Você está na página 1de 25

Dropout prevention measures in the Netherlands, an

evaluation
y
Kristof De Witte So…e J. Cabus
TIER, Maastricht University TIER, Maastricht University

Kapoenstraat 2, 6200 Maastricht Kapoenstraat 2, 6200 Maastricht

and

KU Leuven

Naamsestraat 69

3000 Leuven, Belgium

April 26, 2010

Abstract

In line with the Lisbon Agenda, set by the European Council in the year 2000,
European governments formulated ambitious plans to half the level of early school leavers
by 2012. This paper outlines the dropout prevention measures in the Netherlands and
analyzes their e¤ect at both the individual level and school level. Using a panel probit
model, we …nd little in‡uence of policy at the individual level. By means of quantile
regressions, we observe that schools with a relatively high dropout rate bene…t the most
from dropout prevention measures.
Keywords: E¤ectiveness, Dropout prevention, Secondary education, Logit, Quan-
tile regression
JEL-classi…cation: I21, C35
Corresponding author: kristof.dewitte@econ.kuleuven.be
y We would like to thank Wim Groot, Henriëtte Maassen van den Brink, Chris van Klaveren, Marton
Csillag, participants of the TIER seminar at the University of Groningen, members of the ’feedback committee’
at the Dutch Ministry of Education, Culture and Sciences and ’The Scienti…c Review Commission’at NICIS
Institute for insightful comments.

1
1 Introduction

"All Dutch municipalities should register (potential) dropout students and make sure that
by following a suited educational track, they obtain a starters quali…cation".
(OCW, 2010, p.1)

Over the last decades, societies have developed from an industrial towards a knowledge-
driven economy. “The economic welfare of individuals and the competitive advantage of
nations have come to depend on knowledge, skills and enterprise of the workforce" (Brown et
al., 2003). Investment in human capital plays a key-role in economic prosperity. The human
capital theory suggests that schooling raises productivity and earnings (Becker, 1992, 1993)
and can serve as one’s signal of productivity (Spence, 1973). Nelson and Phelps (1966) and
Schultz (1967) treat human capital of the workforce as a crucial factor for adoption of -
innovative - productive technologies.
Every year, many students drop out of school without obtaining a higher secondary edu-
cation diploma. This is not desirable in a knowledge-driven economy, not only for society’s
productiveness, but also for individual development. These so-called ’dropout students’ or
’early school leavers’constitute a group that is heavily at risk (Psacharopoulos, 2007). They
have a relatively higher risk of (1) entering a vicious circle in which on turn their children ob-
tain lower education levels (e.g., Bowles, 1972; McLanahan, 1985; Anger and Heineck, 2009),
(2) long-term unemployment or failing to secure productive employment (e.g., Rumberger
and Lamb, 2003; OECD, 2008), (3) su¤ering from health problems (e.g., Groot and Maassen
van den Brink, 2007) or (4) lack of social cohesion (e.g., Milligan et al., 2004; van der Steeg
and Webbink, 2006).
At the Lisbon 2000 summit, the European council decided to aim for a lower dropout rate,
among other benchmarks. The average rate of early school leavers should be no more than
10% by 2012. Following the European council, we de…ne an early school leaver (or dropout)
as a person younger than 23 who leaves education without a higher secondary education
degree. Thanks to political eagerness to tackle the problem, the European member states
developed various programs to reduce dropping out at secondary education. From Figure
1, we could deduce a declining trend in the dropout rates in EU countries. Since 1992, the
European dropout rate has fallen from about 35% to about 17%.1
Determining the most e¤ective way of tackling the dropout problem is not straightforward
as students do not dropout at secondary education because of one speci…c drawback. They
often are piling up problems, both at home, in their neighborhood or at school, before they
1 Dropout is also a major issue in other continents. For instance, consider the following citation out of the
inauguration speech of U.S. president Obama. "Every American will need to get more than a high school
diploma. And dropping out of high school is no longer an option " (Obama, 2009).

2
People in EU-12 aged 18-24 with only lower secondary education
40

35

30

not in education (in %)


25

20

15

10

0
92

93

94

95

96

97

98

99

00

01

02

03

04

05

06

07
19

19

19

19

19

19

19

19

20

20

20

20

20

20

20

20
Figure 1: Dropout rates of high schools in the EU

actually make the dropout decision (Rumberger, 2001). The literature indicates, for instance,
that dropout students change school more often (Rumberger and Larson, 1998), have more
retentions in grade (Roderick, 1994; Jimerson, 1999), struggle through their study curriculum
(Garnier et al., 1997), are more often involved in criminal activities (Elliot and Voss, 1974;
Phillips and Kelly, 1979), use more often cannabis, alcohol or other drugs (Fergusson et
al., 2003; ter Borgt et al., 2009), and are more likely to live in disadvantaged neighborhoods
(Bobonis and Finan, 2009) and in poorer families (Nelson et al., 1996). It is the accumulation
of small and large problems which pushes the pupil eventually towards the dropout decision.

This paper discusses the dropout prevention policy in the Netherlands and analyzes its
e¤ectiveness. Numerous measures and actions have been taken nationwide. Because of the
underlying population, di¤erent regions and cities have often di¤erent needs. Therefore, the
Ministry of Education created 39 regional dropout authorities (RMC) in 2002. Each of those
regions can take di¤erent actions towards policy goal settings. As this is not desirable for
working up to an integral approach (Holter and Bruinsma, 2009), the Ministry of Education
decided to outline a general framework, known as the ’covenants’(‘gentlemen agreements’).
A covenant is a signed written agreement between the Ministry on the one hand and the
RMC and the schools at the other hand. The covenants contain a list of measures (‘menu-
items’) and actions to …ght dropout, e.g., to improve the registration of non-attendance and
dropout, to improve ‡exibility of educational participation, to intensify the care for potential
dropouts and to increase attention for a good preparation for apprenticeships. In this way,
there are 39 covenants. The Ministry of Education signed covenants in all RMC-regions in
the Netherlands over the period 2007-2008. Nevertheless, in 2006-2007 14 regions with the

3
highest dropout levels have been signing a previous round of covenants.
The covenant includes a performance bonus to schools that reduce the number of dropouts
at secondary education over the period 2009-2012. The …nancial incentive consists of a …xed
fee of 2,000 euro per dropout less in comparison to base year 2005-2006.2 This should enhance
the proper implementation of the covenants. Van der Steeg et al. (2008) evaluated the 2006
covenant, however, and concluded that the 2006 covenant policy (as a bundle of activities)
was not e¤ective in reducing early school-leaving.

