Inequality, Poverty and Economic Growth. The Case of Chile 1990-2009.

By Jorge Rojas The aim of this article is to provide insight on the relationships between dierent dimensions of inequality, poverty, and economic growth using regression analysis for the Chilean case between 1990 and 2009. This study explores dierent sources of inequality and poverty from a theoretical perspective such as income distribution, wealth distribution and access to information. On the empirical side, the analysis is done using data for household size, schooling levels, household income, nutritional state of minors and government subsidies disaggregated by quintiles and zones. Descriptive statistics are provided and a parsimonious econometric model is estimated being subject to dierent robustness checks. In the literature, the evidence between economic growth and its impact on the reduction of poverty is strong. However, the relationship between growth and inequality is much more weak. There are no conclusive results in this regard. For the case of Chile, there is enough evidence to suggest that the impact of economic growth is not signicant on reducing inequality, but it plays an important role in reducing poverty. These results could be explained for the increase in access to higher education and its hight level of return. Another nding is that the Chilean government has been indeed providing subsidies to the group who needed the most, that is, the rst and second quintiles in rural areas. However, this government support seems not to be enough in the sense that people obtaining it still have very low standards of living relatively to the ones in the two richest quintiles. Keywords: income, wealth, distribution, Chile, poverty, inequality, economic growth. Is equality in income really important to achieve a sustainable economic growth? Is equality in terms of access to quality education, information and training of skills what really matters? Would reducing the household size in poor families reduce inequality and poverty? What are the channels in which economic growth impacts poverty and inequality? What are the dimensions of inequality that matter the most? All these questions have concerned economists in the last decades and probably will keep doing so for the years to come. This introduction is divided into two parts. The rst section is a review of the literature related to economic growth and inequality in a macroeconomic level, namely, the dierent neoclassical approaches to answer the question about whether inequality is harmful for
1

QUARTER PAPER

NOVEMBER 2012

economic growth or not. The second section consists in a review of the literature associated with the micro-determinants of poverty and inequality, and their interactions with economic growth. Informational economics is also looked at as a tool to understand how asymmetries in information create inequality in opportunities and its potential impact on eciency and growth. In modern societies the positive relationship between educational level and income is well established. For instance, Barro and Lee (2001) nd that in 2000, members of advanced economies had an average of 9.80 years of school, as compared with only 4.89 years for those from developing countries. A similar pattern occurs within a given country, Mayer (2010) shows that on average, the greater the family income per-capita, the higher the years of schooling for the children of this family. Mayer (2010) suggests that reducing income inequality would reduce inequality in schooling, but there is not strong evidence that reducing the gap in schooling would reduce the gap in income. However, there are several studies that show the endogeneity of schooling and income through the estimation of Mincer equations. In Lucas (1988), Barro (1991) and Benhabib and Spiegel (1994) there is sucient evidence to claim that human capital and education are important driving forces in the determination of growth. There is much less consensus on the relationship between income distribution and growth. In the 1990s, theoretical and empirical studies emerged showing that inequality would be negatively associated with growth across countries, including Bertola (1993), Alesina and Rodrik (1994) and Persson and Tabellini (1994). On the other hand, researchers such as Li and Zou (1998) and Forbes (2000) report ndings with the opposite sign. At the same time, Barro (2000) nds evidence of a negative eect of inequality on growth for poor countries, whereas a positive relationship for rich countries, with only a weak overall eect of inequality on growth and investment. In contrast, Rehme (2007) nds evidence of a U-shaped relation between education and growth, with increases in education leading to rst increases then decreases in growth as well as income inequality. In other words, this issue has important policy implications, since dierent assumptions about the channels between growth and inequality tend to produce dierent outcomes for the poor, see Deininger and Squire (1997). Alesina and Rodrik (1994) study the eect of inequality in income on the economic growth through a neoclassical framework in which agents maximise utility continuously along their lives and do not dier in their levels of human capital. They nd that the decrease in economic growth is not directly related to income inequality but rather the scal and taxation policies implemented by the governments to address the problem. In their model were taxes the responsible for less economic growth since government with high levels of inequality tend to increase taxes and use them in a redistributional manner. Bertola (1993) develops a model with a representative agent of innite life, focusing on the growth and distributional eects of scal policy through taxes and subsidies. Persson and Tabellini (1994) work with an overlapping generations model with two periods of life and agents who have dierent endowments of skills, but they focus on the political equilibrium of society and conclude that inequality leads to policies that do not protect

