Portugal, EU and US
12
8
percent
0
1970 1975 1980 1985 1990 1995 2000
Portugal United States EU (avg exc LUX)
Source: OECD; Economic Outlook Database
Figure 2 - Natural rate of unemployment (NAWRU)
12
9
percent
0
1978 1983 1988 1993 1998
Over the last 30 years, the performance of unemployment in Europe can be divided into
three phases. The first one, which started after WWII and lasted until the mid-1970’s, was a
period of low unemployment accompanied by high wage growth, fuelled by increases in
productivity and capital accumulation. By contrast, from the mid-1970’s to the mid-1990’s,
the performance of European labor markets was lackluster. This second period was
characterized by a dramatic and prolonged rise in unemployment rates. The upward trend
eased between 1985 and 1990, but regained speed in the early 1990’s, reaching an all-time
high in 1994. This increase in unemployment was due, not only to cyclical factors, but also
to structural elements that augmented the natural rate of unemployment (Figure 2). Finally,
the third period, starting in the second half of the 1990’s saw a recovery in many countries,
with sharp declines in historically high unemployment and in the NAWRU. This recovery
did not happen in all countries (and thus in under-represented in the EU average): the
Netherlands, the UK and Ireland are success stories, while France, Germany and Italy lag
behind.
By contrast, in the US, the unemployment rate increased from the mid-1970’s to the
mid-1980’s, due to cyclical factors. By 1985, it had returned to its historical equilibrium
level of around 6%. In the second half of the 1990’s, unemployment in the US declined to
historical lows. All throughout, the NAWRU has remained relatively stable, just above 5%.
Finally, the evolution of the unemployment rate in Portugal post 1970 is distinct from
the rest of Europe. During the 1970’s, the unemployment rate suffered a dramatic rise,
heavily influenced by the return of Portuguese nationals from its former colonies;
Portugal’s labor supply rising by no less than 14% between 1973 and 1980 (OECD, 1996).
The 1970’s produced also dramatic social, political and economic events, which, along with
previous backwardness, implied that Portugal entered the 1980’s with an economy beset by
severe structural problems. Hence productivity failed to rise, and job growth was
insufficient to absorb the increased labor supply.
Unemployment finally started to decline in the mid 1980’s, as a result of a shift to
market-oriented policies and special factors linked to accession to the EU, like the inflow of
structural funds. By the mid-1980’s structural unemployment in Portugal was already lower
than the EU average, at around 6% (Figure 2). However, the structural reforms and high
growth of the mid-1980’s managed to further reduce unemployment, with the NAWRU
falling consistently, to just above 4% in the late 1990’s.
A striking feature of Figure 1 and Figure 2 is that from the mid-1980’s (starting around
1987 in Figure 1), the Portuguese unemployment rate runs parallel to that of the EU,
capturing the influence of EU-wide cyclical factors in the Portuguese economy, but at a
much lower level: the unemployment rate in Portugal is consistently four points below the
EU average during the 1990’s. Indeed, the average level of the unemployment rates in
Portugal is much closer to that of the flexible US labor market than the EU. In particular,
the rate of structural unemployment - NAWRU - was lower in Portugal than the US by 1%,
in the late 1990’s.
Duration
Another key aspect of unemployment is the duration of unemployment spells. Figure 3

provides data on the compositions of unemployment pools by duration. 2
2
The labor force survey for Portugal suffered a change of coverage in 1992, which generates a statistical
irregularity in the series for that year. In the figure, the values for 1992 were obtained as the averages of 1991
and 1993.
402 Daniel Traça
Figure 3 - Unemployment by duration
a: Less than 1 month

60
% of total unemployment
50
40
30
20
10
0
1986 1990 1994 1998
Portugal United States EU (avg exc AUT)
b: 1 year and over

60
% of total unemployment
50
40
30
20
10
0
1986 1990 1994 1998
Portugal United States EU (avg exc AUT)
Source: OECD (2001), Labor Force Statistics
As expected, given the higher rate of unemployment in Europe, the long term
unemployed (1 year and over) constitute a much higher proportion of the unemployed in
Europe. This fact is a major reason for concern, given its implications for the depreciation
of a worker’s human capital and the subsequent loss of motivation for job search. However,
the striking feature of Figure 3 is that, in terms of duration, Portuguese unemployment is
much closer to the EU average than to US levels. This contrasts sharply with the numbers
for the unemployment rate mentioned above (Figure 1 and Figure 2), where Portugal was
much closer to US levels.
Composition by age and educational attainment
Figure 4 - Unemployment rates for youth (<25 yrs) and older (25 yrs +) workers
20
15
10
-
< 25 25 yrs + < 25 25 yrs + < 25 25 yrs +
Portugal USA EU (avg exc GRE,

AUT,DEU, FIN)
1992-94 1998-00
Source: OECD, Quarterly Labor force Statistics
Figure 5 – Unemployment rate, by educational attainment - aged 25-59
-
Pri Sec Ter Pri Sec Ter
Portugal EU (avg exc LUX)
1992-94 1998-00
Source: EUROSTAT, Labor Force Survey
The incidence of the unemployment in Europe is very diverse, varying by education

levels and age groups. Figure 4 and Figure 5 show the disaggregated unemployment rates
by age and by educational attainment.3 In terms of age distribution, Figure 5 shows that
unemployment is mostly a youth phenomenon, with unemployment rates for the less than
25 years old more than five points higher than older workers both in the US and the EU,
and also for Portugal. Nevertheless, the relative incidence of youth unemployment is
slightly higher in the EU.
By contrast, the decomposition of unemployment by educational attainment (for older
workers) presents a distinct picture between Portugal and the EU. Data for the US was not
available, since the source was EUROSTAT. In the EU, unemployment is mostly an
3
It should be noted that unlike most other data presented here, which is sourced from the OECD, the data by
educational attainment is based on EUROSTAT. For this reason, they are not entirely comparable.
404 Daniel Traça
unskilled issue, with unemployment rates falling sharply as educational attainment rises. In
Portugal, the unemployment rate is highest among those with secondary education,
followed closely by the least skilled, and is lowest for those with tertiary education.
However, the effect of tertiary education on the likelihood of unemployment is much less
pronounced in Portugal than in the EU. Moreover, while in the EU the unemployment rate
for the highly skilled fell in the second half of the 1990’s, it has increased in Portugal. By
contrast, that of the low skilled, already low by EU standards, has fallen further. As we will
see below, this evidence highlights the extraordinary ability of Portuguese labor market to
create jobs for the low skilled.
ii. Employment and job creation
To complement our assessment of unemployment, we review job creation among

OECD countries. Figure 6 presents estimates of net job creation relative to working-age
population, output and capital stock, as reported in Garibaldi and Mauro (1999).
Figure 6 - Job Creation, 1980-97
9.0
6.0
3.0
0.0
Gr. L Ch. L/P Gr. Y/L Gr. K/L
-3.0
Portugal United States EU (avg exc LUX))
Average employment growth (Gr. L); Change in employment to working age population
(Ch. L/P); Average growth of output to employment (Gr. Y/L); and Average growth of
capital to employment (Gr. K/L).
Source: Garibaldi and Mauro (1999); Table 1
Like with unemployment, the net job creation experience has varied widely across
OECD countries, with the Atlantic providing a clear divide: the United States have clearly
outperformed most Continental European countries, both in absolute terms and relative to
working age population, particularly if we ignore some outliers like the Netherlands and
Ireland.4 Moreover, the growth of job creation in the United States has contributed to an
4
A widely held view is that most job creation in the US has taken the form of low quality service jobs. Garibaldi
and Mauro (1999) show that, indeed, a considerable part of employment growth among the fast job creators has
taken place in the retail sector, while countries with unfavorable initial conditions (such as a high share of the
labor force in agriculture) have done worse. However, statistical analysis to account for the role of sectoral
increase in the labor intensity of production, relative to the average EU member. In

Portugal, a relatively high rate of capital accumulation, related to the catch-up process of
the last 20 years, has also contributed for an increase in capital intensity of production.
Educational attainment
Figure 7 – Employment by educational attainment, as share of total adult population

(aged 25-59)
60
50
40
30
20
10
-
Pri Sec Ter Pri Sec Ter
Portugal EU (avg exc LUX)
1992-94 1998-00
Source: EUROSTAT, Labor Force Survey
A striking feature of employment and job creation in Portugal has been the ability to
create jobs for the low skilled, which has produced the low unemployment rates mentioned
above. In fact, Figure 7 shows that in Portugal, over the last decade, job creation by
educational attainment has followed a very distinct pattern from the EU. Relative to total
population, employment in Portugal has increased for workers with primary education and
fallen for those with tertiary education, while the reverse is true in the EU.
Two factors can be behind the unskilled-jobs created in Portugal: a benign interpretation
is that it is due to low disincentives to employment of the low paid; a more pessimistic view
is that it represents a pattern of specialization according to comparative advantage in the
EU context. This view is given strength by the lackluster performance of job creation for
skilled workers. Since productivity growth requires an increased proportion of high skilled
workers, the inability to create jobs for the highly skilled in Portugal is worrisome, if
interpreted in light of the second view. Addressing this issue in further detail is beyond the
scope of this study.
differences show that only one fifth of differences in job creation can be accounted for by differences in sectoral
composition of employment.
406 Daniel Traça
Government employment
Figure 8 - Government Employment
% of Total Employment
15%
5%
1970 1975 1980 1985 1990 1995 2000
Portugal United States

EU (avg exc. LUX)
Source: OECD, Economic Outlook Database
Another element of job creation is the role of the public sector. Figure 8 depicts the
evolution of government employment as a proportion of total employment. Government
employment in the public sector has remained stable in the US, just above 15%, with a
slight decrease after the mid-1970’s. In Europe, the 1970’s and 1980’s saw a dramatic rise
in the share of government employment, depicting the growth of the welfare state. From the
mid 1990’s, the trend was inverted, as the Welfare State began to be scaled down.
For Portugal the rise in government employment is dramatic: from just above 8% in
1970 to over 17% in 2000, approaching the EU average at a fast pace. Two periods of
dramatic increase in the proportion of government employment in total employment are the
years of 1983 and 1991. Public sector jobs are usually highly sheltered from lay-off risk.
Moreover, a recent study by Portugal and Centeno (2001) provides evidence that public
sector wages in Portugal are higher than in the private sector, and that this difference is the
highest among EU countries. As a result of high wages and high employment, Portugal has
the second highest ratio of personnel expenditure in the public sector to GDP among EU
countries (Portugal and Centeno, 2001).
The reigning explanation for the growth of the public sector from cross-country studies
is that governments have used public-sector employment as a tool for generating and
redistributing rents (Gelb et al., 1991). Rodrik (1999) argues that the expansion of safe
public sector jobs arises as a policy response to increased volatility in the private sector,
due to expanded exposure to international trade. In any case, the magnitude of the Welfare
State in Portugal is clearly insufficient to explain that such a high proportion of the labor
force is involved in the delivery of public services. As we will mention below, the need for
reform of the public sector, namely by reducing its size, is a key challenge for Portugal,
which will have implications for its labor market.
iii. Productivity and wages
Figure 9 – Productivity, Compensation and Labor share5

a: Labor Productivity Growth
15%
10%
5%
0%
-5%
-10%
1970 1975 1980 1985 1990 1995
Portugal United States European Union
b: Growth of compensation
20%
15%
10%
5%
0%
-5%
1970 1975 1980 1985 1990 1995 2000
5
OECD data on productivity and compensation differs from that of the Bank of Portugal, for the mid-1970’s, and
seems to overestimate the productivity decline in that period. Nevertheless, in order to allow for comparisons
across different countries from a single source, we use OECD data.
408 Daniel Traça
c: Employee compensation relative to productivity (1985=100)
150
130
110
90
1970 1975 1980 1985 1990 1995 2000
d: Labor Compensation share
70%
60%
50%
40%
1970 1975 1980 1985 1990 1995
Portugal United States EU (avg)
Source: OECD, Economic Outlook Database, National Accounts Database
To conclude, Figure 9 displays the evolution of productivity and wages. The

background for the comparative performance of the US and the EU can be summarized as
follows. Starting from the mid 1970’s, the growth rate of productivity declined from an
average of 5% to a miserly 2%. The slowdown was even more pronounced in Europe, as
many countries were still in a catch up phase, after WWII. Consequently, the growth of
compensation fell in Europe and the US (Figure 9b). However, the slowdown of the growth
of real compensation happened with a lag in the average EU member, causing an increase
in the wages to productivity ratio in Europe vis-à-vis the US (Figure 9c). At the root of the
different speeds of adjustment to the productivity slowdown are differences in labor market
institutions, as we will see below. As argued by Blanchard (1997), a slower wage response
increased the labor share in Europe in the 70’s. However, as firms took the increase in wage
costs as an incentive to become more capital intensive, demand for labor fell, causing
unemployment and reducing the labor share.
Capturing the convergence process, productivity growth in Portugal has remained above
the EU average, after the events of the 1970s, except for the economic crisis of the early
1980’s. Wage and productivity growth in Portugal are extremely volatile, relative to the US
and the EU average. This is due to a well-know cycle of boom-and-bust that characterizes
Portugal’s macroeconomic variables. Figure 9c: shows that the adjustment of Portuguese
wages to productivity, after the wage explosion of the post-revolutionary period, took about
ten years. Since 1985, the ratio of compensation to productivity has remained stable
(possibly a steady-state), apart for a compensation crunch at the end of the 1980’s. Finally,
the labor share in Portugal has followed the adjustment of wages. Surprisingly, it is
considerably lower than the EU average or the level for the US, possibly due to the
country’s lower capital intensity.
2. Labor Market Rigidities and Reform
i. Labor markets and The Welfare State
Overall, most economists agree on the sequence of events that has brought about
Europe’s sustained high unemployment. The story can be summarized as follows. The
initial rise in the rate was triggered by a series of adverse shocks during the 70’s and the
early 80’s, itemized in Box 1. Once these shocks subsided, certain institutional features of
European labor markets kept unemployment rates from falling. 6
Box 1 - Shocks affecting European labor markets in the 1970’s and 1980’s
• Labor force trends (rise in female participation, slowdown in long-term decrease in the
length of the workweek),
• Technological change that is complementary with skills and reduces the need for unskilled
workers, even in services, but whose progress has decelerated considerably, generating a
productivity slowdown;
• The emergence of a number of Third World competitors that gain market shares in the
manufacturing sector;
• Macroeconomic shocks such as the hike in oil prices in the mid 1970’s or the rise in world
interest rates, starting from the early 1980’s.
On the other hand, European labor market rigidities arose over the years, with a series
of rules and regulations aimed at promoting social justice and improving the standard of
living of workers, which became known as the Welfare State.
6
Blanchard and Wolfers (2000) show that the interaction of some labor market institutions with cyclical and
structural shocks contributes to explaining the increase in European unemployment over time as well as the
heterogeneity in unemployment evolutions across EU. Their results indicate that whereas cyclical and structural
shocks contribute to the general increase in unemployment, the interaction of these changes with different
national labor market institutions seems to explain some of the heterogeneity of unemployment evolutions.
410 Daniel Traça
Box 2 - Taxonomy of labor market institutions

• Wage-formation: includes the centralization and coordination of wage-bargaining and the
role of the minimum wage;
• Tax and benefits systems: comprises the characteristics of subsidies to unemployment and
the level of taxation of labor through income taxes and social security contributions;
• Employment protection legislation (EPL): covers the laws regulating the procedures for
separation, including collective dismissals, and alternative forms of employment (e.g.
temporary work);
• Active labor market policies (ALMP): involves the activities of the Public Employment
Service (PES) in its three main functions: the job-brokerage or placement function, the
implementation and monitoring of systems of income support to jobseekers, and the
improvement of the employability of jobseekers through vocational training and skill
upgrading.
Deconstructing labor market institutions into the four key elements summarized in Box
2, a large theoretical and empirical literature (the latter using mostly cross-country studies)
has analyzed the effects of the rigidities introduced on unemployment and labor market
performance (for a recent review of the literature, see Nickel and Layard, 1998). Below we
summarize the main conclusions of this research.
Wage-formation
There is some evidence that different collective bargaining arrangements affect labor
market outcomes. Calmfors and Driffill (1988) suggest that either highly centralized and
fully decentralized bargaining systems lead to somewhat lower structural unemployment
compared with intermediate systems, because they help to restrain the wage claims of
insiders. (For a recent discussion, see Elmeskov at al., 1998).
Tax and benefit systems
Price incentives affect the structural rate of unemployment on both the demand and
supply sides. On the demand side, high tax wedges and social security contributions raise
the cost of labor, causing a decline in the demand for labor and a substitution towards
capital-intensive techniques. Recent studies seem to suggest a significant effect of taxes on
labor on unemployment. The impact is stronger and more robust when wages are negotiated
at the sectoral level with a lack of coordination, and thus, may not fully adjust to the tax-
hike. (See, for example, Elmeskov at al., 1998).
On the supply side, generous unemployment insurance and other non-work benefits
reduce the opportunity cost of unemployment, and reduce the search effort of the
unemployed. Elmeskov at al. (1998) provide evidence that very generous unemployment
benefit systems – in terms of both levels and duration – may contribute to structural
unemployment.
Employment protection legislation (EPL)
The key intent of EPL is to enhance job and income security, benefiting worker satisfaction
and longer-term attachment, which in turn favor the accumulation of firm- and job-specific
skills. The theoretical implications for unemployment are unclear, since EPL should reduce
both lay-off rates and hiring rates. Hence it is not surprising that empirical results are
somewhat mixed, with weak evidence of a positive effect of EPL on structural
unemployment (Elmeskov et al., 1998; OECD, 1999; Heckman and Pagés, 2000).
However, EPL’s has the potential for several other perverse effects on labor markets:
• Enhancing the dualism of labor markets, by favoring “insiders”;

• Shifting firms preferences to adjusting working hours when responding to demand
shocks;
• Biasing firms against “risky” workers, such as young workers and the long-term
unemployed, in order to reduce the likelihood of a bad match;
• Discouraging job creation and job destruction, by taxing work-force adjustments
and raising a “firing” cost;
• Making the unemployment pool more stagnant, due to lower inflows and outflows,
i.e. less employment turnover, and longer durations. (See OECD, 1999 for
empirical evidence).
•
Active labor market policies (ALMP)
The evidence points to a positive and significant impact of expenditure in the Public
Employment Service (PES) in reducing unemployment. The significance rises if Sweden,
with a very high spending on labor market programs, is removed from the panel (Scarpetta,
1996; and Nickell and Layard, 1998).
ii. Directions for reform

In 1994, OECD member countries agreed on a Jobs Strategy that contained a wide-
ranging set of policies to enhance employment growth, reduce unemployment and increase
prosperity. They focused on the areas of labor market reform, education and training
policies and business environment; the latter including macroeconomic stability, innovation
and product market competition. The main recommendations in the area of labor market
reform are summarized in Box 3.
412 Daniel Traça
Box 3 - Labor market reforms from the OECD Jobs Strategy (1994)
• Increase flexibility of working time (both short-term and lifetime) voluntarily sought by
workers and employers
• Make wage and labor costs more flexible by removing restrictions that prevent wages from
reflecting local conditions and individual skill levels, in particular of younger workers
• Reform employment security provisions that inhibit the expansion of employment in the
privates sector.
• Strengthen the emphasis on active labor market policies and reinforce their effectiveness.
• Reform unemployment and related benefit systems – and their interactions with the tax
system – such that societies fundamental equity goals are achieved in ways that impinge far
less on the efficient functioning of labor markets.
• Improve labor force skills and competencies through wide-ranging changes in education
and training systems.
The implementation of labor market reforms in OECD countries has been extremely
uneven. A number of countries with high and persistent unemployment have been hesitant
to implement reforms that affect core workers. For example, few changes have been made
to stringent employment protection for workers with regular contracts, whereas most
countries have eased regulations for temporary contracts (fixed-term and temporary work
agencies). Moves toward decentralized wage bargaining have been implemented in a
number of countries, but most have been reluctant to reform minimum wages or, more
generally, to allow a widening of the earnings distribution. In addition, only a few countries
have significantly reduced the level and maximum duration of unemployment benefits and
other non-employment benefits; while many have preferred to tighten the eligibility criteria.
iii. Successful reformers and recent developments

As mentioned above, the mid 90’s witnessed a change in unemployment trends. In the
US, unemployment reached a 30-year low in 1998, fuelled by one of the most resilient
economic expansions in that country’s history. This included a decline in an already low
rate on natural unemployment. In the EU, unemployment also showed signs of easing, with
a decline in the NAWRU.
However, the experience of individual member countries in the EU has been diverse.
Some countries have achieved significant reductions of structural unemployment: the UK,
Denmark, Ireland and the Netherlands; and more recently Spain. Meanwhile, France, Italy
and Greece, along with Finland and Sweden, recovering from recession, have seen rises in
structural unemployment (OECD, 1999). In assessing the role of labor markets reforms on
the decline in unemployment, OECD (1999) argues that countries that have progressed
most along the lines of the recommendations of the OECD jobs strategy (see Box 3) have
experienced the most significant improvements in labor market conditions.
Figure 10 - Unemployment in theNetherlands
12
percent
0
1970 1975 1980 1985 1990 1995 2000
EU (avg exc LUX) Netherlands
Source: OECD, Economic Outlook Database
Among the successful performers, the reforms in the Netherlands provide some
interesting insights.7 Figure 10 shows the striking evolution of unemployment rate in that
country. Assessing the implementation of Jobs Strategy recommendations, the OECD
estimated that sufficient action had been taken in 30% of its recommendations, the highest
level after the UK with much fewer recommendations.
Notably, the Wassenar agreement of 1982, reached through negotiations between
government, unions and employers resulted in several key reforms, including the end of
wage indexation, and decline in minimum wages and labor taxes. To address the problem
of low-skilled workers the government introduced permanent tax-relief for firms hiring
low-wage workers (up to 115 percent of the minimum wage) and temporary tax reductions
for hiring the long-term unemployed. It also encouraged the creation and use of both
bottom pay scales, which are close to the legal minimum wage, and “opening clauses”,
which allow workers to be paid below minimum wages set in collective agreements. In
1986-7, the unemployment subsidy was significantly curtailed, both in terms of the
replacement rate and duration. Some relaxation of employment security has also been
introduced. Finally, the Wassenar agreement eased rules on part-time work, making part-
time employees eligible for full social security benefits, including unemployment and
disability insurance.
As a result, employment growth as surpassed an already high rate of growth of the labor
force, with job creation focusing on the service sector. Strikingly, part-time employment
has accounted for two-thirds of the net jobs created since the early 1980’s, with the share of
part-time in employment reaching 47 percent (the highest in the EU) in 1997 (OECD,
1998). Box 4 summarizes the main lessons from the experience of the Netherlands and
other countries for the implementation of labor market reforms:
7
In Ireland, the decline in unemployment was driven mostly by structural reforms outside the labor markets,
which have paved the way to an increase in investment rates, financed by the inflow of foreign capital.
However, several reforms, including limits on wage increases, reductions in income taxes, a cutback in
unemployment benefits, and the adoption of active labor market policies in the late 1980’s have also played a
role. The relatively long lag to the beginning of employment growth in 1993 raises some questions about the
role of these reforms on the decline in unemployment.
414 Daniel Traça
Box 4 - Success factors in the implementation of labor market reform

• Complementarities among different policies, namely between macroeconomic adjustment
and structural reform, imply that comprehensive reforms tend to be more successful (see
Coe and Snower, 1997).
• Generally speaking, structural reform generates improved labor market performance with a
relatively long lag. Indeed, some of the countries with significant falls in structural
unemployment began implementing reforms in the early to mid-1980.
• Consensual changes, obtained by co-ordination between employers and workers’
representatives have a higher chance of success. In the Netherlands (1982) improved co-
ordination of wage bargaining via-tripartite agreements between unions, employers and the
government, was key to obtain wage moderation accompanied by macroeconomic
stabilization measures.
• Successful reforms arose as part of a comprehensive package combining wage restraints
with other initiative that helped soften the impacts of the restraints on workers.
3. Comparative Features of Portuguese labor markets
As we have shown in section II.1, the overall performance of labor markets in Portugal
has been much stronger than its European counterparts. Box 5 summarizes the performance
of Portuguese labor markets over the last 30 years, based on the data presented in section
II.1.
Box 5 – Summary of the recent Portuguese labor market performance
• Low unemployment rates, closer to the flexible US benchmark than the high unemployment
in the rigid EU area;
• High proportion of long term unemployment, closer to EU area levels;
• High overall job creation, by European standards, but less impressive than the US miracle;
• High job creation for the low skilled, and sagging job creation for the highly skilled;
• High growth of employment in the government sector, rapidly catching-up with European
levels;
• Relatively strong, but highly volatile, productivity growth, due to economic convergence;
followed closely by compensation over the last 15 years, after a period of slow recovery
from the post-revolutionary excesses.
• The share of labor on national income is lower for Portugal than for the US or the EU
average.
In this context, the recommendations of the OECD jobs strategy for Portugal stressed
the challenges of reforming the education system to match the demand for skills and other
structural reforms in privatization and fiscal convergence, instead of the reform of labor
market institutions (see Figure 11). Overall, seven recommendations were made in the area
of labor market reform, six in the area of education and training and ten for the reform of
the business environment.
Figure 11 - Recommendation from the OECD Jobs Strategy
12
10.56
Number of recommendations
10
10
8 7
6.4
6
6
3.8
4
0
Labor Market Education and Business
Reform Training Environment
Portugal OECD (25) average
Source: OECD (1999)
In terms of labor market reform, the key recommendations were: to reduce the tax
burden on labor income, to improve the efficiency of active labor market policies, and to
improve the labor-force skills and competencies (OECD, 1996). Below, we analyze the
comparative features of Portuguese labor markets along the four dimensions highlighted in
Box 2 (pp 410).
i. Wage-formation
Wage formation in Portugal is characterized by sectoral wage bargaining with limited

coordination between bargaining units (OECD, 1996). There are three bargaining regimes.
First, a company can negotiate its own collective agreement with one or several unions
(single-firm contracts). Second, several companies can form a coalition and negotiate with
trade unions (multi-firm contracts). Third, employers’ associations, normally at the industry
level, negotiate with trade unions (sectoral contracts). Negotiations usually set wage floors,
giving the employer room for manoeuvre, as actual wages significantly exceed industry-
wide agreements (Hartog et al., 2001). Although a worker is covered by a specific
collective agreement if and only if he or she is affiliated with the respective trade union and
the firm has signed the contract, negotiated contracts are widely extended at a sectoral level
by the government. Consequently, 90% of workers in the non-agricultural sector in 1995
were covered by some form of collective bargaining (OECD, 1996).
The rate of unionization has suffered a sharp drop during the 1980’s. There are two
workers’ confederations (CGTP and UGT), organized along political lines, and a significant
number of independent trade unions. The CGTP is based on many small trade unions,
organized mostly at the industry level; while the UGT is organized at the level of profession
or category. Although the confederations can negotiate and sign collective agreements, they
have preferred to play an advisory role to the participating trade unions (Hartog et al.,
2001).
In Portugal, the representation of trade unions and their ability to negotiate depends on
union membership. The ambiguity in the representation of each trade union undermines the
potential for coordination among unions; often creating a situation of competition that
ensures wage flexibility (Portugal and Centeno, 2001). Employers’ associations are able to
416 Daniel Traça
reach agreements with the least demanding unions, while waiting for the government to
extent them nationwide (Addison and Teixeira, 1999). This flexibility is captured in the low
unemployment rate, and in the very high dispersion of labor costs in Table 1. In addition,
empirical research suggests that nominal wage growth responds rather strongly to the rate
of unemployment compared with most other OECD countries (OECD, 1996).
Figure 12 - Labor Cost Dispersion
Dispersion of inter-sectoral Dispersion of labor costs by size of
labor costs in 1989 industrial firms in 1988
Portugal 0.277 Portugal 0.253
United Sates 0.276 Ireland 0.239
Japan 0.268 Belgium 0.237
Spain 0.216 Germany 0.143
Germany 0.181 Netherlands 0.120
United Kingdom 0.170 Italy 0.116
France 0.156 United Kingdom 0.108
Italy 0.142 Denmark 0.030
France 0.133
Source: OECD (1996); Table 25, pp 85
ii. Tax and benefit systems
On the supply side, transfers to out-of-job individuals have historically been low in
Portugal. For unemployment insurance, although the level is near the median for
industrialized countries, the provisions are quite restrictive. Beneficiaries must have been
contributing for at least eighteen months in the previous two years (Bover et al., 2000).
Means-tested unemployment assistance, unrelated to earnings, is also low in international
comparisons. In 1997, a minimum guaranteed income scheme has been introduced nation-
wide, ensuring a family income dependent upon family status and employment income.
Gouveia and Farinha Rodrigues (1999) provide a description of the program, and assess its
implications for the distribution of household incomes and poverty, as well as for the size
of government expenditures. An analysis of its consequences for labor supply and for labor
markets was not available at the time of this study.
Figure 13 - Unemployment benefits by time of benefit receipt, 1997