The contributions of this paper are threefold. Firstly, we examine the e¤ectiveness of the
‘gentlemen agreements’ by analyzing which of the incentives signi…cantly correlate with a
lower probability of student dropout at secondary education. We use an exceptionally rich
panel data set (BRON; Basis Register Onderwijsnummer) which covers all students in the
Netherlands from 2004 to 2008. Thanks to postcode information, the data are enriched with
neighborhood characteristics obtained from Statistics Netherlands. We start by an analysis
at the individual level. We estimate by a panel probit model the probability that a student
drops out. While controlling for student characteristics (e.g., gender, school track, etnicity),
neigborhood characteristics (e.g., income per capita, green areas, employment in the area),
a time trend (controlling for the increased awareness of obtaining a diploma) and region
…xed e¤ects, we correlate the in‡uence of dropout prevention measures to the individual
probability of dropping out.
Secondly, at the individual level, various unobserved exogenous variables can in‡uence
the dropout decision. Therefore, we aggregate all data at the school level. This provides an
indication of schools with many and few dropout rates. Using a quantile analysis, and con-
trolling for the student, neighborhood, time and regional in‡uence, we examine the in‡uence
of dropout prevention policy measures for schools with few (i.e. 25th quantile), average (i.e.,
50th quantile) and many (i.e., 75th quantile) dropouts. As selecting quantiles is rather arbi-
trarily, we estimate also the in‡uence of the dropout measures for a continuum of quantiles.
Thirdly, this paper is to our best knowledge the …rst to describe the dropout prevention
incentives in one of the EU member states. The Dutch Ministry of Education spends 313
million a year (anno 2008) on dropout prevention policy, which implies 0.83% of its total
budget (Dutch Agency for Statistics). It has been foreseen that this budget will increase to
400 million euro a year by 2011 (Ceulenaere et al., 2009 and Statistics Netherlands). Not a
negligible budget, which e¤ectiveness is worth analyzing.
Although this paper focusses on the Netherlands, its impact goes far beyond this speci…c
country. Firstly, dropout policy is high on the political agenda in about all industrialized
countries. Secondly, our analysis reveals some best practice policy, what might be interesting
2 Recently, the fee has been raised to 2,500 euro per dropout less to reduce the Financial risks coupled
with the schools’investments in …ghting dropout.

4
Preven tive
Curative Measures
Measure s
Mentoring & C oaching EVC or Dual Tra cks
Care & Advisory Teams Frequent Intakes
Smoothing Transition
Extended School
Dropout preve ntion
policy

Compulsory
Education
Measures
Registration & Commu nication
Reporting Truants ("verzuimloket")
Apprenticeships

National Measure s
Sta rter's Qualification

Figure 2: Policy measures in the Netherlands to reduce dropout (The broken border line
refers to the measures and actions included in the RMC-region speci…c covenants)

also for other countries.


The paper proceeds as follows. The Dutch dropout prevention policy is described in
Section 2. Section 3.1 brie‡y presents the data, its structure and some descriptive statistics.
We examine the in‡uence of the Dutch dropout prevention measures at the individual level
in Section 3.2 and at the school level in Section 3.3. A …nal section provides some policy
advice.

2 Dropout Prevention Measures in the Netherlands


This section describes brie‡y the dropout policy in the Netherlands. We summarize the policy
measures systematically in Figure 2. Although the Ministry of Education did not make a
distinction between the measures, we can distinguish four types of measures: compulsory
education measures, preventive measures, curative measures and national measures. The
former three arise from the so-called ‘covenant’. We discuss each of the policy measures next
and link them to the international literature.

5
Table 1: Summary: convenant dropout prevention measures
Measure Implementation
A. Compulsory Education measures
1 Reporting truants Reporting and tackling truancy at a very early stage
2 Changing subject A tailored track for students who choose a wrong subject or who
prefers another subject.
3 Guidance towards to the students’op- Work placement, writing a letter of application, apprenticeship pro-
timal track or profession grams, creating a portfolio
4 Apprenticeship Coordination with local private …rms and adhanced apprenticeship
programs for students who prefer to do manual jobs.
B. Preventive measures
5 Mentoring and coaching Students are matched with a coach from public or private organisa-
tions
6 Care and advisory team Coordination of student care by social workers, youth assistance,
school attendance o¢ cers, health services and police.
7 Smoothing the transition from the pre- Intake talks at the vocational school, providing more information on
vocational level to the vocational level the educational tracks, and checking wheter the students e¤ectively
enroll at and start in the new vocational school
8 Extended school Add more sports and culture to schools in order to make school more
attractive.
C. Curative measure
9 Dual track O¤ering the posibility for dropout students to re-enter education by
a tailered educational track.
10 Frequent intakes Increasing the number of moments that students may enter secondary
education.

2.1 Compulsory Education Measures


Compulsory education measures are supported by Dutch law. In this way, every RMC-region
is forced to undertake actions in order to obey the law. We brie‡y discuss three compulsory
education measures: registration and communication, reporting truants, and apprenticeships.

2.1.1 Registration and Communication

A good measurement instrument is indispensable when it comes to the evaluation of dropout


policies. In the past, registration of early school leavers was inaccurate and unreliable. There
was de…nitely a lack of transparency (Expertisecentrum, 2006). Therefore, the program
‘Aanval op de uitval’(or …ghting dropout) has been launched in the year 2006, which implied
an upgrade of the registration system. Nowadays, we can use nearly complete and reliable
data on dropout levels in the Netherlands.
The registration of dropouts takes place as follows. Every pupil who attends school in
the Dutch educational system gets a personal identi…cation number (or education number
record). All schools register students using this personal identi…cation number. In the end, all
registrations end up in one large nationwide database called ‘het Basisregister Onderwijs’or
BRON. Since 2007, BRON can be used to evaluate the regional and national policy measures
for reducing dropout in secondary education.
An incentive based on ’naming and shaming’of schools with superior and inferior dropout
rates can boost competition among schools. Information on early school leavers has been

6
published on the World Wide Web, e.g., VSV-Atlas, VSV-verkenner and the website voorti-
jdigschoolverlaten.nl.
Another important role of communication is to inform youngsters about the relevance
of obtaining a starter’s quali…cation. In the Netherlands, a VSV-jongerenteam3 has been
launched in the summer of 2009 (cf. ‘stayinthegame.nu’). Youngsters who have (a bad)
experience with dropping out at school give information on the relevance of obtaining a
starter’s quali…cation (which is a minimal higher secondary diploma) to other youngsters.

2.1.2 Reporting Truants

A second policy measure aims at reporting non-attendance by registering truants in a central


database, the so-called ’digital o¢ ce’. It o¤ers the opportunity to signal better potential
dropouts (Auditdienst OCW, 2007). There is only one digital o¢ ce in the Netherlands
(Expertise Centrum, 2006).
An important feature can be attributed to the digital o¢ ce of non-attendance: to dis-
courage risk-averse pupils. It is possible that pupils do not want to run the risk "to be
caught" at not attending classes. For example, they don’t want their parents to know about
skipping classes. Therefore, augmenting the chance to be caught can discourage those pupils
to undertake outside school activities during school time. Further research on the e¤ect of
the digital o¢ ce of non-attendance on pupils’ behavior is desirable: does this digital o¢ ce
increase the probability to be caught?
Some previous work on truancy has been done in the literature. Attwood and Croll (2006)
have used the British Household Panel Survey and in-depth interviews to ask persistent tru-
ants about the extent, consequences and explanations for truancy from secondary schools.
Poor relationships with teachers, bullying and a more general dislike of the school’s at-
mosphere are considered as triggers for the dropout decision. In contrast to existing liter-
ature (e.g., Beekhoven and Dekkers, 2005), socio-economic factors, such as status, parental
involvement and the value of education, do not play a key role in non-attendance rates of
pupils. With this, Attwood and Croll suggest a distinction between socio-economic and atti-
tudinal factors. Davis and Lee (2004) also adhere to above …ndings. They …rst collected all
existing material on attending or not attending school in large cities of England, then went
into discussion with truants, as well as attendees and some parents. Davis and Lee argue
that, in contrast to professionals and existing literature, the curriculum is not considered as
a dropout trigger. This fact has been a¢ rmed by Attwood and Croll (2006) and weakened
by Beekhoven and Dekkers (2005) who, in contrast, put emphasis on learning problems, lack
of motivation and problems arising from choosing the wrong vocational track.
3 VSV-jongerenteam or ESL-team composed out of youngsters.