INEQUALITY AND POVERTY

property rights, and hence through this channel there is a decrease in economic growth. Rehme (2007) contends that while most empirical works focus on linear relationships, the growth-inequality relationship may in fact be nonlinear. He argues that both growth and measured inequality can be represented as inverted U-shaped functions of the percentage of high-skilled people in the population (which represents the stage in the development process). This could explain why the growth-income inequality relationship is so dicult to predict, since the sign of the association depends upon where each particular economy lies, in terms of its percentage of high-skilled people relative to the turning points of its growth and measured inequality functions. The above review is addressing the macroeconomic characteristics of the problem. However, some may argue that there is too much aggregation to obtain real insight about the true mechanisms in which economic growth inuences the time path of poverty and inequality. Data from the World Bank (WB), the International Monetary Fund (IMF), and other international organizations show strong evidence of the importance of economic growth in the reduction of absolute poverty. For example, from the World Bank database is possible to observe that Chile has experienced an economic growth in the last decade around 4% per year and the reduction in poverty, measured as the national poverty line, has dropped from 21.6% in 1998 to 15.1% in 2009. However, income inequality has stably remained around a Gini index of 50 points, being 52.1 in 2009. Another good example, and even a more dramatic one of this relationship between growth, poverty and inequality is the case of China. Over the past two decades, China has grown at a rate close to 10% per year, and absolute poverty has dropped from 83.7% in 1987 to 29.8% in 2008. Nonetheless, inequality grew from a Gini index of 32.4 in 1990 to 42.5 in 2005. From these two cases result hard to nd conclusive answers about the links between the issues at hand. In an attempt to understand this seemingly contradictory results is that microeconomic literature may be helpful to review. Lockwood and Manning (1988) explore inequality in information with individuals who have the same level of ability and equality to get into a given occupation, i.e., equality of opportunity. They show that these dierent levels of information lead to inecient economic outcomes. Moreover, they show that under reasonable assumptions with heterogenous individuals with asymmetric information, it may be the least productive individuals who become managers, and hence the decline in productivity and the worsening in inequality. Thus, access to information plays a signicant role not only for economic growth, but also for the distribution of that growth. The last part of this literature review is associated with the interlinkages between poverty, inequality and growth, but at the household level. This part focuses on articles that use empirical analysis to uncover the channels that might play a role in the reduction of absolute and relative poverty. Weller (2007) shows some evidence for industrializing countries that a progressive income taxation system may result in a more equitable income distribution with higher revenues, less nancial and economic volatility, and faster growth, but also with larger decits. The mechanism that explains these ndings is basically a strong formal economy. Wodon (2000) does a micro-oriented analysis of poverty and inequality for Bangladesh using household level data of ve national surveys. He runs the analysis

QUARTER PAPER

NOVEMBER 2012

with poverty data decomposed by regions, i.e., he computes Gini indices at the rural, urban and national level. His main nding is that education in urban areas and land in rural ones are the two variables that better explain inequality. In a similar fashion David and Marouani (2012) study the Syrian case. Krsti and Sanfey (2011) do a similar analysis for c Serbia, but taking into account the informal economy. Goudie and Ladd (1999) write an article that shows the interactions between economic growth, poverty and inequality in all directions. Figure (4) shows the endogeneity nature of inequality.

Figure 1. Linkages between growth, inequality and absolute poverty. Source: This gure is retrieved from Goudie and Ladd (1999, p. 179)

Mayer (2010) and Fleisher et al. (2005) analyse the eects of schooling and the returns to higher education on inequality, respectively. Mayer focuses her analysis in the USA, while Fleisher et al. do it in China. Both achieve a similar conclusion in the sense that higher educational attainment, on average, improves the earnings of that given person, holding everything else constant. Hence, a reasonable, plausible and simple conclusion is that one way to decrease income inequality may be through public policies oriented to increase the higher education attainment of low-income and middle class families. Fan (2008, ch. 2) provides the methodology to study public spending in developing countries explaining in great detail where and how to get the data, and the econometric tools to be used in running the analyses, namely, the generalized method of moments (GMM) and panel data. For the specic case of Chile, Contreras (2003) focuses his analysis on inequality and poverty and its relationship with the rapid economic growth experienced by Chile between 1990 and 1996. He uses parametric and non-parametric methods to address the question at hand. Using the Datt-Ravallion decomposition, Contreras nds that the 85 % of the reduction in poverty may be attributed to economic growth. Then he concludes that although inequality remains the same, poor and rich are better o in absolute terms. Agostini, Brown and Roman (2010) focus their analysis on ethnic groups rather than the

INEQUALITY AND POVERTY

usual regional o zonal approach. They use bootstrapping to simulate values of household income, and this complete set of simulated values is then used to calculate the headcount ratio and Gini coecients for each ethnicity represented in the population. Their analysis is highly technical, but most of their conclusions are very intuitive and descriptive. Much research has been done on the relationships between poverty, inequality and economic growth, and the channels in which they interact with each other. In this paper, there are two questions to be answered: How has inequality in Chile evolved during the last two decades? Would reducing the household size in poor families decrease inequality in income, nutrition and schooling?
I. Descriptive Statistics

According to Schneider (2002), the size of the informal economy in Chile was around 19.8% in 2000. So, keeping this gure in mind, the household data at hand is expected to be reliable for the period of time to be analysed. These data sets have been obtained from the Characterizacin Socioeconmica Nacional (CASEN), a periodic survey undertaken by o o Chiles Ministry of Planning and Cooperation (MIDEPLAN). The details of the variables computed to run the econometric analysis are located at the appendix. Next, in this section there are dierent aggregate data attempting to provide an overview of Chile and its characteristics at the macroeconomic level about inequality, economic growth, education and government spending on education.