Single Person
90
% of previous wage
60
30
0
1st month 2nd year 5th year
Portugal United States EU (avg exc GRE)
Source: OECD (1999), Table A.1, pp112
On the demand side, labor taxation is near the average for EU countries. However, like
in most other European countries, it is considered too high, and an obstacle to job creation.
In its 1996 recommendations, the OECD suggested the reduction in overall taxes on labor
income, with the view of reducing labor costs and fostering employment. In its 1999
assessment it acknowledged that, although some action had been taken, still more was
necessary.
Figure 14 - Tax wedges and employer social
security contributions8
Single worker without dependents
50
% of gross labor costs
40
30
20
10
0
1991 1994 1996 1991 1994 1996
Total Tax Wedge Employer´s Soc. Sec. Contr.
Source: OECD (1999), Table A.4; pp119
Self-employed
One factor that has minimized the effects of labor taxation, with perverse effects on other
features of the labor market, is the preferential treatment of the self-employed, in terms of
8
Total tax wedge includes income taxes, employer and employee social security contributions, but not indirect
taxes.
418 Daniel Traça
contributions to social security. Figure 15 shows that in Portugal, the social security
contributions by self-employed imply much lower labor costs than regular employment.
Consequently, arms-length employment relationships have gained ground relative to
contractual links to employers, captured in the high proportion of the self-employed in total
employment depicted in Figure 16. However, given that the social security contributions for
the self-employed in Portugal are not abnormally low, vis-à-vis other European countries,
the relatively high proportion of self employed in Portugal requires an additional
explanation.
Centeno (2002) finds that for 18 OECD countries, the cost of self-employment, measured
by the ratio of social security contributions by the self-employed to nominal GDP per
capita, is an important determinant of the share of self-employment. He finds also, that
accounting for the latter, labor market rigidity (e.g. hiring and firing restrictions) expands
the share of self-employment. In this light, we will see below that self-employment in
Portugal arises also as a response to the strictness of employment protection legislation.
Figure 15 - Social Security Contributions: self-employed and regular employment
60
Proportion of Labor Cost
40
20
0
om
s
l
ly
ce
en
ga
nd
Ita
iu
an
ed
rtu
rla
lg
ng
Fr
Sw
Be
Po
he
Ki
et
d
N
te
ni
Regular employment Self- employment

U
Source: OECD (1996); Table 29, pp 99
Figure 16 - Self-employed as a proportion of total employment
40%
% of Total Employment
30%
20%
10%
0%
1971 1976 1981 1986 1991 1996

iii. Flexibility of employment and working time
Figure 17 - Strictness of EPL for individual and collective dismissals (regular

employment)
Late Late Late Late Late Late Late Late Late

1980s 1990s 1980s 1990s 1980s 1990s 1980s 1990s 1990s
Regular procedural Notice and Difficulty of Overall EPL for EPL:

inconveniences severance pay for dismissal individual dismissalscollect/indiv.
Source: OECD Employment Outlook (1998); Tables 2.2

(pp57) and 2.4 (pp65)
The Achilles heal of Portuguese labor markets lies in the obstacles to the adjustment of
employment on the part of firms, embodied in an extremely restrictive employment
protection legislation (EPL). Figure 17 shows the comparative rankings of Portuguese EPL,
in each of the three dimensions: procedural requirements that must be followed from the
decision to dismiss to the actual termination, notice and severance pay requirements,
difficulties of dismissal related to the requirements faced by employers in the case of
unjustified or unfair termination, as well as the relative strictness of collective dismissals.
Even more striking, despite some liberalization in 1990’s, Portugal had the most restrictive
EPL in the EU (and among OECD countries) in the late 1990’s, according to the OECD.
Collective dismissals require administrative authorization, and are a minority in Portugal,
with 14.5% of total dismissals in 1996 (Bover et al., 2000).
The strictness of EPL has implications for the turnover in Portuguese labor markets.
Blanchard and Portugal (2001) find strong evidence of reduced turnover, compared with the
US market, despite the similarity of unemployment rates (see section II.1.i). Figure 18
presents their estimates of quarterly job and worker flows for Portugal and the US, obtained
from household data surveys.
420 Daniel Traça
Figure 18 - Job Flows and Worker Outflows in Portugal and US

12.6
% of employment
7.8
6.8 6.8
6.0
4.0 3.9 4.0

2.8
2.2
1.8
0.8 1.1 1.0
Entry Expansion Job Exit Contraction Job Worker

Creation destruction Outflows
(sum) (sum)
Portugal US
Source: Blanchard and Portugal (2001)
As the figure shows, quarterly job creation is Portugal stands just above 50% of the
corresponding levels for the US, with the entry/exit margin capturing a disproportionate
amount of the labor turnover in Portugal.9 Blanchard and Portugal (2001) argue that the
lower turnover in the Portuguese labor market, particularly with respect to the
expansion/contraction margin, arises due to the higher EPL in Portugal. Instead, Portuguese
firms, characterized by their extremely small sized may have be using the entry/exit margin
in adjusting to shocks. In Figure 19, we present a table that provides a summary of the
different elements of employment protection legislation in Portugal. For comparison, we
show also the conditions in the United States.
9
Interestingly, comparable annual data provides a different picture, with job creation and destruction levels
slightly higher in Portugal than in the US. This difference captures the intra-year adjustments of establishments
in the US, along the contraction/expansion margin.
Figure 19 - Employment Protection Regulation and Practice, Portugal

and United States
Component Portugal United States
Notification procedures: Written
notice to employee and employee
representative justifying the reasons
for dismissal and the lack of suitable
alternatives. In case of individual
termination for unsuitability, a
Administrative No prescribed procedures. Only a
replacement must be hired. In case of
few states prescribe a “service
procedures for economic redundancy, employee
letter” a certain period after
individual notice and representatives can call in the Labor
dismissal, noting the reason for
inspectorate to verify justification of
dismissal termination.
dismissal
Delay: After initial notification, a
minimum of two weeks for employee
or works council to present their
views, and a further 5 days before
final notice is issued
Required notice and Notice: 2 months
No legal regulations (but can be
severance pay for Severance pay: 1 month per year of
included in collective agreements)
individual dismissal service
Fair: Disciplinary reasons. After
1989 and 1991 revisions: economic Fair: With the exception of the
grounds and lack of professional or public sector, it is generally fair to
technical ability. Dismissals for terminate an open-ended
individual redundancy must be based employment relationship without
on urgent needs and must not involve justification or explanation, unless
Conditions under which posts manned by workers on fixed the parties have placed specific
individual dismissals term contracts. Dismissals for lack of restrictions on terminations.
are fair or unfair competence are only possible after Unfair: Equal employment
introduction of new technology or opportunity principles and dismissal
change to job functions. of employees with physical or
Unfair: Dismissals where employees mental impairment, if work can be
could have been reasonably, in view preformed with appropriate
of their skills and abilities, transferred workplace adjustment.
of retrained
Employee can choose between
Compensation and reinstatement with full back pay
related remedies counting from the date of dismissal to
following unjustified the actual court sentence; or
compensation of one month of pay
dismissal
per year of service
422 Daniel Traça
Figure 19 - Employment Protection Regulation and Practice, Portugal

and United States(cont)
Component Portugal United States
Definition: 2+ workers, in firms < 51
employees; 5+ workers, otherwise.
Notifications: Duty to inform and
Definition: In firms >100
consult with works council or trade
employees, 50+ workers in case of
union delegation. Notification of
plant closure; 500+ workers in case
Labor inspectorate required.
Procedures and of layoffs; 50-499 workers, if
Negotiation: Consultation on
greater than 1/3 of workforce.
standards for collective alternatives to redundancy, selection
Notifications: Duty to inform
dismissal standards and ways to mitigate
affected workers and labor unions,
effects. Written agreement to be
where they exist, as well as state and
reached, if necessary via conciliation
local authorities.
of Labor inspectorate.
Delays: 60 day notice period.
Delays: 75 days if agreement on
procedures can be reached; 90 days
otherwise.
Source: OECD Employment Outlook, June 1999
Fixed term contracts
An increasingly important way to circumvent strict EPL, in Portugal and the rest of the
EU, is the use of fixed-term contracts. Temporary employment, which includes also a fast
growing but relatively small number of workers employed in temporary work agencies, is
also the object of severe legal restrictions in most EU member countries, including
Portugal. Figure 20 shows the relative strictness of legislation on temporary employment.
Among 26 OECD countries, Portugal ranks 19th in the strictness of its regulations of fixed-
term contracts. Fixed-term contracts are permitted, inter alia, for business start-ups,
launching a new activity of uncertain duration, and recruiting “risky” workers.
Figure 20 - Strictness of regulation for fixed-term contracts
Late 1980s Late 1990s
Source: OECD Employment Outlook (1998); Table 2.3, pp63

Figure 21 - Fixed term contracts as a proportion of total employment
40
% of total employees
30
20
10
-
1986 1990 1994 1998
Portugal Spain EU (avg exc AUT, FIN, SWE)
Source: Eurostat, Labor force survey
As a consequence of high firing costs for regular workers, due to EPL, fixed-term
contracts, with a traditional duration of six-months, have gained increasing weight in
Portugal. Figure 21 captures the relative incidence of fixed term contracts as a share of total
employment. Fixed-term contracts remained near the EU average between 1992 and 1997,
with around 10% of total employment. However, despite legislative changes in Portugal in
1989 that made fixed-term contracts more restrictive (Bover et al., 2000), fixed-term
contracts have suffered a dramatic increase, in recent years, reaching 20% in 2000.
Figure 22 - Job turnover and worker turnover, by type of contract (1991-98)
40
% of employment
30
20
10
0
All workers Permanent Temporary
workers workers
Job Creation Job Destruction Job Turonver
Worker Turnover Separation rate Hiring rate
Source: Varejão and Portugal (2001)
The relevance of temporary workers is even more striking in terms of flows. Figure 22
presents evidence for job and worker turnover in Portugal for 1991-98, i.e. before the recent
(post-1997) surge, from Varejão and Portugal (2001). Job turnover captures the creative-
destruction of work posts, while worker turnover captures the hiring and separation of
workers, regardless of whether the post was created or extinguished. Overall, fixed-term
contracts accounted for 62% of all accessions and 43% of all separations, while the
turnover of temporary workers was about three times larger than that of permanent workers
(29 and 10 percent of total worker turnover, respectively).
424 Daniel Traça
Several features from Figure 22 are worth noting. First, fixed-term contracts are related
to temporary tasks, as indicated by the higher job turnover in this category. Second, worker
turnover, relative to job turnover, for workers with temporary contracts is much higher than
for permanent workers. Third, the hiring rate is higher than the separation rate for
temporary workers, while the opposite is true for permanent workers.
Varejão and Portugal (2001) argue that the last two facts are related to the notion that
these fixed-term contracts are used as a screening device, with four to five percent of
workers initially hired with fixed term contracts actually receiving a permanent contract.
However, a more pessimistic view is presented in Blanchard and Landier (2000), who
present evidence of higher relative worker turnover for temporary workers in France. They
argue that this arises as a perverse effect of the differences in firing costs between regular
employment and fixed-term contracts, as this asymmetry makes firms more reluctant to
keep workers that were hired with fixed-term contracts on regular jobs, even if a match
turns out to be quite productive. From this perspective, fixed-term contracts provide a poor
substitute for reform of employment protection legislation and may result in more low-
productivity entry-level jobs, with the negative implication that firms will reduce the
investment in the worker, which is key to expand productivity.10
In addition to temporary contracts, Portuguese firms have tended to circumvent
employment regulations with arms-length relationships through self-employment
procedures (see Figure 16). Furthermore, Portugal’s small firms being much smaller that in
other European countries, probably find it easy not to comply fully with employment
regulations, as they do not pay taxes and social security contributions. Perhaps for this
reason, according to OECD (1997, pp. 68) employer surveys show that, in 1994, only 33%
of employers (36% for companies with 500+ workers) claimed that hiring/firing practices
were very important or important reasons for not employing more people. Of course, the
extent of selection bias in the sample must be taken into account.
iv. Active labor market policies
The Institute for Employment and Vocational Training (Instituto de Emprego e

Formação Profissional, IEFP), which operates under the supervision of the Ministry of
Labor and Solidarity, is responsible for the public employment service in Portugal. Charana
and Rodrigues (2001) provide a description of its activities. Figure 23 and Figure 24
provide some evidence on the activities of the IEFP in the fields of job placement and
training, vis-à-vis several other countries.
10
Blanchard (2000) presents evidence that the easing of fixed-term contract legislation in France has increased
turnover, without substantially increasing unemployment duration. This intuition is also consistent with the high
labor flows between unemployment and fixed-term contracts in Spain documented by Bover et al. (2000).
Figure 23 - Vacancies and Placements in the PES, relative to Total Hiring
60
40
20
0
BEL DEN FIN GER IRL NED POR ESP UK
PES Vacancies/Total Hirings PES Placements/ Total Hirings
Source: OECD (2001), pp 37; Table 1
Figure 24 - PES Expenditure in Training, relative to GDP
1.2
1
Percent of GDP
0.8
0.6
0.4
0.2
0
US DEN FIN SWE AUT BEL FRA GER IRL LUX NED UK POR ESP
Unemployed Employed
Source: OECD (2001), pp 51; Table 2
In terms of job brokerage, the placement activities of the IEFP in Portugal play a much
smaller role in terms of labor market turnover, than in most European countries. Vacancies
posted in the IEFP and subsequent placements through the service constitute only 14 and 8
percent, respectively, of total hirings. On the other hand, the expenditure with training is
about average for the sample of countries in the table. However, Portugal is the only
country with a larger share of training expenditure on the employed than on the
unemployed. An important determinant may be the low unemployment rate in the country –
note that the expenditure is training for the unemployed is similar to the US, with a similar
unemployment rate. However, the extraordinarily high expenditure with training of the
employed raises questions about the matching of that training to the needs of the specific
job of the employed worker.
In its 1996 report, the OECD concluded that the overall performance of ALMP in
Portugal has been unsatisfactory, and suggested as key areas for reform: more evaluations
426 Daniel Traça
of the existing policies and targeting of the long-term unemployed. In a recent review
(OECD, 1998), the OECD pointed out that no progress had been made in these directions.
In sum, the labor market institutions in Portugal can be summarized in the following points:
Box 6 - Summary of labor market institutions in Portugal
• Wage formation: decentralized and flexible.
• Labor taxes: high, but in line with other EU countries; biased toward the self-employed.
• Unemployment subsidies: Moderate levels but very strict criteria.
• Employment Protection: Extremely strict, including legislation of temporary work.
• Active labor market policies: Ineffective job placement role; low level of vocational
training for the unemployed, with perversely large role for training of the employed
III. Challenges for the next 30 years

Section III of this study addresses the challenges facing labor markets in Portugal, and
more generally in Europe, over the next 30 years. Part 1 identifies the fundamental drivers
of change, and Part 2 assesses the implications for labor markets. Two key trends for labor
markets are outlined: the continuation of the shift in labor demand toward the highly skilled
and an increase in the volatility of labor demand. Focusing on the latter, we speculate on
the implications for labor markets and on the critical factors of success in coping with this
new trend in Part 3. Finally, Part 4 looks at the Portuguese labor markets in comparison
with the success factors outlined, and suggests directions for reform.
1. Drivers
At the same time that it pushes through with its continuing structural convergence to
Europe, the Portuguese economy will be affected by some new fundamental changes, in the
medium to long-term. These changes, which will have an effect also on the rest of Europe,
are predicated on three key drivers, summarized in Box 7. Although they represent a vast
potential for the improvement of the European competitiveness, these changes pose
important challenges to the functioning of labor markets. Moreover, their impact for the
different member countries will be differentiated. Next, we address each of these drivers in
detail.
Box 7 - Drivers of structural economic change in Europe, in the medium term
• The expansion of trade and capital flows, at a global and European scale;
• The increase in the speed of technological change and institutional change in capital
markets.
• The implementation of the European Monetary Union (EMU), and the fiscal constraints of
the stability pact.
i. Trade and capital flows.
The increased integration of goods and capital markets is a well-known feature of the
last twenty years. At a global level, the volume of international trade and capital flows has
expanded at a much faster rate than the growth of output. At the same time, global
integration has expanded in geographical depth, with the increasing role of developing
countries in the world economy. At a European level, the completion of the Single
European Market has increased economic integration, both in terms of trade and flows of
capital.
These trends are likely to continue, and gather increasing pace. On the one hand, the
increased awareness that developing countries have gotten the shorter end of stick in the
Uruguay Round has focused the new round of WTO negotiations in improving the
conditions for those countries. A key element of these negotiations will be improving the
market access of developing countries to industrialized countries, namely the European
Union. Particularly noteworthy is the accession of China to the WTO. On the other hand, at
an intra-European level, the deepening of the Single European Market and the eastward
expansion of the EU will also augment the competitive pressure on product markets, which
will trickle down to the labor market.
The increased pressure will have two implications for labor markets. First, the factor
endowments of Eastern European and other developing countries imply that poorer
countries in the EU, like Portugal, will be the most affected in terms of unskilled labor
competition. Second, the increased competition and economic integration will enhance
volatility and uncertainty in labor markets. We address each of these aspects in turn.
a Unskilled-labor Competition
The low level of skills and capital abundance in Portugal has produced a pattern of
specialization and comparative advantage that will suffer most from the increased
competitive pressure from Eastern European and other developing countries, with a similar
factor endowment. As these countries obtain improved access to the currently protected EU
markets in sectors such as textiles, clothing and footwear, the Portuguese terms of trade are
likely to deteriorate.
It should be noted that the increased competitive pressure from developing countries,
outside the EU, and from Eastern European countries, within the EU, will arise not only
due to trade, but also due to the increased mobility of capital, with the flow of investment to
lower wage countries. In addition, Eastern European countries will also imply increased
rivalry in the demand for EU structural funds.
Overall, the competitive pressure from Eastern European and other developing countries
is likely to hurt mostly those with less skills, and working in labor-intensive industries in
direct competition with low-wage countries. This is clearly the case of Portugal. For
countries with a larger skill base, or for individuals with higher skill attainment, such
competition will imply increased benefits, due to the gains form trade.
b Volatility and uncertainty of labor demand
Another consequence of the increased integration of trade and capital markets for labor
markets is the increase in the volatility of labor demand. As described in Traca (2001),
increased international competition increases the volatility of labor demand, in the presence
of local/industry specific productivity shocks for two reasons: first, the increased elasticity
of demand reduces the compensating price-effect that occurs when industries are shielded
from foreign competition; second, international exposure allows the pass-though of external
productivity shocks to domestic labor markets. Bhagwati and Deheija (1998) have
428 Daniel Traça
addressed this as kaleisdoscopic comparative advantage. Using data from the United States,
Traca (2001) shows that indeed industries more open to international competition have
higher wage volatility.
The increased integration of product markets at a global and European scale is likely to
generate a similar process of increased volatility of labor demand. This will not only arise
due to trade with Eastern European and other developing countries, like the structural
adjustment of the previous section, but also due to the increased integration within the EU
and with other developed countries.
ii. Technological and institutional change

A key aspect of the end of the 1990’s, which is likely to carry on in the medium- to long-term
future, are the changes to technology and capital market institutions.
a Technological change
The computer revolution of the last 20 years has produced a dramatic change in society.
Although this has generated an unprecedented surge in productivity growth, fuelling the
longest expansion in history in the United States, technological change, in particular of the
type that has arisen from the computer and Internet revolutions imposes added pressure on
labor markets.
• The complementary between the new technology and the skills of the labor force.
One of the key features of the technology revolution has been the extent to which
it has substituted the unskilled worker in the factory floor and, on the other hand,
increased the demand for skilled workers, able to handle complex machines.
• The “riskiness” of innovation. The process of innovation is risky and fraught with
uncertainty. The computer and Internet revolutions have taken place in the context
of micro-enterprises, with many of them going under in a short period of time.
Even for large companies, the risk of being replaced by upcoming companies and
losing competitive advantage means high risk of bankruptcies and contractions of
economic activity. Easy adjustments of labor costs become key to survival.
• Increased competition due to the reduction in the cost of information. The

expansion of the Internet will increase (product and labor) market competition. As
mentioned above, this increase in competition is likely to expand the volatility in
labor markets, as firm- and industry-specific shocks will no longer be passed on to
consumers, being absorbed instead by the labor margin in the firm.
b Institutional change
Another important driver of labor market changes will be the institutional changes in
product and capital markets. In product markets, increased de-regulation will lead to greater
competition, with the already mentioned effects on the volatility of labor demand. On
capital markets, a less noticed change has been happening across Europe, with the interests
of the shareholder gaining increased preeminence in the minds of managers – the
presumption of shareholder value maximization is that all changes that increase the value of
the stock are the sole objective of the shareholder. This represents a dramatic change from
traditional forms of management in Europe, which were focused on the interest of the
stakeholders, including the firm’s labor force. The cross-national aspect of value-based
management has also increased in the last decade, and is likely to intensify within Europe,
with the advent of the EMU.
From this perspective, and given the problems of short-termism in capital markets, the
increased focus on shareholder value will reduce the commitment of the firm to the worker
and the willingness to ensure workers through labor hoarding. Firms will be more likely to
expand and contract their labor force, according to the needs of profitability. This trend has
become increasingly clear in countries like France and Germany, where traditional forms of
stakeholder management were more entrenched, and the changes are more striking.
iii. European Monetary Union

Last, but not least, a dramatic change will occur in the European economies with the
advent of the EURO, in the context of European Monetary Union (EMU). The implications
of monetary unification in the literature are well established, and relate to the loss of the
exchange rate as an adjustment mechanism to countrywide shocks. Less attention has been
paid to the role of the EURO in increasing product market competition and cross-border
trade and specialization within Europe, and the consequences for labor markets. We address
each of these in turn
a The loss of the exchange rate flexibility
A key aspect of monetary unification is that the nominal exchange rate will no longer
constitute an adjustment margin. Adjustments in the nominal exchange rate are useful to
minimize the price adjustments necessary to reestablish the equilibrium of the real
exchange rate, in response to asymmetric countrywide shocks.
Friedman (1953), in a well-know analogy, best captures the role of nominal exchange
rate flexibility. He compares the gains from flexible exchange rates to those from setting all
clocks back one hour in the fall and forward in the spring: it is more efficient to change the
nominal time standard (the nominal exchange rate) that it is to require millions of
individuals to adjust their daily time schedules (nominal domestic prices), waking up and
getting to work an hour earlier or later, to the annual solar cycle (shocks to a country’s
demand and supply).
This analogy highlights the substitutability between exchange rate adjustments and
nominal and real rigidities. If prices and wages do adjust easily, then the loss of the
exchange rate as an adjustment margin will matter little. Otherwise, in the presence of
nominal or real rigidities (and if mobility is low), the adjustment mechanism will start to
falter and full-employment will be at risk.
Asymmetric shocks in Portugal
As mentioned above, a key element for the consequences of the loss of exchange rate
flexibility is the incidence of asymmetric countrywide shocks. Numerous empirical studies
have dealt with the incidence of asymmetric shocks in the EMU area, at the level of regions
and countries, in the pre-EMU period. They find that variations of output are more
synchronized, and variations of real exchange rates smaller, among EU member states than
430 Daniel Traça
among regions within individual member states (see, for example, Fatas, 1997 and
Decressin and Fatas, 1995). In other words, country shocks are less important than regional
shocks, which roughly average out at the country level.
However, there are differences among member countries. In a core group of countries –
Germany, France, the Benelux, Austria and Denmark – asymmetric shocks have been
relatively rare as variations of output and exchange rates were quite similar; by contrast, the
incidence of asymmetric shocks has been rather high for Portugal, Greece, Spain, Italy,
UK, Ireland, Sweden and Finland. The higher exposure of these countries to asymmetric
shocks can be explained by their less diversified nature, due to their smaller size, and higher
exposure to shocks from outside the EU.
A key issue is whether these shocks are indeed country/region-specific, affecting all
industries within a region, or if they result from the interaction between industry-specific
shocks and a diversified industry-structure. Bayoumi and Prasad (1997) find that the
importance of aggregate, industry-specific and country- or region-specific shocks is roughly
similar in Europe and the US, as each of these shocks plays an important role. Moreover, a
large portion of the variance in their study is unexplained, and should be attributed to
shocks that are specific to a country/region-and-industry. They show also that region-
specific disturbances in the US are more important in the non-traded goods sectors, while
country-specific disturbances in the EU are more prevalent for traded goods.
These historical patterns may not persist with the EMU, as, for example, monetary
shocks will no longer be country-specific. Two possibilities are raised:
• Frankel and Rose (1998) argue that an increased synchronization of business

cycles within the EMU may be expected, because trade linkages between
European countries are likely to increase further. This outcome is predicated in the
notion that shocks are indeed country-specific, affecting all industries within a
country. This it is more relevant for the case of demand shocks, or if intra-industry
trade is predominant.
• A different view is presented by Krugman (1993). He argues that increased trade
specialization, arising from monetary integration, will increase the asymmetry of
shocks at the country level. This position is predicated in the notion that shocks are
industry-specific, namely productivity shocks, and their country effects are
determined by the country’s specialization.
The empirical evidence is scant. Krugman (1993) supports his view on empirical data
for the monetary union in the US. This is in accordance with recent estimates in Traca
(2001) showing that the industry-specific component constitutes the bulk of productivity
shocks in the United States. Frankel and Rose (1998) present evidence on synchronization
of business cycles in industrializes countries over the last 30 years.
In a recent study, Barbosa et al. (1998) presents some evidence that after 1985, with the
increasing integration of the Portuguese economy in the EU, the Portuguese cycle has
become increasingly correlated with the European one. They argue that this convergence is
the result of increased trade, as well as ever more similar macroeconomic policies.
Therefore, although scant, evidence seems to point that the integration of monetary is likely
to lead to a convergence of economic cycles.
b Increased trade and competition
A less noted aspect of the effects of monetary union on labor markets is the extent to
which it may increase volatility and uncertainty due to its effect on trade and capital flows.
As mentioned in section III.1.i, increased international competition implies an amplified
volatility of labor demand, if shocks are industry-specific, due to the magnification and
transmission effects.
The evidence that a currency union affects trade is presented in Glick and Rose (2001).
Using a large annual panel data set covering 217 countries, from 1948 to 1997, they show
that countries that left currency unions experienced economically and statistically
significant declines in bilateral trade, after accounting for other factors. For the symmetric
case, they estimate that a pair of countries that starts to use a common currency experiences
a doubling in bilateral trade. In sum, it is likely that the EMU will intensify trade and
competition in the Single market, raising the potential for increased volatility of labor
demand in the presence of industry-specific shocks.
c The pressure to reduce the public sector
Another key element of EMU are the limits to budgetary deficits and debt imposed by
the Stability Pact, which have led to increased pressure to reduce the size of public sector
employment. As was clear in section II.1.ii, the number of public sector workers and the
expenses that they entail is extremely high in Portugal. Since public sector jobs are safer,
the decline in public employment’s share will expand private sector’s more volatile jobs.
2. Labor market trends
Now, we look at the implications for labor markets of the three drivers of change
mentioned above. We identify two main trends: the continuation of a shift in labor demand
toward skilled workers, against the unskilled, and the expansion in the volatility and
uncertainty of labor demand. Below, we address each of these trends in detail.
i. Shift in the relative demand for skills

Several shocks are likely to contribute to an expansion of the relative demand for
skilled, educated workers. These shocks include the increased competition from Eastern
European and other developing countries, which are abundant in unskilled labor, and the
pace of technological change, which requires complementary labor force skills. If
unchecked by a corresponding change in supply, this demand shift will lead, in a best-case
scenario, to an increase in the relative wage of skilled workers; and in the worst-case, to a
decline in the real wage of unskilled workers.
As became clear in section II.1, such trends have already been present in the labor
market developments. The last 30 years have been characterized by low wage growth in the
United States, particularly for unskilled workers, and increased unemployment in Europe,
again for unskilled and young workers.
There has been a myriad of empirical studies addressing the causes of these trends,
particularly attempting to identify the role of low-wage competition from developing
countries vis-à-vis the effects of technological change. For the case of the US, these studies
432 Daniel Traça
have overwhelmingly concluded that the role of trade with developing countries amounts to
no more than 20-30% of the increase in the wage-skill gap (see Cline, 1997 for a survey).
In terms of European unemployment, the evidence also points to a reduced role of trade
with the LDC’s (Neven and Wyploz, 1999; Dewatripont et al., 1999). Technological
evolution and domestic policy choices are much more important factors behind Europe’s
labor market problems. In fact, trade with LDC’s should be even less harmful for Europe
than for the US, since Europe has a current-account surplus in the aggregate and is less
open than the US to LDC trade.
Since the paces of technological change and exposure to low-wage competition are
likely to increase, the pressures on unskilled workers are likely to mount. Responses to this
challenge can be of two types: defensive, like slowing the speed of technological change
and reversing the process of Globalization; and accommodative, such as adjusting supply
by increasing the skill attainment of the labor force. In reality only the latter constitute real
alternatives. Defensive policies are highly impractical, for two reasons: they are very
difficult to implement and they would yield extremely high aggregate costs (undermining
the gains from trade and the benefits of technological progress).
ii. Increased in the volatility and uncertainty of labor demand
A second element of change in labor markets will be an increase in labor demand

volatility (LDV). Box 8 summarizes the factors contributing to this trend. It should be
mentioned that, even if they expand volatility, and may thus impact workers adversely,
these shocks represent a vast potential for value creation, which will boost output per
capita.
Box 8 – Sources of higher labor demand volatility (LDV)
• The enhanced exposure to international trade and competition, not only from low-wage
countries but also from other industrialized countries, due to Globalization and the EMU;
• The increased pace and riskiness of technological change;
• The institutional changes in product and capital markets, with de-regulation and increased
market competition raising the randomness of market conditions for specific firms; and
shareholders increasing the pressure for quick adjustments on labor costs;
• The contraction in the public sector reducing the supply of safe, secure jobs.
a Implications for labor markets
The implications of an increase in LDV to labor markets depend on the specific labor
market institutions and the characteristics of the workforce; namely, the extent of labor
mobility and adaptability, the wage flexibility, and the constraints on contracts (see Box 9
for definitions).
Box 9 - Defining Adaptability and Flexibility