7
2.1.3 Apprenticeships

Apprenticeships are shown to be interesting learning methods to develop transferable, inter-


personal skills (cf. Lucas and Lammont, 1998) and to increase employability, especially in
vocational tracks. A lack of workplaces for apprentices is an important trigger of the dropout
decision. Finding better matches between apprenticeship and labor organization and improv-
ing information and support for pupils can make the problem less persistent (Onstenk, 2004;
Onstenk and Blokhuis, 2007).
More evidence from abroad can be found in the book of Bosch and Charest (2009). It
deals with various aspects of national vocational training systems and o¤ers an in-depth com-
parative analysis of the following countries: Australia, Canada, Denmark, France, Germany,
Korea, Mexico, Morocco, the United Kingdom and the United States of America.

2.2 Preventive measures


Preventive measures try to keep youngsters in school, i.e., before the dropout decision has
been taken. Four preventive measures will dealt with: mentoring and coaching, care and
advisory teams, smoothing transition, and extended school.

2.2.1 Mentoring and Coaching

One of the main reasons of dropping out at secondary education is making the wrong study
choice. About 20% of all dropouts in the Netherlands indicate to leave school because of a
bad study choice (ROA, 2009/1). A professional approach of managing the study curriculum
can enhance school attendance. The Dutch government provides subsidies to stimulate the
realization of common educational programs over the period 2008-2011.
‘Back 2 Your Future’is an example of improving study choice. It is a course that aims at
discovering your (study or work) capacities through, for instance, workshops or e-learning.
In international literature, curriculum di¤erentiation, or streaming and/or ability group-
ing is the most persistent issue in managing the study curriculum (for some examples, see
Keitel, 1987; Oakes et al.,1992; Gamoran et al., 1995; Terwel, 2004, 2005).

2.2.2 Care and Advisory Teams

In the Netherlands, care and advisory teams have been set up to connect internal and external
care for potential dropouts. It is the ambition to have full coverage in all schools by August
2011 (OCW). Although di¤erent settings are possible, a care and advisory team typically
consists of a pshycologist, pedagogues, social workers of the school, a representative of the
region and a policy o¢ cer.

8
2.2.3 Smoothing Transition

A long summer break of four months (‘het zomerlek’or summer leakage) is often referred to
as the main reason of di¢ cult school enrolment in the …rst year track of vocational educa-
tion. Besides losing connection with their teachers and schools, in the Netherlands, students
have to physically go to another school in the transition from pre-vocational education to
vocational education. During this transition, students often lost track, which suggests a tran-
sition problem between pre-vocational and vocational education (van der Steeg and Webbink,
2006). This covenant measure tries to smooth the transition from pre-vocational to vocational
education by staying in touch with the students.
Evidence has been found in the literature by Felner et al. (1981, 1982) who conducted
a randomized experiment in the US, called the Transition Project. The experiment had
two goal settings: restructuring of the role of teachers and reorganizing the school’s envi-
ronment. Students with improved transition reported signi…cantly higher levels of teacher
support, teacher a¢ liation and involvement than students without the additional transition.
As a result, students belonging to the treatment group had better scores on the assessments
instruments. In sum, the experiment indicates that primary preventive community based
programs may help pupils during school transitions and may actually reduce dropout rates.

2.2.4 Extended School

Extended school(-time) refers to a range of services and activities, often beyond the school
day, to meet the needs of pupil and their families. For instance, schools o¤er sport and leisure
activities to augment their attractiveness for pupils.
‘Time-In’is a such a project in the Netherlands (de Zwart et al., 2009). The initial goal
of the project is to teach skills to pupil by means of education or work, such that pupils do
not leave school without a diploma or job. It also o¤ers the chance to motivate youngsters to
do sport activities, to combat the problem of overweight and to develop talented youngsters
to professional sport careers.

2.3 Curative Measures


Curative measures aim at taking dropouts back in school, i.e. after the dropout decision
has been taken. We discuss two educational measures in this …eld: EVC or dual tracks and
frequent intakes.

2.3.1 EVC or Dual Tracks

Curative interventions deal with early school leavers already working in the labor market
(ROA, 2009/4). The Dutch government wants to support these youngsters in attaining
their starter’s quali…cation by means of EVC or dual tracks. EVC stands for a learning

9
certi…cate which can be obtained if a student passes a learning module. Dual tracks refer
to the combination of labor and study. For instance, part-time learning on construction
techniques and part-time working in construction.
Borghans et al. (2000) o¤er more insight into the demand and supply of Dutch vocational
students on the labor market. Both measures aim at (unemployed) dropouts in times of eco-
nomic crisis. Unemployment rates have been increasing at the end of 2008: for pre-vocational
dropouts from 6% in 2007 to 9% in 2008 and for dropouts in the …rst year of vocational edu-
cation from 10% in 2007 to 16% in 2008 (ROA, 2009/4). Temporary arrangements can o¤er
those youngsters a perspective on a (long term) job.

2.3.2 Frequent Intakes

With frequent intakes, there is more than one moment during the academic year to enter
secondary education. This actually means that pupils can enter the academic year after
October 1th (which is one month after the o¢ cial start of the school year). The ‘Centraal
Toegangsloket voor het Onderwijs’ (ROC) or Central Entry O¢ ce for Education organizes
frequent intakes for pupils following vocational tracks. Moreover, early school leavers can
also enter a ’reception class’, which is a special class for previous dropout students. After
a possible revision of the study choice, youngsters can continue another track as soon as
possible.

2.4 Non-covenant measures


The dropout prevention measures of the previous subsections arise from the ’covenant’(i.e.,
the agreement between the Ministry of Education on the one hand, and the regions and
schools on the other hand). However, the policy measures go beyond the covenant. The
Ministry of Education created a so-called starter’s quali…cation, which is a minimal degree
before one can leave school. In other words, Dutch pupils are compelled to go to school
until the age of 18 or until a starter’s quali…cation is obtained. A starter’s quali…cation is
considered as the minimum level of education a student needs to be equipped for labor market
entrance (Eimers et al., 2009). In the literature, a starter’s quali…cation corresponds to a
higher secondary diploma. Students who only obtained a lower secondary diploma do hence
not ful…ll the requirements for obtaining a starter’s quali…cation. There are three types of
quali…cations one can obtain before entering the labor market (Dutch abbreviation between
brackets): pre-university education (vwo), senior general secondary education (havo) and
vocational secondary education (mbo). This paper examines the in‡uence of the starter’s
quali…cation, but only summarizes below some evidence from the literature.