Table 1Annual growth of macroeconomic indicators for Chile, 1990-2009. 1990 GDP growth 3.7 Population growth 1.8 Ination rate 26.0 Unemployment rate 5.7 Mining production growth 0.8 Source: International Monetary Fund. 1992 12.3 1.9 15.4 4.4 6.0 1994 5.7 1.8 11.4 5.9 5.8 1996 7.4 1.5 7.4 5.4 22.8 1998 3.2 1.3 5.1 7.2 6.5 2000 4.5 1.2 3.8 8.3 4.4 2003 3.9 1.1 2.8 7.4 6.9 2006 5.7 1.0 3.4 7.7 0.7 2009 -1.0 1.0 1.5 9.8 0.4

From table (1) is easy to note that the Chilean macroeconomy responds very quickly to external shocks. For example, in 1998 the GDP growth observe was 3.2% while two years before was 7.4%. This decrease in growth for this period of time can be explained through the Asian crisis. Next, in 2009 there was a negative GDP growth, once again this may be explained by the Global Financial Crisis (GFC) started in 2008. From this table is also possible to observe the strong link between economic growth and employment, in which for the period 1990-2009, the lowest unemployment was reached in 1992 when Chile was growing at the very high level of 12.3%, while the highest unemployment level corresponds to the year after the GFC when Chiles economic growth was negative one percent. Mining production growth is also very responsive to external demand for minerals, mainly, copper.

QUARTER PAPER

NOVEMBER 2012

Table 2Income and health for Chile, 1990-2009. GDP per capita PPP constant 2005 $ 6,936 8,108 8,848 10,186 10,906 11,029 11,698 13,248 13,832 Life expectancy at birth Total years 73.6 74.2 74.8 75.3 76.0 76.8 77.9 78.4 78.8

1990 1992 1994 1996 1998 2000 2003 2006 2009 Source: World Bank.

Table (2) shows the direct results of the rapid economic growth experience for Chile in the last two decades. GDP per capita grew from a modest seven thousand 2005 dollars in purchasing power parity in 1990 to almost fourteen thousand 2005 dollars PPP in 2009. This represents an overall increase in welfare of two-folds in only two decades. Life expectancy also increased from 73.6 years to 78.8 years for the same period. Although, these aggregated numbers look very well. There is a reality in which Chile has not performed as well as it wanted. This weakness in the overall economy is inequality. Table (3) provides three measures of inequality. The rst one is the Gini index, that as we know is subject to multiple criticisms. Nonetheless, this index is a useful to get an approximate idea of the inequality problems that a capitalist and open to international trade country as Chile may have. From table (3) is clear that the aggregate level of inequality in Chile has not changed signicantly during the last two decades. Despite the eorts by the dierent governments through progressive scal policies, the inequality levels has remained absolutely static. However, as it was mentioned before, the absolute poverty has declined signicantly showing how economic growth plays an important role to alleviate poverty, but lacks of impact in matters of inequality. The same can be concluded for income share held by the

Table 3Measures of Inequality for Chile, 1990-2009. 1990 55.3 45.2 1.27 1992 54.8 45.1 1.42 1994 55.1 44.6 1.29 1996 54.9 44.4 1.33 1998 55.5 45.0 1.25 2000 55.3 45.3 1.31 2003 54.6 45.0 1.37 2006 51.8 42.0 1.55 2009 52.1 42.8 1.53

Gini index Income share held by highest 10% Income share held by lowest 10% Source: World Bank.

highest 10% and the lowest 10%. These shares are practically unchanged. Another striking feature from table (3) is not only the static gure of inequality but is also its degree. For example, in 2009 the highest 10% obtained the 42.8% of income, while the lowest 10%

INEQUALITY AND POVERTY

obtained only the 1.53%. This is a clear sign that either the highest 10% is extremely productive or that there are important market failures such as high bargaining power by the business class, networks based on family background not merit or talent, moral hazard by the regulators, or some social structure inequality from colonial times. Moreover, notice that inequality within the highest 10% is also very signicant. It basically replicates the structure for the whole country but within this decile.

Table 4Expenditures on Education, 1998-2009. Public (Pb.Exp.) Private (Pr.Exp.) Pb.Exp. annual growth Pr.Exp. annual growth % of GDP % of GDP % % 1998 3.4 2.7 1999 3.8 3.0 11.0 13.6 2000 3.9 3.2 6.3 11.1 2001 4.5 3.4 19.1 8.3 1.5 7.7 2002 4.5 3.6 2003 4.1 3.8 -3.9 9.1 -1.3 -2.0 2004 3.9 3.5 2005 3.6 3.1 -2.5 -8.3 4.3 -8.3 2006 3.6 2.6 2007 3.9 2.7 16.5 5.8 14.4 7.1 2008 4.4 2.8 2009 4.8 3.0 8.9 8.1 Notes: Authors computation based on dierent sources. Source: UNESCO-OECD-Eurostat (UOE) data collection on education statistics and World Bank.