• Adaptability captures the ability of workers and their employers to adjust to the shock by
moving to another country (in the case of a country-specific shock) or another industry (for
an industry-specific shock).
• Flexibility depicts the ability of wages and prices to adjust to the shock, bringing the real
wage in line with marginal productivity in the different industries and/or countries. It
encompasses also the facility to adjust working hours and shifts.
In order to address the positive and normative implications of an increase in labor

demand volatility, Figure 25 presents a diagrammatic exposition of the effects of a decline
in the productivity in sector/country Y in spot labor markets, under different assumptions
on adaptability and flexibility.
Figure 25 - Effects of volatility under different labor market scenarios
a) Adaptability and Flexibility b) Flexibility

Losses relative to
Adaptability
LX LY LX LY
L*Y L*Y
L1 L2 L2 L1
Sector/ Country X Labor force Sector /Country Y Sector/ Country X Labor force Sector /Country Y
c) No Adaptability, No Flexibility
Losses relative to
Adaptability
LX LY
L*Y
unemployment
L1 L2
Sector/ Country X Labor force Sector /Country Y
The best response, i.e. the one entailing maximum welfare, is the movement of labor to
country/industry X. This is depicted in diagram a) by the expansion in sector X
employment from L1 to L2. As workers flow from countries/industries where shocks were
adverse to those where shocks were favorable, their productivity and standard of living will
increase. This is possible only in the presence of adaptability and flexibility. The latter is
required, as the equilibrium competitive wage will also fall. In this case, an increase in
LDV raises job and worker turnover, and job insecurity.
A second best response, in the absence of adaptability, is that prices adjust in order to
ensure full-employment. This is captured in diagram b). The relative wage of the
country/region with an adverse shock falls, ensuring the country/industry remains
competitive. Note that the wage decline in sector/country Y is smaller than in the case of
adaptability; - from this perspective, there is some substitutability between adaptability and
434 Daniel Traça
flexibility. In this case, higher LDV means higher wage volatility, instead of job turnover.
This adjustment is always worse than the mobility of labor, since workers will be stuck in
low productivity countries/industries. It should also be mentioned that, in the case of a
countrywide shock, the decline in the relative wage arises as a depreciation of the real
exchange rate.
Finally, diagram c) captures the worst-case scenario, where there is no adaptability and
no flexibility. In this case, the inability to bring down real wages in industry/country Y will
hurt its competitiveness and lead to layoffs and bankruptcies. Since workers cannot move to
X, the result will be unemployment, and an extremely high deadweight loss.
If shocks are temporary, workers and firms may choose to respond to increased LDV by
transferring the wage and employment uncertainty to the less risk-averse employer. As
shown in Traca (2000), given the limitations on labor market contracts and the nature of the
shocks mentioned above, complete insurance of the worker in impossible.
Also in the case of contracts, the consequences of LDV depend on the adaptability of
the labor market. The main difference in this case is that for low adaptability, i.e. if
mobility costs are high, higher LDV implies not only higher wage volatility but also a
decline in wage growth and in the labor share, as workers hedge against the probability
becoming unprofitable in the event of low productivity. Alan Greenspan (1997), in a well-
known testimony, suggested that: “a restriction on compensation increases has been evident
for a few years now and appears to be mainly the consequence of greater work insecurity”.
In addition to increased separation rates, job turnover and wage volatility, and slower
wage growth, other implications of the increase in labor market volatility are also
important. Below, we speculate on some of these.
First, the end of the life long jobs has implications for the investment in the human
capital of the workforce. As mentioned above, increased LDV will expand the odds of
voluntary or involuntary separation, as firms are forced to contract or workers find new,
more profitable opportunities in other industries and/or regions. As a result, the attachment
between the worker and the firm will become more tenuous. Hence the firm will have fewer
incentives to invest in its labor force, and training will become increasingly a responsibility
of the worker. Given the well know shortcomings of markets for human capital, this may
hurt productivity growth. Moreover, half-hazardous responses, like the expansion of short-
term contracts, are likely to be compounding factors.
Second, the slower wage growth and the decline in the labor share will put pressure on
the income distribution. From this perspective, the traditional form of social contract
between labor and capital that has emerged in the post-war period will be reassessed. With
workers moving from job to job, the firm will no longer provide for the long-term needs of
the worker’s household.
Third, the end of job security has also psychological and sociological implications, as
the distress and social pressure on workers increases. Recurring periods of joblessness and
search, for example, will be costs associated with volatility.
Finally, fourth, the higher mobility of the labor force will cause difficulties for trade
unions, namely those operating at a sectoral/regional level. Since the passage of each
worker by the group will become shorter, his/her commitment will decline.
b Recent empirical evidence
There is increasing evidence of a trend toward increased volatility. Gottschalk and

Moffit (1994) present evidence of a significant increase in the short-term volatility of wages
in the US. Farber (1997) concluded that “after controlling for demographic characteristics,
the fraction of workers reporting more than 10 and more than 20 years of tenure fell
substantially after 1993. Looking at displacement rates, Aaronson and Sullivan (1997) show
that “overall displacement rates were higher in 1995 than at any time since the data began,
in 1979”. They show also that displacement rates for highly educated, white-collar and
service workers have risen especially fast. Aaronson and Sullivan (1997) present an
extensive survey of the research on turnover, job tenure and displacement. Katz and
Krueger (2000) argue that national data show a slight decline in job tenure and increase in
displacement rates in the mid-1990’s.
For the UK, Nickell et al. (2000) shows that there has been a rise in job insecurity for
British men since the early 1980’s, which has come from a rise in the costs of job loss and a
probability of substantial year-on-year fall in real wage. Although similar studies for
continental Europe are not available, the role of temporary work and short-term contracts in
the current surge in employment shows that similar trends are at work. In Portugal, the
recent growth of fixed-term contracts captured in Figure 21 suggests also that the trend to
increased LDV has arrived.
3. Policy Responses to Increased Labor Demand Volatility
As was clear from section III.2.ii.a, the policy responses to increased LDV must rely on a
two-pronged approach to enhance adaptability and increase flexibility. This will constitute a
key challenge in the medium-term, at the policy level, as well as for the individual worker.
This section addresses the main determinants of adaptability and flexibility, and looks at the
Portuguese case, suggesting some directions for reform. Recall that both these concepts
were introduced in Box 9 (pp 433).
i. Adaptability
The determinants of the adaptability of the worker depend on the nature of the shock.
For country-specific shocks, adaptability implies migration. For the case of industry
shocks, the ability of workers to move across industries (inter-sectoral mobility) is key. In
this case, workers will respond to industry shocks by moving to higher productivity
industries. If industry structures are diverse across countries, inter-sectoral mobility must be
coupled with regional mobility, as workers must move to expanding industries in different
countries.
Geographical mobility
The importance of geographical mobility is likely to increase with the EMU, as the loss
of exchange rate flexibility will enhance nominal rigidities in the response to country-
specific shocks. Contrary to the United States, the geographical mobility of labor in Europe
is extremely reduced, even at the regional level (Decressin and Fatas, 1995; Bayoumi and
Prasad, 1997). Difficulties related to language barriers and other institutional factors make
the process extremely costly in Europe. Although Portuguese nationals have historically
436 Daniel Traça
represented one of the most mobile group, with large waves of out-migration in the 1960’s
responding to the dramatic wage differentials between Portugal and the European core, it is
unlikely that such mobility will be enough to operate as an adjustment margin.
Education and Adaptability
We speculate that a key driver of the adaptability is education, and that the increase in
LDV is likely to contribute to an increase in the education premium. We think so for two
reasons. First, education provides general human capital, transferable across sectors and
firms, while on-the-job learning of unskilled workers is specific to firm and industry
(Grossman and Shapiro, 1982). Second, from an informational perspective, it improves the
ability to assess market conditions, and to track down opportunities.
There is increasing evidence of a positive correlation between education and
adaptability. Pissarides and Wadsworth (1989) provide evidence that education facilitates
geographical mobility. Mauro and Spilimbergo (1998) show that, in Spain, the highly
skilled are found to migrate very promptly in response to a decline in local labor demand,
whereas the low skilled drop out of the labor force or stay unemployed for a long time. In
terms of sectoral mobility, Fallick (1993) finds that education increases the probability that
a displaced worker moves to a different industry.
In this context, Traca (2000) argues that less skilled workers are more likely to respond
to labor demand volatility by accepting wage cuts, for the sake of increased job security;
while the skilled (more adaptable) will suffer higher displacement rates and move to other
industries.
Labor market institutions and Adaptability
In terms of labor market institutions, a key aspect of adaptability is the employment
protection legislation. EPL works as a barrier to entry, since it increases the costs of
adjustment in case of a negative shock. In a world of high volatility, firms must be able to
open and close with minimum costs, or to contract and expand the labor force to respond to
markets. As mentioned above, fixed term contracts provide only an imperfect substitute.
Another factor that enhances the adaptability of the labor force are the active labor
market policies. Since periods of unemployment will be more frequent, the challenge is to
make them shorter and more productive. Three elements are key. First, a reduction in
search costs: using technology to reveal information about opportunities, facilitating
procedures for unemployment insurance, providing facilities for workers to develop job
search activities (telephones, computers, etc.). Second, an intensive program and
opportunities for retraining while unemployed and continuing education, which will
enhance the ability to adjust to new tasks. Finally, third, elements that are crucial for
regional mobility, such as the housing market, should also be addressed.
ii. Flexibility
Flexibility of real wages to follow changes in productivity is another key factor of

success is responding to LDV. It is directly related to wage-setting mechanisms, and the
role and power of “insiders”. As has become clear from the previous sections, European
labor market performance in this domain is unsatisfactory. Although some institutional
changes have been made, attempts at reform have failed to produce the necessary overall.
Flexibility of real wages is also likely to suffer from the increase in nominal rigidities
due to the loss of exchange rate as an adjustment margin. Hence, structural reform to
increase wage-flexibility in Europe becomes paramount. According to Soltwedel et al.
(1999), countries like Finland, Italy and Spain, subject to highly asymmetric shocks, with
little labor mobility and extremely inflexible labor markets face a high risk of increasing
unemployment.
Another important aspect of flexibility is the ability to adjust working hours to the
demands of the marketplace. Periods of low demand should permit shorter working hours,
allowing workers to devote time to other activities, including searching for alternative jobs
and vocational training
To summarize, Box 10 presents the key policy responses to increase adaptability and
flexibility
Box 10 – Responses to enhance adaptability and flexibility
Adaptability Flexibility
• 11
Education and
Increase skills
retraining
• Decentralized wage
Labor market de- • Lenient EPL
setting
regulation • Efficient ALMP
• Adjust of working hours
iii. Portugal: directions for reform
Finally, we address the extent to which Portuguese labor markets are ready for the
challenges of the next 30 years, in terms of adaptability and flexibility, and present some
directions for reform. Our conclusion is that while the Portuguese labor markets show high
flexibility, and thus are unlikely to see increased unemployment as a result of increased
LDV, they fare much worse in terms of the adaptability of the labor force. Recall that the
adaptability margin entails larger benefits, since it permits workers to move to
industries/regions where productivity is higher.
a Adaptability in Portugal
The key challenge for Portugal lies in the low adaptability of its labor force and
markets. As we mentioned above, the three determinants of adaptability are: the education
of the labor force, the regulation of employment protection and the active labor market
policies It is quite striking that these are also the three poorest aspect of labor market
performance in Portugal, and the ones with strongest recommendations for reform in OECD
Jobs Study (see section II.3).
Recent evidence shows that the labor market in Portugal is extremely stagnant, with
labor flows that are closer to countries like Spain, with unemployment rates 10 points
higher, than to the US, where unemployment rates are similar. Comparing data for Portugal
and the US, Blanchard and Portugal (2001) show that worker flows (as a proportion of
employment) stand in Portugal at 21 to 28 percent of US levels, while relative job
11
This aspect is also key, given the shift in the relative demand for skills mentioned in section III.2.i.
438 Daniel Traça
destructions number are much higher. Interpreting job destruction primarily as layoffs, and
worker outflows in excess of job destruction as quits, the data shows that not only layoffs
are lower in Portugal, but so are quits. In other words, US workers are much more likely
than Portuguese worker to voluntarily move out of their current position in search for better
opportunities, without waiting for a dismissal.
Figure 26 – Job destruction and Worker Outflows,
15
10
Inactivity
Exit
Unemployment
Employment
Contraction
Destruction
Outflows Total
Total
Worker
Job
Job destruction, from… Worker Outflows, to …
Portugal United States
Source: Blanchard and Portugal (2001)12
Bover et al. (2000) compare the labor market flows in and out of unemployment in
Portugal and Spain. They conclude that flows into unemployment are higher in Spain,
particularly out of temporary contract, while flows out of unemployment are similar in both
countries. What is more striking, given the much higher unemployment rate in Spain, is that
for the first nine months, the hazard rate (i.e. the probability of leaving unemployment in
quarter) is higher for Spain. The reason is the high rate of exit to fixed-term employment in
Spain.
We interpret the results of both these studies as evidence of an inability of the
Portuguese labor force and markets to adapt continuously. The reason for the low
adaptability of the labor market in Portugal can be found in many of the features mentioned
above. First, extreme EPL undermines that ability of firms to react in time to shocks, and to
take advantage of high productivity industries. Second, the low education level of the labor
force, hinders its ability to adapt to new industries, creates the incentives for the creation of
only job specific skills, which the worker cannot bring to new industries. Finally, third,
despite the recent formation of Instituto de Emprego e Formação Profissional, active labor
market policies are still not satisfactory, with training for the unemployed.
b Flexibility in Portugal
By contrast to its lack of adaptability, several forces have ensured the flexibility of
Portuguese labor markets, and kept wages sensitive to unemployment rates (see section
II.3). From this perspective, labor markets in Portugal are unlikely to suffer increased
12
Flows to employment in the US shown as average of the 2.4 – 5.4 estimate interval presented in Blanchard and
Portugal (2001).
unemployment as a result of any of the shocks described in III.1. Instead, the implication
will be an increase in wage volatility and a slowdown in wage growth, as workers
(particularly the unskilled) try to hedge the increased LDV. This may be hindered by capital
accumulation and convergence to the EU, which is likely to expand wages and productivity.
c Directions for reform
Enhancing adaptability in labor markets, through reform of labor market institutions and
of the education system, constitutes the key challenge for labor market policy in Portugal,
over the next 30 years. Key areas include the relaxation of employment protection and the
increase in the effectiveness of active labor market policies.
As shown in Figure 19, employment protection legislation in Portugal is extremely
rigid, acting as a deterrent to the creation of new firms and preventing the adjustment of
established firms to demand conditions. This factor is ever more significant, since the
development of new opportunities is the best response to the rise in LDV. Sunk costs, due
to required long term commitments to workers and poor flexibility in risk-sharing
mechanisms betweens employers and workers, are likely to work as barriers to entry.
Key elements are the procedural requirements, and the severe limitations on collective
bargaining. Although the reforms of 1989 and 1991 have greatly eased firing restrictions
through a wider range of admissible lay-off motivations, the possibility of collective lay-
offs and easier resolution of severance pay disagreements, further attention should be paid
to the following areas:
• Allow more flexibility for workers and their representatives to negotiate contracts,
including termination clauses, in order the make the mix of job-security and pay
more responsive to the characteristics of the labor force and industry.
• Reduce relative barriers to collective dismissal - rises in LDV are likely to increase
the importance of collective dismissals, for economic reasons, relative to
individual dismissals on personal grounds.
• Liberalize procedural barriers to individual dismissals and increase flexibility of
pay arrangements, including stock options.
Another key component of labor market institutional reform is the implementation of

effective active labor market policies. Commitment to active labor market policies in
Portugal has increased dramatically in recent years. However, the need for increased
monitoring is evident, from past and recent experience. Furthermore, key elements
deserving heightened attention include:
• Contestability of PES and exposure to market forces, to increase the effectiveness
of the job placement function – One example tells the story. In Australia, for
example, as from 1994, job seekers have a choice to be served by either public or
private agencies, the latter being paid a set fee for successful placements that vary
with the degree of disadvantage of the clients.
• Experimentation and monitoring – New and creative initiatives should be
promoted. Competition and experimentation among different regional offices and
governments, with a quick dissemination of success stories, is key to find new,
440 Daniel Traça
effective formulas. It is also important to ensure that initiatives are properly

monitored and evaluated on a cost-benefit basis. Evidence shows that the degree of
efficiency of different polices vary considerably, and that even the same policy
may have different effectiveness, according to the environment.
• Focus training on the unemployed – Evidence shows that firms are better place, in
terms of information and incentives, to provide for the acquisition of firm specific
skills on the part of the worker. The activities of the PES should focus on general
skill improvement for the unemployed. The current trend to the provision of
training to the already employed workers seems an unwise expense of funds.
Contestability and the interaction with the private sector are also a trend to be
encouraged.
• Separate social policies from labor market policies – Exclusion and other social
problems should be handled in a separate institutional setting from the public
employment service, at the risk of diverting attention from PES workers.
Below, in Box 11, we summarize the main suggestions for reform.
Box 11 - Directions for reform in Portugal
Employment protection legislation
• Flexibility to negotiate conditions of termination on an individual, firm basis
• Liberalize collective dismissals
• Liberalize procedural barriers to dismissals and increase flexibility of pay arrangements
Active labor market policies
• Contestability of PES
• Experimentation and monitoring, and institutional competition
• Focus training on the unemployed
• Separate social policy from labor market policies.
IV. Conclusion
This study has started by addressing the evolution of Portuguese labor markets over the
last thirty years. Against the background of developments in Europe and the US,
Portuguese labor markets have shown a very distinct pattern. On one hand, very flexible
wage negotiating institutions and relatively low price disincentives to job search have
ensured a relatively low rate of unemployment, in contrast to the towering unemployment
rates in Europe. On the other, very rigid employment protection legislation and ineffective
active labor market policies have produced a very sclerotic and lifeless labor market, with
very low turnover and a high proportion of long term unemployed, in contrast to the
dynamic US market.
We have argued that, for the next thirty years, an expansion in the volatility of labor
demand is likely to be the key transformation in labor markets in industrialized countries,
such as Portugal. Developments in international trade and capital flows, technological and
institutional change in product and capital markets, and the implementation of the EURO
all will work to raise the volatility of labor demand. Another important trend will be the
increase in the relative demand for skilled labor.
The ability to respond to the increased volatility of labor demand will be determined by
the flexibility and adaptability of labor market institutions and of the labor force. Flexibility
depicts the ability of wages and prices to adjust to the shock, bringing the real wage in line
with marginal productivity in the different industries and/or countries. Adaptability captures
the ability of workers and their employers to adjust to the shock by moving to another
country or industry. We have shown that while flexibility ensures that higher labor demand
volatility does not generate additional unemployment, the first-best outcome is obtained
only through adaptability, which allows workers and entrepreneurs to move to expanding
industries, where value-creating opportunities can be found.
Finally, we addressed the extent to which labor markets in Portugal are ready to handle
the increase in labor demand volatility, form the perspective of flexibility and adaptability.
We concluded that the flexibility of Portuguese labor markets, particularly at the level of
the decentralization of wage formation mechanisms, provides some reassurance that labor
demand volatility will not generate additional unemployment.
However, in terms of adaptability, Portuguese labor markets are likely to under perform,
preventing the economy from reaching the first-best outcome. We identified three main
determinants of adaptability: the level of educational attainment of the labor force, the ease
of hiring and firing restrictions, and the effectiveness of active labor market policies,
particularly in terms of the job-brokerage and training-provision functions. In all these three
components, we found the performance of Portuguese labor market institutions sub-
standard. Hence, we finished our study by suggesting specific measure and directions for
reform, with the aim of expanding adaptability in Portuguese labor markets. In general,
these recommendations focus on the relaxation of employment protection legislation, and
increased experimentation and monitoring of active labor market policies, in interaction
with the private sector.
442 Daniel Traça
V. References
Aaronson, D. and D. Sullivan (1998): “The Decline of Job Security in the 1990’s:
Displacement, Anxiety and their effect on Wage Growth”; Economic Perspectives,
Federal Reserve Bank of Chicago
Addison and Teixeira (1999): “Is Portugal really so arteriosclerotic? Results from a cross-
country analysis of labor adjustment” Discussion Paper n. 99-30, Zentrum fur
Europaische Wirtschaftsfor schung
Barbosa, A. P. and others (1998): “O impacto do EURO na Economia Portuguesa”,
Biblioteca de Economia & Empresa, Publicações Dom Quixote
Bayoumi, T. and Prasad, E. (1997): “Currency Unions, Economic Fluctuations and
Adjustment: Some New Empirical Evidence”; IMF Staff Papers, Vol. 44, n.1,
March 1997
Bhagwati, J. and Deheija (1994): “Freer Trade and the Wages on the Unskilled - Is Marx
Striking Again?” in Trade and Wages, Bhagwati and Kosters ed.s, American
Enterprise Institute, Washington, 1994
Blanchard, O. (1997): “The Medium Run”; Brookings Papers on Economic Activity, Vol.
2, 1997
Blanchard, O. and Landier, A. (2000): “The Perverse Effects of Partial Labor Market
Reform: Fixed Duration Contracts in France”; Mimeo
Blanchard, O. and Wolfers, J. (2000): “The Role of Shocks and Institutions in the Rise of
European Unemployment: The Aggregate Evidence; Economic Journal, Vol. 110
Blanchard, O. and Portugal, P. (2001): “What Hides Behind an Unemployment Rate:
Comparing the Portuguese and US Labor Markets” American Economic review,
Vol. 91, n. 1, March 2001
Burda, M. (1999): “European Labor Markets and the EURO: How Much Flexibility Do We
Really Need?” Mimeo
Calmfors, L. and Driffil, J. (1988): “Bargaining Structure, Corporations and
Macroeconomic Perfromence”; Economic Policy, Vol. 3, n. 6
Centeno, M. (2002): “Self-employment and Labor Market Regulation” mimeo, Banco de
Portugal and ISEG- Universidade Técnica de Lisboa
Charana, A. and Rodrigues, V. (2000): “A “Model” Local Employment Office in Portugal”
in Labor Market Policies and the Public Employment Service; OECD, Paris
Coe, D. and Snower, D. (1997): “Policy Complementarities: The Case for Fundamental
Labor Market Reform”; IMF Staff Papers Vol. 44, n. 1, March 1997
Bover, O., Garcia-Perea and Portugal, P. (2000): “Labor Market outliers: lessons from
Portugal and Spain”; Economic Policy, Vol.31, October 2000
Cline, W. (1997): Trade and Income Distribution; Institute of International Economics,
1997, Washington, DC.
Decressin, J. and Fatas, A. (1995): “Regional Labor Market Dynamics in Europe”;
European Economic Review, Vol. 39, n. 9
Dewatripoint, Sapir and Sekkat (1999): “Labor Market Effects of Trade with LDC’s in
Europe” in Dewatripoint, Sapir and Sekkat eds.: Trade and Jobs in Europe: Much
Ado about Nothing; Oxford University Press
Elmeskov, J., Martin, J.P. and Scarpetta, S. (1998): “Key Lessons for Labor Market
Reform: Evidence from OECD Countries’ Experiences”; Swedish economic
Policy review, Vol. 5, n. 2
Fallick, B. (1993): “The Industrial Mobility of Displaced Workers”; Journal of Labor

Economics, 1993, Vol 11, n. 2
Fatas, A. (1997): EMU: Countries or Regions? Lessons from the EMS Experience”; CEPR
Discussion Paper n. 1558
Farber, H (1996): “The Changing Face of Job Loss in the United States, 1981-1993”;
NBER Working Paper #5596, May 1996
Frankel, J. and Rose, A. (1998): “The Endogeneity of the Optimum Currency Area
Criteria”; Economic Journal, Vol. 108, n. 449
Friedman, M. (1953): “The Case for Flexible Exchange Rates” in Essays in Positive
Economics, University of Chicago Press
Garibaldi, P. and Mauro, P. (1999): “Deconstructing Job Creation”; IMF Working Paper
WP/99/109
Gelb A., Knight, B. and Sabot R. (1991): “Public Sector Employment, Rent Seeking and
Economic Growth”; Economic Journal, Vol. 101, September 1991
Glick, R. and Rose, A. (2001): “Does a Currency Union Affect Trade? The Time Series
Evidence”; NBER Working Paper #8396, July 2001
Gottschalk, P. and R. Moffit (1994): “The Growth of Earnings Instability in the U.S. Labor
Market”; Brookings Papers on Economic Activity 2, 1994, Brookings Institution,
Washington DC.
Gouveia, M and Farinha Rodrigues, C. (1999): “The impact of a “Minimum Guaranteed
Income Program” in Portugal”, ISEG/DE Working Papers, CISEP
Greenspan, Alan (1997): “Monetary policy” - testimony and report before the U.S. House
Committee on Banking, Housing and Urban Affairs, 104th Congress, 1st session,
February 26, 1997
Grossman and Shapiro (1982): “A Theory of Factor Mobility”, Journal of Political
Economy, Vol. 90 (5); Oct 1982
Hartog, J., Pereira, P. and Vieira, J.(2001): “Bargaining Regimes and Wages in Portugal”,
Mimeo
Heckman, J. and Pages, C. (2000): “The Cost of Job Security Regulation: Evidence from
Latin America”; NBER Working Paper #7775
Katz, L. and A. Krueger (2000): “The High-Pressure US Labor Market in the 1990’s”;
Brookings Papers on Economic Activity 1: 1-88; Washington: Brookings
Institution
Krugman, P. (1993): “Lessons of Massachusetts for EMU” in Adjustment and Growth in
the European Monetary Union, Francisco Torres and Francesco Giavazi eds.;
Cambridge University Press
Nickell, S and Layard, R. (1998): “Labor Market Institutions and Economic Performance”
in Handbook of Labor Economics, Vol. 3C; O. Ashenfelter and D. Card ed.s,
North-Holland
Mauro, P. and Spilimbergo, A. (1998): “How do the Skilled and the Unskilled Respond to
Regional Shocks? The case of Spain”; IMF Working Paper WP/98/77
Neven, D. and Wyploz, C. (1999): “Relative Prices, Trade and Restructuring in European
Industry” in Dewatripoint, Sapir and Sekkat eds.: Trade and Jobs in Europe: Much
Ado about Nothing; Oxford University Press
Nickell, S., Jones, P. and Quintini, G. (2000): “A Picture of Job Insecurity Facing British
Men” Center for Economic Performance, London School of Economics,
December 2000
444 Daniel Traça
OECD (1994): L’étude de l’OCDE sur l’emploi; OECD, Paris

OECD (1996): Economic Survey, Portugal; OECD, Paris
OECD (1997): Implementing the OECD jobs strategy – Member Countries’ Experience,
OECD Paris
OECD (1998): Economic Survey, Netherlands; OECD, Paris
OECD (1999): Implementing the OECD jobs strategy - Assessing Performance and Policy;
OECD, Paris
OECD (1999): Employment Outlook; OECD, Paris
OECD (2000): Labor Market Policies and the Public Employment Service; OECD
Secretariat, Paris
Pissarides, C. and J. Wadsworth (1989): “Unemployment and the Inter-Regional Mobility
of Labor”; Economic Journal, 99, September 1989
Portugal, P. and Centeno, M. (2001): “Os Salários Da Função Publica”; Boletim
Económico, Banco de Portugal, Setembro 2001
Rodrik, D. (1998): “Why do More Open Economies have Bigger Governments?” Journal of
Political Economy, 1998, vol. 106, n. 5
Scarpetta, S. (1996): “Assessing the Role of Labor Market Policies and Institutional
Settings on Unemployment: A Cross-Country Study”; OECD Economic Studies,
26
Soltwedel, R., Dohse, D. and Krieger-Boden, C. (1999): “EMU Challenges and European
Labor Markets”; IMF Working Paper, Washington DC
Soskice, D. (1990): “Wage Determination: The Changing Role of Institutions in Advanced
Industrialized Countries”; Oxford Review of Economic Policy, Vol. 6, n. 4
Tile, C. and Yi, K.M. (2001): “Curbing Unemployment in Europe: Are There Lessons
From Ireland and the Netherlands?”; Current Issues in Economics and Finance,
Federal Reserve Bank of New York, Vol. 7, n. 5, May 2001
Traca, D. (2000): “International Trade, Job-insecurity, Wages and Labor Contracts”,
Mimeo
Traca, D. (2001): “Trade exposure, Export intensity and Wage Volatility: Theory and
Evidence”, Mimeo
Varejão, J. and Portugal, P. (2001): “Why Do Firms Use Fixed-Term Contracts”; Mimeo
presented at the 16th Annual Congress of the European Economic Association,
Lausanne, Switzerland
A JUSTIÇA E SEU IMPACTO SOBRE AS EMPRESAS PORTUGUESAS ∗
Célia da Costa Cabral