Indications on its e¤ectiveness A starter’s quali…cation implies mandatory school at-


tendance. The e¤ect of compulsory schooling on dropout rates in school has only little

10
attention in evidence-based literature (for some examples, see Angrist and Krueger, 1991;
Oosterbeek and Webbink, 2004; Pischke and von Wachter, 2005; Oreopoulos, 2006, 2007).
van der Steeg and Webbink (2006) stress the importance of the additional years of education
gained through compulsory school attendance – even if the pupil drops out of school too
early. They promote the use of a sliding scale: more education leads to better labor market
outcomes irrespectively of actually obtaining the school-leaving certi…cate. They also o¤er
di¤erent labor market perspectives to di¤erent types of vocational education and orientation.
The level of starter’s quali…cation is hence no critical boundary for labor market success.
Nonetheless, to leave secondary school without a diploma means to have fewer chances
on the labor market to be successful (Spence, 1973; Becker, 1992; Card, 1999; Rumberger
and Lamb, 2003; among others). Providing pupils education and skills needed to enhance
success on today’s labor market, has long been a key policy goal in the Netherlands (OCW).
In further research, we want to …nd out the schooling and labor market consequences of a
starter’s quali…cation on dropping out at secondary education in the Netherlands.

3 Results: E¤ects of dropout policy


In the previous section, we have described educational measures within the scope of dropout
prevention policy in the Netherlands. We have distinguished two broad categories: covenant
and non-covenant measures. The research results of this paper only deal with analyses on
the e¤ectiveness of covenant measures. We perform these analyses at the individual level (cf.
section 3.2) and the school level (cf. section 3.3), but start with a description of the data in
the next section 3.1.

3.1 The data


This paper uses the BRON-data, an unique registered data set of all Dutch students. The
database contains pupil speci…c information about his/her personal characteristics (e.g., gen-
der, ethnicity and receives special care at school), schooling (e.g., school type, school track
and major subjects) and about the parents (e.g., single parent household). Because of match-
ing variables (as postcode or identi…cation numbers) government institutions can further link
this BRON data to neighborhood characteristics, tax information on the parents, work status,
and basically any other o¢ cial Dutch data set.
Thanks to the Ministry of Education, this research could bene…t from the full BRON
data set, which comprises all Dutch students enrolled between 2004 and 2008. Using post-
code information, we linked this data set to neighborhood characteristics as obtained from
Statistics Netherlands. By carefully analyzing the agreements between the government on
the one hand and the regional dropout authorities and schools on the other hand, we con-
structed dummy variables which capture the particular implementation of the agreements

11
at regional level. As schools and municipalities are collaborating extensively within each of
the 39 regions, and due to lack of information at the local level, we assume that all schools
within a region are implementing the agreements in a similar way. Some summary statistics
on the data are presented in Table 2.
To examine the e¤ectiveness of the dropout prevention, we analyze the data at two levels:
(1) at individual level and (2) at school level.
We follow the Ministry of Education in de…ning dropout students as follows. In the
Netherlands, dropout students are determined by comparing the students younger than 22
on the …rst of October of a given year, with the students at the …rst of October of the next
year. Students who did not obtain a diploma (i.e., a starter’s quali…cation) and left the
database are considered as dropouts.4 As such, students between the age of 12 and 23 who
do not have a starter’s quali…cation at the time that they drop out at secondary education,
are de…ned as early school leavers.
We estimate by means of a panel probit model the probability that a student will dropout
at secondary education. With this, we estimate two speci…cations: a random e¤ect panel
probit model and population averaged panel probit model. Controlling for (1) student, (2)
neighborhood, and (3) regional characteristics, we relate the dropout probability to the pre-
vention mechanisms for dropout reduction. We remark, however, that due to data limitations
(e.g., it is to our best knowlegde impossible to …nd proper instrumental variables), we do
not attempt to estimate the causal impact of the dropout prevention measures. Hence, we
rather look for strong correlations within the sample. As we rigorously control for various
background characteristics of the students, the neighborhood and the schools, and as we
allow for a time trend in the data (i.e. control for potential time e¤ects), our results give a
clear indication on the impact of menu-items within the covenants.

3.2 Analysis at individual level


We start analyzing the e¤ectiveness of the Dutch dropout prevention policy by considering
the impact on the individual student. In other words, while controlling for various in‡uences,
we correlate by a panel probit model the probability of dropping out with the introduction
of particular prevention measures. The results are presented in Table 3.
Following the literature, gender, ethnicity and family background are indicated as triggers
of the individual dropout decision (Rumberger, 1983; Astone and Mclanahan, 1991, 1994;
Mayer, 1991; Steinberg et al., 1992; Berktold et al., 1998; Pong and Ju, 2000; among others).
We observe that neighborhood characteristics play an important role. Students living in
poor and high density areas have a higher probability of dropping out, as well as students
4 We consider all students living in the Netherlands, who are taking courses in secondary education (includes
vwo, havo, vmbo, mbo) and vavo. We do not consider students living abroad, in an international Baccalaureat,
or in an apprenticeship program (praktijkonderwijs).

