EDUCATION POLICY AND EQUALITY OF
OPPORTUNITY
GABRIELA SCHUETZ
H
EINRICH W. URSPRUNG
L
UDGER WOESSMANN
CESIFO WORKING PAPER NO. 1518
CATEGORY 3: SOCIAL PROTECTION
A
UGUST 2005
An electronic version of the paper may be downloaded
•
from the SSRN website: www.SSRN.com
•
from the CESifo website: www.CESifo-group.de
CESifo Working Paper No. 1518
EDUCATION POLICY AND EQUALITY OF
OPPORTUNITY
Abstract
We provide a measure of equality of educational opportunity in 54 countries, estimated as the
effect of family background on student performance in two international TIMSS tests. We
then show how organizational features of the education system affect equality of educational
opportunity. Our model predicts that late tracking and a long pre-school cycle are beneficial
for equality, while pre-school enrollment is detrimental at low levels of enrollment and
beneficial at higher levels. Using cross-country variations in education policies and their
interaction with family background at the student level, we provide empirical evidence
supportive of these predictions.
JEL Code: I21, J62, H52.
Keywords: equality of opportunity, educational production, family background, student
performance, tracking, pre-school, efficiency-equity tradeoff.
Gabriela Schuetz
Ifo Institute for Economic Research
at the University of Munich
Poschingerstr. 5
81679 Munich
Germany
schuetz@ifo.de
Heinrich W. Ursprung
Department of Economics
University of Konstanz
Box D-138
78457 Konstanz
Germany
Heinrich.Ursprung@uni-konstanz.de
Ludger Woessmann
Ifo Institute for Economic Research
at the University of Munich
Poschingerstr. 5
81679 Munich
Germany
woessmann@ifo.de
For helpful comments and discussion, we would like to thank Steve Machin, Winfried
Pohlmeier and Joachim Winter, as well as participants at the joint SOLE/EALE World
Conference on Labor Economics in San Francisco.
1
1. Introduction
Equality in educational outcomes is a crucial determinant of the extent of equality of
opportunity and intergenerational mobility achieved by societies. Nickell (2004), for example,
shows that a large part of the existing cross-country variation in earnings inequality can be
attributed to cross-country variation in skill dispersion. It is, therefore, of prime policy interest
to understand the effects of education policies such as ability tracking, pre-school education,
length of the school day and educational spending on the educational success of children from
various family backgrounds. The direction of these effects is, however, by no means
straightforward from a theory point of view, and empirical evidence is limited. The lack of
empirical evidence derives from the fact that variation in the organization of education
systems is largely lacking within countries, and where it is not, it is unlikely to be exogenous
to students’ performance and family backgrounds.
This state of affairs is the starting point of our paper which makes three contributions.
First, it provides a comparable measure for 54 countries of how strongly children’s
educational performance is related to their family background (Section 2). We interpret this
measure as a proxy for the extent of inequality of opportunity. Second, the paper develops a
theoretical model that traces the effect of different education policies on the equality of
opportunity (Section 3). Third, we present empirical evidence, using cross-country variations
in education policies and their interaction with family background at the individual student
level to identify the impact of education policies on equality of opportunity (Section 4).
The database used combines two related extensive international student achievement tests,
the Third International Mathematics and Science Study (TIMSS) and its replication for a
partly different set of countries (TIMSS-Repeat) (Section 2.1). These datasets provide
information on students’ educational performance, their family background and relevant
control variables for individual students in each participating country. As our main indicator
of family background, we use the number of books in the students’ home. As suggested in the
sociological literature, books at home provide a powerful proxy for the educational, social and
economic background of the students’ families. Moreover, previous research on the same and
other datasets suggests that in most countries, books at home are the single most important
predictor of student performance, even surpassing parental education (Wößmann 2003, 2004;
Fuchs and Wößmann 2004). Furthermore, data coverage on this indicator is superior to
parental education, and we argue that it is more readily comparable across countries.