Universidade Nova de Lisboa
Armando Castelar Pinheiro

BNDES e Fundação Getúlio Vargas
(Rio de Janeiro)
Fevereiro de 2002
∗
Este trabalho beneficiou de financiamento do Gabinete de Política Legislativa e Planeamento do Ministério da
Justiça bem como de um financiamento da Fundação para a Ciência e Tecnologia do MCT (proj
nº3786/ECO/99)
446 Célia da Costa Cabral, Armando Castelar Pinheiro
1. Introdução
Porquê a preocupação em inquirir junto das empresas portuguesas a sua percepção do

funcionamento da justiça em Portugal? Porquê tentar compreender especificamente quais as
decisões das empresas que o actual funcionamento da justiça afecta? Qual o interesse deste
tema? A resposta a essas questões passa pela constatação, cada vez mais generalizada, de
que o bom funcionamento das empresas e dos mercados, e por conseguinte da economia,
depende da existência de instituições sólidas e eficientes. Neste sentido, as questões que
aqui tratamos inserem-se no âmbito de um leque de temas mais alargado e que têm
começado a receber uma atenção crescente nos últimos anos: qual o papel que as
instituições desempenham no desenho do desenvolvimento económico, e qual a sua
importância quantitativa?
North (1990, p. 97) define as instituições como “restrições humanamente concebidas
que estruturam as interacções políticas, económicas e sociais. Elas consistem em restrições
informais (sanções, tabus, costumes, tradições e códigos de conduta) e regras formais
(constituições, leis, direitos de propriedade). (…) Em conjunto com as restrições
económicas padrão, elas definem o conjunto de escolha que determinam os custos de
transacção e de produção e por sua vez a rendibilidade e a possibilidade de desenvolver a
actividade económica”. De entre os trabalhos que relacionam instituições e
desenvolvimento, destacam-se os trabalhos de North (1981) e de Olson (1996).1 Estes
autores encaram as instituições como um dos principais determinantes do crescimento
económico e argumentam que as diferenças institucionais explicam uma parte importante
das diferenças de rendimento entre países. Olson (1996) vai mesmo ao ponto de defender
que reformar a política económica e as instituições é suficiente para uma nação rapidamente
conseguir alcançar uma trajectória de elevada taxa de crescimento. Estes argumentos são
reforçados pelos resultados dos trabalhos de Barro e Lee (1994) e Scully (1988), entre
outros, sendo que este mostra que os países onde as instituições funcionam correctamente
são duas vezes mais eficientes e crescem três vezes mais rapidamente.2
De entre essas instituições, assume particular relevância o Sistema de Justiça (o
conjunto do Sistema Legal e do Sistema judicial) pois, como colocado por Olson, são “os
sistemas legais que garantem os contratos e protegem os direitos de propriedade”. De facto,
um correcto funcionamento do sistema económico depende fundamentalmente do Sistema
de Justiça instituído – quer da própria legislação existente em vigor, quer do sistema
judicial que assegura o cumprimento dessa legislação.3 É o funcionamento do conjunto
1
Contribuições fundamentais a esse tema foram dadas também por R. Coase (1988), para quem é, de facto, a
noção de custos de transacção que permite compreender o funcionamento do sistema económico, analisar a
maioria dos seus problemas de forma útil e ter uma base para a determinação de políticas. Coase foi
particularmente influente em mostrar a influência das leis na atividade económica. Coase (1988) apresenta uma
coletânea dos principais trabalhos desse autor. O World Development Report 2002, produzido pelo Banco
Mundial, e que tem como título “Building Institutions for Markets”, traz ampla descrição do papel das
instituições na promoção do desenvolvimento.
2
Esta conclusão é retirada através do estudo do impacto das instituições sobre níveis e taxas de crescimento em
115 economias e conclui que países com boas instituições são duas vezes mais eficientes e crescem três vezes
mais em termos per capita, enquanto que países com instituições precárias apresentam um declínio contínuo da
eficiência. Ver Castelar (2000) e Aron (2000) para referências adicionais de estudos empíricos nesse tema.
3
A interdependência entre os dois sistemas é enfatizada por Castelar (2000), que defende que “a capacidade de os
tribunais julgarem com presteza, justiça e previsibilidade depende de as leis terem sentido claro e serem bem
escritas e consistentes com outras leis e com as práticas comerciais. Da mesma forma, para que os tribunais
A Justiça e seu Impacto sobre as Empresas Portuguesas 447
destes dois elementos que determina a facilidade ou a dificuldade de obtenção de uma

correcta afectação dos recursos de um País dados os próprios mecanismos de mercado: para
que as trocas entre agentes económicos funcionem de forma correcta e eficiente, permitindo
um aproveitamento eficaz da especialização e uma exploração eficiente das economias de
escala é essencial a existência de uma adequada protecção dos direitos de propriedade e dos
direitos contratuais. Vale dizer, o mau funcionamento do Sistema de Justiça prejudica o
desempenho económico estreitando a abrangência da actividade económica,
desestimulando a especialização e dificultando a exploração de economias de escala;
desencorajando investimentos e a correcta utilização do capital disponível; distorcendo o
sistema de preços, ao introduzir fontes de risco adicionais nos negócios; e diminuindo a
qualidade da política económica. A magnitude desses efeitos pode ser grande, dependendo
do tamanho dos problemas encontrados no Sistema de Justiça de um país. Em termos
históricos, North (1992) constata que “no mundo ocidental, a evolução dos tribunais, dos
sistemas legais e dum sistema judicial relativamente imparcial tem desempenhado um papel
preponderante no desenvolvimento de um complexo sistema de contratos capaz de se
estender no tempo e no espaço, um requisito essencial para a especialização económica”.
Um trabalho que reforça e quantifica esta idéia é o de Castelar e Cabral (2001), que
demonstra que a eficiência do sistema judicial tem um forte impacto no desenvolvimento
dos mercados de crédito.
Este trabalho insere-se neste esforço de compreensão do papel das instituições no
desempenho económico. Tendo o sistema jurídico-legal um papel fundamental como uma
das instituições que mais influenciam o desempenho económico de uma nação, é sobre o
impacto do sistema judicial que nos debruçamos. O nosso objectivo é o de determinar o
impacto do actual funcionamento das instituições do sistema judicial no desempenho da
economia portuguesa. Para o fazer, partimos da constatação de que era indispensável ter em
conta os processos de decisão dos agentes que mais decisivamente contribuem para esse
desempenho, os intervenientes no processo produtivo: as empresas. Assim, avalia-se o
impacto do funcionamento do sistema judicial nas decisões das empresas portuguesas e, a
partir dessa medida, procura-se obter uma medida desse impacto sobre o desempenho
económico do país. Com esse objetivo, procedemos a um inquérito junto das empresas
portuguesas dos diversos ramos de actividade, a nível nacional. A abordagem passa, assim,
pela compreensão dos factores que mais afectam as decisões de produção, de investimento
e de pricing das empresas portuguesas, dado o enquadramento institucional em que se
inserem e que fornecem as relevantes restrições para o seu processo de escolha individual.
A perspectiva assumida é aquela da teoria económica – partindo do pressuposto que cada
agente económico defende o seu interesse próprio (maximiza a sua função objectivo)
assumindo como dadas certas restrições que delimitam as suas possibilidades: as
preferências da sociedade (a procura), a tecnologia da produção (a oferta) e as regras que
lhes são impostas (as leis e a eficácia do cumprimento destas).4
funcionem de forma eficiente, os contratos, quer se refiram a partes privadas, quer envolvam o Estado, devem
estar correctamente elaborados, ser consistentes com a legislação e conter cláusulas passíveis tanto de
verificação como de aplicação”. O sistema judicial, por sua vez, pode desempenhar um papel importante de
melhorar a qualidade das leis e dos contratos.
4
As regras servem para criar e garantir as condições necessárias à manutenção de transacções entre agentes
económicos; é preciso ter-se em conta que, na defesa do interesse próprio, alguns indivíduos ou agentes
económicos pretenderão apropriar-se da riqueza criada por outros membros da sociedade – é a ideia da
“metáfora criminosa” de Olson (2000). Assim, se as regras legais funcionam, por um lado, como uma restrição,
elas têm também um papel de protecção. Porque a apropriação acaba por afectar negativamente tanto quem
Vale observar que os problemas do campo institucional são frequentemente

reconhecidos pelas entidades do Sistema de Justiça dos países em que este não funciona
bem, e Portugal não é uma excepção. Esses problemas são de diversas naturezas: o perfil
das instituições, o formato das estruturas institucionais existentes, problemas da esfera
legislativa e dos rituais dos procedimentos instituídos (que muitas vezes acentuam os
primeiros), instabilidade legislativa e um excessivo grau de formalidade (que contribuem
ainda mais para retardar o trabalho da justiça ao tornar os processos demasiado
burocratizados), entre outros.
Em Portugal, a insatisfação com o desempenho do sistema de justiça é elevada, e tem
desaguado numa vontade de combater algumas das dificuldades do sistema, traduzida em
numerosas medidas recentemente implementadas pelo Ministério da Justiça, que vão desde
a desburocratização, a alteração dos códigos de processo e a simplificação de
procedimentos, até a desjudicialização de certos procedimentos. Parece pois haver um
elevado esforço e empenhamento em colmatar muitas das dificuldades que se têm vindo a
sentir nesta área. Importa chamar a atenção para o facto que o conhecimento sobre os
custos económicos que decorrem de um sistema judicial ineficiente poderá permitir uma
mais clara definição das áreas prioritárias de intervenção, de forma a efectivamente se
conseguir melhorar a administração da justiça, na prática. Uma quantificação destes custos
económicos poderá, adicionalmente, indicar o grau de urgência deste tipo de intervenção.
Assim, este trabalho pretende ser também uma contribuição para o avanço na solução dos
problemas enfrentados pelo Sistema de Justiça português.
O artigo está estruturado em quatro secções. A secção 2 analisa o sistema judicial
enquanto instituição económica, revendo conceitos e introduzindo o ferramental analítico
que orientou a montagem do inquérito. A secção 3 discute os resultados do inquérito
realizado junto às empresas. Uma última secção resume as principais conclusões do estudo.
2. O Sistema judicial como Instituição Económica5
2.1. Como avaliar a qualidade do sistema judicial?

Caracterizar o que é um bom sistema judicial, do ponto de vista do funcionamento da
economia, é menos imediato do que pode parecer à primeira vista. Enquanto que, a nível
das decisões individuais, é razoável pensar na existência de um critério ético que permita
distinguir entre um comportamento produtivo e um de apropriação da produção alheia, a
nível de decisões colectivas essa separação torna-se mais complexa – quer esse conjunto de
indivíduos seja uma empresa ou seja uma instituição pública. O problema complica-se
adicionalmente com relação à definição do objectivo da instituição em questão – a função
de bem-estar que a instituição procura ou deveria procurar maximizar.
Assim, uma corrente de literatura tem-se debruçado sobre a questão da definição
daquilo que, do ponto de vista da eficiente afectação dos recursos, um bom sistema judicial
deveria ser e fazer. As abordagens são variadas: Hay, Shleifer, Vishny (1996) defendem
que um bom sistema legal é um a que os indivíduos recorrem para estruturar as suas
actividades económicas e solucionar as suas disputas; isto inclui uma aprendizagem das
produz riqueza como quem dela se apropria (porque a própria riqueza diminui), é do interesse comum que as
regras legais existam e que elas sejam cumpridas. Como ferramentas essenciais que são para uma correcta
afectação dos recursos, elas têm um papel fundamental no crescimento económico de uma nação.
5
Esta secção é baseada parcialmente em Castelar (2001).
regras legais, a estruturação das transacções económicas de acordo com essas regras,
procurando penalizar ou obter compensação de quem não as cumpre e recorrendo às
entidades públicas para fazer cumprir as regras”. Shihata (1995) defende que “um bom
Sistema judicial é aquele que assegura que a justiça seja acessível e aplicada a todos, que
direitos e deveres sejam respeitados, além de aplicados com um baixo custo para a
sociedade.” Há, é claro, um trade-off entre baixo custo, rapidez de decisões e amplo direito
de defesa. Hay et al. (1996) acrescentam que “para ser aceite, o sistema legal tem que
expulsar completamente outros mecanismos, tipicamente privados, de cumprimento de
contratos e de resolução de disputas”. Isto pressupõe mecanismos que consigam garantir o
cumprimento da lei e das decisões dos tribunais em tempo útil. Castelar (2000) argumenta
também que “um sistema judicial que leve a muitos litígios não está sendo eficiente por
duas razões: uma porque consome muitos recursos (...). Outra porque litígios em excesso
indicam que as leis e os direitos não se acham suficientemente bem definidos e/ou
respeitados. (...) Um número escasso de litígios é também sinal de que o sistema judicial
não está funcionando bem (...) indica que as firmas e os indivíduos não confiam que o
sistema judicial vá proteger os seus direitos de maneira eficiente. Pode também indicar que
os custos de se recorrer ao sistema judicial são muito altos, impedindo, na prática, o acesso
universal à justiça pelas partes”. Finalmente, Hay et al. (1996) argumentam que para que o
sistema possa funcionar para mais pessoas seria necessário que “em primeiro lugar, as
regras más – que impedem as pessoas de usar o sistema legal porque impedem ou falham
na protecção de actividade económica legítima – devem ser eliminadas; em segundo lugar,
as novas regras devem seguir, na medida do possível, as práticas comerciais (…). Em
terceiro lugar, as novas regras deveriam ajudar os tribunais a solucionar disputas dizendo-
lhes o que fazer nos casos em que leis existentes são mais claramente incompletas. Em
particular, os tribunais, com os seus muito limitados recursos, deveriam ser capazes de
verificar se violações das regras legais tiveram ou não lugar”.6
Para se avaliar a importância quantitativa do sistema judicial enquanto instituição
económica é preciso ir além dos conceitos gerais e definir indicadores que permitam
“pensar” a qualidade do desempenho do sistema judicial em termos dos seus reflexos sobre
o funcionamento da economia. Ou seja, é necessário ter critério objetivos e mensuráveis
para avaliar o desempenho do sistema judicial. Definições genéricas, como a de Shihata
(1995), acima, embora capturem a essência do problema, são de difícil utilização. Neste
sentido, três alternativas são propostas na literatura. Sherwood et al. (1994, p.7) sugerem
que o desempenho do sistema judicial seja avaliado considerando-se os serviços que ele
produz, em particular, em termos de “garantia de acesso, previsibilidade e presteza dos
resultados, além de remédios adequados”. Ou seja, pensar a justiça enquanto uma entidade
que presta serviços à sociedade e considerar a qualidade dos serviços oferecidos. Isto
permitiria não apenas estabelecer comparações entre diferentes jurisdições, como também
avaliar o desempenho de um determinado sistema judicial, ou duma parte dele, ao longo do
tempo. Além disso, associando-se indicadores de “produção” aos custos incorridos pela
justiça podem-se derivar indicadores de eficiência, que também podem ser comparados
com benchmarks internacionais ou em outras jurisdições no mesmo país, ou acompanhados
no tempo. De facto, há países em que os tribunais já utilizam indicadores de produtividade
6
Ver também Olson (1965), que fornece o instrumental analítico que permite compreender as condições sob as
quais as Instituições Políticas e Judiciais administram as regras legais.
para monitorizar o trabalho dos seus magistrados.7 A dificuldade com esta metodologia é
que a produção do sistema judicial depende tanto da quantidade de serviços como de sua
qualidade, sendo a importância desta última maior do que em outros setores, estando além
disso sujeita a grande subjetividade. É isto que torna atraente a sugestão de Hay et al
(1996), de que a qualidade do sistema judicial seja medida pela frequência com que os
indivíduos recorrem ao sistema e não a mecanismos concorrentes de resolução de conflitos
e de aplicação da lei. Ou seja, que se pode medir o desempenho do sistema judicial não pela
sua produção, mas por sua competitividade frente a outras instituições que prestam os
mesmos serviços. Também com esta medida há, porém, um problema: o pouco uso do
sistema judicial pode reflectir não o seu mau desempenho, mas a qualidade superior de
outros mecanismos de resolução de conflitos e fazer com que os contratos sejam
respeitados. Ou, no extremo oposto, não apenas o sistema judicial mas também essas
outras instituições podem ser percebidas como fornecendo serviços caros e de má
qualidade, sendo a procura baixa para todas elas. Uma maneira de corrigir este efeito é
utilizar um meio ainda mais indirecto de avaliar o desempenho da justiça, olhando para o
perfil das transacções que efectivamente têm lugar na economia e, em particular, a
frequência com que ocorrem transacções dependentes do bom funcionamento da justiça:
transacções complexas, escalonadas no tempo, com elevado grau de especificidade, e
envolvendo agentes económicos autónomos.8 Castelar (2000) desenvolve um modelo que
permite avaliar o impacto da qualidade dos serviços fornecidos pelo sistema judicial (ou
outro mecanismo de solução de disputas) sobre a utilidade das partes e, portanto, sobre a
sua propensão a litigar. Esta utilidade, por sua vez, funciona como um valor de reserva
para os agentes económicos quando esses se envolvem em actividades produtivas, de
investimento, concessão de crédito etc. Ou seja, o mínimo que se espera “salvar” se a outra
parte no contrato o desrespeitar. Quanto maior for esse valor, maior a propensão dos
agentes económicos a desempenhar essas actividades. Em Castelar (2000), a utilidade
esperada de recorrer à justiça depende, positivamente, do valor líquido que se espera
receber e, negativamente, da variância desse ganho, que reflete a incerteza quanto a ganhar
ou perder a disputa e ao tempo até que uma decisão seja tomada. Assim, a utilidade
advinda da utilização de um mecanismo específico de resolução de conflitos, como o
sistema judicial, é função do valor do direito em causa, dos custos envolvidos, da rapidez
com que uma decisão é alcançada, da imparcialidade do árbitro, da taxa de juro e da
previsibilidade das decisões e do tempo até que estas sejam alcançadas. Neste sentido, um
sistema que funciona bem deve ostentar quatro propriedades: 1) baixo custo,9 2) decisões
justas, 3) rápidas e 4) previsíveis (em termos de conteúdo e de prazo). Um sistema de
resolução de conflitos caracteriza-se como justo quando a probabilidade de vitória é
7
Esta é, em certo sentido, a visão adoptada pelo Banco Mundial no seu Relatório sobre o Desenvolvimento
Mundial de 1997, em que o Banco enumera as três características que, a seu ver, caracterizariam um bom
sistema judicial: independência; força (i.e. instrumentos para implementar suas decisões) e eficiência na sua
gestão.
8
Conforme colocado por Williamson (1995, pp. 181-2): “O resultado é que se pode inferir a qualidade do sistema
judicial de forma indirecta: uma economia com alto desempenho (expresso em termos de governance) irá
permitir mais transacções numa faixa intermediária [i.e. contratos de longo prazo estabelecidos fora de
organizações hierarquizadas] do que uma economia com um sistema judicial problemático”
9
O custo esperado de recorrer ao sistema judicial não depende apenas das taxas pagas à justiça, mas também das
despesas incorridas durante o processo de litígio, da probabilidade de se vencer (probabilidade que pode ela
própria depender do quanto é gasto) e de como os custos do litígio são distribuídos entre quem ganha e quem
perde a causa.
próxima a um para o lado que tem a razão e a zero para o lado que não a tem.
Relativamente à previsibilidade, as decisões são previsíveis quando a variância ex-ante do
ganho líquido de custos é pequena.10 A previsibilidade é alta quando a probabilidade de se
vencer se aproxima de zero ou um e a variância do tempo gasto para se tomar a decisão é
pequena.11 A parcialidade é claramente indesejável, e difere da imprevisibilidade porque
distorce o sentido da justiça de uma forma intencional e determinística. Os tribunais podem
ser tendenciosos devido à corrupção, por serem politizados (favorecendo certas classes de
litigantes, como membros da elite, trabalhadores, devedores, residentes, etc.), ou por não
gozarem de independência em relação ao Estado, curvando-se à sua vontade quando o
governo é parte na disputa. O insucesso em se produzir decisões com presteza é
frequentemente citado como um problema importante dos sistemas judiciais em todo o
mundo. A morosidade reduz o valor presente do direito em disputa, significando que o
sistema judicial só em parte protege os direitos de propriedade.
O nosso inquérito mostra que o principal problema do sistema judicial português, de
acordo com as empresas, é a sua falta de agilidade: praticamente todos as empresas que
responderam ao inquérito consideraram a justiça má ou muito má em relação a este
atributo. A avaliação é também bastante negativa relativamente aos custos de acesso, ainda
que menos do que no que respeita à agilidade – nove em cada dez empresas disseram que
este custo é elevado ou muito elevado – negativa também em relação à previsibilidade das
decisões judiciais, mas relativamente positiva em relação à imparcialidade das decisões
judiciais. No conjunto, uma larga maioria das empresas (88,0%) apontou o desempenho do
sistema judicial português como mau ou muito mau, uma proporção suficientemente
elevada para não deixar margem para dúvida de que, do ponto de vista da classe
empresarial portuguesa, há muito que precisa de ser feito para melhorar o funcionamento
do sistema judicial.
Com relação à imprevisibilidade das decisões judiciais, observa-se que a maioria dos
entrevistados (53,1% das empresas, e quase dois terços das que emitiram opinião) apontou
que constante ou frequentemente os tribunais emitem decisões diferentes para disputas
semelhantes. Também por larga maioria (80% das entrevistadas e 90% das que emitiram
opinião a esse respeito), as empresas afirmaram que a possibilidade de melhor prever o
resultado de um caso em tribunal seria útil ou muito útil ao planeamento das suas
actividades.
2.2. Sistema judicial e Crescimento

A ineficiência do sistema judicial não preocupa apenas pelas injustiças que causa,
particularmente entre os mais pobres. A literatura mostra que dela também resultam custos
económicos elevados, em especial para uma economia que, como a portuguesa, tem
10
Note-se que essa variância é formada tanto pela variância do resultado em si (i.e., perde ou ganha), como do
tempo necessário para se alcançar uma decisão. Ambas representam factores indesejáveis e atuam como
desincentivos ao recurso ao sistema judicial.
11
Os tribunais podem ser imprevisíveis porque as leis e/ou contratos são escritos precariamente, porque os juizes
são incompetentes ou mal informados, porque a decisão sobre um mesmo caso varia muito de um juiz para
outro, ou devido à incerteza quanto ao tempo que será necessário aguardar até que uma decisão seja tomada.
Métodos alternativos de resolução de conflitos podem ser preferidos, consequentemente, não só porque são
mais rápidos, mas também porque os árbitros podem estar mais bem preparados para interpretar a questão em
disputa. Por exemplo, a arbitragem internacional, apesar de cara, é por vezes preferida para resolver conflitos
em transacções internacionais pelo facto dos árbitros terem maior conhecimento técnico sobre a questão em
disputa.
passado por transformações profundas em direção a uma maior liberalização e à diminuição

da influência do Estado. Como é sabido, à medida que uma economia vai sendo
liberalizada, muitas das transacções que se processavam no interior das grandes
organizações estatais ou sob coordenação do sector público, passam a ter lugar no mercado.
Neste, é do sistema legal e judicial de que dependem, em última instância, a protecção dos
direitos e dos contratos assim celebrados. À medida que as transacções se viraram cada vez
mais para o mercado (para o que contribuiu também largamente a entrada de Portugal na
EU e as privatizações que vieram a ter lugar), estas alterações contribuíram para uma
profunda alteração do ritmo da actividade económica e foi sendo acompanhada, por
motivos diversos, de um crescente recurso ao sistema de justiça. No entanto, o sistema de
justiça tem revelado grandes dificuldades em responder a este aumento crescente de
solicitações e isso tem-se traduzido num crescente acumular de processos nos tribunais. O
resultado, por todos conhecidos, é o de uma enorme morosidade processual que tem como
consequência uma diluição temporal dos direitos reclamados em tribunal e que a lei deveria
consagrar.
A morosidade dos processos desacredita o sistema judicial como mediador e
solucionador dos conflitos inerentes à livre actuação dos agentes económicos nos mercados.
Os efeitos da morosidade são bem conhecidos: os direitos e as garantias deixam de estar
assegurados; as partes lesadas aceitam frequentemente acordos menos do que “justos”
(porque a alternativa, a de recurso ao sistema de justiça, não lhe garante uma solução
melhor); quando a Justiça é lenta, o valor esperado do ganho ou da perda das partes reduz-
se substancialmente; os custos de recorrer à justiça aumentam. O comportamento racional
dos agentes incorpora o conhecimento destes resultados nas suas acções – torna-se possível
rentabilizar comportamentos oportunistas; torna-se assim também importante a protecção
de potenciais comportamentos oportunistas das partes com quem se contrata; há que
calcular o risco acrescido de incumprimento quando os custos de recorrer a justiça são
elevados e há que os compensar. É desta forma que as empresas, como agentes racionais,
são afectadas pelo funcionamento da justiça e se vêm obrigadas a alterar os seus
comportamentos. O resultado é, pois, um em que há uma distorção das decisões das
empresas, que poderão reduzir os seus níveis de investimento, evitar certas áreas de
negócio, cobrar preços mais altos 12... Assim se criam obstáculos ao crescimento do
investimento e se limita a competitividade internacional das empresas. Finalmente, o
próprio crescimento e desenvolvimento económico do país pode ser posto em causa.
Talvez a melhor forma de pensar nos benefícios de uma melhoria da eficiência do
sistema judicial seja recorrendo a um raciocínio contrafactual: se o sistema for ineficiente,
os custos de transacção entre os indivíduos aumentam significativamente uma vez que os
litígios terão mais difícil resolução, quanto mais não seja por esta ser mais prolongada, o
que desencorajará os agentes económicos a aceitar contractos sem pesadas cláusulas penais
em caso de incumprimento com o objectivo de desincentivar condutas fraudulentas de uma
12
as empresas, de uma maneira ou de outra, calculam os seus preços de forma a incorporar o custo do tempo
necessário a recuperar judicialmente as quantias que os seus devedores não pagam pontualmente. É caso dos
spreads bancários - a visão das empresas portuguesas confirma esta percepção acerca funcionamento da justiça
sobre os spreads bancários - mais de metade (52,2%) das empresas entrevistadas considera que entre “sempre”
e frequentemente”, as instituições financeiras incorporam um “prémio de risco judicial” na taxa de juro. Para
uma discussão de como a forma de funcionamento do sistema judicial afecta o mercado de crédito ver Castelar
e Cabral (2001).
parte incumpridora. Este simples facto desincentiva o estabelecimento de contratos com

agentes com os quais ainda não tenha havido negócio e em quem não se tenha estabelecido
uma relação de confiança, dada a impossibilidade de resolução rápida de possíveis conflitos
emergentes, o que retardará receitas e aumentará custos.
O sistema judicial tem, assim, um forte impacto no desempenho económico, actuando
por vários canais. Quatro desses canais são: o progresso tecnológico13, a eficiência das
empresas, o investimento e a qualidade da política económica.
Quando os contratos não são eficientemente garantidos, as empresas podem decidir não
realizar determinados negócios, deixar de explorar economias de escala, combinar inputs
ineficientemente, não afectar sua produção entre clientes e mercados da melhor forma,
deixar recursos ociosos, etc. Além disso, tendem a verticalizar-se, trazendo para o seio da
empresa atividades que poderiam ser melhor desenvolvidas em empresas especializadas –
gerando perdas de eficiência nas empresas. Um sistema judicial eficiente é essencial
também para que empresas e indivíduos se sintam seguros para fazer investimentos
específicos, sejam eles físicos ou em capital humano.14 O impacto da qualidade do sistema
judicial sobre o investimento será tão maior quanto mais especializada e específica for a
natureza desse investimento.15 A produção especializada requer frequentemente activos
específicos a uma transacção e os agentes privados só farão investimentos altamente
especializados se estiverem seguros de que os contratos que garantem as suas actividades
serão correctamente implementados. A ausência de um sistema judicial eficiente faz com
que este tipo de investimento não ocorra ou que tenha que ser realizado pelo Estado. O
sistema judicial também pode estimular o crescimento reduzindo a instabilidade e
melhorando a qualidade da política económica. Políticas económicas voláteis e arbitrárias,
ao desestabilizarem as “regras do jogo”, desencorajam o investimento e a produção. Um
bom sistema judicial contribui para reduzir a instabilidade das políticas ao garantir o
cumprimento de compromissos legislativos e constitucionais e ao limitar o arbítrio
governamental. A efetividade da política económica também depende do desempenho do
sistema judicial. Quando definindo que restrições se devem impor à actuação do governo,
os países têm de trabalhar com um trade-off básico. Por um lado, o estímulo ao
investimento crescerá conforme se restrinja o poder discricionário do Estado. Por outro
lado, porque as circunstâncias em que a economia opera mudam com o tempo, a política
económica só pode ser eficiente se os governos tiverem alguma flexibilidade na aplicação
da lei. Num ambiente em constante mutação, um sistema legal que permita a ampla
adaptação permitirá uma eficiência da política económica que não será viável em quadros
legais muito rígidos. A solução para este dilema é a presença de um sistema judicial que
13
Um bom sistema judicial contribui directamente para o crescimento económico estimulando o desenvolvimento
e a difusão da tecnologia quando protege a propriedade intelectual e ao fomentar o investimento em I&D no
país e facilitando a aquisição de tecnologia avançada de outros países. Contribui também indirectamente, ao
contribuir para a redução dos custos de transação, estimulando os agentes económicos a aumentar o número e a
dispersão geográfica de seus negócios – o que aumenta a difusão do conhecimento (não apenas científico mas
também de gestão, de marketing, financeiro, etc.)
14
Entende-se por activo ou investimento específico uma aplicação de capital cujo aproveitamento em outra
actividade é impossível ou, se realizada, implique grande perda de valor. Para uma discussão mais aprofundada
sobre a especificidade de activos ver Williamson (1985).
15
Isto porque, uma vez realizado um investimento específico, é natural que a outra parte num negócio tentar agir
oportunisticamente e expropriar o proprietário do investimento, procurando pagar apenas o custo variável de
provisão do serviço contratado.
coíba o oportunismo do executivo, ao mesmo tempo que lhe dá maior flexibilidade na

condução da sua política.
2.3. Evidência Empírica de Estudos Cross-Country

A maior parte da literatura sobre o impacto das instituições, em geral, e dos sistemas
judiciais, em particular, sobre o crescimento baseia-se em regressões com cross-sections de
países. O trabalho nesta área fundamenta-se usualmente em modelos de convergência
condicionada, nos quais se toma por hipótese que quanto menor o capital institucional de
um país – e, em particular, a eficiência de seu sistema judicial – menor o seu PIB per capita
de equilíbrio e, como consequência, menores as taxas de crescimento económico (ver
Barro, 1991, e Barro e Sala-i-Martin 1992, 1995).
Muitos dos estudos iniciais nesta área usaram como proxies da qualidade dos sistemas
judiciais e legais medidas de instabilidade política, ou, com menor frequência, a natureza
do sistema político. Esses trabalhos supõem, por um lado, que a instabilidade política reduz
a segurança dos direitos de propriedade e, em particular, que os sistemas judiciais e legais
de países envoltos em guerras, revoluções ou outras formas violentas de transição política
estejam menos aptos a assegurar os direitos de propriedade do que aqueles de países não
expostos a essa sorte de eventos. Por outro lado, supõem que os regimes democráticos
sejam mais capazes de garantir os direitos de propriedade. Enquanto o efeito de rupturas
políticas violentas sobre o crescimento é significativo, ainda que amplo demais para poder
ser atribuído ao funcionamento da justiça, a evidência empírica a respeito da influência do
sistema político sobre a economia é, na melhor das hipóteses, ambígua.16
Mais recentemente, diversos estudos tentaram avaliar o impacto de sistemas
legais/judiciais sobre o crescimento económico focando variáveis mais proximamente
relacionadas com a segurança dos direitos de propriedade, a estabilidade das políticas e o
desempenho dos sistemas judiciais. Knack e Keefer (1995) e Mauro (1995) avaliam a
qualidade do sistema judicial utilizando medidas de risco-país geradas por instituições
privadas, obtendo um impacto sobre a taxa de investimento, o PIB per capita de steady
state e a taxa de crescimento do PIB maior do que o estimado anteriormente utilizando
medidas alternativas da qualidade do sistema judicial e das leis, como o grau de violência
política ou índices de liberdades políticas e civis. Brunetti e Weder (1995) utilizam dados
de 310 empresas, em 28 países em desenvolvimento, e concluem que a instabilidade das
leis e das políticas reduz as taxas de crescimento económico. Clague et al. (1995) usam a
proporção de “moeda contratualmente intensiva” (definida como um menos a razão entre o
total de papel moeda fora dos bancos e M2) como uma medida da intensidade de
transacções na economia potencialmente dependentes do sistema judicial, obtendo um
efeito significativo desta medida sobre o crescimento.
Estes estudos indicam que o efeito do sistema judicial sobre o crescimento económico,
seja através do investimento, seja através do crescimento da produtividade (isto é, do
progresso tecnológico e da eficiência) é quantitativamente importante. Mas, como
discutido em detalhe por Castelar (2000) e Aron (2000), eles sofrem de problemas
econométricos potencialmente relevantes: endogeneidade das variáveis explicativas,
sensibilidade à amostra utilizada e à especificação funcional, e má qualidade das proxies
são alguns deles. Neste sentido, devem ser vistos mais como uma indicação da importância
do sistema judicial para a economia do que como medidas precisas desse efeito.
16
Castelar (2000) apresenta e discute em mais detalhe esta literatura.
Uma abordagem semelhante, mas que evita pelo menos parte destes problemas é o
estudo de Castelar e Cabral (2001), que trabalha com uma cross-section de estados
brasileiros para avaliar o impacto da qualidade do sistema judicial sobre o volume de
crédito como proporção do PIB. O trabalho explora o facto de que a legislação que regula
o mercado de crédito é única em todo o país (e portanto não explica diferenças
interestaduais no funcionamento desse mercado) e utiliza como medida da qualidade do
sistema judicial a avaliação da justiça feita directamente pelos empresários de cada estado
brasileiro. Os autores consideram individual e conjuntamente três características do
sistema judicial – morosidade, custos e parcialidade – concluindo que as três afectam
negativamente o volume de crédito rural e não-rural concedido pelo sistema financeiro.
Castelar e Cabral concluem que diferenças no desempenho da justiça são tão importantes
como diferenças no rendimento per capita para explicar a variação interestadual no rácio
volume de crédito/PIB.
3. A evolução do sector da Justiça em Portugal e descrição do estudo a efectuar