12
Table 2: Descriptive statistics - total number, unless otherwise stated in second column
2004 2005 2006 2007
Student characteristics
dropout 58,600 52,700 50,900 46,800
School type General education 911,421 922,062 928,563 945,605
Vocational 373,770 417,257 427,954 431,998
Adult education 6,462 12,288 10,869 9,566
Higher education 5,820 78,538 158,392 236,206
schtyp pro 9,628 10,133 10,195 27,080
First class 245,707 153,946 150,761 149,098
Student care 92,910 98,616 99,966 101,812
pre-vocational training 234,773 302,784 294,166 280,997
higher general sec- 152,089 160,903 166,213 170,406
ondary
pre-university training 176,314 195,680 204,025 211,676
Care for student vmbo, geen lwoo 234,773 302,784 294,897 281,912
care 92,910 98,616 99,966 101,812
City Amsterdam 47,816 50,916 56,882 62,653
Rotterdam 47,488 50,355 54,235 57,217
The Hague 31,408 35,596 38,621 41,248
Utrecht 15,879 17,451 23,051 27,912
Average sized city 313,125 340,236 378,668 414,049
Gender Female 643,450 701,762 752,802 802,909
Etnicity Autochton 1,025,540 1,111,011 1,188,187 1,265,127
Suriname 39,802 42,440 44,506 46,363
Aruba 16,806 17,753 18,721 19,870
Turkey 44,914 49,237 52,796 56,355
Moroco 42,425 45,806 48,314 50,817
non-western migrant 64,802 70,050 74,853 79,071
Western migrant 84,958 90,768 96,568 102,183
Unknown 3,676 4,376 2,702 3,589
Generation of migrant Autochton 1,025,540 1,111,011 1,188,187 1,265,127
First generation 90,849 94,085 95,104 94,383
Second generation 202,858 221,969 240,654 260,276
Unknown 3,676 4,376 2,702 3,589
Living in poor area 212,858 227,814 252,192 273,349
Characteristics munici-
pality
Number of inhabitants mean (std) 3387 (3776) 3380 (3763) 3393 (3771) 3390 (3758)
Population density mean (std) 3934 (3951) 3939 (3953) 4022 (4012) 4069 (4043)
% one person house- mean (std) 28.888 (13.892) 28.762 (13.806) 29.371 (14.415) 29.801 (14.815)
hold
% Allochton mean (std) 9.194 (13.256) 9.140 (13.183) 9.243 (13.155) 9.298 (13.139)
Average income mean (std) 16.395 (2.566) 16.412 (2.569) 16.399 (2.579) 16.392 (2.586)
Green areas (km2) mean (std) 40.649 (5.365) 40.663 (5.350) 40.547 (5.382) 40.462 (5.390)
Number of households mean (std) 95.996 (24.035) 95.882 (24.012) 96.542 (24.248) 96.956 (24.351)
moving
Average house value mean (std) 134.998 (32.472) 135.255 (32.483) 135.039 (32.387) 134.911 (32.292)
Employment in the mean (std) 64.107 (109.889) 63.030 (108.497) 64.996 (110.175) 66.289 (111.327)
area
Dropout prevention
Initial implementor 0 0 774,770 774,770
Number of prevention mean (std) 0 0 2.354 (2.458) 4.886 (1.443)
items
Care and advisory 0 0 667,198 1,406,188
team
Smoothing the transi- 0 0 646,830 1,305,316
tion
Mentoring and coach- 0 0 397,911 752,125
ing
Changing subject 0 0 128,053 276,347
optimal track or profes- 0 0 346,551 766,340
sion
Apprenticeship 0 0 127,940 246,567
Frequent intakes 13 0 0 452,063 835,898
Extended school 0 0 0 202,601
Reporting truants 0 0 657,507 1,228,348
Dual track 0 0 0 149,715
Number of schools 594 606 596 690
living in neighborhoods with more one person households and migrant families. More op-
portunities to …nd work in the area, in addition, also increase the student’s probability to
dropout (this fact has also been found in, McNeal, 1997).
To control better for unobserved heterogeneity, we have estimated …xed e¤ects at region
and school type level. In addition, we have also estimated a time trend. Firstly, school type
…xed e¤ects relates to the educational track a pupil follows in the Dutch educational system
(e.g., pre-university, pre-vocational or vocational education). The estimates (available upon
request) indicate that students following tracks with lower ability levels (i.e. ‘lwoo’or care
students in vocational track) have a relatively higher probability to drop out at secondary
education. These results are in line with van der Steeg and Webbink (2006), who argue
that early school leavers are concentrated in the lowest level of pre-vocational and vocational
education.
Secondly, we also consider region …xed e¤ects. As expected (referring to above analyses on
neighborhood characteristics), we deal with signi…cant region …xed e¤ects. Further research
is needed to disentangle these e¤ects.
Thirdly, the time trend takes sensitization into account. As pupils get more informed on
the relevance of obtaining a higher secondary diploma, the dropout rate can fall over time
without any in‡uence of dropout prevention measures. The results, presented in Table 3,
con…rm that this kind of sensitization takes place in the Netherlands. This was also clear
from Figure 1, which indicated a decrease in dropout since 1992. We remark that, to our
best knowledge, no further research can be found on the impact of time e¤ects on dropout
rates.
Focussing on the correlation coe¢ cients of the ten menu-items used in the dropout pre-
vention policy delivers interesting insights. Out of 10 prevention measures, only 3 turn out to
have a signi…cant impact on the individual’s dropout decision: (1) mentoring and coaching,
(2) optimal track or profession, and (3) dual tracks. Not unexpectedly, those 3 measures
correspond to preventions which regions cannot implement overnight. They are innovative,
in a way that it is not possible for the school to re-label existing procedures, and require a
clear follow-up of the student. Moreover, the number of items that regions are implementing
does not have a signi…cant impact.
Finally, we observe that the individual dropout decision did not alter in regions that
implemented dropout prevention programs one year before other regions. These ’early im-
plementor regions’were the 14 regions with the highest dropout rate in 2005-2006 (as such,
there was not a random selection). Our results are in line with previous results of van der
Steeg et al. (2008), who analysed the e¤ectiveness of the covenant based on a di¤erence in
di¤erence approach in those two regions.

14
Table 3: Menu-items at individual level
coe¤. St. error t-statistic p-value
Student characteristics
Gender -0.1426 0.0064 -22.4000 0.0000 ***
Etnicity (native = reference)
Suriname 0.7051 0.0276 25.5800 0.0000 ***
Dutch Antilles 0.7165 0.0284 25.2300 0.0000 ***
Turkey 0.6408 0.0270 23.7400 0.0000 ***
Marocco 0.6930 0.0268 25.8300 0.0000 ***
Non-western migrant 0.6874 0.0212 32.3600 0.0000 ***
Western migrant 0.6686 0.0238 28.1200 0.0000 ***
Unknown 3.2450 0.1173 27.6600 0.0000 ***
Generation of migrant -0.2629 0.0122 -21.5500 0.0000 ***
Postcode characteristics
Poor area 0.0804 0.0106 7.5600 0.0000 ***
Number of inhabitants 0.0000 0.0000 -0.8200 0.4120
Population density 0.0000 0.0000 -2.2100 0.0270 **
One person household 0.0007 0.0003 2.3600 0.0180 **
Number of migrants 0.0008 0.0004 2.1800 0.0290 **
Income per capita -0.0047 0.0016 -2.8500 0.0040 ***
Green areas -0.0007 0.0011 -0.7100 0.4770
Frequency of moving 0.0014 0.0003 4.9800 0.0000 ***
Average housing cost -0.0005 0.0002 -2.8100 0.0050 ***
Employment in the area -0.0003 0.0001 -5.3000 0.0000 ***
Dropout prevention measures
Early implementor 0.0095 0.0117 0.8100 0.4150
Number of implemented prevention items 0.0203 0.0172 1.1800 0.2380
Care and advisory team -0.0083 0.0280 -0.3000 0.7670
Smoothing the transition -0.0130 0.0382 -0.3400 0.7320
Mentoring and coaching -0.0403 0.0244 -1.6500 0.0990 *
Changing subject -0.0275 0.0299 -0.9200 0.3580
Optimal track or profession -0.0434 0.0226 -1.9200 0.0550 *
Apprenticeship -0.0264 0.0347 -0.7600 0.4470
Frequent intakes -0.0243 0.0207 -1.1700 0.2410
Extended school 0.0315 0.0345 0.9100 0.3620
Reporting truants -0.0221 0.0246 -0.9000 0.3700
Dual tracks -0.0626 0.0329 -1.9000 0.0570 *
Constant 147.9593 10.8351 13.6600 0.0000 ***
Region …xed e¤ects Yes
Time …xed e¤ect Yes
School type …xed e¤ects Yes

