ArticlePDF Available

The socio-spatial dimension of educational inequality: A comparative European analysis


Abstract and Figures

Given recent evidence of rising levels of social segregation in European countries, this study uses standardized data from the Program for International Student Assessment (n = 171,159; 50.5% male) to examine the extent to which education systems in Europe are socially segregated and whether social segregation in the school system affects achievement gaps between students of different social origin. Results suggest that the degree of social segregation within education systems varied substantially across countries. Furthermore, multilevel regression models indicate that the effect of socioeconomic status on student achievement was moderately but significantly stronger in more segregated education systems, even after controlling for alternative system-level determinants of social inequality in student achievement. These findings provide original evidence that social segregation in education systems may contribute to the intergenerational transmission of educational (dis)advantage and thus serve to exacerbate wider problems of socioeconomic inequality in Europe.
Content may be subject to copyright.
Author manuscript (07/09/2019)
Published in final edited version in:
Studies in Educational Evaluation.
The socio-spatial dimension of educational inequality: A
comparative European analysis
Kaspar Burger
University of Minnesota, University College London
Given recent evidence of rising levels of social segregation in European countries, this study uses
standardized data from the Program for International Student Assessment (n = 171,159; 50.5% male)
to examine the extent to which education systems in Europe are socially segregated and whether
social segregation in the school system affects achievement gaps between students of different social
origin. Results suggest that the degree of social segregation within education systems varied
substantially across countries. Furthermore, multilevel regression models indicate that the effect of
socioeconomic status on student achievement was moderately but significantly stronger in more
segregated education systems, even after controlling for alternative system-level determinants of
social inequality in student achievement. These findings provide original evidence that social
segregation in education systems may contribute to the intergenerational transmission of educational
(dis)advantage and thus serve to exacerbate wider problems of socioeconomic inequality in Europe.
Cross-national comparison; Social segregation; Standardized assessment; European education systems;
This study is part of a project that has received funding from the European Union’s Horizon 2020 research and
innovation program under the Marie Skłodowska-Curie Grant Agreement No. 791804.
Correspondence should be addressed to Kaspar Burger, 1014 Social Sciences Building, Department of
Sociology, 267 19th Avenue South, University of Minnesota, Minneapolis, MN 55455, USA, E-mail:
The Socio-Spatial Dimension of Educational Inequality: A Comparative
European Analysis
1. Introduction
In recent years, the level of social segregation in European countries has increased (Marcinczak,
Musterd, van Ham, & Tammaru, 2016). It is therefore crucial to examine whether, and to what
extent, education systems in Europe are also segregated along social lines, and whether social
segregation between schools shapes individual student achievement and social inequality in
educational outcomes.
Social segregation in education systems refers to the uneven distribution across schools
of students from different socioeconomic backgrounds (Jenkins, Micklewright, & Schnepf,
2008). Where students are highly segregated by socioeconomic origin between schools,
resources that contribute to students’ educational success—such as social, economic, and
cultural capitalare more unequally distributed (Owens, 2018; Reardon & Owens, 2014). An
unequal distribution of such resources among student populations typically leads to disparities
in educational opportunities, because schools draw on these resources informally in educating
their students (Chiu, 2015; Croxford & Paterson, 2006). For instance, schools serving
socioeconomically advantaged students receive more support from parents (Lee & Burkam,
2002). Their students come from families who tend to have higher educational expectations for
their children (Davis-Kean, 2005; Neuenschwander, Vida, Garrett, & Eccles, 2007); and these
families often have more knowledge of the education system (e.g., about its written and
unwritten rules, what and how students should learn, or what educational decisions to take;
Crosnoe & Muller, 2014; Jackson, Erikson, Goldthorpe, & Yaish, 2007). Moreover, schools
that serve advantaged students benefit from the fact that their students typically attach great
value to education and use similar forms of communication and interactions in school and in
their family environment (Lareau & Weininger, 2003). As a result, their student populations
constitute functional communities particularly conducive to learning (Lee & Bowen, 2006).
Children learn from each other, and peer achievement affects achievement growth (Hanushek,
Kain, Markman, & Rivkin, 2003; Lavy, Paserman, & Schlosser, 2012). Consequently, an
uneven distribution of students of diverse social origins within an education system may affect
not only student achievement but also social disparities in achievement.
So far, research on the relationship between social segregation within education systems
and social class gradients in student achievement across European countries is scarce. This is
despite researchers and policymakers increasingly acknowledging the need to address common
challenges, such as ensuring social cohesion and fairness in education, at a European level (e.g.,
Pépin, 2011).
In light of the above, this study pursues two main objectives. First, it assesses the links
between social segregation and socioeconomic gradients in student achievement within
European education systems. Second, it examines whether social segregation in these education
systems moderates the micro-level associations between socioeconomic status and educational
achievement, when controlling for further country- as well as school-, and individual-level
variables. The study thereby extends cross-national comparative research on the mechanisms
underlying socioeconomic inequality in educational achievement, which (for brevity) we also
refer to simply as educational inequality.”