3.1. O recurso a um inquérito às empresas Portuguesas
Até ao trabalho de Castelar (2000) não exista uma metodologia que permitisse
claramente compreender como certas restrições institucionais produzem um determinado
resultado em termos do produto nacional, uma medida objectiva do desempenho
económico. Assim, os trabalhos empíricos que pretenderam abordar este problema no
passado reduziram-se a buscas de uma relação estatística entre crescimento económico e
algum indicador representativo das variáveis económicas, institucionais e políticas. Castelar
(2000) desenvolve pela primeira vez uma metodologia que permite medir os custos
económicos de um deficiente funcionamento do sistema judicial, obtendo estimativas
numéricas destes custos. O mesmo procedimento foi depois adoptado pelo Instituto Apoyo,
pelo Foro para la Administración de Justicia e pelo Fraser Institute para estudar o impacto
do sistema judicial nas economias do Perú, da Argentina e do Canadá, respectivamente. É
esta metodologia que utilizamos neste trabalho.
Assim, de forma a conseguir quantificar os efeitos nas decisões das empresas do
funcionamento do sistema judicial, elaborou-se um questionário com o objectivo de obter
informação relativa aos vários aspectos que se prendem com a tomada de decisão e com o
comportamento das empresas.
Os aspectos que se procuram captar no inquérito utilizado são diversos - se por um lado
se pretende saber qual a visão geral das empresas acerca do funcionamento do sistema
judicial, procura-se também saber em que medida o funcionamento do sistema judicial é ou
não um factor fundamental nos processos de tomada de decisão das empresas. Pretende-se
determinar quais os principais factores na determinação dos custos das empresas e, nestes,
qual a posição relativa do factor “justiça”. Além disso, procura-se saber se as decisões de
produção, investimento e de pricing são afectadas pelo funcionamento do sistema judicial
e, em caso afirmativo, tenta-se obter uma quantificação aproximada do efeito. Saber quais
os tipos de processo em que as empresas se vêm mais frequentemente envolvidas é também
um dos objectivos deste inquérito, bem como a duração média dos processos em que estas
se vêm envolvidas e a percentagem de processos que acabam por ser negociados fora dos
tribunais.
Mais detalhadamente, um conjunto de objectivos do questionário pretende captar a

visão geral das empresas acerca do funcionamento do sistema judicial e a importância
atribuída às recentes alterações legislativas introduzidas nos códigos de processo
(perguntas 1 e 41). Relativamente à visão das empresas acerca do desempenho do Sistema
judicial, procura-se obter uma avaliação individual das três vertentes que consideramos ser
essenciais para o desempenho deste sistema: a celeridade, a imparcialidade da decisão final,
bem como a previsibilidade da mesma, e o custo de recorrer à justiça (perguntas 2 e 3). A
questão da previsibilidade da decisão final é ainda abordada no sentido de procurar saber
até que ponto uma melhor previsibilidade poderia ter utilidade nas actividades de
planeamento da empresa (perguntas 5 e 6). Com vista a perspectivar a importância do
desempenho do sistema de justiça relativamente a outras restrições que a empresa enfrenta
e que mais afectam os seus custos e as suas decisões, solicita-se que estas façam uma
ordenação destes factores (pergunta 4). Finalmente, procura-se avaliar o grau de
importância do impacto que o funcionamento do Sistema judicial tem no desempenho da
empresa e qual a percepção da empresa do impacto do mesmo sobre o desempenho da
economia nacional (pergunta 19).
Outro conjunto de objectivos do questionário prende-se com a compreensão da
extensão do efeito da percepção acerca do funcionamento do sistema judicial sobre certos
comportamentos-resposta por parte das empresas no que respeita às suas decisões de
investimento, ao desenho de contratos elaborados com elevada protecção, e que acabem por
influenciar o nível das taxas de juro. Pretende-se saber se a empresa adopta este tipo de
comportamentos-resposta e se ela se sente afectada por estes comportamentos-resposta por
parte de outras empresas (perguntas 7 a 11). Por outro lado, procuram-se avaliar os
aspectos relacionados com a organização dos negócios das empresas (como as decisões de
contratação de trabalhadores permanentes, a sub-contratação de actividades produtivas e a
sub-contratação de actividades não essenciais são afectadas pelo desempenho do Sistema
judicial) e com as oportunidades de negócio (fornecedores e clientes conhecidos têm
preferência sobre outros que possam oferecer melhores condições? Existe uma necessidade
acrescida de recurso a serviços de empresas de informações que permitam qualificar
potenciais parceiros de negócios? Evitam as empresas trabalhar em países que tenham um
Sistema judicial ineficiente? Receiam as empresas contratar com o sector público ou
governamental ou apresentam esses contratos dificuldades acrescidas para as empresas?)
(perguntas 12 a 18). A compreensão da profundidade dos efeitos do desempenho do
Sistema judicial pode ainda ser obtida pedindo às empresas que tentem quantificar os
efeitos que resultariam se o Sistema de Justiça melhorasse o seu desempenho e se tornasse
comparável ao dos países mais avançados da EU – em termos de investimentos adicionais,
recurso ao crédito, redução dos custos dos contratos, criação de novos postos de trabalho,
variação no nível de sub-contratação de actividades produtivas e de actividades não
essenciais, no nível dos preços dos seus produtos e na redução das actividades da chamada
“economia informal” (perguntas 20 a 26). No âmbito destas questões, uma adicional foi
introduzida no questionário depois de no pré-teste do mesmo se ter determinado existir um
factor frequentemente referido pelas empresas como fundamental nas suas decisões – a
questão da legislação laboral e do desempenho do tribunal de trabalho. Assim, e apesar de
se afastar um pouco do âmbito do proposto neste estudo, pede-se às empresas que
quantifiquem o impacto nos preços que podem atribuir a uma falta de flexibilidade da
legislação laboral (perguntas 27 e 28a), não descurando a vertente mais directamente
associada com o âmbito deste estudo: o desempenho do tribunal de trabalho (pergunta

28b).
Um terceiro conjunto de objectivos do questionário prende-se com a experiência da
empresa com a Justiça. Assim, procura-se saber em quantos processos a empresa esteve
envolvida nos últimos anos e quais os tipos de processos mais frequentes – quer iniciados
pela empresa, quer dirigidos à empresa. Destes, pede-se que estimem a duração média de
cada tipo de processo e quantos processos foram concluídos por acordo (perguntas 29 a
36). Finalmente, procura-se saber se a empresa também beneficia da própria morosidade
dos processos e em que sentido (perguntas 39 e 40). Uma última questão, aberta, permite à
empresa expressar-se acerca dos problemas relativos ao sistema de justiça que afectam a
sua actividade, que carecem de resolução e que não foram abordados no questionário
(pergunta 42).
Para todas as empresas inquiridas, é recolhida informação acerca do seu ramo de
actividade (com o seu código CAE), se o capital é nacional ou estrangeiro, se a empresa é
privada ou pública, qual o seu número de empregados, o seu volume de vendas e os seus
resultados líquidos. Estes elementos da tipologia da empresa permitirão verificar até que
ponto determinadas características da empresa, tais como a sua estrutura de capitais e a sua
dimensão conduzem a determinado padrão de respostas – fundamentalmente, permitirá
responder a questões tais como: serão as empresas mais fortes do ponto de vista económico
as que mais recorrem ao sistema judicial (por eventualmente poderem suportar melhor os
custos de uma eventual ineficiência do sistema)?
Sendo o questionário aplicado a empresas dos mais diferentes sectores de actividade,
procurar-se-á observar se estes são afectados diferencialmente e como, em particular.
3.2. Aspectos Metodológicos

Os dados utilizados para a definição da amostra (as empresas a inquirir) baseiam-se no
Universo das empresas registadas nos Quadros de Pessoal de Departamento de Estatística
do Emprego e Formação Profissional do Ministério do Trabalho e da Solidariedade. Deste
universo, foram eliminadas algumas CAE (Classificação de Actividade Económica) que
considerámos representar actividades cuja relação com o sistema judicial é menor –
nomeadamente os sectores primários e a administração pública. O Quadro 1 apresenta o
Universo parcial das empresas assim definido.
Quadro 1
Universo parcial das Empresas Registadas nos Quadros de Pessoal
(n.º de trabalhadores)
CAE Total 1a4 5a9 10 a 19 20 a 49 50 a 99 100 a 200 a 400 a 500 a >= 1000
199 399 499 999
C Ind. Extractivas 1090 404 287 206 149 34 8 1 0 0 1
D Ind. Transformadora 44678 18436 10525 7071 5360 1916 853 341 64 87 25
E Prod. Dist. Electricidade 649 160 169 120 106 53 24 14 2 1 0
F Construção 29523 16456 7191 3523 1689 436 150 67 3 7 1
G Comércio por grosso 91164 63782 17600 6330 2610 596 186 41 9 10 0
H Alojamento e restauração 29051 19936 5995 2008 787 225 73 25 2 0 0
I Transportes e armazenagem 9365 5308 1827 1064 702 225 153 53 10 14 9
J Actividades financeiras 6934 2616 2576 1164 378 106 58 24 1 6 5
K Actividades imobiliárias, alugueres 20773 14470 3724 1405 668 231 161 63 15 21 15
Total 233227 141568 49894 22891 12449 3822 1666 629 106 146 56
Fonte: Quadros de Pessoal, DETEFP – MTS, 1998.

Para a definição da amostra, utilizou-se a CAE e a dimensão de pessoal ao serviço.

Optou-se por não se considerar a variável volume de vendas por não se julgar muito fiável.
Dado o universo parcial a considerar, construíram-se duas amostras independentes (para um
nível de confiança de 95%): (i) uma para as empresas com menos de 100 trabalhadores; (ii)
uma segunda para as empresas com 100 e mais trabalhadores. Esta solução foi considerada
necessária de forma a que se evitasse uma sub-representação das empresas de maior
dimensão, tal como aconteceria se fosse utilizada uma amostra única para o universo
parcial considerado.17 Esta solução permite ultrapassar o problema da subrepresentividade
das as empresas de maior dimensão (já que, dada a problemática em análise, consideramos
pertinente a opinião das empresas de maior dimensão dado que o seu posicionamento face
ao sistema judicial apresentará especificidades relativamente ao das empresas de menor
dimensão) evitando-se, por outro lado, eliminar da amostra as empresas de menor dimensão
(com 4 ou menos trabalhadores) – o que seria indesejável, já que estas são numerosas. A
amostra, assim definida, está descrita no Quadro 2:
Quadro 2
Amostra tendo por base o Universo das Empresas Registadas nos Quadros de Pessoal
CAE 100 e mais

Até 100 trabalhadores Total
trabalhadores
C Ind. Extractivas 6 9 15
D Ind. Transformadora 234 543 777
E Prod. Dist. Electricidade 6 6 12
F Construção 171 84 255
G Comércio por grosso 432 138 570
H Alojamento e restauração 156 42 198
I Transportes e armazenagem 42 54 96
J Actividades financeiras 9 30 39
K Actividades imobiliárias, alugueres 108 102 210
Total 1164 1008 2172
Foi feito o envio postal do inquérito para 2172 empresas (o triplo da dimensão da
amostra, tendo sido aleatoriamente seleccionadas entre as que constituem o Universo), às
quais se acrescentam ainda as 500 maiores empresas do ano transacto – totalizando 2672
empresas.
Os pré-testes ao questionário foram realizados através de entrevistas presenciais.
Durante estes contactos pessoais com os responsáveis das empresas contactadas, verificou-
se que os entrevistados tinham alguma dificuldade na interpretação de duas perguntas e
que, nalguns casos, certas perguntas não se aplicavam à empresa em análise – não sendo no
entanto possível responder “não se aplica” mas apenas “não sabe/não responde”. Por outro
lado, muitos dos entrevistados, ao lhes ser perguntado se consideravam haver outras
17
O Anexo 1 descreve, em maior detalhe, as alternativas consideradas para o processo de amostragem e explica a
escolha tomada.
questões não abordadas no questionário mas que afectassem as decisões da empresa de

forma importante, revelaram que a questão da legislação laboral e do funcionamento do
tribunal do trabalho era uma questão que particularmente os afectava. Dados estes
elementos, as duas perguntas em questão foram reformuladas, introduziu-se a possibilidade
de responder “não se aplica” a várias das questões (já que de facto “não se aplica” e “não
sabe” representam duas situações distintas) e introduziram-se no questionário duas
perguntas relativas à questão da legislação laboral e ao funcionamento do tribunal de
trabalho.
A metodologia a seguir é a seguinte: estima-se a participação de cada sector na
economia (tendo em conta que o sector primário e da administração pública não foram
incluídos na amostragem) a partir de dados sobre produção / valor acrescentado. Assim, no
que respeita à estimação dos efeitos do funcionamento do sistema judicial sobre o nível de
produto, sobre o investimento e sobre o nível de preços, é feita uma análise por sector
(onde se determinará, através de uma média ponderada, os valores respeitantes ao sector).
Uma vez obtidos os valores correspondentes a cada sector, estes valores são utilizados
como peso do respectivo sector no total da economia, obtendo-se assim uma estimativa
acerca do efeito sobre o desempenho da economia nacional.
A divisão da amostra em dois grupos, segundo o número de trabalhadores (como
medida de aproximação à dimensão da empresa), permitirá ainda identificar se o grau de
dificuldade das empresas devido ao funcionamento do sistema judicial difere de forma
significativa ou não entre as maiores empresas e as pequenas empresas. Permite ainda, em
caso afirmativo, localizar quais os pontos onde as divergências existentes são significativas
e quais são aqueles que não apresentam diferenças segundo a dimensão.
Dependendo dos resultados entretanto observados, poderá ou não justificar-se a
apresentação dos resultados não só agregados por CAE mas também divididos segundo a
dimensão da empresa (para as empresas com menos de 100 trabalhadores e para as
empresas com mais de 100 trabalhadores) – procedendo-se à análise deste tipo de
cruzamentos.
O peso relativo dos dois grupos de empresas dentro de cada classificação CAE terá que
ser estimado para que seja possível a obtenção dos dados agregados. Para tal, recorreremos
à contribuição dos dois grupos para o total da produção ou das vendas de cada CAE.
4. Resultados do Estudo para Portugal (Resultados Preliminares)
Apresentamos alguns dos resultados preliminares dos inquéritos conduzidos junto de

empresas portuguesas, dos vários sectores de actividade. Os dados tratados, nesta fase,
dizem respeito a 209 empresas – grupo que constitui apenas parte da amostra a ser utilizada
no estudo que nos propomos conduzir mas que nos permitem já retirar algumas conclusões,
ainda que preliminares, acerca do impacto do funcionamento do sistema judicial sobre as
decisões das empresas portuguesas.
Observamos que a grande maioria das empresas portuguesas – 88% - classifica o
funcionamento do sistema de justiça como “Mau” ou “Muito Mau”.
P1 - Classificação do funcionamento do sistema judicial

português
Percentagem
Frequência % Acumulada
2 Bom 16 7,7 7,7
3 Mau 125 59,8 67,5
4 Muito 59 28,2 95,7
mau 100,0
5 Não sabe/Não 9 4,3
responde
Total 209 100,0
Entendendo que a classificação do desempenho do sistema de justiça em geral depende

da percepção do funcionamento de quatro componentes principais – a morosidade dos
processos, a imparcialidade das decisões, a previsibilidade das decisões e o custo de
recorrer aos tribunais – notamos que a morosidade parece ser o principal factor de
descontentamento. 67 % das empresas classificam esta vertente do funcionamento do
sistema de justiça como “Muito Mau” e 32% como “Mau”. O segundo factor de
descontentamento é o custo de recorrer aos tribunais, com 23,9% das empresas a classificar
este custo como “Muito Alto” e 60,8% como “Alto” – ou seja, a grande maioria das
empresas portuguesas considera caro recorrer à justiça.
P2a - Morosidade dos processos
Percentagem
ent
2 Bom 1 ,5 ,5
3 Mau 61 29,2 29,7
4 Muito 140 67,0 96,7
mau
100,0
responde
Total 209 100,0
Notamos, adicionalmente, que a percepção do custo de recorrer aos tribunais é mais

pessimista nas empresas menores. De facto, procedemos à divisão das observações em três
grupos de acordo com a dimensão da empresa: um grupo composto por empresas
pertencentes ao universo das 500 maiores empresas portuguesas; um outro grupo composto
por empresas com 100 ou mais trabalhadores e um terceiro grupo composto por empresas
com menos de 100 trabalhadores. No grupo das 500 maiores empresas, 20% consideram
este custo "Muito Elevado"; no grupo de empresas com mais de 100 trabalhadores esta
percentagem aumenta para 23,8% e no grupo das empresas com menos de 100
trabalhadores para 29%. Estas diferenças acentuam-se mais ainda quando tomamos em
consideração a percentagem depois de excluídas as empresas que respondem "Não

sabe/Não responde". Neste caso, as percentagens alteram-se para 20,5%, 25% e 34%,
respectivamente.
P3 - Avaliação do custo de recorrer aos tribunais
Percentagem
1 Muito elevado 50 23,9 23,9

2 Elevado 127 60,8 84,7
3 Baixo 18 8,6 93,3
5 Não sabe/Não responde 14 6,7 100,0
Total 209 100,0
Classificação das empresas 2º o nº de trabalhadores ao serviço * P3 - Avaliação do custo de recorrer aos tribunais Crosstabulation
P3 - Avaliação do custo de recorrer aos tribunais

5 Não
1 Muito sabe/Não
elevado 2 Elevado 3 Baixo responde Total
Classificação das 1 500 maiores Count 9 31 4 1 45
empresas 2º o nº de % within Classificação
trabalhadores ao serviço das empresas 2º o nº de 20,0% 68,9% 8,9% 2,2% 100,0%
trabalhadores ao serviço
2 100 e + Count 25 63 12 5 105
% within Classificação
das empresas 2º o nº de 23,8% 60,0% 11,4% 4,8% 100,0%
3 - de 100 Count 16 29 2 8 55
Total Count 50 123 18 14 205
A percepção relativa à imparcialidade é bastante positiva, com mais de metade das

empresas a atribuir a esta dimensão classificações de “Bom” (47.4%) e “Muito Bom”
(2,9%). São as maiores empresas aquelas que consideram o sistema relativamente mais
imparcial - no grupo das 500 maiores empresas, 55,6% consideram o sistema como "Bom"
ou "Muito Bom" no que respeita à imparcialidade; já nas do grupo das empresas com mais
de 100 trabalhadores esta percentagem é de 51,4% e nas empresas com menos de 100
trabalhadores ela é de 43,6%.
P2b - Imparcialidade da decisão final
Percentagem
1 Muito 6 2,9 2,9

bom
2 Bom 99 47,4 50,2
3 Mau 67 32,1 82,3
4 Muito 7 3,3 85,6
mau
Total 209 100,0
Já no que respeita à previsibilidade da decisão final os resultados são menos optimistas,

com 30,6% das empresas a atribuírem a esta componente uma classificação de “Bom” e
42,6% a atribuírem uma classificação de “Mau”. Poucos consideram que seja quer “Muito
Bom” (0,5%) quer “Muito Mau” (5,3%).
P2c - Previsibilidade da decisão final
Percentagem
1 Muito 1 ,5 ,5
bom 31,1
2 Bom 64 30,6
3 Mau 89 42,6 73,7
4 Muito 11 5,3 78,9
mau 100,0
responde
Total 209 100,0
O facto da maioria das empresas considerar este aspecto relativamente negativo é ainda
perceptível na resposta a uma outra pergunta do questionário: quando perguntado se é
frequente os tribunais emitirem decisões diferentes para disputas semelhantes, mais de
metade das empresas considera esta situação ocorre “Frequentemente” (46,0%) ou “Sempre
ou quase sempre” (6,2%). Ainda nesta questão, observamos um pessimismo relativamente
mais elevado por parte das maiores empresas: o grupo das 500 maiores esta percentagem é
de 62,2%, no grupo com mais de 100 trabalhadores é de 55,2% e no grupo das empresas
com menos de 100 trabalhadores e de apenas 40%. Em parte, estas diferenças são
explicadas por um maior número de respostas "Não sabe/Não responde" nas empresas
menores.
P5 - Os tribunais emitem decisões diferentes para disputas semelhantes
Percentagem
1 Constantemente 13 6,2 6,2
2 Frequentemente 98 46,9 53,1
3 Raramente 55 26,3 79,4
4 Muito raramente ou nunca 6 2,9 82,3
Total 209 100,0
Classificação das empresas 2º o nº de trabalhadores ao serviço * P2c - Previsibilidade da decisão final Crosstabulation
P2c - Previsibilidade da decisão final

5 Não
sabe/Não
1 Muito bom 2 Bom 3 Mau 4 Muito mau responde Total
Classificação das 1 500 maiores Count 13 23 2 7 45
trabalhadores ao serviço das empresas 2º o nº de 28,9% 51,1% 4,4% 15,6% 100,0%
2 100 e + Count 1 34 47 6 17 105
das empresas 2º o nº de 1,0% 32,4% 44,8% 5,7% 16,2% 100,0%
3 - de 100 Count 16 16 3 20 55
Total Count 1 63 86 11 44 205
das empresas 2º o nº de ,5% 30,7% 42,0% 5,4% 21,5% 100,0%
A previsibilidade das decisões é importante para o planeamento e a tomada de decisão

por parte das empresas, conforme manifesta a quase totalidade das empresas do grupo em
análise, com 38,8% das empresas especificando este factor como “Muitíssimo importante”
e 46,9% das empresas como “Muito Importante” (enquanto que apenas 7,2% o consideram
pouco importante, 1,9% muito pouco importante e 10,5% não sabem ou não respondem à
questão).
P6 - A possibilidade de melhor prever o resultado de um caso em tribunal seria útil ao

planeamento das actividades da empresa
Percentagem
1 Muito útil 81 38,8 38,8
2 Útil 87 41,6 80,4
3 Pouco útil 15 7,2 87,6
4 Inútil 4 1,9 89,5
Total 209 100,0
Dado o desempenho da vertente morosidade, convém no entanto referir que existe

algum optimismo no que respeita às simplificação dos processos decorrentes de alterações
legislativas relativamente recentes (incluindo as alterações ao regime de notificações)
quanto a uma resultante redução na demora na resolução dos processos em tribunal – ou
seja, redução da morosidade. Embora 20% das empresas tenha respondido “Não Sabe/Não
Responde”, 13,9% das empresas classificam a importância destas alterações “Muito
Importante”, 30,1% “Importante”, 25,8% classificam-na como “Pouco Importante” e
apenas 11% a consideram “Irrelevante”.
P41 - Que importância atribui às recentes alterações legislativas no sentido da simplificação