15
3.3 E¤ectiveness of dropout incentives at school level
Various unobserved exogenous variables may in‡uence the dropout decision at the individual
level. Therefore, we examine the e¤ectiveness of the dropout prevention measures by aggre-
gating all data at the school level. As the BRON database is a registered dataset of all Dutch
students (and not a sample), the model is not vulnerable to any selection e¤ect.5 We estimate
the e¤ectiveness of the dropout prevention measures by means of a nonparametric quantile
regression (Koenker and Bassett, 1978). Quantile regressions are convenient to estimate the
impact on other levels than the mean (i.e., other quantiles). In this way, we can estimate the
whole conditional distribution of the dependent variable y.
In the school level analysis, we consider three kinds of schools: (1) schools with a low
dropout rate, (2) schools with a median dropout rate, and (3) schools with a high dropout
rate. They are decided on the …rst (25%), second (50%) and third quantile (75%), respec-
tively. Besides a time trend, we control for school type and region …xed e¤ects. Table 4
report the results of the quantile analysis.
We see that schools with relatively high dropout rates bene…t most from dropout preven-
tion measures. All dropout prevention measures, except for advisory team and dual track
projects, are associated with lower dropout rates. In contrast, we do not …nd a signi…cant
impact of the dropout prevention measures on schools with low or median dropout rates.
Obviously, schools cannot simply be divided into three groups. The distribution of
dropout rates is more a continuum (Rumberger and Thomas, 2000). Therefore, we esti-
mate the impact of each dropout prevention measure on the dropout level of schools for all
quantiles. We plot the corresponding graphs in Figure 3. We see a negative slope in about
all graphs (but not all of them are signi…cant). This indicates that the higher the dropout
level of the school, the larger the impact of the dropout prevention measure.

5 We observe all Dutch students and all Dutch schools.

16
Table 4: E¤ect of dropout prevention in …rst, second and third quartile of school
25% quartile 50% quartile 75% quartile
coe¤. St. error t-statistic coe¤. St. error t-statistic coe¤. St. error t-statistic
% of females at school -0.0037 0.0024 -1.5400 -0.0073 0.0021 -3.4800 *** 0.0004 0.0013 0.3000
% of care students at school 0.0646 0.0016 40.5100 *** 0.0755 0.0015 51.1200 *** 0.0825 0.0009 95.7100 ***
% of natives at school -0.0044 0.0019 -2.2900 ** -0.0138 0.0017 -8.0500 *** -0.0249 0.0010 -25.3100 ***
% of single parents at school -0.0721 0.0084 -8.6300 *** -0.0498 0.0077 -6.4400 *** -0.0063 0.0039 -1.6400 *
school in poor area 0.0077 0.0017 4.4800 *** 0.0034 0.0017 2.0000 ** -0.0035 0.0010 -3.5300 ***
Number of inhabitants 0.0000 0.0000 1.0500 0.0000 0.0000 -1.4200 0.0000 0.0000 -5.2300 ***
Population density 0.0000 0.0000 -1.2400 0.0000 0.0000 -0.7700 0.0000 0.0000 1.7500 *
% one person household 0.0000 0.0000 0.8600 0.0000 0.0000 2.2500 ** 0.0000 0.0000 3.5300 ***
% Migrant 0.0001 0.0000 4.3100 *** 0.0001 0.0000 3.4100 *** 0.0001 0.0000 4.2000 ***
Average income 0.0000 0.0001 0.1100 0.0003 0.0001 3.0400 *** 0.0003 0.0000 5.9900 ***
Green areas (km2) -0.0001 0.0000 -1.9000 * 0.0000 0.0001 0.4400 0.0000 0.0000 -0.3600
Number of households moving 0.0000 0.0000 -1.8900 * 0.0000 0.0000 -0.1900 0.0000 0.0000 2.0100 **
Average house value 0.0000 0.0000 0.1300 0.0000 0.0000 2.6000 *** 0.0001 0.0000 15.5300 ***

17
Employment in the area 0.0000 0.0000 2.5600 ** 0.0000 0.0000 1.1000 0.0000 0.0000 1.2500
Initial implementor 0.0005 0.0024 0.2100 0.0018 0.0025 0.7200 0.0095 0.0014 6.6600 ***
Number menu-items 0.0001 0.0009 0.1300 0.0007 0.0009 0.7900 0.0022 0.0005 4.4100 ***
Care and advisory team 0.0015 0.0016 0.9500 0.0020 0.0017 1.1600 -0.0005 0.0010 -0.4900
Smoothing the transition -0.0029 0.0022 -1.2900 -0.0041 0.0023 -1.8400 * -0.0110 0.0013 -8.7500 ***
Mentoring and coaching 0.0002 0.0012 0.1500 -0.0006 0.0012 -0.4700 -0.0020 0.0007 -2.9900 ***
Changing subject 0.0002 0.0015 0.1300 -0.0002 0.0016 -0.1400 -0.0023 0.0009 -2.5800 ***
Optimal track or profession -0.0004 0.0012 -0.3000 -0.0005 0.0012 -0.4500 -0.0018 0.0007 -2.7900 ***
Apprenticeship -0.0025 0.0019 -1.2700 -0.0053 0.0020 -2.6200 *** -0.0112 0.0011 -10.3000 ***
Frequent intakes -0.0003 0.0011 -0.2300 -0.0008 0.0012 -0.6600 -0.0016 0.0006 -2.5500 ***
Extended school 0.0027 0.0017 1.5900 0.0011 0.0018 0.6100 -0.0052 0.0010 -5.2800 ***
Reporting truants 0.0008 0.0013 0.6400 -0.0020 0.0014 -1.4800 -0.0053 0.0007 -7.1700 ***
Dual tracks 0.0012 0.0015 0.8300 0.0002 0.0015 0.1100 0.0004 0.0008 0.4400
Constant 0.0085 0.0042 2.0200 ** 0.0121 0.0043 2.7900 *** 0.0119 0.0024 5.0400 ***
Region …xed e¤ects Yes Yes Yes
Time …xed e¤ect Yes Yes Yes
School type …xed e¤ects Yes Yes Yes
4 Conclusion
In line with the targets of the Lisbon Agenda, the Dutch goverment created a broad policy
agenda to reduce dropout at secondary education. This paper analyzed the impact of the
policy measures at both the individual level (i.e., do the menu-items of the covenant change
the dropout decision of the student) and at the school level (i.e., do the menu-items change
the number of students dropping out at schools?). Firstly consider the individual impact.
While most of the menu-items correlate negatively with the individual dropout decission,
only ’mentoring and coaching’, ’optimal track or profession’and ’dual track’have a sign…cant
negative impact on the individual dropout decision. Secondly, we …nd that the number of
menu-items implemented by the RMC does not have a signi…cant impact. Thirdly, regions
that implemented the covenant one year before the other regions are not obtaining lower
dropout rates. We remark, however, that the early implementors were regions with the highest
dropout levels in the Netherlands.