2. Prior research on social segregation and educational inequality
2.1. System-level links between social segregation and educational inequality
Prior research indicated a positive correlation between social segregation within education
systems and socioeconomic disparities in student achievement (Felouzis & Charmillot, 2013).
However, this research compared education systems at subnational levels in Switzerland. To
date, there is no research analyzing specifically whether, across Europe, more socially
segregated education systems are those in which student achievement is more closely linked to
socioeconomic status.
2.2. Effects of social segregation on educational inequality
Some studies sought to examine whether social segregation in education systems affects
educational inequality. McPherson and Willms (1987) found that moving from a selective to a
comprehensive secondary school system in Scotland minimized social class segregation
between schools and improved the educational achievement, in particular, of poor children. A
more recent study suggests that educational inequality was more pronounced in OECD
countries whose education systems exhibited higher levels of social segregation (Holtmann,
2016). However, this study did not control for any other country-level determinants of
educational inequality, thus making it difficult to conclude that segregation was the actual driver
of this inequality. Furthermore, evidence from the United States indicates that income
segregation between school districts exacerbated achievement gaps between privileged and
underprivileged students (Owens, 2018; Reardon, 2011). However, it remains unclear whether
social segregation also increases socioeconomic inequality in educational outcomes in
European countries where the levels of social segregation are estimated to be substantially
lower (Marcinczak et al., 2016; see also Sortkær, 2018).
More generally, there is relatively little cross-national comparative research on the
consequences of system-level segregation on educational inequalities. Prior research on socio-
spatial inequalities in education typically focused on school social composition effects (Borman
& Dowling, 2010; Dumay & Dupriez, 2008; Fekjær, & Birkelund, 2007; Opdenakker & van
Damme, 2007; Palardy, 2013; Rumberger & Palardy, 2005), rather than system-level
segregation effects. In fact, in a review of research, Reardon and Owens (2014) concluded that
“much of the research purporting to assess the links between segregation and student outcomes
instead measures the association between school composition and student outcomes” (p. 200).
Research on school composition effects tests the impact of segregation in only a limited sense,
under the assumption that segregation affects educational achievement and/or inequality
predominantly through school composition mechanisms, rather than through other mechanisms
such as the uneven distribution of resources and the corresponding disparities in learning
opportunities on a broader system level. Moreover, research on school composition effects does
not allow for analyzing system-wide segregation effects. Within a country, a given set of
schools may exhibit low levels of social segregation, although the degree of segregation at the
overall system level might be substantial. Cross-national comparative research allows for
distinguishing between school composition and system-wide segregation effects and thus may
provide a more comprehensive picture of the consequences of socio-spatial clustering of
students. In addition, cross-national research provides the opportunity to examine systematic
patterns of covariation between social segregation and educational inequality across countries
by taking into account potential system-level confounders. Prior research focusing on school
composition effects was conducted in diverse countries that differed not only in the overall level
of social segregation within the system, but also in other macro-level variables (e.g., Belfi et
al., 2014; Driessen, 2002; Lauen & Gaddis, 2013; Strand, 2010; Televantou et al., 2015; Van
Ewijk & Sleegers, 2010). In this research, effects of the socio-spatial clustering of students may
have been confounded with those of further, unmeasured, country-specific influences.
Specifically, this prior research may have overlooked alternative country-level explanations of
educational inequality, such as the overall level of national inequality (Chmielewski & Reardon,
2016), the economic development of a country (Yaish & Andersen, 2012), or the
comprehensiveness of the education system (Burger, 2016a).
Given that standardized cross-
national data on student achievement are now available, it is now possible to analyze effects of
social segregation within comparative designs that also consider further potential country-level
determinants of educational inequality. We develop such a design here.
3. Contribution to the literature
This study extends knowledge of social segregation and inequality in European countries
(Benito et al., 2014; Bernelius, & Vaattovaara, 2016; Böhlmark, Holmlund, & Lindahl, 2016;
Musterd, Marcińczak, van Ham, & Tammaru, 2017; Yang Hansen, & Gustafsson, 2016, in
press; Yang Hansen, Rosén, & Gustafsson, 2011). First, it uses cross-national standardized data
to analyze the link between social segregation within education systems and socioeconomic
gradients in student achievement across European countries. Second, because socioeconomic
gradients in achievement could be a consequence of further system-level influences (rather than
the result of segregation within the education system), the study investigates whether
segregation moderates these gradients when alternative system-level influences are considered.
Our strategy is to examine major system-level influences comprehensively while keeping the
models parsimonious. Thus, we concentrate on five economic and education policy dimensions
that have been identified as major system-level determinants of educational inequality in prior
research: (1) economic development, (2) population-level socioeconomic inequality, (3) annual
schooling time, (4) preschool enrollment rate, and (5) public expenditure on education.