dos processos em tribunal, incluindo as alterações ao regime de notificação, no efeito que
terão na redução da demora na resolução dos processos em tribunal?
Percentagem
1 Muito 29 13,9 13,9
importante 44,0
2 Importante 63 30,1
3 Pouco 54 25,8 69,9
importante 80,9
4 Irrelevante 23 11,0
5 Não sabe/Não 40 19,1 100,0
responde
Total 209 100,0
Importa questionar qual a importância que o desempenho do sistema judicial tem sobre
os custos e as decisões das empresas. Uma das perguntas do questionário pretende,
precisamente, posicionar esta variável relativamente a outras que representam também
restrições que afectam os custos e as decisões das empresas, procurando estimar a sua
importância relativa entre estes outros factores – os outros factores referidos nesta questão
são: Impostos; Infra-estruturas; Custos salariais; Acesso a mão-de-obra especializada;
Custo de cumprir padrões de produção impostos pela legislação; Custo de cumprir a
regulação industrial; e Outros.
Mais de metade das empresas do grupo sob análise consideraram ser os impostos um
dos dois principais factores que afectam os seus custos e as suas decisões e mais de dois
terços das empresas incluíram esta variável entre as três principais. O segundo e terceiros
factores que surgem como os mais importantes nos seus efeitos sobre os custos e decisões
das empresas são os custos salariais e o acesso a mão-de-obra especializada. 43% das
empresas enumeram os custos salariais entre os dois principais e mais de metade o referem
entre os três principais factores. 32,5% das empresas apontam o acesso a mão-de-obra
especializada como um dos dois principais factores, e 60,7% referem este factor entre os
três principais.
Surgem então a Justiça e as Infra-estruturas. As infra-estruturas são apontadas como o
quarto factor mais importante, em média, seguido da Justiça – a Justiça não parece afectar
de forma homogénea as empresas, aparecendo este factor apontado como o 4º, 5º, 6º ou 7º
mais importante, com uma distribuição relativamente homogénea para cada uma destas
posições.
O Custo de cumprir com padrões de produção impostos pela legislação parece ter
também uma importância desigual com resultados dispersos quanto ao seu grau de
importância, enquanto que o custo de cumprir a regulação industrial é, em geral,
considerado pouco importante.
O desempenho do sistema de justiça não deixa de ter um impacto significativo nas
decisões das empresas – em particular, nas suas decisões de investimento. Mais de metade
das empresas indicam que, em função da sua experiência, as empresas rejeitam
oportunidades de investimento que envolvem um alto risco de vir a Ter que lidar com os
tribunais – 20,1% afirma que isto acontece sempre e 36,8% afirma que isto acontece
frequentemente. Relativamente a esta questão da redução do investimento, são as menores
empresas as que revelam que mais rejeitam oportunidades de investimento - de facto,
27,7% das empresas com menos de 100 trabalhadores que respondem a esta questão,
afirmam que este fenómeno acontece "Sempre", e 34% "Frequentemente". Nas empresas
com mais de 100 trabalhadores estes valores são de 20,4 e 40,8%, enquanto que no grupo
das 500 maiores empresas esses valores são da ordem dos 16,6 e 31,1%, respectivamente.
Classificação das empresas 2º o nº de trabalhadores ao serviço * P7b - As empresas rejeitam oportunidades de investimento
que envolvam risco de ter que lidar com os tribunais - Crosstabulation
P7b - As empresas rejeitam oportunidades de investimento que envolvam risco de ter que lidar
com os tribunais
1 Sempre 2 4 5 Muito 6 7 Não
ou quase Frequente 3 Por Pouco raro Nunca sabe/Não
sempre mente vezes frequente responde Total
Classificação das 1 500 maiores Count 1
7 16 11 7 3 45
das empresas 2º o nº de 15,6% 35,6% 24,4% 15,6% 2,2% 6,7% 100,0%
2 100 e + Count 21 42 30 5 5 2 105
3 - de 100 Count 13 16 15 1 2 8 55
Total Count 41 74 56 12 2 13 205
das empresas 2º o nº de 20,0% 36,1% 27,3% 5,9% 3,4% 1,0% 6,3% 100,0%
Além disso, mais de metade dos inquiridos deste grupo entende que as empresas tomam
demasiadas precauções (tais como exigir cauções elevadas, garantias de terceiros ou
transferindo a responsabilidade legal para jurisdições offshore) para evitar quebras de
contrato – 16,7% afirma que sempre e 45,5% que frequentemente. Cerca de metade das
empresas inquiridas neste grupo reconhece seguir este tipo de comportamentos – cerca de
metade rejeita oportunidades de investimento e cerca de um quarto toma demasiadas
precauções para evitar quebras de contrato. Mais de metade das empresas revela ainda ser
afectada por este tipo de comportamento por parte de outras empresas (cerca de um quarto
das empresas no que respeita a cada uma destas vertentes). (depois de mais de metade
seguem...)
As percentagens referidas, apesar de ultrapassarem, em média, os 50%, variam com a
dimensão da empresa. Cerca de 2/3 das empresas no grupo das 500 maiores admitem seguir
este tipo de comportamento e 60% das que integram o grupo com mais de 100
trabalhadores, mas apenas 40% das empresas que integram o grupo das empresas com
menos de 100 trabalhadores admitem seguir estes comportamentos. Já no que respeita à
questão "a empresa é afectada por este tipo de comportamento por parte de outras
empresas", já a não se assiste a uma dispersão dos resultados segundo a dimensão da
empresa, com todos os grupos a apresentar uma resposta afirmativa de cerca de 65%.
A incerteza acerca a capacidade de resolução justa e atempada de disputas por parte dos
tribunais faz com que 27,8% negoceiem “Sempre ou Quase Sempre” com fornecedores ou
clientes conhecidos ou com aqueles sobre os quais têm boas referências e 45,5% o façam
“Frequentemente”. Pelos mesmos motivos, as empresas sentem uma necessidade acrescida
de recorrer a serviços de informações, tais como os fornecidos por empresas como a “Dunn
& Bradstreet”, com cerca de 60% das empresas a seguir este tipo de comportamento
“Sempre ou Quase Sempre” ou “Frequentemente”. Nas empresas do grupo das 500 maiores
empresas, 25% recorre "Sempre ou Quase sempre" e 40% frequentemente, enquanto que
nas empresas no grupo das que têm menos de 100 trabalhadores apenas 7,2% recorre
"Sempre ou Quase Sempre" e 36,4% "Frequentemente".
Os factores mencionados nos dois últimos parágrafos revelam que existem factores que
contribuem para um aumento significativo dos custos de transacção das empresas e que,
adicionalmente, podem implicar ineficiência na afectação dos recursos – nomeadamente ao

existir um peso importante no factor “conhecer com quem se vai negociar” que poderá
dominar outros efeitos resultantes da saudável concorrência no preço ou na qualidade entre
diferentes fornecedores/clientes.
Desta forma, não é de estranhar que quase ¾ das empresas considerem que o impacto
do actual funcionamento do sistema judicial tenha um impacto “Negativo” (53.6%) ou
“Muito Negativo” (19.6%) no desempenho da economia nacional. (apenas cerca de 15%
das empresas considera que esse impacto é positivo ou inexistente). Metade das empresas
considera que o impacto na sua própria empresa é “Negativo” (42.1%) ou “Muito
Negativo” (7,7%).
P19b - Impacto do funcionamento do sistema judicial no desempenho da economia
nacional
Frequência % Percentagem
Acumulada
1 Muito positivo 9 4,3 4,3
2 Positivo 15 7,2 11,5
3 Inexistente 9 4,3 15,8
4 Negativo 112 53,6 69,4
5 Muito negativo 41 19,6 89,0
Total 209 100,0
P19a - Impacto do funcionamento do sistema judicial no desempenho da sua empresa
Percentagem
1 Muito positivo 4 1,9 1,9
2 Positivo 18 8,6 10,5
3 Inexistente 66 31,6 42,1
4 Negativo 88 42,1 84,2
5 Muito negativo 16 7,7 91,9
Total 209 100,0
Adicionalmente, mais de metade das empresas consideram que a economia informal é

consequência do funcionamento do sistema judicial: considerando este factor como tendo
uma influência "Muito Grande" ou "Grande". Apenas cerca de 1/4 das empresas considera
esta influência "Ligeira" e cerca de 5% considera que a relação entre as duas variáveis seja
"Inexistente". As restantes empresas afirmam "Não saber / Não responder"
Em particular, procurou avaliar-se qual seria o resultado duma melhoria no desempenho
do sistema judicial – pedindo às empresas que imaginassem os resultados de uma alteração
neste desempenho que o colocasse ao nível do da maioria dos países da União Europeia (no
que respeita a celeridade, justiça das decisões finais, custo e capacidade de fazer cumprir os
contratos) e que tentassem estimar os efeitos que esta alteração produziria (1) no nível de
novos investimentos da empresa, (2) em contratação de novos empregados, (3) no nível de
preços cobrados pela empresa e (4) na evolução da sua facturação anual.
Observando as respostas a estas perguntas, podemos ver que os resultados estimados
pelas empresas são duma magnitude importante:
(1) No que respeita a mais e novos investimentos, embora 26,8% das empresas afirmem
que não haveria qualquer tipo de aumento decorrente da melhoria das condições do sistema
judicial, 60% das empresas afirma que o investimento aumentaria. Embora exista alguma
dispersão na quantificação do aumento esperado, em termos médios o aumento médio do
investimento ronda os 8,33%. É interessante verificar que são as empresas de dimensão
média as que estimam um maior aumento desta variável – no escalão de volume de vendas
de 500.000 contos a 1.000.000 contos, o aumento estimado é de 20,7% e no escalão entre
100.000 contos e 200.000 contos é de 20,7%. Os aumentos estimados esbatem-se quer para
volumes de venda superiores a um milhão de contos, quer para níveis inferiores a 100.000
contos – com uma excepção interessante: as menores empresas (com volumes de vendas
inferiores a 10.000 contos) estimam, em média, que o investimento aumentaria 15,5%.
Tomando a totalidade das empresas que responderam a esta questão, vemos que o aumento
estimado com maior frequência de respostas (19% das respostas) é um de 5 a 10%, seguido
de um aumento de 10 a 20%, indicado por 10,6% das empresas, enquanto que também
10,6% estimam que esse aumento fosse inferior a 2% e 7,8% estimam que estivesse
compreendido entre 2 e 5%. Ainda assim, 6,7% das empresas acreditam que o investimento
aumentaria entre 20 e 50% mas já apenas 2,8% acreditam que o seu aumento superasse os
50%.
(2) Em relação a novas contratações, 22% afirma que não haveria qualquer efeito mas
58% afirma que estas aumentariam. A dispersão das respostas é grande nos intervalos
“menos de 2%”, “2 a 5%”, “5 a 10%” e “10 a 20%”, dividindo-se as respostas de forma
relativamente homogénea por estes quatro intervalos de aumento estimado, sendo a média
de cerca de 5%.
(3) Os preços praticados sofreriam também algumas alterações com a melhoria do
sistema de justiça sugerida. Mais de metade das empresas afirmam que, se o sistema
judicial melhorasse o seu desempenho (reduzindo os riscos associados com o
incumprimento de contratos), os preços actualmente cobrados pelas empresas do sector
seriam afectados. Das que respondem afirmativamente, a redução estimada média desses
preços é de 4,75%. O valor mais frequentemente indicado (com maior frequência) é 5%.
(4) Relativamente à facturação, apenas 19% das empresas afirma que uma melhoria das
condições do sistema de justiça não teria qualquer efeito. Na avaliação do aumento da
facturação média estimada, os dois intervalos onde registamos maior número de respostas
(correspondendo a cerca de 40% das respostas) são: “2 a 5%” e “5 a 10%”. É interessante
verificar que cerca de metade das empresas indica valores inferiores a 5% e cerca de
metade aumentos superiores a 5%. O intervalo com maior frequência de resposta é o
intervalo correspondente a um aumento da facturação anual de 5 a 10%. A média simples

dos aumentos estimados pelas empresas que responderam a esta pergunta ronda os 9%. É
ainda interessante observar que são as empresas com níveis de facturação mais baixos as
que estimam maiores reduções de preços, tendendo sugerir valores de 5% e de 10%. Já as
empresas com maior volume de facturação estimam aumentos ligeiramente inferiores –
concentrando-se as respostas mais em torno dos 2%, 3% e 5% (poucas referem os 10%).
Os resultados aqui obtidos podem ser comparados com os de outros estudos
semelhantes realizadas no Brasil, no Peru, no Canadá e na Argentina, utilizando uma
metodologia semelhante:
Impacto Estimado do Aumento da Eficiência do Sistema judicial (%)*
Aumento médio em cada variável Portugal Brasil Peru Argentin Canadá

a
Volume anual de investimento 8,33 13,7 9,5 28,0 2,0
Volume de negócios 7,7 18,5 20,5 19,0 2,0
Número de empregados 5 12,3 8,2 18,0 -
Investimento em outros estados n.a. 6,2 n.a. 23,0 -
Volume de negócios em outros estados n.a. 8,4 n.a. - -
Proporção de atividades terceirizadas 13,9 13,8 15,0 -
Volume de negócios com o sector público - 13,7 17,5 23,0 1,4
Redução de preços 2,4 - - - -
Fontes adicionais: Castelar (2000), Eyzaguirre, Andrade e Salhuana (1998), e Foro para la
Administración de Justicia (2000), Lippert (2001).
(*) Resultados preliminares.
Observamos que as reacções das empresas portuguesas seriam na mesma direcção das
empresas neste grupo de países mas menos significativas em termos quantitativos – com
excepção do caso canadiano, onde o sistema judicial já é percebido como de boa
qualidade.18
Um factor que parece ser importante na variável preços parece ser a Lei Laboral. A
nível agregado, 2/3 das empresas afirmam que as actuais condições da legislação laboral
tem um impacto importante nos preços praticados no sector. Cerca de 55% das empresas
integradas quer no grupo das 500 maiores quer no grupo com menos de 100 trabalhadores,
sendo no entanto esta percentagem de 76% nas empresas no grupo daquelas que têm mais
de 100 trabalhadores. De facto, 2/3 das empresas afirmam que uma maior flexibilidade da
Lei do trabalho permitiria uma redução de preços – 21,6% das empresas estimam que essa
redução seria da ordem dos 5%, 26% da ordem dos 10% e 9,4% estimam essa redução em
cerca de 20%. Já um melhor funcionamento do tribunal do trabalho (mantendo-se a Lei
existente) teria um impacto menor sobre os preços, mas ainda importante: 11% das
18
De facto, a maior parte dos empresários canadianos que participaram no inquérito indicou que as suas decisões
empresariais não seriam afectadas por melhorias no sistema judicial do país.
empresas estima que a redução seria de cerca de 2%, 16% das empresas aponta para os 5%
e 12% das empresas para os 10%.
Finalmente, uma observação que retiramos dos resultados deste estudo é que são as
maiores empresas as que mais recorrem aos tribunais. Quando se pergunta se já recorreram
aos tribunais, 91,1% das empresas do grupo das 500 maiores responde afirmativamente,
84,4% das empresas do grupo com mais de 100 trabalhadores também, mas apenas 58,2%
das empresas com menos de 100 trabalhadores afirmam tê-lo feito.
4. Conclusões
Os resultados do inquérito conduzido indicam que a justiça tem um papel importante no

desempenho económico português, indo ao encontro dos estudos que procuram relacionar a
justiça com o papel que esta desempenha sobre a actividade económica. O estudo sobre
Portugal seguiu a linha de outros estudos anteriores e mostrou que o desempenho do
sistema judicial tem uma avaliação bastante negativa por parte dos empresários
portugueses. A morosidade dos processos em tribunal é, de longe, o factor que mereceu
pior avaliação por parte das empresas. Mostrou ainda que estas encaram a justiça
portuguesa como cara e como imprevisível, imprevisibilidade que consideram impor um
pesado ónus às empresas portuguesas. Em termos quantitativos, permitiu avaliar que o
desempenho do sistema judicial pode ser considerado uma causa importante para a
contracção do Investimento em Portugal, servindo de obstáculo ao crescimento do País;
resulta numa redução do emprego; em maiores spreads – que por sua vez resultam em mais
altas taxas de juro; e em preços mais elevados. Em suma, mostra que o desempenho do
sistema judicial provoca uma distorção nas decisões das empresas. Desta forma, o estudo
realça a teoria acerca do papel das instituições no desempenho económico de um País.
Permite ainda concluir que melhorias no desempenho do sistema judicial permitiriam uma
melhor inserção concorrencial de Portugal no espaço europeu e beneficiariam a sua
posição.
De maneira geral, as respostas indicaram que, com um melhor desempenho do sistema
judicial, haveria uma mudança nas práticas empresariais, mas que essa mudança não seria
dramática. A produção, medida como o volume de negócios cresceria cerca de 7,7%.
Haveria também um aumento de 8,33% no volume de investimentos, sendo esta a variável
mais significativamente afectada, o que sinaliza um impacto importante de uma melhoria
do desempenho da justiça sobre o crescimento. O emprego também seria positivamente
afectado, aumentando em cerca de 5%.
Estes resultados são semelhantes, qualitativamente, a resultados de estudos conduzidos
para outros países, embora a sua ordem de grandeza seja inferior.
ANEXO 1
A definição da amostra para o inquérito às empresas:
Do universo de empresas registadas nos Quadros de Pessoal, determinou-se o número

de empresas a inquirir (a amostra) para um nível de confiança de 95%, depois de
eliminadas algumas CAE que considerámos representar actividades cuja relação com o
sistema judicial é menor – nomeadamente os sectores primários e a administração pública.
Dado o universo parcial a considerar, a amostra resultante do Universo das empresas
apresentado no Quadro 1 (secção 3.2) está apresentada no Quadro A1:
Quadro A1
Amostra de empresas a inquirir
CAE Total 1a4 5a9 10 a 19 20 a 49 50 a 99 100 a 200 a 400 a 500 a >= 1000
199 399 499 999
C Ind. Extractivas 4 2 1 1 0 0 0 0 0 0 0
D Ind. Transformadora 125 52 30 20 15 5 2 1 0 0 0
E Prod. Dist. Electricidade 0 0 0 0 0 0 0 0 0 0 0
F Construção 82 46 20 10 5 1 0 0 0 0 0
G Comércio por grosso 258 180 50 18 7 2 1 0 0 0 0
H Alojamento e restauração 82 56 17 6 2 1 0 0 0 0 0
I Transportes e armazenagem 26 15 5 3 2 1 0 0 0 0 0
J Actividades financeiras 18 7 7 3 1 0 0 0 0 0 0
K Actividades imobiliárias, alugueres 58 41 10 4 2 1 0 0 0 0 0
Total 653 399 140 65 34 11 3 1 0 0 0

No entanto, notamos que quase 2/3 da amostra seria constituída por empresas com
menos de 4 trabalhadores, mais de 80% da amostra seria constituída por empresas com
menos de 10 trabalhadores enquanto que apenas um número muito diminuto de empresas
situadas no maior escalão de dimensão de pessoal ao serviço seria abrangido.
Para tentar obviar a este inconveniente, pensámos em eliminar empresas com menos de
10 trabalhadores. O Quadro A2 representa a amostra obtida com base no universo parcial
de empresas anterior, depois de eliminadas as empresas com menos de 10 trabalhadores.
Ainda assim, constatamos que a representatividade das empresas de maior dimensão
continuava a ser diminuta na amostra.
Quadro A2
Amostra de empresas a inquirir depois de retiradas as empresas < 10 trabalhadores
100 a 200 a 400 a 500 a

CAE TOTAL 10 a 19 20 a 49 50 a 99 >= 1000
199 399 499 999
C Ind. Extractivas 6 3 2 1 0 0 0 0 0
D Ind. Transformadora 243 110 83 30 13 5 1 1 0
E Prod. Dist. Electricidade 5 2 2 1 0 0 0 0 0
F Construção 91 55 26 7 2 1 0 0 0
G Comércio por grosso 152 98 41 9 3 1 0 0 0
H Alojamento e restauração 48 31 12 3 1 1 0 0 0
I Transportes e armazenagem 34 17 11 3 2 1 0 0 0
J Actividades financeiras 27 18 6 2 1 0 0 0 0
K Actividades imobiliárias, 40 22 10 4 3 1 0 0 0
alugueres
Total 646 356 193 60 25 10 1 1 0

A representatividade das maiores empresas mantém-se, no entanto, muito reduzida. O

mesmo ocorre após a eliminação do escalão 10-20 trabalhadores. Assim, pensou-se numa
solução alternativa: a construção de duas amostras independentes: (i) uma para as empresas
com menos de 100 trabalhadores; (ii) uma segunda para as empresas com 100 e mais
trabalhadores. Esta partição da amostra evita que as empresas de maior dimensão surjam
sub-representadas já que, dada a problemática em análise, é de considerar pertinente a
opinião das empresas de maior dimensão dado que o seu posicionamento face ao sistema
judicial apresentará especificidades relativamente ao das empresas de menor dimensão. Por
outro lado, evita-se eliminar da amostra as empresas de menor dimensão, o que seria
indesejável, já que estas são numerosas. Foi esta a solução final escolhida, relativamente à
amostra. O Quadro 2, na secção 2.3 descreve a amostra, assim definida.
5. Referências
Aron, J. (2000), Growth and Institutions: A Review of the Evidence, World Bank Research
Observer (15);
Barro, Robert, J., “Economic Growth in a Cross Section of Countries,” Quarterly Journal of
Economics, Maio 1991, Vol. 106, No. 2, 407-43;
Barro, R e Lee, J. (1994), Loosers and Winners in Economic Growth, Proceedings on the
World Bank Conference on Development Economics 1993-1994;
Barro, Robert J. e Sala-i-Martin, Xavier, “Convergence”, Journal of Political Economy,
Vol. 100, No. 2, Abril 1992, 223-51;
Barro, Robert J. e Sala-i-Martin, Xavier, Economic Growth, Boston, MA: McGraw-Hill,
1995;
Castelar Pinheiro, A. e Cabral, C. (2001), Credit Markets in Brazil: The Role of the
Judiciary and Other Institutions, in Defusing Default: Incentives and Institutions,
Pagano M. (ed), Banco Inter-Americano de Desenvolvimento.
Castelar Pinheiro, A. (Org.) (2000), Sistema judicial e Economia no Brasil, Editora Sumaré,
Brasil;
Castelar Pinheiro, A. (2001), Economia e Justiça: Conceitos e Evidência Empírica, Estudos
IFB (www.ifb.com.br);
Coase, R. (1988), The Firm, the Market and the Law, the University of Chicago Press;
Cooter, R., Rubinfeld, D. (1989), Economic Analysis of Legal Disputes and Their
Resolution, Journal of Economic Literature (27) 1067-1097;
Haussman, R. (1996), La Economia Política de la Reforma Judicial en América Latina,
fotocópia de documento de trabalho;
Hay, Shleifer, Vishny (1996), Toward a Theory of Legal Reform, European Economic
Review (40), 559-567;
Lippert, Owen (2001), Are There Economic Consequences of Judicial Performance?, Fraser
Institute, mimeo (www.fraserinstitute.ca);
North, D. (1981), Structure and Change in Economic History, New-York, W. W. Norton;
North, D. (1990), Structure and Change in Economic History, Cambridge University Press;
North, D. (1992), Transaction Costs, Institutions and Economic Performance, Economic
Center for Economic Growth, Occasional Papers n.º 30;
Olson, M (1965), The Logic of Collective Action. Public Goods and the Theory of Groups,
Harvard University Press;
Olson, M. (1996), Distinguished Lecture of Economics in Government – Big Bills Left on

the Sidewalk: Why Some Nations are Rich and Others Poor, Journal of Economic
Perspectives, 10(2);
Olson, M. (2000), Power and Prosperity – Outgrowing Communism and Capitalist
Dictatorships, Basic Books, New Yok;
Scully, G. (1988), The Institutional Framework and Economic Development, Journal of
Political Economy (96, 3) 652-662;
Shihata, I (1995), in Legal Framework for Development: The World Bank’s Role in Legal
and Judicial Reform, Rowat et al. (eds);
Sherwood, R. et al. (1994), Judicial Systems and Economic Performance, The Quarterly
Review of Economics and Finance, 34;
Stigler, G. (1992), Law or Economics?, The Journal of Law and Economics, 35(2).
Williamson, O. (1985), The Economic Institutions of Capitalism, Free Press.
Williamson, O. (1995), The Institutions and Governance of Economic Development and
Reform, Proceedings of the World Bank Annual Conference on Development
Economics 1994.
World Bank (1996), From Plan to Market, World Development Report.
World Bank (1997), The State in a Changing World, World Development Report.
FIXED-TERM CONTRACTS, EMPLOYMENT FLOWS, AND PRODUCTIVITY
José M. Varejão
Faculdade de Economia do Porto
varejao@fep.up.pt
Julho 2002
This research was funded by the Fundação para a Ciência e Tecnologia (research grant
PRAXIS/PCSH/C/CEG/13/96). I am gratefull to the Departamento de Estatística do
Trabalho, do Emprego e da Formação Profissional for permission to use the micro-data
from the Inquérito ao Emprego Estruturado and the Balanço Social.
480 José Manuel Varejão
1. Introduction
Labor market institutions affect long-run growth via their impact on human and
physical capital accumulation, on innovation, and on restructuring (Nickell and Layard,
1999). Institutions that affect the costs of adjustment of the labor input are of particular
interest because their effects cannot be easily undone by the appropriate wage response.
Even if they could, they would change the intertemporal pattern of employment adjustment.
Job security regulations and the use of fixed-term contracts both change the costs of
adjusting labor quantities.
High adjustment costs hinder employment adjustment. Depending on the structure of
adjustment costs, adjustment is less intense and/or less frequent. On the firing margin, high
firing costs reduce the number of separations, but also that of new hires. Hence, the
turnover of jobs and workers is unequivocally reduced as is match creation and destruction
at continuing positions (churning). Employment protection also reduces the separation rate
from employment into unemployment as well as the exit rate from unemployment into
employment, both resulting in less short-term and more long-term unemployment.
Employment protection regulation inhibits the process of reallocation of jobs across
firms, regions or industries. Reallocation of jobs from declining sectors to new sectors
becomes more difficult in large adjustment costs environments. Because, restructuring and
factor reallocation are the core mechanisms that drive growth in market economies (e.g.,
Caballero and Hammour, 2000), high adjustment costs end up reducing economies’ growth
prospects (Hopenhayn and Rogerson, 1993).
However, prospects for reform of rigid institutions are dim. Rigidity is self-sustaining
because it originates rents that deliver support of the employed and, certainly, that of
workers of the least productive firms that remain in business only because those rigid
institutions are in place. Two-tier reform of the labor markets – granting that new
arrangements apply to new contracts only, not to incumbent employees - is expected to
overcome the status quo bias as long as incumbents are not insulated from market forces,
although they are so from new legislation (Saint-Paul, 2000). Hence, two-tier reform is
expected to succeed in shifting the economy from a rigid to a flexible equilibrium.
However, restrictions on the use of fixed-term contracts may also arise as an
equilibrium outcome if holders of flexible contracts prefer the restoration of employment
protection, which they will if there is a credibility problem associated with two-tier reform.
The prohibition of renewal of fixed-term contracts for more than a limited period
(conversion clauses) is a typical real-world example of such restrictions. Conversion
clauses, although harming future job prospects of incumbent employees and reducing their
outside option in bargaining, may be necessary to gain the support of incumbent employees
if incentives for imposing a complete reform after some critical time are strong.
In fact, having to choose between across-the-board and two-tier reforms of labor market
institutions, many European countries preferred the second alternative. Most accepted
restrictions on the use of flexible contracts which, typically, include conversion clauses.
The importance of these clauses is manyfold.
Low firing cost contracts, such as fixed-term contracts, are useful to employers, in the
first place, because they facilitate employment adjustment to shocks. However, the
empirical evidence does not show any significant increase in the speed of aggregate
adjustment following the reduction of firing costs or the facilitation of the use of fixed-term
Fixed-term Contracts, Employment Flows and Productivity 481
contracts (Hunt, 2000). This result indicates that, for employers, the usefulness of fixed-
term contracts is severely reduced by conversion clauses.
The focus on employers’ demand of flexibility may be too narrow. High costs of firing
workers with standard contracts also originate the demand for worker screening (Dertouzos
and Karoly, 1998). One way employers can overcome their initial informational
disadvantage is by offering newly hired workers up-front training. Training becomes an
ability screen which, coupled with an appropriate wage policy, also induces self-selection
of workers (Autor, 2000). Fixed-term contract fit into hiring strategies of this type as they
offer employers’ the opportunity to screen workers for permanent positions keeping,
throughout, the option of laying them off at a lower cost. Conversion clauses do not conflict
with fixed-term contracts being used for this purpose if the maximum duration of the
contract is sufficiently long to permit employers eliciting the necessary information, which
is usually the case.
However, the coexistence of high and low firing cost contracts may itself originate
unnecessary churning of workers. Conversion clauses offer yet another reason for firms
using fixed-term contracts – good temporary matches may be destroyed and replaced by
new matches of an uncertain value only to bypass the legal obligation of converting fixed-
term contracts into open-ended ones (Blanchard and Landier, 2001). There is, indeed,
evidence that churning is a structural component of some firms’ personnel policies and not
only the inevitable response to unfortunate matches (Burgess et al., 2000).
The implications of fixed-term contracts for long-term productivity growth depend
crucially upon the reasons why employers use them.
If fixed-term contracts are used as buffer stocks, implications are mixed. Flexible
contracts facilitate firing in downturns, reducing labor hoarding and fostering productivity.
But, because they reduce job stability, the use of fixed-term contracts as buffer stocks also
hinders match-specific learning-by-doing and investments in training, and harms long-run
growth prospects.
On the contrary, if fixed-term contracts are used as screening devices, they generate
better growth prospects due to better learning about match quality which translates into
better job matches and, therefore, more stable employer-employee relationships (Nagypál,
2001).
Finally, if fixed-term contracts are used for churning workers they unequivocally have
adverse effects in terms of productivity growth, again because they reduce match-specific
learning-by-doing and investments in training, and because more, otherwise good, matches
are terminated and replaced with new ones of an uncertain value (Blanchard and Landier,
2001).
Assessing the role fixed-term contracts play in employers’ staffing policies is essentially
an empirical problem for which only limited evidence is available. Typically, although not
exclusively, studies that address this issue do it from the employee’s perspective, using data
from national Labor Force Surveys to compute transitions in and out of temporary
employment. The results available for Italy (Adam and Canziani, 1998), France (Abowd et
al., 1999), and the United Kingdom (Booth et al., 2000), all indicate that fixed-term
contracts are stepping stones to permanent forms of employment rather than dead-end jobs.
This is consistent with the hypothesis that fixed-term contracts are a mechanism of
screening workers to permanent positions more than they are buffer-stocks or instruments
of churning policies, which would rather lead to labor market segmentation. However, all
studies that use Spanish data (Alba-Ramirez, 1998, Adam and Canziani, 1998, Amuedo-
Dorantes, 2000), indicate the contrary. In Spain fixed-term contracts seem to be very much
a vehicle of labor market segmentation: employment with fixed-term contracts is largely
involuntary, less well paid, and offers limited advancement opportunities. In the U.S. labor
market, temporary forms of work, which include part-time work, temporary agency
employment, independent contracts, and short-term hires, are mostly used to accommodate
workload fluctuations and to fill temporarily vacant positions, although, some employers
report using flexible forms of employment to screen workers for regular positions (Autor,
2000, Houseman, 2001).
The purpose of this paper is to investigate the reasons why fixed-term contracts are used
in the Portuguese labor market, and derive policy implications thereof. The paper is
organized as follows. Section 2 describes the data. Section 3 presents evidence on the
magnitude of job and worker flows, discriminating between types of contracts. In section 4,
the reasons why employers use fixed-term contracts are investigated via regression
analysis. Section 5 looks at the employee’s side and presents evidence on the magnitude of
transitions between labor market states (their determinants and consequences) for workers
with fixed-term contracts. Section 6 concludes.
2. Data description
The Employment Survey (“Inquérito ao Emprego Estruturado”)

The “Inquérito ao Emprego Estruturado” (IEE) is a quarterly survey run by the Portuguese
Ministry of Employment that collects data on job and worker turnover at the establishment
level. Establishments sampled are statistically representative of three-digit industries (as
defined by the SIC code), regions and size classes.
The IEE sample used in this study spans over twenty quarters, from the first quarter of 1991
to the last quarter of 1995. Units in the sample were selected from the 1990 QP file. On
average 6,960 establishments responded each quarter, making a total of 139,203 records
(establishments × quarters) over the entire 5-year period.
The Social Audit (Balanço Social)