According to the OECD, the average public expenditures of its members in 2007 were 4.6% of the GDP. Chile was below this benchmark in this year, and during the last decade the average for Chile has been around 4%. Even more, in 2007 Chile was the country in the OECD with the lowest public expenditure per student1 . Table (4) also shows the private expenditure on education which is comparable to the public expenditures. This is due to the high level of private-subsidized educational institutions in Chile.
II. The Framework

The idea behind the mathematical model described below can be seen in gure (2). This vision is conceived under the implications of family size, education and the political environment. Before getting into the details of this framework, the motivation for this will be exposed taking fragments from The Economics of Poverty and Discrimination, chapter 7, written by Schiller (1984). If the poor are mindlessly, even willfully, propagating beyond their nancial means, policy intervention will be more dicult and less available; but if the poor are excluded from
1 OECD Family Database. Social Policy Division at the Directorate of Employment, Labour and Social Aairs. www.oecd.org/social/family/database

QUARTER PAPER

NOVEMBER 2012

access to birth control, intervention is easier and more broadly acceptable. The former case comes closest to the position that excessive family size leads to poverty, while the latter case implies the reverse. We must also ponder the accumulation of evidence suggesting that the poor continue to have little access to birth control information. Public schools, especially in low-income areas, are still reluctant to provide birth control information, while public welfare authorities are, in most cases, prohibited from providing it... Because of these constraints, poor families often end up with more children than other families, not because they want them, but because they are unable to prevent them. Single-parent, or broken, families do not t easily into traditional (American) patterns and are often regarded as a threat to community morality, not to mention the whole institutional character of marriage... For instance, a twenty-six-year-old widow, may command more social support and economic resources than an unwed mother of the same age... The single parent is unable to devote all time to labor market activity, at least not without paying someone else to assume household responsibilities. Hence, the potential net income of a one-parent family is often closer to one-third rather than to one-half of a two-parent familys income. Poverty and social injustice were more likely to cause family breakup, not the reverse... Daniel P. Moynihan found that the rates of separation and divorce followed economic events very closely... Family stability was apparently the result, not the cause, of economic events... A seven-year study of 5,000 families conducted by the University of Michigan Survey Research Center shows that in general, average family income turned out to be the best predictor of family stability, for both whites and blacks. The lines above provide insight to think of the model in gure (2). Inequality, economic growth and the political institutions are highly integrated. We could describe them as coupled dynamic systems in which each isolated system depends on what is happening in the other ones. Of course, the links between the sectors show in gure (2) may be present in all directions, but the ones show are the ones that will be highlighted in this model since there are sucient studies and evidence to attempt a sensical and plausible explanations. I will start with the node at inequality. For historical reasons we may argue an initial inequality distribution. In the case of Chile, this initial inequality distribution can be measured in either income or wealth, and its origin could be related to migration from Spain at the beginning of the colonial era in which native Chileans were subjugated by the military power of the invaders who took their land, natural resources and slaved these populations. Through consecutive historical events such as revolutions, civil wars and civil rights movements, the initial inequality evolved to the one observed in recent times in which native Chileans are still behind non-native groups, but there are also a large middle class, although not a middle class society. This inequality in turn has an

INEQUALITY AND POVERTY

impact on eciency through the mechanisms described by Lockwood and Manning (1988) in the sense that people with higher natural level of skills than others may not achieve the managerial positions because of asymmetries of information.

Ineciency
f

Inequality
g

Political Institutions

b d

Economic Growth
c

Poverty

Figure 2. Main interactions in the socio-economic system. Notes: Authors own gure. Notice that the relative size of the circles show relative importance of the variable, while dashed lines means feedback relations.

Therefore, this eciency loss would impact on productivity and this onto the overall economy. This micro shock would aect dierent sectors in the economy slowing down growth, or in dierent words, not letting the economy realize its full potential. As discussed in the introduction, there is enough evidence that economic growth tends to decrease poverty, but its impact on inequality is much less clear. This articial decrease in economic growth has two eects. First, it makes harder to eliminate extreme poverty. Second, it limits the amount of available resources to alleviate the inequality problems. Nevertheless, a caveat is in place. China and other developing economies such as Chile and Brazil have experienced high rates of economic growth, decreasing absolute poverty, but increasing inequality in the case of China, and leaving unchanged the inequality levels for Chile and decreasing them very slightly for Brazil over the last decade. This could be seen as a sign that market forces are not enough to solve the inequality problem, and hence some institutional reforms are needed. What type of institutional reforms? This issue will be address below. At the same time, poverty also impacts on inequality because of the unequal access to education and health care. In a capitalist society 2 in which most services are traded as
2 The

meaning of capitalist society in this context is related to a society who advocate for free-market but in