As an analysis at the individual level might be intricate due to unobserved heterogeneity


in the estimates, we have analysed the e¤ect of the covenants at an aggregated school level.
By means of quantile regressions, we have estimated the correlation between the menu-items
and the percentage of dropouts in school. It has been observed that for di¤erent quantiles
of schools (e.g., the schools with the 25% lowest or 25% highest percentage of students) also
di¤erent impacts of menu-items arise. While only few menu-items have a signi…cant e¤ect in
schools with a relatively low percentage of dropouts, schools with a relatively high percentage
of dropouts indicate to bene…t from all but two dropout prevention measures. These two
educational measures are advisory teams and dual tracks. We observe that schools with a
relatively higher dropout level bene…t the most from dropout prevention measures as outlined
in the covenant.
Given the importance of the theme, and the availability of larger and more detailed data
sets, early school leaving will de…nitely attract further research. Potential research avenues
could arise from the policy on truancy, the additional years of compulsory education, or from
long term policy evaluations.

References
[1] Anger, S. and Heineck, G. (2009). Do smart parents raise smart Children? The inter-
generational transmission of cognitive abilities. Working paper series, SOEPpaper, No.
156.

[2] Angrist, J.D.and Krueger, A.B. (1991). Does compulsory school attendance a¤ ect school-
ing and earnings? The Quarterly Journal of Economics, vol. 106, No.4, pp. 979-1014.

18
Figure 3: Quantile estimates of dropout prevention measures

19
[3] Astone, N.M. and McLanahan, S.S. (1991). Family structure, parental practices and high
school completion. American Sociological Review, 56,309-320.

[4] Astone, N.M. and McLanahan, S.S. (1994). Family Structure, Residential Mobility, and
School Dropout: A research note, Vol.31, No.4, pp.575-584.

[5] Attwood, G.and Croll, P. (2006).Truancy in secondary school pupils: prevalence, trajec-
tories and pupil perspectives. Research papers in Education, Vol.21, No.4, pp.467-484.

[6] Auditdienst Ministerie OCW. Rapport van bevindingen over Thema-Onderzoek VSV
2007, auditdienst OCW, Inspectie van het Onderwijs, Den Haag.

[7] Becker, G.S. (1992), Human Capital and the Economy. Proceedings of the American
Philosophical Society Vol.136, No.1, pp.85-92.

[8] Becker, G.S. (1993). Human Capital. A theoretical and empirical analysis with special
reference to education Third Edition. National Bureau of Economic Research, The Uni-
versity of Chicago Press, Ltd. London.

[9] Beekhoven, S. and Dekkers, H. (2005). Early school leaving in the lower vocational track:
triangulation of qualitative and quantitative data. Adolescence, Vol.40, No.157, pp.197-
213.

[10] Berktold, J., Geis, S. and Kaufman, P. (1998). Subsequent educational attainment of
high school dropouts. Washington, DC : U.S. Dept. of Education, O¢ ce of Educational
Research and Improvement, National Center for Education Statistics.

[11] Bobonis, G.J.and Finan, F. (2009). Neighborhood peer e¤ ects in secondary school enroll-
ment decisions. Review of Economics and Statistics, Vol. 91, No. 4, pp.695-716.

[12] Borghans , L., Smits, W., Vlasblom, J.D.and Jacobs, A. (2000). Leren en werken in het
Nederlandse beroepsonderwijs. Vraag- en aanbodontwikkeling voor de BBL 1999-2004.
ROA, Maastricht.

[13] Bosch, G. and Charest, J. (2009). Vocational training: international perspectives. Rout-
ledge, pp.324.

[14] Bowles, S. (1972). Schooling and inequality from generation to generation Journal of
Political Economy, Vol.80, pp.219-251.

[15] Brown, P., Hesketh, A. and Williams, S. (2003). Employability in a knowledge-driven


economy. Journal of Education and Work, Vol.16, No.2, pp.107-126.

[16] Card, D. (1999). The causal e¤ ect of education on earnings. Handbook of labor eco-
nomics, Vol.3, pp.1801-1863.

20
[17] Ceulenaere, B., Willemsen, A., van der Aa, R., van Zuthpen, F. and Groen, C. (2009).
Case studies MKBA voortijd schoolverlaten. Analyse van Rebound Centre Rotterdam,
Delfshaven en de Amstersamse School. Ecorys.

[18] Davies, J.D. and Lee, J. (2004). To attend or not to attend? Why some students chose
school and others reject it. Support for Learning, Vol.21, No.4, pp.204-209.

[19] de Zwart, S., Zwaneveld, F., Hamdam, Y., Schenk, S. and Hamdam, M. (2009).
Maatschappelijke kosten-batenanalyse Time In en Sportklassen. Berenschot and Rebel-
Group

[20] Elliot, D.and Voss, H. (1974). Delinquency and drop out. Toronto-Londen, Lexington.

[21] Eimers, T., van Amelsvoort, J., Roelofs, M. and Schuit, H. (2009). Tussentijdse invoering
van de kwali…catieplicht. Kenniscentrum Beroepsonderwijs Arbeidsmarkt, pp.17.

[22] Expertise Centrum (2006). Eén meld- en registratiepunt voor leerplicht en voortijdig
schoolverlaten. ’s-Gravenhage.

[23] Felner, R.D., Primavera, J. and Cause, A.M. (1981). The impact of school transitions: A
focus for preventive e¤ orts. American Journal for Community Psychology, Vol.9, pp.449-
459.

[24] Felner, R.D., Ginter, M. and Primavera, J. (1982). Primary prevention during school
transitions: social support and environmental structure. American Journal of Commu-
nity Psychology, Vol.10, No.3, pp.277.

[25] Fergusson, D.M., Horwood, L.J. and Beautrais, A.L. (2003). Cannabis and educational
achievement. Society for the study of addiction to alcohol and other drugs, Vol.98,
pp.1681-1692.

[26] Gamoran, A., Nystrand, M., Berends, M. and Lepore, P.C. (1995). An organizational
analysis of the e¤ ects of ability grouping. American Educational Research Journal,
Vol.32, No.4, pp.687-715.

[27] Garnier, H., Stein, J. and Jacobs, J. (1997). The process of dropping out of high school:
a 19-year perspective. American Educational Research Journal, Vol.34, pp.395-419.

[28] Groot, W. and Maassen van den Brink, H. (2007). The health e¤ ects of education. Eco-
nomic of Education Review, Vol.26, pp.186-200.

[29] Holter, N. and Bruinsma, W. (2009). Wat werkt bij het voorkomen van voortijdige
schoolverlaters? Nederlands Jeugdinstituut (NJI).

21
[30] Jimerson, S.R. (1999). On the failure of failure: Examining the association between
early grade retention and education and employment outcomes during late adolescence.
Journal of School Psychology, Vol.37, pp.243-272.

[31] Keitel, C. (1987). What are the goals of mathematics for all? Journal of Curriculum
Studies, Vol.19, No.5, pp.392-407.

[32] Koenker, R. and Basset, G. (1978). Regression Quantiles. Econometrica, Vol.46, No.1,
pp.33-50.

[33] Lucas, R. and Lammont, N. (1998) Combining Work and Study: An Empirical Study of
Full-Time Students in School, College and University. Journal of Education and Work,
1469-9435, Vol.11, No.1, pp.41-56.

[34] Luyten, H., Bosker, R., Dekkers, H. and Derks, A. (2003). Dropout in the lower secondary
tracks of Dutch secondary education: predictor variables and variation among schools.
School e¤ ectiveness and school improvement. Vol.14, No.4, pp.373-411.