The Social Audit ("Balanço Social") is an annual survey run by the Portuguese Ministry
of Employment. When it was first introduced (1986) it covered state-owned firms only.
Since then its coverage spread, first to firms with at least 500 employees and, since 1992, to
firms with at least 100 employees. Answering to this survey is mandatory.
Each year, a respondent establishment reports data on a large variety of topics
concerning the workforce composition and labor costs. This is organised in six major areas:
company details, employment, labor costs, occupational safety, vocational training, and
social expenditures.
The employment block, the largest in this survey, collects detailed information on the
characteristics of the firm's workforce, including the number of workers (year average and
end-of-year count), skill composition of the workforce, age structure, tenure, educational
level, number of foreign workers and workers with disabilities. Total employment is also
decomposed by type of contract and skill level. Extensive information on the stock and
flows of workers with fixed-term contracts is also available. Reasons why permanent
workers left the firm during the course of each calendar year are also reported. The bulk of
data used in this paper comes from this block. Other data used below refers to wages (wage
level and wage dispersion), training costs, and other social expenditures.
Four waves of the survey were available to this study, covering the period from 1993 to
1996. On average 2030 firms responded to the survey each year – the whole sample
contains information on 8121 firms and a total of 3.1 million workers.
The Household Employment Survey (Inquérito ao Emprego)

The Household Employment Survey is a quarterly employment survey run by the
Instituto Nacional de Estatística (INE). It surveys every quarter approximately 40,000
individuals randomly sampled. One sixth of the sample is rotated out every quarter. Seven
spells of the survey, from the first quarter of 1998 to the fourth of 1999, were available to
this study, covering 56140 individuals.
Individual identification numbers permit tracking individuals over consecutive spells of
the survey (5/6 from quarter to quarter). This makes computation of quarterly transition
rates across labor market states (employment, unemployment, and inactivity) possible.
Because information on the type of contract held by the employed is available it is possible
to compare the pattern of transitions for workers on fixed-term and open-ended contracts.
The survey also contains detailed information on the individual’s demographic
characteristics, his or her labor market background, current job characteristics, current
wage, and job-search efforts. The data available permit a full characterization of the
temporary workforce and extensive control for other relevant job and workers
characteristics in regression analysis in section 5.
3. Job and Worker Flows
Between 1991 and 1998, fixed-term contracts represented an average of 14 percent of

total employment. However, their true importance is best measured by the proportion of
fixed-term contracts in employment flows (i.e., accessions and separations), which amounts
to 62 percent of all accessions and 43 percent of all separations.
Quarterly measures of job and worker turnover (Davis and Haltiwanger, 1990) indicate
very different patterns of employment adjustment for permanent and temporary workers
(Table 1). Temporary positions are highly volatile and temporary employment is highly
unstable. The average turnover rate of jobs filled by temporary workers is 21.7 percent
indicating that, on average, every quarter, about one in five temporary positions is either
created or destroyed. The rate of turnover of temporary workers is 30.6 percent implying
that about one in three workers with a fixed-term contract either joins or leaves his
employer every quarter. The corresponding ratios for permanent contracts are one in
twenty-five (for jobs) and one in twenty (for workers). These results are consistent with
other available (Serrano, 1998, Arai and Heyman, 1999) and indicate very clearly that
workers with fixed-term contracts bear most of the burden of employment adjustment.
Table 1. Job Turnover and Worker Turnover, by Type of Contract
Job Job Job Hiring Separation Worker

Creation Destruction Turnover Rate Rate Turnover
Rate Rate
All Workers 2.3 3.1 5.4 4.0 4.8 8.9

Permanent 1.9 2.6 4.6 1.9 3.2 5.1
Workers
Temporary 9.8 12.0 21.7 16.4 14.2 30.6
Workers
Source: IEE.
All measures of job and worker turnover were computed using the methodology of Davis et al. (1996). Figures for
each category are referred to the establishment’s average number of workers in that category.
The hiring rate of temporary workers (16.4 percent) exceeds the corresponding
separation rate (14.2 percent), implying a net increase in the total number of temporary
workers despite the reduction in the number of temporary positions. The contradiction is
only apparent. What these figures tell us is that, on average every quarter, four to five
percent of the total number of number of workers hired with fixed-term contracts actually
receive a permanent contract, which is direct evidence of fixed-term contracts being, to a
non-negligible extent, used as a device of screening workers for permanent positions.
Churning of temporary workers is about three times larger than that of permanent
workers (29 and 10 percent of total worker turnover, respectively), which is as expected.
Churning occurs because of simultaneous hiring and firing or because of quit replacement.
Both reasons imply that churning should be more frequent among temporary workers,
which is what we observed. It is only upon the revelation of the true match value that the
firm optimally decides to maintain or destroy the match. Given the fact that most newly
admitted workers have a fixed-term contract, it is only natural that match destruction for
failure to meet the critical match value hit temporary workers disproportionately.1
As expected, establishments that do not employ temporary workers exhibit less hiring
and firing activities (Table 2).
Every quarter, 9.1 percent of all establishments that employ permanent workers only
report some hiring activity (at least one worker admitted). This figure compares with 12.4
percent for establishments that use both types of workers. The corresponding figures for
separations are 12.6 and 21.0, respectively. These results are consistent with wider inaction
ranges for units that face larger (gross) adjustment costs, i.e., that employ permanent
workers only. However, the interpretation of these results is partly clouded by the fact that
establishments select themselves to each group in response to the degree of uncertainty they
face. This implies that, on average, establishments with a greater share of temporary
1
Think of firms choosing between alternative hiring technologies that combine in different degrees the amount of
pre-hiring screening and the rate of successful matches, with more pre-hiring screening intensive technologies
producing better matches on average. Higher firing costs bias the firm’s choice of the optimal hiring technology
toward those more intensive in pre-hiring screening and should, therefore, be associated with less match
replacement at continuing positions. This is yet another reason why we expect less churning of permanent
workers.
workers are also those that operate in more uncertain environments. They are, therefore,
more likely to be observed active, hiring and firing workers both because it is less costly for
them to do so and because they receive larger and/or more frequent shocks.
Table 2. Percentage of Establishments Reporting Positive Hires or Separations
Establishments without Establishments with

Temporary Workers Temporary Workers
Hiring Permanent Workers 9.1 12.4
Hiring Temporary Workers 29.6
Separating Permanent Workers 12.6 21.0
Separating Temporary Workers 26.2
No such problems arise if we focus on those establishments that may choose between
the adjustment of the stock of temporary workers and the stock of permanent workers. In
this case we find that these establishments predominantly choose to adjust the temporary
workforce – on average 29.6 percent of these establishments report some hiring of
temporary workers, whereas only 12.4 percent hire permanent workers. For separations the
evidence is similar – in 21.0 percent of these establishments at least one separation of
permanent workers was recorded whereas in 26.2 percent of them there are reports of
separations of temporary workers.2
4. The Determinants of Employer’s Use of Fixed-term Contracts
4.1. Empirical Model
Further insight into the determinants of the use of fixed-term contracts was obtained via
regression analysis. Two regression models were estimated. In the first model – the stock
model – the dependent variable is the firm-level number of workers with fixed-term
contracts. In the second model – the transitions model – the dependent variable is the firm-
level number of fixed-term contracts that were converted into permanent contracts. A
Poisson regression model is appropriate in both cases because the number of events (the
2
All separations, including those into retirement, are counted here. This fact largely inflates the proportion of
establishments in the two groups reporting at least one separation in the group of permanent workers.
number of workers with fixed-term contracts and the number of contract conversions) result
from a counting process which exhibit a preponderance of zeros and small values.3
The Poisson regression model, for grouped data, writes as:
µ ( x) = {N }{g ( β | x)}
where µ(x) is the expected value of the number of events (in this case, the number of
workers with a fixed-term contract or the number of contract conversions), x is the vector
of explanatory variables and β the vector of parameters. N denotes the size of the risk set
(the firm’s total workforce or the total number of fixed-term contracts at the firm-level).
The loglinear model corresponding to this specification (for n counts with independent
Poisson distribution), is:
 µ ( x) 
ln   = x' β
 N 
µ ( x)
where = λ ( x) is the rate of incidence of the event (the proportion of workers with
N
fixed-term contract or the proportion of fixed-term contracts that were converted into
permanent).
With s independent groups (s firms) each with a vector xi=(xi1, xi2,…,xit) of k explanatory
variables, the likelihood function is given by
s n
L(n | µ ) = ∏ µ i i {exp(− µ i )} / ni !
4
i =1
and the corresponding log likelihood estimating equation by:
ln L = ni x' β + ni ln N − exp( x ' β ) N − ln( ni ! )
Because of overdispersion, the variance function (V) is allowed to have a multiplicative

factor φ (V(µ)=φµ), which is obtained dividing the deviance by the degrees of freedom.5
The model is estimated by quasi-likelihood techniques.6
3
7 percent of all firms in this dataset do not use fixed-term contracts and 20 percent report a number of fixed-term
contracts inferior to 10.
4
For details, see Cameron and Trivedi (1998).
5
For the Poisson distribution, the deviance is computed as
2∑ wi {ni ln(ni / µ i ) − (ni − µ i )} ,
i
where wi is the known dispersion weight.
6
Allowing for overdispersion has no effects on the parameter estimates. However, the procedure adopted implies
that the covariance matrix is multiplied by φ and the scale deviance and the log likelihoods used in likelihood
ratio tests are divided by φ.
4.2. The Intensity of Use of Fixed-term Contracts
As part of our interest in the role fixed-term contracts play in the Portuguese labor
market, we want to consider first the determinants of the proportion of this type of contracts
at the firm level, which is the dependent variable in the stocks equation. The explanatory
variables in this equation were constructed considering the three major alternative reasons –
buffer stocks, screening or churning. Results are reported in table 3.
Including the variables “wages”, “fringe benefits”, and “training expenditures” as
regressors enables us to discuss the role fixed-term contracts may play as screening devices
or as part of a strategy of churning.
The relationship between wages (taken in their logarithmic form) and the proportion of
workers with fixed-term contracts is not straightforward. If we think of firms choosing a
package of wage level and turnover (with one varying contrary to the other), we will find
that firms facing larger costs of turnover are more likely to choose a low churning – high
wage strategy and, therefore, use less fixed-term contracts (Burgess et al., 2000). In this
case, a negative sign for the wage variable is expected. A negative sign is also expected if
firms use the probability of a fixed-term contract to become permanent, instead of the wage
level, as a motivator (Guell, 2000). By doing so, firms receive the same levels of effort and
turnover without having to pay higher wages. However, a positive sign for the wage
variable is expected if firms use fixed-term contracts as screening devices. In this case, we
expect that firms that pay higher wages are also the ones engaging more intensively in
screening activities. To the extent that fixed-term contracts fulfill this role, then a positive
association between the two is expected.7 This corresponds to a high wage – low churning
strategy that has an obvious human capital interpretation and is consistent with the results
of the standard dynamic labor demand model. However, a positive sign for the wage
variable may also be interpreted as firms paying compensating differentials to workers
employed with fixed-term contracts because of their reduced job security (Hamermesh and
Wolfe, 1990), again if these do not bear the promise of a likely transition to a permanent
contract.
The estimate obtained for Wages indicates that a one percent increase in the average
wage increases by 0.22 percent the proportion of employees with fixed-term contracts at
firm-level (0.33 if we control for tenure). This result is consistent with fixed-term contracts
being used for screening as well as with a compensating wage differential interpretation,
although the estimated wage elasticity of the proportion of fixed-term contracts seems too
large to be properly accounted for by compensating differentials alone.8
The variable Fringe Benefits measures the firms’ per capita total expenditure on items
such as private social security plans, recreational activities for employees, assistance to
student employees, among others.
7
The alternative to this on-the job screening strategy would be pre-hiring screening, in which case no statistically
significant relationship between wages and the proportion of workers with fixed-term contracts is expected,
especially if we control for the qualifications of the firm’s workforce.
8
Evidence reported in section 5 indicates that, controlling for workers and firms’ relevant characteristics as well
as for selectivity, workers with fixed-term contracts, compared with permanent workers, actually start out with
higher wages, but the differential gradually erodes with tenure. For tenure around six months the differential
actually becomes negative.
If we think of fringe benefits as investments in human capital, then we will expect that
firms investing the most also have the most stable workforce. If reinforcing worker
attachment is the main reason why firms may be willing to pay this kind of benefits, large
groups of temporary workers are not expected in firms that actually pay them benefits.
However, we could expect the contrary if firms developed dual internal labor markets,
paying large amounts of benefits to workers in the primary sector, but not to those in the
secondary. In this case, a positive sign for the coefficient of the Fringe Benefits variable is
expected.9 The results obtained here indicate otherwise. This variable was found to have a
strong and statistically significant negative impact on the proportion of employees holding
fixed-term contracts, which is as expected and has an obvious human capital interpretation.
9
Houseman (2001) actually found evidence of firms distinguishing between flexible workers (short-term hires,
part-time workers, and on-call workers) and full-time regular workers in determining benefit eligibility.
Table 3. Intensity of Use of Fixed-Term Contracts, Poisson Regression Estimates

(N=8813)
Independent Variables (1) (2)
Constant -1.396 (0,165)* -3.840 (0.143)*
Seasonality 0.004 (0.000)* 0.002 (0.000)*

Sales volatility 0.080 (0.026)* -0.041 (0.023)*
Wage level 0.215 (0.025)* 0.330 (0.020)*
Wage dispersion 0.003 (0.001)* 0.002 (0.001)*
Fringe Benefits -0.613 (0.037)* -0.392 (0.030)*
Training expenditures 0.024 (0.055) -0.209 (0.049)*
Firm age
< 2 years 0.067 (0.052)* 0.049 (0.044)*
2-5 years 0.109 (0.034)* -0.074 (0.029)
Firm size
500-999 employees 0.016 (0,025) -0.017 (0.021)
1000 employees and more -0.016 (0.022) -0.025 (0.018)
Employment Trend 0.752 (0.070)* 0.417 (0.062)*
Workers age
25 – 44 -1.217 (0.078)* -0.444 (0.067)*
45 – 64 -2.626 (0.072)* -0.526 (0.073)*
Tenure
<1 year 2.653 (0.047)*
1-2 years 2.014 (0.058)*
2-5 years 1.320 (0.062)*
Qualification level
Manager -4.533 (0.948)* -2.927 (0.761)*
Highly Professional -2.825 (0.285)* -2.385 (0.231)*
Professional -2.182 (0.204)* -2.235 (0.171)*
Skilled and Highly Skilled 0.197 (0.204) 0.179 (0.151)
Supervisors -0.686 (0.079)* -0.392 (0.066)*
Semi-skilled -0.206 (0.072)* -0.189 (0.060)*
Unskilled 0.319 (0.075)* 0.035 (0.062)
Industry Dummies Yes Yes
Year Dummies Yes Yes
Log-Likelihood 73222.9 106037.8
Firm age is a set of dummy variables representing the firm’s age (omitted category: over 5 years). Firm size is a
set of dummy variables representing the firm’s size (omitted category: more than 100 and less than 500
employees). Workers age is defined as the proportion of the firm’s employees that fall in each age group (the
omitted category correspond to age below 25). Tenure is defined as the proportion of the firm’s employees with
tenure within each interval (the omitted category is tenure longer than 5 years). Qualification level represents the
firms’ employment structure by qualifications (omitted category is apprentices). All the remaining variables are
defined in the main text.
* denotes significant at 1 percent; otherwise estimates are not significant at 10 percent.
The variable Training Expenditures measures the costs firms pay per employee because
of formal training. Standard human capital theories imply that firms investing more in
training offer (and desire) more job security and, because of that, should employ less
workers with temporary contracts. Thus, a negative sign for the coefficient of Training
Expenditures should be expected. This is contrary to what results in Table 3 (column 1)
show – the estimate obtained for the coefficient of Training is positive, although not
significant. However, if we control for the structure of the firm’s workforce by tenure
(Table 3, column 2), the parameter estimate becomes negative and highly significant, which
we interpret as indicating that training is to a large extent offered to newly admitted
workers. An explanation why firms would want to offer training to workers that are not
offered job security is offered by Autor (2000). This explanation highlights the fact that it is
through training that firms acquire information on the true value of newly-formed matches,
which they need to make the decision of maintaining or destroying the match. Using
training as a screening device also has the effect of inducing self-selection, attracting to the
firm workers with higher perceived ability in anticipation of a steeper earnings profile. For
firms, this interpretation implies that training is actually a substitute to pre-hire screening.
To account for the possibility that firms use fixed-term contracts to fill temporary
positions it is essential to be able to account for the importance of seasonal variations of
employment at the firm level. A measure of quarterly fluctuations of employment was thus
included as an independent variable. This variable (denoted “Seasonality”), measures the
squared deviation of employment in each calendar quarter from its annual average
computed over a five-year period (1991-95). Data from the Employment Survey was used
to compute this variable. Admittedly this is an imperfect way of measuring seasonality but
the only one that the datasets available permit.10 A positive sign for this variable, indicating
that firms facing larger seasonal variations of its workforce are expected to use fixed-term
contracts proportionately more, was anticipated. That is, in fact, what we obtained. The
effect of seasonality, as measured, on the proportion of employees with fixed-term
contracts is positive and significant – a one unit increase in the seasonality measure, raises
by 0.4 percent the proportion of workers with fixed-term contracts.
More uncertain environments are also expected to increase the proportion of employees
with fixed-term contracts because of the firing cost differential between temporary and
permanent employees. To control for this effect, a measure of the volatility of annual sales
was also included as a regressor. “Sales volatility” is a measure of the deviation of annual
sales from its annual average computed over a three to five year period (depending on data
availability) ending on the year before the one the observation is referred to. For firms with
no information on sales for at least three years, the value of this variable was set to missing.
The estimate of the corresponding parameter is significant and equal to 0.08 indicating that
10
The Employment Survey is a quarterly survey whose observation unit is the establishment. Hence, the
seasonality variable was constructed with plant data with no guarantee that they refer to all establishments
affiliated with each firm. Besides, for some firms, none of its establishments were present in the Employment
Survey dataset. In these cases, the value of the seasonality variable was set to missing and a dummy variable
(not reported) indicating that the variable was not observed set equal to one. Notice that, even though a control
for the firm’s age was included, this was done through a set of dummy variables that correspond to different age
intervals. The length of these intervals were defined to match the maximum permitted duration of fixed-term
contracts and does not correspond to the criteria used to compute the “sales volatility” variable.
for every unit increase in sales volatility the firm’s proportion of employees with fixed-term
contracts increases about 8.0 percent.11
Further to this set of regressors a number of controls for both the firm and workforce
characteristics were also included in the model. The results are globally consistent with
what is expected. In most cases, the corresponding coefficients are significant and have the
right sign. This is the case of the variables representing firms’ age and size (larger and older
firms using fixed-term contracts less), and the age and qualification profile of the firm’s
workforce (firms with younger and less qualified workers using fixed-term contracts more).
Firms undergoing persistent growth or decline also use more fixed-term contracts than
those that switch from growing to shrinking (the Employment Trend variable measures the
absolute value of the firm’s average annual employment growth over the sample period). A
measure of wage dispersion was also included in the set of regressors.12 This measure is
interpreted as a proxy for union strength at the firm, as unions are known to reduce wage
dispersion both at the firm level and in the aggregate economy. Because unions also oppose
the use of non-standard employment contracts, such as fixed-term contracts, the coefficient
of “wage dispersion” is expected to be negative. This is, indeed, what results show,
although the effect is very small – raising wage dispersion by one percentage point
increases the proportion of workers with fixed-term contracts by 0.3 percent.
4.3. Conversion of Fixed-term Contracts into Permanent Contracts
Further insight into the reasons why firms use fixed-term contracts can be obtained
through analysis of the determinants of the rate at which firms offer their temporary
workers open-ended contracts.
The model was first run with the exact same regressors used in the empirical model in
the previous section. The corresponding results are reported in columns 1 and 2 (with and
without controls for tenure) of Table 4.
The estimate of the coefficient of the Wage variable indicates that a one percent
increase in the average wage increases by 0.2 percent the proportion of fixed-term contracts
that are converted into permanent. This result indicates that firms paying higher average
wages use temporary workers mostly for screening reasons. Remember that we have seen
before that wages and the proportion of individuals employed with fixed-term contracts are
positively associated at the firm level and that this could indicate either screening or
compensating differentials. However, the compensating differentials interpretation is
weakened by the fact that firms that pay higher average wages are also the ones that more
frequently offer permanent positions to temporary workers. The expectation of obtaining a
permanent position would, of course, reduce the magnitude of the compensating differential
11
Controlling for tenure (table 3.4., column 2), the estimate of the coefficient of the Sales Volatility variable
becomes negative. Again this result may be due to the way the variable was measured. Newly created firms do
not have a sales record long enough to compute the corresponding volatility. In these cases, this variable was set
to zero and, as with the Seasonality variable, a dummy variable equal to one when volatility is not observed, is
included. Hence, zeros are associated with younger firms and this could explain the changing sign of this
variable from one specification to the other. The control for the firm’s age captures this effect only imperfectly
as the age intervals considered do not match those required to compute the measure of volatility.
12
Wage dispersion is the ratio between the highest and the smallest regular wages paid by the firm, after the top
and bottom five percent wages are excluded.
required, as workers, particularly those more able, are even willing to accept reduced wages
in exchange for better prospects of getting a permanent position.13
This is also consistent with the results obtained for “training expenditures”. If we look
at training as a mechanism of screening, then it becomes a complement of fixed-term
contracts. This is the result we obtained in section 2.3. But this approach to training also
implies that a larger share of workers with fixed-term contracts eventually become
permanent. Hence the training variable is expected to show a positive sign in the equation
of transitions from fixed-term to open-ended contracts. This is what results actually show –
raising by one monetary unit the firm’s per capita spending on training increases by some
0.5 percent the proportion of temporary workers that actually become permanent (0.3 if we
do not control for tenure). This result is also consistent with standard human capital
approaches to training, because firms paying for training are the most willing to retain
trained workers and are, therefore, the ones that offer more job security. In both views,
following a training period firms would offer permanent contracts to workers receiving
training either because of their newly-acquired qualifications or because some of them were
identified as good matches. The estimates reported in Table 4 are consistent with both
interpretations. However, because the size of the training effect comes largely reduced
when we control for tenure in the stocks equation, we know that training is largely
associated to training of newly admitted employees more than it is to training of permanent
workers.
13
Aggregation of wages across workers at the firm-level does not permit to test the (modified) efficiency wage
hypothesis which implies that lower wages are consistent with higher rates of conversion of temporary
contracts. This hypothesis could imply a negative sign in the transitions model. However, Guell’s (2000) paper
says nothing about how temporary workers must be motivated after they receive a permanent contract.
Presumably it is through higher wages. If it is, then the relationship between average wage levels and
conversion of fixed-term contracts could actually turn out to be positive.
Table 4. Transitions from Temporary to Permanent Contracts, Poisson Regression

Estimates (N=7545) 14
Independent Variables (1) (2) (3)
Constant -3.402 (0.242)* -2.008 (0.244) * -2.048 (0.245) *

Seasonality -0.002 (0.000)* -0.002 (0.000) * -0.002 (0.000) *
Sales volatility 0.116 (0.039)* 0.049 (0.037) 0.038 (0.037)
Wage level 0.212 (0.037)* 0.083 (0.036)** 0.081 (0.036)**
Wage dispersion -0.009 (0.002)* -0.006 (0.002) * -0.006 (0.002) *
Fringe Benefits 0.093 (0.015)* 0.068 (0.015) * 0.066 (0.015) *
Training expenditures 0.254 (0.066)* 0.461 (0.063) * 0.469 (0.063) *
Attrition 4.243 (1.048) *
Quits 0.528 (0.249)**
Job Turnover -0.138 (0.043) *
Firm age
< 2 years -0.296 (0.087)* -0.202 (0.081)** -0.211 (0.081) *
2-5 years -0.019 (0.051) -0.003 (0.048) -0.008 (0.048)
Firm size
500-999 employees 0.094 (0.037) ** 0.138 (0.035) * 0.141 (0.035) *
1000 employees and more -0.009 (0.031) -0.048 (0.030) * -0.049 (0.030)
Employment Trend 0.033 (0.099) 0.175 (0.088) ** 0.208 (0.088)**
Workers age
25 – 44 0.106 (0.122) -0.443 (0.115) * -0.405 (0.116) *
45 – 64 -0.929 (0.113)* -0.981 (0.129) * -1.068 (0.131) *
Tenure
<1 year -1.910 (0.094) * -1.847 (0.095) *
1-2 years -0.145 (0.107) -0.138 (0.104)
2-5 years 1.422 (0.092) * 1.424 (0.093) *
Qualification level
Manager 2.734 (1.168) ** 0.552 (1.123) * 0.609 (1.119)
Highly Professional 0.213 (0.350) -0.272 (0.339) -0.240 (0.339)
Professional 1.111 (0.238) * 0.685 (0.226) 0.712 (0.226)*
Skilled and Highly Skilled 0.033 (0.299) -0.619 (0.282) * -0.622 (0.283)**
Supervisors 0.450 (0.120) * 0.010 (0.116) ** 0.095 (0.116)
Semi-skilled 0.213 (0.110)*** -0.025 (0.108) -0.018 (0.107)
Unskilled -0.542 (0.119) * -0.449 (0.115) -0.451 (0.115) *
Industry Dummies Yes Yes Yes
Year Dummies Yes Yes Yes
Log-Likelihood 17192.5 20292.2 20377.9
*, **, *** denote significant at 1, 5, and 10 percent; All variables as defined in section 4.2 or in main text.
14
The remaining firms, up to the total 8,113 used to generate the results reported in Table 3, do not employ any
Standard human capital theories are also consistent with the estimates of the coefficients
of “fringe benefits”. Firms incurring in higher non-wage labor costs are the ones that
convert more fixed-term contracts into permanent contracts. For all these three variables –
“wages”, “fringe benefits”, and “training expenditures” – negative signs would be required
if internal dualism were the main reason for firms to employ temporary workers. In this
case, the stock of temporary workers varies so as to accommodate the fluctuations of
business conditions and fixed-term contracts are not expected to become permanent.
Temporary workers hired to fill short-lived positions because of seasonal fluctuations in
the workload are expected to leave firms as workload falls back to its off-peak level. Hence,
we expect the intensity of seasonal fluctuations to be negatively associated with the
proportion of the firm’s temporary workers who become permanent. The same result is
expected if temporary workers are hired to meet uncertain economic conditions.
The results for “seasonality” are as expected. A one point increase in the index of
seasonality reduces by 0.2 percent the proportion of temporary workers receiving a
permanent contract. Things are less straightforward with the measure of the volatility of
sales that is taken as a proxy for uncertainty. The estimate obtained with specification (1) is
positive and significant, although, as discussed before, this may be due to measurement
problems. Controlling for tenure (column 2) the parameter estimate actually becomes very
small and not significant.
An unequivocal sign of the importance of screening for the use of fixed-term contracts
is obtained by adding three extra regressors measuring the extent of job and worker
turnover at the firm level (column 3). Here, “job turnover” is defined as the sum of the
squared deviation of the two employment counts (as off the beginning and end of each
calendar year) from its annual average. The two variables measuring worker turnover are
the proportion of permanent workers that each year either quit voluntarily (Quits) or leave
the firm for retirement or because of other natural causes, such as death (Attrition). Notice
that these two latter variables, that refer to permanent workers only, permit us focusing on
the firms’ policies to fill permanent positions. If the destruction of permanent matches is
actually an occasion for firms to destroy the corresponding positions, no relation is
expected between attrition or quits and the rate of conversion of fixed-term contracts. Such
relation is also not expected if permanent positions were maintained and filled by workers
hired as permanents. The hypothesis of fixed-term contracts being used as part of a strategy
of screening implies a positive sign for both attrition and quits in the equation of transitions.
On the contrary, if firms use separations of permanent workers as opportunities to replace
them with temporary workers as part of a strategy of churning no transitions to permanent
contracts occur and a negative sign is expected.
The estimates obtained indicate that both “attrition” and “quits”, but especially
“attrition”, have a strong positive effect on transition rates. A one percentage point increase
in the attrition rate of permanent workers increases by 4.2 percent the proportion of
temporary workers that get a permanent contract; for quits the corresponding figure is 0.5
percent. As expected the intensity of job turnover (measured by the deviation of both
beginning and end-period counts of the firm’s total employment from its annual average)
has a negative impact on transitions.
temporary worker (i.e, for them N=0).