10

QUARTER PAPER

NOVEMBER 2012

consumption goods, and taking into account that many developing countries provide poorquality public services compared with the ones provided by private sector, a small income dierence among the poor may create large dierence in the longer run. For instance, a poor family with one child may provide a slightly better education than a neighbor family with two children. This results in one family with double disposable income for education for a child than the other family, holding anything else constant. Usually the returns to education, particularly, higher education are very large. Thus, a young person who make it to college may achieve a much better standards of living in his o her adulthood than his or her childhood neighbor who did not get into college education. Beyond this contribution of poverty to inequality there is another crucial factor to be analyzed. Politics. Poor families in their struggle for survival tend to ask their children to work. These children might attend primary school and even high school, of course, public schools with low quality standards at any international measure. Thus, these children devote little time to their learning process and hence their critical thinking is poorly developed. Once they enter into their adulthood become citizens with little understanding of the impacts of public policies in their life, and therefore, they lack the necessary tools to promote change in their societies. This lack of knowledge and skills in turns help to maintain the status quo. In other words, there are no profound institutional reforms. Unless, there is an exogenous shock, i.e., revolutions, civil wars or coup dtats that are able e to introduce important structural reforms. Political institutions with economic growth play a major role to eliminate unfair inequality. By unfair inequality should be understood the idea of being paid an amount smaller that the marginal productivity of labor, capital or any other factor of production as well as not developing the full potential and skills of a person because of lack in opportunities. The globalized world has observed that economic growth by itself does not solve the unfair inequality. Economic growth is a necessary condition, but not sucient. Moreover, the legal framework and the initial conditions of a given economy are essential to eectively address the inequality problem. The next section shows a game theory based model in which the importance of the institutional framework is highlighted in terms of how the political structural may play a role in alleviating poverty and decreasing inequality or stopping inequality to rise. Section IV introduces a simple econometric model. Section V shows the econometric results for the Chilean data. Section VI concludes the project and section VII describes some extensions and renements for future work.
III. The Political Model

The model develops in this section is one of political cooperation for legal reforms. The model is developed in general terms, but its insight may be applied to policies related to poverty and/or inequality.
reality there are many oligopolies rather than being close to high levels of competition.

INEQUALITY AND POVERTY

11

A.

Assumptions

(a) There are two political parties i = A, B, (b) The game is innitely repeated, (c) Both parties have the same discount factor [0, 1],
t (d) Both of them want to minimise their lifetime utility given by: t=0 E[Li (yt , t )] where yt is the policy at time t and t is an exogenous shock at time t,

(e) For simplicity, the loss function is dened as Li (yt , t ) = [(yt yi ) t ]2 where yi is the preferred policy of party i, (f) The economic shock t has zero mean and is iid, (g) yA = yB captures the elements of conict, and t captures the idea of common interest. The recognition rule is dened as t = i in the sense that party i decides yt in period t. The probabilities to this rule may be assigned as: (1a) P(t = A) = P(t = B) = 1

(1b)

where is the probability of being elected at time t and hence being the decision or policy maker. Of course, (0, 1). Further assume that at t = 0 the parties can make some agreements. This is the so-called contracting moment. With this setup the next step is to analyse cooperation and non-cooperation between these two political parties or political coalitions.
B. Cooperation

The rst best is obtained as follows:

(2)

min
t=0

E(LA (yt , t )) +
t=0

t E(LB (yt , t ))

min

t [E(LA (yt , t )) + E(LB (yt , t ))]

t=0

min
t=0

t E(LA (yt , t ) + LB (yt , t ))

12

QUARTER PAPER

NOVEMBER 2012

Next, LA (yt , t ) + LB (yt , t ) = = = = (yt yA t )2 + (yt yB t )2 2 2 (yt yA )2 2(yt yA )t + t + (yt yB )2 2(yt yB )t + t 2 2 2 2 2 yt 2yt yA + yA 2yt t + 2yA t + 2 + yt + 2yt yA + yA 2yt t 2yA t + t 2 2 2 2yt + 2yA 4yt t + 2t

Thus, the minimization problem becomes:

min
t=0

2 2 2 t E(2yt 4yt t + 2t + 2yA )

2 2 t E(yt 2yt t + t ) + 2 t E(2yA ) t=0

min
2 E(2yA ) 2 2yA

2
t=0

is a constant. This value is not important for the minimization = Since problem. The same is true for the number 2 that appears in front of the summation. Therefore, after dropping the constant term the problem is reduced to:

min
t=0

2 2 t E(yt 2yt t + t )

min
t=0

t E[(yt t )2 ]

t ) = t t this value minimises the expected loss through the whole game if and only if the two players cooperate. In this case, with dierent recognition rules this result is unchanged.
C. Non-cooperation

y (

The one-shot Nash Equilibrium (NE) is given by non-cooperation. Notice that the oneshot NE is always an equilibrium in the innite repetition of the game. For the noncooperation path the recognition rule plays a role. P(t = A) = P(t = B) = 1 The equilibrium strategy for party i in the non-cooperation path is:
i yt = yi + t

yt = yt + t

INEQUALITY AND POVERTY

13

Without loss of generality, let i be equal to A, and using the assumption that yA = yB . This could be interpreted that one party wants a progressive tax reform, while the other party wants elimination of any tax. Thus, LA (yt , t ) = LA (yt , t |t = A) = 0 A and LB (yt , t ) = LA (yt , t |t = B) = (2yB )2 A The expected loss at time t is: E(LA (yt , t )) = LA (yt , t |t = A) + (1 )LA (yt , t |t = B) E(LA (yt , t )) = 0 + (1 )(2yB )2 E(LA (yt , t )) = (1 )(2yB )2