[35] Mayer, S. (1991). How much does a high school’s racial and socioeconomic mix a¤ ect
graduation and teenage fertility rates? In C. Jencks & P. Peterson (Eds.), The Urban
Underclass, pp.321-341. Washington, D.C.: Brookings Institution.

[36] Mclanahan, S.S. (1985). Family structure and the intergenerational transmission of
poverty. American Journal of Sociology, Vol.90, pp.873-901.

[37] McNeal, R.B. (1997). Are students being pulled out of high school? The e¤ ect of adoles-
cent employment on dropping out. Sociology of Education, Vol.70, pp.206-220.

[38] Milligan, K., Moretti, E. and Oreopoulos, P. (2004). Does education improve citizenship?
Evidence from the United States and the United Kingdom. Journal of Public Economics,
Vol.88, No.9, pp.1667-1695.

[39] Ministry of Education (2010). Voortijdig schoolverlaten: cijfers en beleid. Letter to the
Chairman of the Tweede Kamer der Staaten-Generaal. 17.02.2010.

[40] Ministry of Education, Kerncijfers 2003-2007 OCW, Ministerie van Onderwijs Cultuur
en Wetenschap, Den Haag.

[41] Nelson, P.S., Simoni, J. M., and Adelman, H. S. (1996). Mobility and school functioning
in the early grades. Journal of Educational Research, Vol.89, No.6, pp.365-369.

[42] Nelson, R. and Phelps, E. (1966). Investment in humans, technological di¤ usion, and
economic growth. American Economic Review, Vol.51, No.2, pp. 69-75.

22
[43] Oakes, J., Gamoran, A. and Page, R.N. ( 1992). Curriculum di¤ erentiation. Opportuni-
ties, outcomes and meanings. Handbook of Research on Curriculum, pp.570-608.

[44] OECD (2008). Jobs for youth. The Netherlands.

[45] Onstenk, J. (2004). Innovation in vocational education in the Netherlands. Vocal, Vol.5,
pp.17-21.

[46] Onstenk, J. and Blokhuis, F. (2007). Apprenticeship in the Netherlands: connecting


school- and work-based learning. Education + Training, Vol.49, No.6 pp.489-499.

[47] Oosterbeek, H. and Webbink, D. (2004). Wage e¤ ects of an extra year of lower vocational
education: Evidence from a simultaneous change of compulsory school leaving age and
program length. Department of Economics, University of Amsterdam and NOW-pragram
“Scholar”.

[48] Oreopoulos, P. (2006). The compelling e¤ ects of compulsory schooling: evidence from
Canada. Canadian Journal of Economics, Vol.39, No.1, pp.23-52.

[49] Oreopoulos, P. (2007). Do dropouts drop out too soon? Wealth, health and happiness
from compulsory schooling. Journal of Public Economics, Vol.91, Issues 11-12, pp.2213-
2229.

[50] Phillips, J. and Kelly, D. (1979). School failure and delinquency: which causes which?
Criminology, pp.194-207.

[51] Pischke, J-S. and von Wachter, T. (2005). Zero returns to compulsory schooling in Ger-
many: evidence and interpretation. Working Paper 11414, National Bureau of Economic
Research, Cambridge: pp.48.

[52] Pong, S.-L. and Ju, D.B. (2000). The e¤ ects of change in family structure and income
on dropping out of middle and high school. Journal of Family Issues, 21,147-169.

[53] Psacharopoulos, G. (2007). The cost of school failure – A feasibility study. European
Expert Network on Economics of Education.

[54] ROA (2009/1). Zonder diploma. Aanleiding, kansen en toekomstintenties. Researchcen-


trum voor Onderwijs en Arbeidsmarkt, Universiteit Maastricht.

[55] ROA (2009/4). Schoolverlaters tussen onderwijs en arbeidsmarkt. Researchcentrum voor


Onderwijs en Arbeidsmarkt, Maastricht University.

[56] Roderick, M. (1994). Grade Retention and School Dropout: Investigating the Association.
American Educational Research Journal, 31, 729-759.

23
[57] Rumberger, R.W. (1983). Dropping out of high school: The in‡uence of race, sex, and
family background. American Educational Research Journal, 20, pp.199-220.

[58] Rumberger, R.W. and Larson, K.A. (1998). Student mobility and the increased risk of
high school dropout. American Journal of Education, Vol.107, No.1, pp.1-35.

[59] Rumberger, R.W. and Thomas, S.L. (2000). The distribution of dropout and turnover
rates among urban and suburban high schools. Sociology of Education, Vol.73, pp.39-67.

[60] Rumberger, R.W. (2001). Who drops out of school and why. Paper prepared for the
National Research Council, Committee on Educational Excellence and Testing Equity
Workshop, School Completion in Standards-Based Reform: Facts and Strategies, and
incorporated into their report, Understanding Dropouts: statistics, strategies, and high-
stakes testing, edited by A. Beatty, U. Neiser, W. Trent, and J. Heubert, Washington:
National Academy Press.

[61] Rumberger, R.W. and Lamb, S.P. (2003) The Early Employment and Further Education
Experiences of High School Dropouts: A Comparative Study of the United States and
Australia. Economics of Education Review, Vol.22, pp.353-356.

[62] Rumberger, R.W (2004). Why students drop out of school. In Gary Or…ed (Ed.),
Dropouts in America: Confronting the Graduation Rate Crisis, Cambridge, MA: Har-
vard Education Press, (pp.131-155).

[63] Schultz, T. (1967). The economic value of education. Columbia University Press, New
York

[64] Spence, M. (1973). Job Market Signaling. The Quarterly Journal of Economics, Vol.87,
No.3, pp.355-374.

[65] Steinberg, L., Dornbusch, S.M. and Brown, B.B. (1992). Ethnic di¤ erences in adolescent
achievement. American Psychologist, Vol.47, pp.723-729.

[66] ter borgt, T., van Lieshout, M., Doornwaard, S. and Eijkemans, Y. (2009). Middelenge-
bruik en voortijdig schoolverlaten. Twee onderzoeken naar de actuele en gepercipieerde
rol van alcohol en cannabis in relatie tot spijbelen, schoolprestaties, motivatie en uitval.
Universiteit Utrecht and Trimbos-instituut.

[67] Terwel, J. (2004). Curriculum and curriculum di¤ erentiation. In Terwel, J., Walker,
D.F., (2004) Curriculum as a shaping force: toward a principle approach in curriculum
theory and practice. Nova Science, pp.33-50.

[68] Terwel, J. (2005). Curriculum di¤ erentiation: multiple perspectives and developments in
education. Journal of Curriculum Studies, Vol.37, No.6, pp.653-670.

24
[69] van der Steeg, M., van Elk, R. and Webbink, D. (2008). Did the 2006 covenants reduce
school dropout in the Netherlands? CPB document 177.

[70] van der Steeg, M. and Webbink, D. (2006). Voortijdig schoolverlaten in Nederland: om-
vang, beleid en resultaat. CPB Document, No.107.

25

Você também pode gostar