Together, even if there are signs of fixed-term contracts being used to fill temporary
positions and as buffer stocks, these results indicate that the primary reason why firms use
fixed-term contracts is to screen workers for permanent positions.
5. The Employment Record of Temporary Workers
If firms use fixed-term contracts predominately for screening, for workers the contracts
are stepping-stones to a permanent form of employment. On the contrary, if fixed-term
contracts are used for churning or to deal with workload fluctuations, they become a
transitory form of employment with poor prospects of career advancement.
Employment with fixed-term contract is a transient state by design. Several destination
states are admissible. Here we consider transitions out of temporary employment into
unemployment, inactivity and another job, as well as transitions from temporary (FTC) to
permanent contracts (OEC). As a benchmark, the corresponding transition rates originating
in permanent employment were also computed. Data from the Household Employment
Survey, for the period from the first quarter of 1998 to the third quarter of 1999 were used
for this purpose (as well as to produce all the other results discussed in this section). Results
are in table 5. 15
Table 5. Quarterly Transition Rates from Employment, by Type of Contract (%)
Unemployment Inactivity Other Job OEC FTC
From OEC (1) 0.39 0.81 0.75 0.33
From FTC (2) 3.36 3.34 3.39 5.48
(2)/(1) 8.6 4.2 4.5

* Transition rates are computed as the ratio between the number of individuals that in each period were employed
with one type of contract and moved to any destination state over the course of one quarter, and the total number
of individuals with the same type of contract in the beginning of the period.
Transitions out of employment are less frequent in Portugal than in other countries.16
However, because of their very nature we expect that workers with fixed-term contracts
change their status more frequently, which is exactly what we see in table 5. Workers with
fixed-term contracts are about five times as likely to loose their jobs, and eight times as
15
The sample period covers an upturn only. This could, of course, influence the results reported here. However, a
comparison with similar evidence for the 1992-1997 period (Portugal, 2000), is reassuring. Except for
transitions to unemployment, which are, as expected, higher (0.61 and 4.70 for open-ended contracts and fixed-
term contracts), the magnitude of transition rates is remarkably similar for the two periods. More important, for
all types of transitions, the differential between the two types of contracts is about the same in the two periods.
16
Blanchard and Portugal (2001) estimate that the magnitude of worker flows out of employment for Portugal is
about one fourth of the corresponding figure for the United States.
likely to become unemployed, as workers on open-ended contracts. This is of course what

one could expect whatever the reasons why firms use fixed-term contracts may be. The
second point is that about 5.5 percent of all workers on a fixed-term contract are likely to be
given an open-ended contract every quarter. This is suggestive of the importance of
screening as a motivation for hiring fixed-term contracts.
Further analysis of transitions across labor market sates was implemented using
multivariate analysis. The complete sample was used to study the determinants of
transitions from employment (both types of contracts) to unemployment, inactivity and
employment with a different employer. Regression models for transitions from fixed-term
contracts to open-end contracts and vice-versa were also estimated. A dummy variable
representing the type of contract (taking the value 1 in case of a fixed-term contract) was
included as a regressor in all regression equations.
The dataset used covers six potential transitions (seven quarters). Overall it contains 56140
observations, 12.5% of which (7057) are workers with fixed-term contracts. Results are
reported in table 6.
Results clearly indicate that workers with fixed-term contracts face higher probabilities of
transition to all the destination states considered. All the estimates for the FTC dummy in
the three equations are positive and statistically significant at 1 percent and indicate non-
trivial effects of the type of contract on their probability of changing status.
Transitions out of employment are significantly influenced by human capital variables.
Workers with more years of education are less likely to leave their jobs to become
unemployed or inactive but are more likely to leave to another job. Tenure and wages also
reduce the probability that the worker makes a transition to unemployment or drops out of
the labor force. Consistent with a matching interpretation, longer tenures and higher wages
are also associated to fewer transitions to other jobs.
Table 6. Transitions Out of Employment

E-U E-I E-E’
Coef. Mg. Coef. Mg. Coef. Mg.
Effect Effect Effect
Constant -1.1377 -1.2484 -1.0456

(0.3993) (0.3141) (0.3547)
Sex 0.0379 *** 0.0005 -0.1528 -0.0043 0.1381 0.0017

(0.0407) (0.0338) (0.0358)
Age 0.0067 *** 0.0001 0.0041*** 0.0001 -0.0224 -0.0003

(0.0041) (0.0032) (0.0049)
Voc. Training 0.0505 *** 0.0008 0.0046*** 0.0001 0.0272 *** 0.0004
(0.0681) (0.0612) (0.0603)
Schooling -0.0031 *** -0.00004 -0.0162 -0.0004 0.0072 *** 0.0001

(0.0062) (0.0052) (0.0056)
First Job -0.0581 *** -0.0008 0.0583*** 0.0017 -0.1063 ** -0.0013

(0.0684) (0.0530) (0.0572)
Labor Market Experience -0.00631 *** -0.0001 0.0025*** 0.0001 0.0113 * 0.0001
(0.0040) (0.0031) (0.0046)
Left Previous Job because
Dismissal 0.0912 *** 0.0014 -0.094** -0.0024 -0.1110 * -0.0013

(0.0607) (0.0509) (0.0556)
End of FTC 0.1784 0.0030 0.0034*** 0.0001 0.0290 *** 0.0004

(0.0602) (0.0528) (0.0541)
Quit -0.0229 *** -0.0003 -0.1078* -0.0027 0.0162 *** 0.0002

(0.0655) (0.0528) (0.0514)
Type of Employer
Private 0.0997 ** 0.0013 0.0589*** 0.0016 0.2412 0.0027

(0.0550) (0.0431) (0.0562)
State-Owned -0.0390 *** -0.0005 0.1503 0.0048 0.1094 *** 0.0016

(0.0991) (0.0648) (0.0941)
Fixed-Term Contracts 0.6966 0.0204 0.6642 0.0334 0.3902 0.0075

(0.0557) (0.0493) (0.0522)
Log Hourly Wage -0.1930 -0.0027 -0.1505 -0.0042 -0.1086 * -0.8388

(0.0557) (0.0437) (0.0494)
Tenure -0.0018 -0.00003 -0.0001*** -0.000003 -0.0034 -0.4277

(0.0003) (0.0002) (0.0004)
Tenure (FTC) -0.0065 -0.00009 -0.0043* -0.0095 -0.0080 -0.0178

(0.0023) (0.0017) (0.0027)
Number of Obs. 56140 56140 56140
Log Likelihood -8832.02 -13112.27 -11841.67

χ2
610.98 398.77 764.38
*, **significant at 5 percent and 10 percent, respectively; *** not significant at 10 percent; otherwise significant at
1 percent. Standard errors in brackets. Industry and quarterly dummies present.
Table 6.A. Variables Definition
Variables Definition
Sex Dummy variable (1=males; 0=females).

Age Dummy variable that equals one if the individual belongs to each age
group out of the four considered: less than 21 (omitted), between 21 and
35, between 36 and 55, more than 56.
Vocational Training Dummy variable (1 if the individual attended some vocational training
program; 0 otherwise).
Schooling Years of schooling. Alternatively, this is measured by a set of categorical
variables each corresponding to a certain schooling level. Six levels were
considered: less than 4 years (omitted), 4 to 6 years, 9 years, high school,
technical college, and college.
First Job Dummy variable (1 if current job is the first job; 0 otherwise)
Labor Market Experience Years since the individual got his first job.
Tenure Months with current employer.
Left Previous Job Because Set of dummy variables for the reasons why individual who are not in
their first job left the previous one. Four reasons were considered:
dismissal (individual or collective dismissal, including if because of plant
closing mutually agreed termination of contract), end of fixed-term
contract, quit, and others (omitted). All equal 1 if the corresponding
situation holds.
Type of Employer Set of dummies for the type of employer. Three categories were
considered – private company, state-owned company, and Public
Administration and others (omitted); all equal one if the corresponding
category holds.
Fixed-term Contracts Dummy variable (1 if the individual is on a fixed-term contract, 0
otherwise)
Log Hourly Wage Logarithm of the individual hourly wage (hourly wages are computed as
the ratio between the wage per month and the usual hours of work).
Things are less clear-cut with labor market experience. The positive sign in the
employment to out-of-the-labor force equation captures retirements, which are imperfectly
captured by the age variable. In the employment-to-unemployment equation the negative
sign is as expected and has an obvious matching interpretation. However, the same
interpretation would imply a negative sign in the employment to employment equation
which is contrary to the parameter estimated obtained. More experienced workers changing
jobs could be the result of job loss due to plant closings that imply termination of good and
bad matches alike. This would be consistent with the positive sign of the labor market
experience variable in the open-ended to the fixed-term contract equation, but inconsistent
with the negative sign of the tenure variable in the employment to employment equation.
Taking the estimates of the labor market experience and tenure variables in the
employment-to-employment equation jointly suggests that there is a non-trivial pool of
experienced workers still switching jobs. These workers may either have acquired
experience through a succession of temporary jobs or in a single stable job that at some
point was extinguished. However, if we drop the age variable from the set of regressors in
both equations in table 7, the parameter estimates for experience actually become negative.
In the OEC-FTC equation the negative sign is as expected because more experienced
workers are more likely to have found an adequate match. The negative sign in the FTC-
OEC equation indicates that older workers that at some point receive a temporary contract
are less likely to switch to a regular position.
The signs of the coefficients of the End of FTC variable (that indicates individuals who
left their previous job because of the expiration of a fixed-term contracts) in all the
equations clearly indicate that a series of temporary contracts with intervening
unemployment or inactivity episodes are the rule for some workers at least.
The equation for the transitions from fixed-term to open-ended contracts (table 7)
highlights the role of some human capital variables.17 Workers with more schooling, more
experience, and longer tenures are more likely to receive a permanent contract.
17
Transitions from open-ended contracts to fixed-term contracts are rare events – every quarter 0.33% of all
workers with a permanent contract move to a fixed-term contract. This reduces the quality of the results
obtained for the equation of transitions between these two types of contracts.
Table 7. Transitions between Fixed-Term and Open-Ended Contracts

FTC-OEC OEC-FTC
Coef. Mg. Effect Coef. Mg. Effect
Constant -1.2864 ** -1.2293 *
(0.0521) (0.6088)
Sex 0.1017 ** 0.0181 -0.0639 *** -0.0001
(0.0521) (0.0605)
Age -0.0124 ** -0.0013 -0.0333 -0.0001
(0.0069) (0.0092)
Voc. Training 0.0521 *** 0.0057 -0.0775 *** -0.0001
(0.0841) (0.1143)
Schooling 0.0100 *** 0.0011 0.0191 * 0.00003
(0.0077) (0.0097)
First Job -0.0535 *** 0.0055 0.0747 *** 0.0001
(0.0892) (0.0986)
Labor Market Experience 0.0049 *** 0.0005 0.0237 * 0.00004
(0.0064) (0.0087)
Left Previous Job because
Dismissal 0.0993 *** 0.0111 -0.0484 *** -0.0001
(0.0841) (0.0959)
End of FTC -0.0135 *** -0.0014 0.1150 *** 0.0002
(0.0790) (0.0975)
Quit 0.1273 *** 0.0145 0.0240 *** 0.00004
(0.0854) (0.0902)
Firm’s Ownership
Private 0.2033 * 0.0120 0.1382 *** 0.0002
(0.071) (0.0970)
State-Owned -0.0890 *** -0.0088 -0.0921 *** -0.0001
(0.1267) (0.1921)
Log Hourly Wage -0.0521 *** -0.0055 -0.0725 *** -0.0001
(0.0691) (0.0855)
Tenure 0.0039 0.0004 -0.0068 -0.00001
(0.0009) (0.0008)
Number of Obs. 7057 490863
Log Likelihood -5893.8 -3770.18

χ2
52.67 301.02
Industry and quarterly dummies included.
*, **significant at 5 percent and 10 percent, respectively;*** not significant at 10 percent.
Standard errors in brackets.
Wages are inversely related to the probability of moving from a temporary to a fixed-
term contract, which means that there is a wage differential compensating workers for job
insecurity, but not necessarily for having one fixed-term contract. Estimates are significant
only for age, professional experience, and tenure variables.
The negative sign obtained for the age and tenure variables are consistent with a
matching interpretation. Older workers and those with longer tenures are less likely to leave
their permanent positions to take a temporary contract (the effect is the same for job
switching independently of the type of contract). Younger workers, even if they have found
a permanent match, are more likely to engage in on-the–job search and, eventually, switch
jobs, even if to one with less formal job security. Schooling has a positive effect on the
probability of both job switching and switch from an open-ended to a fixed-term contract.18
The negative sign for the estimate of the coefficient of the tenure variable is also consistent
with reverse seniority criteria applied to dismissals.
As discussed, most of the consequences of fixed-term contracts for long-run economic
growth are transmitted via changes in productivity that, eventually, reflect themselves in
earnings profiles. Hence, Mincerian-type wage equations for workers on fixed-term
contracts and open-ended were estimated. Maximum likelihood methods, adjusting for self-
selectivity, were adopted. The two-step Heckman correction method for selectivity bias was
adopted. The set of variables indicating the worker’s age and the reasons why he or she left
the previous job were included in the probit and dropped from the wage equations. The
assumption is that the age of the individual has an adverse effect on his or her probability of
receiving a permanent contract but has no effect on his wage conditional on the type of
contract and work experience. It is also reasonable to assume that the reason why the
individual left the previous job influences the probability of receiving a temporary contract
but not the wage he is offered, controlling for the workers’ relevant characteristics. Results
are also reported in table 8.
Results for the probit equation confirm the influence of individual characteristics on the
type of contract the individual holds and are as expected. From the results of the probit
equation we can see that each additional year of labor market experience reduces the
probability the individual holds a temporary contract. This is consistent with fixed-term
contracts being used as screening devices as, in this case, the longer individuals have been
in the labor market the more likely it is that they have found an adequate (permanent)
match. Three additional dummy variables indicating the reasons why individuals who are
not on their first job left the previous one were also included as regressors in the probit
equation. All the three reasons considered raise the probability that the worker is on a fixed-
term contract.19 Because in the Portuguese labor market more than 60 percent of all
admissions are on fixed-term contracts it is not surprising to find that a temporary contract
is the most likely outcome of job loss episodes. Being dismissed increases 8.7 percent the
probability that the individual receives a fixed-term contract. Hence, this type of contract
18
The coefficient of the vocational training variable is not significant but has a negative sign, indicating that
general and specific human capital, as measured by schooling and professional experience, and tenure and
vocational training, respectively, have the standard effects in terms of employee-employer attachment.
19
Individuals who previously had another job are asked to choose one among a comprehensive list of reasons to
have left the previous job. These were grouped in four categories – dismissals, voluntary quitting, end of a
temporary contract, and others (omitted in the regression). The omitted category pertains essentially to personal
reasons.
appears as a response to job loss, as temporary jobs reportedly do in other countries

(Farber, 1999).
However, workers who left their previous positions because of the expiration of a
temporary contract are the ones that face the highest probability of being employed with a
fixed-term contract one quarter ahead (the probability exceeds that of the omitted category
by 21.7 percent). Although a more definite conclusion would require the possibility of
tracking workers over longer periods, that suggests that for some individuals, at least, fixed-
term contracts are not bridges to a permanent contract, but rather a persistent work
arrangement that implies frequent job changes.
Without controlling for any employee or employer characteristics, fixed-term contracts
are clearly associated to lower pay - the average hourly wage of employees with fixed-term
contracts is about 82% of the corresponding measure for employees with open-ended
contracts. But, looking at the results of the estimation of the wage equations we see that
returns to labor market experience and tenure differ across contractual arrangements
(figures 1.A and B).
For workers on open-ended contracts, returns to experience follow the usual pattern,
increasing at a decreasing rate until thirty years of experience in the labor market and
declines onwards. However, for workers on fixed-term contracts returns to experience are
smaller at any point of their working lives and over the life-cycle. Individuals that at some
point of their professional careers move to a fixed-term contract experience a long-term fall
in returns to human capital investments (figure 1.A).
Returns to tenure also differ significantly across types of contracts. Up to six months,
wages of workers with fixed-term contracts are above those of workers with open-ended
contracts do (figure 1.B). For tenures longer than six months the reverse is true. Because
fixed-term contracts are easier to terminate, bad matches are likely to be quickly undone.
This implies more turnover and better average match quality among temporary than
permanent workers, therefore higher wages for temporary workers with short tenures.
However, a compensating differential mechanism may also be at work here, with temporary
workers being compensated for less job security. If the probability of receiving an open-
ended contract increases with tenure then the compensating differential should gradually
fade away generating a wage profile similar to the one in panel B.
The fact that this relationship is reversed within a six-month period may be the result of
good temporary matches becoming permanent well before the legal maximum duration of
fixed-term contracts is reached.
Estimates of the coefficient of the part-time variable as well as firm’s ownership
dummies all change signs from one wage regression to the other. For workers on fixed-term
contracts, being on a part-time job reduces hourly wages whereas for permanent contracts
the opposite is true. Private firms pay permanent workers above Public Administration but
below in the case of workers on fixed-term contracts. This suggests that fixed-term
contracts may be used to a non-negligible extent as a means of insulating permanent
workers from business fluctuations along an insider-outsider mechanism. Fixed-term
contracts permit firms saving on direct wage costs only for part-time workers. If anything,
for full-time workers wage costs are higher than the corresponding for permanent workers,
particularly for shorter tenures (less than one year).
Table 8. Wage Equation, by Type of Contract
Probit Equation Wage Equation

(dependent variable: FTC=1)
Coefficient Marginal Fixed-Term Open-Ended

Effects Contracts Contracts
Constant -0.4184 (0,0296) 7.0471 7.0672
(0.0129) (0.0326)
Sex -0.1059 (0.0149) -0.0202 0.1703 0.1390
(0.0037) (0.0093)
Age
20-34 -0.3681 (-0.0269) -0.0669
35-54 -0.5326 (0.0374) -0.0688
55-64 -0.4223 (0.0613) -0.0637
Schooling -0.0015* (0.0019) -0.0003
Primary 0.1810 0.0682*
(0.0065) (0.0244)
Mandatory 0.4463 0.1862
(0.0078) (0.0264)
Secondary 0.6130 0.3587
(0.0081) (0.0266)
Technical School 1.1012 0.8457
(0.0108) (0.0354)
College 1.2919 1.1387
(0.0087) (0.0273)
Vocational Training 0.1018 0.0670
(0.0055) (0.0139)
Labor Market Experience -0.0281 (0.0011) -0.0053 0.0195 0.01361
(0.0006) (0.0015)
Labor Market Experience – -0.0003 -0.0003
Squared (0.00001) (0.00004)
Tenure 0.0159 0.0172
(0.0005) (0.0036)
Tenure-Squared -0.0002 -0.0001
(0.00001) (0.0002)
First Job -0.0499 -0.0799
(0.0047) (0.0145)
Part-time -0.0557 0.3122
(0.0056) (0.1306)
Type of Employer
Private -0.0520 0.0011
(0.0053) (0.0148)
State-Owned 0.0579 -0.0127
(0.0064) (0.0142)
Left Previous Jobs because
Dismissal 0.3954 (0.0207) 0.0874
End FTC 0.8244 (0.0203) 0.2170
Quit 0.1990 (0.0213) 0.0407
Industry Dummies Yes Yes
σ 0.3414 0.3436
ρ 0.3922 -0.3101
Number of observations = 56140; Log Likelihood = -37932.45. See table 6.A for definitions.
The omitted category for Age is “less than 20” and for schooling “less than 4 years”.
Figure 1. Wage Profiles, by type of Contract
W age Tenure Profile
7,8
7,7
7,6
log hourly wage
7,5
7,4
7,3
7,2
7,1
7,0
6,9
0 5 10 15 20 25 30 35
m onth
open-ended ftc
W age-Ex perienc e P rofile
7,7
7,6
log hourly wage
7,5
7,4
7,3
7,2
7,1
7,0
0 10 20 30 40 50
y ears
open-ended ftc
6. Conclusions
Consequences of fixed-term contracts for growth depend crucially on the use employers
make of them. In this paper I considered three possibilities. Fixed-term contracts may be
used to (i) deal with fluctuations of the workload, (ii) screen workers for permanent
positions, or (iii) churn workers in continuing jobs.
Using firm-level data on the proportion of fixed-term contracts and the number of fixed-
term contracts converted into permanent, I found that the major reason firms use these
contracts is for screening. What firms actually do is to hire new workers with temporary
contracts, offer them training thereby eliciting the information they need about the quality
of the match. When this information arrives, employers decide either to destroy the match
or convert it into permanent.
The use of fixed-term contracts and the amount of training firms offer to their
workforce are complements. This is similar to what we observe in other labor markets
concerning the relation between training and some temporary forms of employment, such
as temporary agency work. To some extent, these two contractual types – fixed-term
contracts and temporary agency work – may be thought of as substitutes. This permits
anticipating that enforcing more restrictions on the use of these contracts may lead to their
replacement by other forms of temporary work, instead of stimulating the use of open-
ended contracts.
Consistent with fixed-term contracts being used as screening devices, the rates of
transition of individuals from temporary to permanent contracts also indicate that fixed-
term contracts are stepping-stones towards a permanent job rather than dead-end jobs. This
result is particularly strong for young workers who use fixed-term contracts as ports of
entry into the labor market. In this context, the observed magnitude of the rates of
transitions from fixed-term to open-ended contracts becomes an indicator of the success of
this joint process of search and screening.
However, for some workers, at least, a sequence of temporary contacts with or without
intervening episodes of job loss is the rule. For these workers, the results obtained from the
estimation of wage equations raise some serious policy concerns. If we compare the age-
earnings profiles of workers with temporary and permanent contracts we will find that
temporary workers receive less returns to experience. This occurs either because they
receive less training due to the very nature of their employment contracts, or because
frequent spells of unemployment or inactivity imply a complete depreciation of whatever
human capital they may have accumulated. These results should not obscure the fact that
employer and employee data alike indicate that fixed-term contracts are predominantly used
as screening/searching devices that eventually lead to the formation of better, therefore
more stable, matches.
Unlike what would happen if they were mostly used as buffer stocks or instruments of
churning policies, I found no reasons supporting concerns over fixed-term contracts having
prejudicial effects in terms of the economy’s growth prospects, via harmful productivity
effects.
The role fixed-term contracts play as part of employers and employees’ strategies in the
Portuguese labor market is akin to what has been reported for Italy, France and the United
Kingdom, although very different to the role they play in Spain. But, we already knew that
the Portuguese and the Spanish labor markets differ more than they coincide.
References
Abowd, John M., Patrick Corbel, and Francis Kramarz (1999), “The Entry and Exit of
Workers and the Growth of Employment: An Analysis of French
Establishments”, Review of Economics and Statistics, 81: 170-187.
Adam, Paula, and Patrizia Canziani (1998), “Partial De-regulation: Fixed-term Contracts in
Italy and Spain”, Centre for Economic Performance Discussion Paper nº 306.
Alba-Ramírez, Alfonso (1998), “How Temporary is Temporary Employment in Spain?”,
Journal of Labor Research, 19: 695-709.
Amuedo-Dorantes, Catalina (2000), “Work Transitions Into and Out of Involuntary
Temporary Employment in a Segmented Market: Evidence from Spain”,
Industrial and Labor Relations Review, 53: 309-325.
Arai, Mahmood, and Fredrik Heyman (1999), “Permanent and Temporary Labour: Job and
Worker Flows in Sweden, 1989-1998”, Paper presented to the XI Conference of
the European Association of Labour Economists, Regensburg (Germany), 23-26
September.
Autor, David H. (2000), “Why Do Temporary Help Firms Provide Free General Skills
Training?”, NBER Working Paper 7637.
Blanchard, Olivier, and Augustin Landier (2001), “The Perverse Effects of Partial Labor
Market Reform: Fixed Duration Contracts in France”, NBER Working Paper
8219.
Blanchard, Olivier, and Pedro Portugal (2001), “What Hides Behind an Unemployment
Rate: Comparing Portuguese and U.S. Labor Markets”, American Economic
Review, 91: 187-207.
Booth, Alison L., Marco Francesconi, and Jeff Frank (2000), “Temporary Jobs: Stepping
Stones or Dead Ends?”, IZA Discussion Paper nº 205.
Burgess, Simon, Julia Lane, and David Stevens (2000), “Job Flows, Worker Flows, and
Churning”, Journal of Labor Economics, 18: 473-502.
Caballero, Ricardo J., and Mohamad Hammour (2000), “Institutions, Restructuring, and
Macroeconomic Performance”, NBER Working Paper 7720.
Cameron, A. Colin, and Pravin K. Trivedi (1998), Regression Analysis of Count Data,
Cambridge: Cambridge University Press.
Davis, Steven J., and John Haltiwanger (1990), “Gross Job Creation and Destruction:
Microeconomic Evidence and Macroeconomic Implications”, NBER
Macroeconomics Annual 5: 123-168.
Davis, Steven J., John Haltiwanger, and Scott Schuh (1996), Job Creation and Destruction,
Cambridge (Ma): MIT Press.
Dertouzos, James N., and Lynn Karoly (1998), “The Implications of Employment Liability
for Welfare Reform”, Journal of Labour Research, 19: 295-311.
Farber, Henry S. (1999), “Alternative and Part-time Employment Arrangements as a
Response to Job Loss”, Journal of Labor Economics, 17: S142-S169.
Guell, Maya (2000), “Fixed-term Contracts and Unemployment: An Efficiency Wage
Analysis”, Centre for Economic Performance.
Hamermesh, Daniel S., and John R. Wolfe (1990), “Compensating Wage Differentials and
the Duration of Wage Loss”, Journal of Labor Economics, 8: S175-S197.
Hopenhayn, H. and R. Rogerson (1993), “Job Turnover and Policy Evaluation: A General
Equilibrium Analysis”, Journal of Political Economy, 101: 915-938.
Houseman, Susan N. (2001), “Why Employers Use Flexible Staffing Arrangements:

Evidence from an Establishment Survey”, Industrial and Labor Relations
Reviews, 55: 149-170.
Hunt, Jennifer (2000), “Firing Costs, Employment Fluctuations and Average Employment:
An Examination of Germany”, Economica, 67: 177-202.
Nagypál, Éva (2001), “Fixed-term Contracts in Europe: A Reassessment in Light of the
Importance of Match-specific Learning”, Stanford University (mimeo).
Nickell, Stephen, and Richard Layard (1999), “Labor Market Institutions and Economic
Performance”, in Orley Ashenfelter and David Card (eds.), Handbook of Labor
Economics, vol.3: 3029-3084.
Portugal, Pedro (2000), “Volatilidade do Emprego, Protecção ao Emprego e Desemprego”,
Banco de Portugal (mimeo).
Saint-Paul, Gilles (2000), The Political Economy of Labour Market Institutions, New York:
Oxford University Press.
Serrano, Carlos García (1998), “Worker Turnover and Job Reallocation: The Role of Fixed-
term Contracts”, Oxford Economic Papers, 50: 709-725.

Desenvolvimento Económico Português No Espaço Europeu 2002

Enviado por

Dados do documento

Descrição original:

Direitos autorais

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Desenvolvimento Económico Português No Espaço Europeu 2002

Enviado por

Direitos autorais:

Conferência do Banco de Portugal

Economic Research Department

Printing number: 750

Duty copy no. 208266/04

A Conferência “Desenvolvimento Económico Português no Espaço Europeu:

Procurou configurar-se esta iniciativa com alguns traços relativamente inovadores. Em

Os estudos que estão na base da Conferência provieram quer de um concurso público

The Conference “Portuguese Economic Development in the European Context:

Nota de Apresentação / Foreword

Nota de Apresentação ................................................................................................... V

Conferência do Banco de Portugal sobre “Desenvolvimento Económico

Conference Held by the Banco de Portugal on “Portuguese Economic Development

Sessão 1 – Convergência Real Portuguesa / Session 1 – Portuguese Real

Portuguese Economic Growth Re-Examined: an Anti-Fado Manifesto ....................... 33

Productivity Convergence: Portugal and the European Union ..................................... 61

Sessão 2 – Geografia Económica e Infra-estruturas / Session 2 – Economic

Public Investment in Transportation Infrastructures and Economic Performance

Production Technologies and Technical Efficiency: Evidence from Portuguese

Transport Policies in Light of the New Economic Geography:

Acessibilidade Económica e Bem-Estar: Evidência da Península Ibérica.................... 189

Sessão 3 – Capital Humano e Crescimento / Session 3 – Human Capital and Growth

Education and Earnings in Portugal ............................................................................. 219

Evaluating Efficiency in the Portuguese Health and Education Sectors....................... 253

Qualification Requirements and Educational Attainment in Portugal, 1985-1997....... 303

Sessão 4 – Instituições, Reformas e Crescimento / Session 4 Institutions,

Labor Markets in Portugal: Recent Performance and Challenges for Development

A Justiça e seu Impacto Sobre as Empresas Portuguesas ............................................. 445

Fixed-Term Contracts, Employment Flows, and Productivity ..................................... 479

Conference Held by the Banco de Portugal on “Portuguese

José A. Ferreira Machado

3. Opening the black box

3.1. Transport infrastructures

Transport infrastructures are important for development in as far as accessibility to

The contribution by Armando Pires (“Economic Accessibility and Welfare: Evidence in

3.2. Human Capital

3.3. Labour Market

As Traça remarks in his contribution “Labour Markets in Portugal: Recent Performance

Today there is a consensus among economists on the importance of institutions for

4. Concluding remarks and policy implications

i. It is necessary to pursue the effort of improving transport infrastructures. However, I

iv. It is necessary to evaluate the results of public investment in vocational training.

Gylfason, T. “Principles of Economic Growth”. Oxford U. P., 1999.

Session 1 - Portuguese Real Convergence

Paper prepared for submission to the Conference on

2. The comparative performance of the Portuguese economy

GRAPH 1 ABOUT HERE

Portugal Spain Greece Ireland Average 9

Notes and sources: see Table 1

GRAPH 2 ABOUT HERE

Portugal Spain Greece Ireland

Notes: convergence defined according to:

GRAPH 3 ABOUT HERE

3. Accounting for economic growth

Labour Human capital Capital GDP

1910-1934 1.00 2.08 1.25 2.17

Annual growth rates As percent of output growth

4. Structural change and economic growth slowdown

Graph1 - Growth of income per capita: Portugal and Average 9, 1910-98

Sources and notes: see Table 1.

Graph 2 -Portuguese income per capita as per cent of Average 9, 1910-1998

Sources and notes: see Table 1.

Sources: See Table 1.

Miguel Lebre de Freitas

This paper provides an empirical assessment of the Portuguese economic growth,

THOMAS MALTHUS (1766-1834)1

1 Thomas Malthus (1817). The quotation is borrowed from Gylfason (1999).