(3) For party B: (4)

E(LB (yt , t )) = (2yB )2

Therefore, the expected loss for the innite game V N is given by:
N VA N VA N VA N = (1 )(2yB )2 + VA (1 )(2yB )2 = 1 1 = (2yB )2 1

In a similar fashion for party B is obtained:

N VB =

(2yB )2 1

Assuming that the players are cooperating, but one of them deviates at time t and the other applies the Grim-trigger strategy dened as follows: < t

i yt =

t if y = t yi + t

In general, the payo along the equilibrium path of cooperation will be dierent for the two parties. Given that cooperation implies yt = t this entails to the following payos

14

QUARTER PAPER

NOVEMBER 2012

from cooperation:
2 VA = yA + yA + 2 yA + 3 yA + ... 2 yA 2 VA = 1

For party B is basically the same, that is:

2 VB = 2 yB 1

Let ViD be dened as the present value of the expected payo from an opportunistic deviation.
D N VA = 0 + VA D N VB = 0 + VB D VA = D VB

1 (2yB )2 1 = (2yB )2 1

Next, compare the expected loss for the case in the equilibrium path of cooperation and the situation in which there is an opportunistic deviation that leads the game to noncooperation. For party A:
C VA D VA cooperation 1 4(1 )

AC For party B:
C VB

AC

D VB cooperation 1 4

Notice that if = 1/2, then AC = BC = 1/2. This result is coherent with the one in the political economy literature, namely, Alesina, Spiller, Tommasi (S&T). Given that the game is assumed to be with perfect information, there are only two regions: cooperation and non-cooperation. The region where only one player cooperates becomes a non-cooperation region because cooperation is not a best response when the other player does not. However, in a game with imperfect information, cooperation might be seen as a signal to the other player. This situation is not analysed in this simple model. Figure (3) shows the cooperation region between the two players or political parties. is the recognition rule or probability of party A being elected. In a competitive political

INEQUALITY AND POVERTY

15

Figure 3. Diagram to define cooperation region. Source: Authors numerical simulation.

environment it may be reasonable to assume that bot parties have similar probabilities of winning. Hence, may be in a neighborhood of 1/2. For this recognition rule for party A the discount factor is exactly 1/2. Therefore, if the political parties are patient enough, that is, with discount factors greater than 1/2 the political cooperation will occur. This implies that the parties will compromise in this innite game, and therefore there will be gradual reform. The common interest of the two coalitions may be economic growth, while the dierence in policy could be one in which one party advocates for progressive tax reform and the other one for tax cuts. In the long run with two political coalitions with the same probability of being elected the status quo should be the outcome. However, in the transition dynamics in which one party gets elected for a few consecutive terms could be allowed to implement gradual progressive reforms aecting the distribution of inequality in one direction or the other. In the inequality problem initial conditions are essential, so if one party achieves small reforms may have a larger impact in the long run. The restriction to be in a competitive political system encourages cooperation for degrees of patience that do not need to be extremely high, and hence, alleviation of poverty and inequality may be more likely. On the other one, suppose that the recognition rule is extremely high, say = 0.8. That is, the probability of party A to be elected is 80 %. In this case, the level of patience required to cooperate is very high, almost a discount factor of 1. This would imply that party A would try to implement its most preferred policy every time that is elected without compromising with the other party. This excess of power may lead to corruption of the ruling elite and emphasize growth-oriented-policies without paying to much attention to social issues such as poverty or inequality. This model explains relatively well the phenomenon observe in the USA, Chile and China. In the USA and Chile, there are competitive political system in which the two main parties have to

16

QUARTER PAPER

NOVEMBER 2012

compromise in the long run, but with one common interest, not to harm economic growth. On the other hand, China has a unique party system in practice, and therefore, they may implement their most preferred policy. Data suggest that this policy was and is focused on economic growth since inequality has increased in the last decade. Chile and China have experienced similar economic booms, but while inequality grew in China, inequality remained unchanged in Chile. This outcome could be explained by the political framework.
IV. The Econometric Model

The econometric model is a simple regression of the type: (5) yit = x it +

it

where i = 1, . . . , n for individuals, and t = 1, . . . , T for time periods. For this research, i is equal to 5 (the ve quintiles), and t is as dened above. The pooled cross-sectional data is balanced. If xit is correlated with it , then there is a multi-equation system with common coecients and endogenous regressors. Thus, it is necessary to use an estimation procedure that takes into account the endogeneity problem. In the panel data context, under certain assumptions, there is a procedure to deal with endogenous regressors without using instrumental variables. This is the so-called xed eects (FE) estimator. Assume the following: = i : E[xit it ] = E[ i i ] =
it

i + it unobserved xed eect 0

The xed eect component i is an unobserved random variable that captures unobserved heterogeneity across individuals that is xed over time. To eliminate the endogeneity problem rst dierencing the regression model will eliminate the endogeneity without using IVE. The specic model is given by: (6) where: Yaut,i,t : autonomous real income for quintile i in period t, Ysubsidy,i,t : real income from subsidies for quintile i in period t, schoolingi,t : average years of schooling for quintile i in period t, Num.PPi,t : number of people per household for quintile i in period t, log(Yaut,i,t ) = 0 + 1 log(Ysubsidy,i,t ) + 2 schoolingi,t + 3 Num.PPi,t + i + i,t