2
This database allows us to estimate an index of equality of educational opportunity in 54
countries (Section 2.2). More precisely, the index measures the inequality of educational
outcomes for children from different family backgrounds. Given the strong relationship
between education and economic outcomes, our measure can thus also serve as an index of
the inequality of opportunity later in life for children from different family backgrounds. To
our knowledge, no previous evidence on the inequality of opportunity across countries has
been available on a comparable scale. In estimating our index at the micro level, we make
sure that it is not affected by cross-country differences in the immigrant population, but only
reflects performance differences associated with socio-economic background.
Our results show that equality of opportunity as measured by our index varies considerably
across countries (Section 2.3). Among OECD countries, the impact of our family-background
measure on student performance is largest in England, Scotland, Hungary and Germany and
lowest in France, Canada, Portugal and Flemish Belgium. The family-background effect in
the former countries is on average 2.7 times as large as in the latter countries. The United
States falls in the top quarter of the most unequal OECD countries.
To understand the substantial cross-country variation in the extent of equality of
educational opportunity, we develop a model to explain how key organizational features of
school systems affect the extent to which equality of educational opportunity is achieved.
These features include the duration of and enrollment in pre-school education (Section 3.1)
and the timing of the change from comprehensiveness to ability tracking in the school system
(Section 3.2). Numerical solution of the model suggests that late tracking and a long pre-
school cycle are beneficial for equality of opportunity, while pre-school enrollment has a
detrimental influence at low levels of enrollment and a beneficial influence at higher levels
(Section 3.3).
Combining the observed variation in the index of educational inequality with country-level
data on features of the education systems, we can test the predictions of our model
empirically. Our preferred empirical identification strategy is to estimate how the different
country-level features of the school systems interact with the family-background measure at
the student level in determining student performance, while at the same time controlling for
unobserved country heterogeneity by country fixed effects (Section 4.1).
The empirical results strongly support our theoretical model. We find that the family-
background effect is larger (i.e. equality of opportunity is lower), the earlier a country tracks
its students into different school types by ability (Section 4.2). Also, the family-background
3
effect is larger in countries with shorter pre-school education. With respect to pre-school
enrollment, we find an inverted U-shaped relationship, with educational inequality increasing
up to an enrollment of roughly 60 percent and decreasing thereafter. These results prove
robust to more extensive model specifications, in which we do not find a statistically
significant difference in the equality of opportunity by school starting age or between half-day
and whole-day school systems. Neither does the observed equality of opportunity differ with
average educational spending, nor with the country’s level of economic development. At least
in the OECD sample, there is also no statistically significant relationship between equality of
opportunity and a country’s mean test score. Finally, the family-background effect is larger in
countries with a larger share of private funding, but at the same time, it is smaller in countries
with a larger share of private provision.
Our theoretical and empirical results thus show how school systems can accelerate
intergenerational mobility. The results suggest that education policies such as comprehensive
school systems and extensive early-childhood education can increase the equality of
educational opportunity for children from different family backgrounds.
1
The empirical
evidence on the other hand also suggests that extending the school day into the afternoon,
bringing forward the age at which compulsory education begins or increasing educational
spending do not appear to have a significant effect on the equality of educational opportunity.
There also does not appear to exist an efficiency-equity tradeoff in education in the sense that
more equal systems would systematically affect the mean performance of their students, at
least not among OECD countries. Relying on private spending to finance education varies
positively with increased inequality of educational opportunity, relying on private production
of schooling on the other hand varies positively with equality of educational opportunity.
2. Estimating Equality of Opportunity across Countries
2.1 Data
2.1.1 The Two TIMSS International Student Achievement Studies
To derive estimates of the equality of educational opportunity, we employ student-level micro
data from two extensive international student achievement tests. The first test is the Third
1
For selected previous analyses of ability tracking, cf. Betts and Skolnick (2000), Figlio and Page (2002),
Epple et al. (2002), Galindo-Rueda and Vignoles (2004), Meghir and Palme (2005) and Hanushek and
Wößmann (2005). For recent evidence on the effects of early-childhood education, cf. Garces et al. (2002),
Magnuson et al. (2004) and Schweinhart et al. (2005); Barnett (1992) and Currie (2001) provide surveys of
previous analyses.