INEQUALITY AND POVERTY

17

i : xed eect for quintile i, i,t : uncorrelated random error. Notice that the regression was run for data for urban areas only because of time constraints to calculate the inputs for the regression for rural areas. This extension is suggested for future work.
V. Results and Discussion

The regressions show in this section were estimated using EViews c . Table (5) shows that the greater the income the lower the government subsidies. Although this coecient is relatively small, the sign is intuitively correct and statistically signicant at the same time. This could be interpreted as an indication of well-focused scal policies to alleviate poverty problems. Likewise, the sign for the average years of schooling is right since the higher the educational attainment the higher the expected autonomous income for the people in that quintile, holding anything else constant. One of the questions that this

Table 5Panel Data Estimation for Households, Fixed Eects. log(autonomous income) (1) Constant log(income subsidy) average schooling Number of people Number of malnourish children R2 Notes: Authors computation. * Signicant at 10 percent. ** Signicant at 5 percent. *** Signicant at 1 percent. 0.944 15.014*** (1.687) -0.237*** (0.088) 0.226*** (0.066) -0.576** (0.247) log(autonomous income) (2) 15.170*** (1.695) -0.227*** (0.085) 0.224*** (0.066) -0.638** (0.265) 3.53E-06 (4.22E-06) 0.945

study attempts to answer is related to how important is the family size in its relation to poverty and inequality. Table (5) shows that the coecient for the number of people per household in absolute value is more important than schooling and is also statistically signicant at the 5% signicance level. Notice that the other two variables, income from subsidies and schooling, are signicant at the 1%. Therefore, there is sucient evidence to support a correlation between autonomous real income and family size. Despite this, it is hard to suggest the causality direction without doing further research. The literature suggests though that the direction could be from poverty to family size rather than from

18

QUARTER PAPER

NOVEMBER 2012

family size to poverty. See Schiller (1984, ch.7) for further discussion. As a robustness test the variable number of malnourished children per household was included, but this did not change the results in any signicant way. Concerning the issue of inequality, the
Table 6Cross-section Fixed Eects for regression (1). Quintil I II III IV V Notes: Authors computation. Fixed Eect -1.355 -0.539 -0.080 0.385 1.589

xed eects that are all statistically signicant at the 1% suggest that there would be some structural inequality in the case of Chile since the three poorest quintiles show evidence of a handicap while the two richest quintiles would have some advantage over the other groups. Even more, the poorest quintile exhibits the most negative xed eect, while the richest quintile the most positive one. This provides enough evidence to argue that there is indeed a trap of poverty, and at the same time there is a safety net for the rich either family connections or capital stocks.
15 14 13 12 .08 .04 .00 -.04 -.08 -.12
I - 90 I - 92 I - 94 I - 96 I - 98 I - 00 I - 03 I - 06 I - 09 II - 90 II - 92 II - 94 II - 96 II - 98 II - 00 II - 03 II - 06 II - 09 III - 90 III - 92 III - 94 III - 96 III - 98 III - 00 III - 03 III - 06 III - 09 IV - 90 IV - 92 IV - 94 IV - 96 IV - 98 IV - 00 IV - 03 IV - 06 IV - 09 V - 90 V - 92 V - 94 V - 96 V - 98 V - 00 V - 03 V - 06 V - 09

11

Residual

Actual

Fitted

Figure 4. Fitted and actual autonomous income for regression (1). Source: Authors computation.

Figure (4) shows the tted and actual data from the panel data regression, and table (7) shows the xed eects test.

INEQUALITY AND POVERTY

19

Table 7Redundant Fixed Eects Test for regression (1). Eects Test Cross-section F Cross-section 2 Notes: Authors computation. Statistic 251.510 150.254 d.f. (4,37) 4 p-value 0.000 0.000

VI.

Conclusions

The empirical results show that although the impact of the number of people per household on autonomous income is not large is absolute terms, it is statistically signicant with a negative coecient. Thus, eorts to increase the economic opportunity of the poor should promote smaller family size. This may be done through direct eorts to make birth control available to the poor such as public sexual education in primary schools, legalization of abortion and access to abortion in public healthcare centers. These reforms will do much to alleviate present poverty and retard its intergenerational transmission. From the institutional perspective the simple political model develops in this paper shows that political competition is more likely to address the issues related to poverty and inequality since the political parties need to compete for the votes of these social sectors. Whereas in the unique party system this need is not found and hence the ruling elite may focus only in economic growth since it is the only eect that may expand their own welfare. The econometric results provide strong evidence of some sort of structural inequality that deserves attention and need to be further explained. In addition, the average years of schooling and the number of people per household show to have a high explanatory power in determining real autonomous income. Finally, there is a strong link between economic growth and poverty, but this link between inequality and growth is unclear and its direction inconclusive. One of the major problems encountered to study inequality in this exercise was the steady behaviour of inequality. In other words, inequality was almost time-invariant, and this makes hard to understand the variables aecting it. Further work is needed in the theoretical and econometrical sides.
VII. Extensions

First, some possible improvements are related to the microeconomic data in the sense that it would be an advantage to divide the population in deciles rather than quintiles. As well as dividing the data by regions and zones, rather than just zones. Further improvements are related to the variables to be used. This analysis was limited to autonomous income, income from subsidies, average schooling per household, number of people per household, and the number of malnourished children per household. The statistical analysis could include more variables or indices composed by variables. This due to the richness of the dataset.

20

QUARTER PAPER

NOVEMBER 2012

The econometric exercise was done with xed eects estimation. This could be enrich with other econometric techniques such as quantile regression, and Oaxaca decomposition. Another extension could be related to robustness tests. Although, there was one robustness test showed in section V. This is clearly not sucient and further work is needed. Finally, the political economy model could be attempted to be more general rather than simply using a quadratic function for analytical simplicity. Nevertheless, it does provide insight about the problem at hand.
REFERENCES

Agostini, Claudio A., Philip H. Brown, and Andrei C. Roman. 2010. Poverty and Inequality Among Ethnic Groups in Chile. World Development, 38(7): 10361046. Alesina, Alberto, and Dani Rodrik. 1994. Distributive Politics and Economic Growth. The Quarterly Journal of Economics, 109(2): 465 490. Barro, Robert. 1991. Economic Growth in a Cross Section of Countries. The quarterly journal of economics., 106(2): 407 443. Barro, Robert. 2000. Inequality and Growth in a Panel of Countries. Journal of Economic Growth, 5(1): 532. Barro, Robert, and Jong-Wha Lee. 2001. International Data on Educational Attainment: Updates and Implications. Oxford Economic Papers, 53(3): 541563. Benhabib, Jess, and Mark M. Spiegel. 1994. The role of human capital in economic development: Evidence from aggregate cross-country data. Journal of monetary economics., 34(2): 143173. Bertola, Giuseppe. 1993. Factor Shares and Savings in Endogenous Growth. The American Economic Review, 83(5): 11841198. Contreras, Dante. 2003. Poverty and inequality in a rapid growth economy: Chile 1990-96. The Journal of Development Studies, 39(3): 181200. David, Anda Mariana;, and Mohamed Ali Marouani. 2012. Poverty Reduction and Growth Interactions: What Can Be Learned from the Syrian Experience? Development Policy Review, 30(6): 773 787. Deininger, Klaus, and Lyn Squire. 1997. Economic Growth and Income Inequality: Reexamining the Links. Finance & Development, 34(1): 3841. Fan, Shenggen. 2008. Public expenditures, growth, and poverty : lessons from developing countries. Baltimore:Johns Hopkins University Press. Fleisher, Belton M, Haizheng Li, Shi Li, and Xiaojun Wang. 2005. Sorting, Selection, and Transformation of Return to College Education in China.

INEQUALITY AND POVERTY

21

Forbes, Kristin. 2000. A Reassessment of the Relationship between Inequality and Growth. American Economic Review, 90(4): 869887. Goudie, Andrew, and Paul Ladd. 1999. Economic growth, poverty and inequality. Journal of International Development, 11(2): 177195. Krsti, Gorana, and Peter Sanfey. 2011. Earnings inequality and the informal econc omy. Economics of Transition, 19(1): 179199. Li, Hongyi, and Heng-fu Zou. 1998. Income Inequality is not Harmful for Growth: Theory and Evidence. Review of Development Economics, 2(3): 318334. Lockwood, Ben, and Alan Manning. 1988. Inequality And Ineciency In A Model Of Occupational Choice With Asymmetric Information. Journal of Public Economics, 37(2): 147169. Lucas, Robert. 1988. On The Mechanics Of Economic Development. Journal of Monetary Economics, 22(1): 342. Mayer, Susan. 2010. The relationship between income inequality and inequality in schooling. Theory and Research in Education, 8(1): 5 20. Persson, Torsten, and Guido Tabellini. 1994. Is Inequality Harmful for Growth? The American Economic Review, 84(3): 600621. Rehme, Gunther. 2007. Education, Economic Growth and Measured Income Inequality. Economica, 74(295): 493514. Schiller, Bradley. 1984. The economics of poverty and discrimination. . 4th ed., Englewood Clis NJ:Prentice-Hall. Schneider, Friedrich. 2002. Size and Measurement of the Informal Economy in 110 Countries around the World. Weller, Christian. 2007. The Benets of Progressive Taxation in Economic Development. Review of Radical Political Economics, 39(3): 368 376. Wodon, Quentin T. 2000. Microdeterminants of consumption, poverty, growth, and inequality in Bangladesh. Applied Economics, 32(10): 13371352.