Economic development and socioeconomic inequality have long been recognized as
potential drivers of educational inequality (Heyneman & Loxley, 1983; Jerrim & Macmillan,
2015). Specifically, research has shown that the level of economic development correlates
negatively with educational inequality because more economically developed societies tend to
be more open societies in which the importance of ascriptive (“non-merit) factors such as
social origin for individual educational attainment gradually decreases (Ferreira & Gignoux,
2014; Gustafsson, Nilsen, & Yang Hansen, 2018; Marks, 2009; van Doorn, Pop, & Wolbers,
2011). Moreover, evidence suggests that socioeconomic inequality is related positively to
educational inequality (Campbell, Haveman, Sandefur, & Wolfe, 2005; Chmielewski &
A few studies used cross-national comparative designs, but they did not specifically consider country-specific
determinants of educational achievement and inequality (Alegre & Ferrer, 2010; Benito, Alegre, & Gonzàlez-
Balletbò, 2014; Yang Hansen, Gustafsson, & Rosén, 2014).
Reardon, 2016; Kearney & Levine, 2014). One explanation for this is that schools may
reproduce or even exacerbate the inequalities that children bring with them (Downey &
Condron, 2016).
In addition, the comprehensiveness of education systemsin terms of the annual
schooling time, preschool enrollment rate, and public expenditure on educationmay affect
educational inequality (Burger, 2016a; Pfeffer, 2008; Schütz, Ursprung, & Wössmann, 2008;
Stadelmann-Steffen, 2012). A longer annual schooling time can reduce educational inequality
because children from all social classes share similar learning environments at school, benefit
from similar learning opportunities, and thus make similar learning progress (Ammermüller,
2005; Schlicht, Stadelmann-Steffen, & Freitag, 2010). Preschool enrollment may equalize
educational outcomes among children because children of low socioeconomic status, who often
lag behind in their academic development, typically make greater developmental progress in
preschool programs than their more advantaged peers (Burger, 2010, 2013, 2015, 2016b;
Cebolla-Boado, Radl, & Salazar, 2017). Finally, public expenditure on education is commonly
thought to reduce educational inequality (OECD, 2012; Schütz et al., 2008). Where public
expenditure on education is low, a shift in responsibility from the public to the private sector
may occur, resulting in diverging educational opportunities among social classes, with more
advantaged families being likely to spend more on their children’s education (Schlicht et al.,
2010; Schmidt, 2004).
To identify the unique contribution of social segregation to educational inequality, the
current study distinguishes between social segregation and the above-mentioned economic and
education policy dimensions as potential country-specific sources of educational inequality.
Furthermore, it is essential to recognize that social segregation in education systems
may be related in part to educational tracking (Felouzis & Charmillot, 2013; Pfeffer, 2015), or
allocation of students to different types of schools or curricula that are vertically structured by
student performance and typically prepare students either for further academic or for vocational
programs. This is because a student’s likelihood of transitioning to a given track is to some
extent associated with family background characteristics (Brunello & Checchi, 2007; Lucas,
2001). However, associations between tracking and social segregation differ considerably
across education systems (Alegre & Ferrer, 2010; Chmielewski, 2014; Maaz, Trautwein,
Lüdtke, & Baumert, 2008). Moreover, the degree to which education systems are socially
segregated varies significantly, even among those systems that use comparable tracking regimes
(see Appendix A). For instance, several education systems display comparatively high levels
of social segregation, although they use little or no tracking, which is in part explained by the
fact that social segregation is often a result of choices made, whether consciously or
unconsciously, by families who tend to live in socially homogeneous school catchment areas,
or may decide to enroll their children in particular high-performing or private schools
(Lockheed, Prokic-Bruer, & Shadrova, 2015; Saporito & Sohoni, 2007). In addition, research
also suggests that de-tracking schools may lead to an increase in residential segregation (De
Fraja & Martinez Mora, 2012). Consequently, school tracking might actually have a de-
segregating effect, or at least prevent further increases in segregation. In a similar vein, a study
from Japan found that de-tracking reforms can yield unintended consequences, as they may
drive better-performing students out of public schools, and thus exacerbate the divide between
students from different socioeconomic backgrounds (Kariya & Rosenbaum, 1999). In
conclusion, these findings suggest that social segregation within education systems can affect
educational inequality independent of tracking (Esser & Relikowski, 2015; Waldinger, 2006).
Nevertheless, the educational track that a student attends should be considered in any study
designed to assess social disparities in educational outcomes. Thus, we consider whether a
student attended a general academic program (designed to give access to further academic
studies at the next educational level), or a pre-vocational or vocational program (designed to
give access to vocational studies or the labor market).
To conceptualize segregation effects, we draw on the distinction between “Type A” and
“Type B” effects (cf., Raudenbush & Willms, 1995). Type A effects refer to the effects that
school systems have on individual student achievement through both mechanisms they control
(e.g., educational resources) and mechanisms they do not control (contextual effects such as
peer influences). By contrast, Type B effects refer to the controllable effects alone (Castellano,
Rabe-Hesketh, & Skrondal, 2014). We study Type A effects of school system segregation,
which represent both controllable and uncontrollable influences on student achievement. This
allows us to assess the net effect of segregation, which corresponds to the sum of positive and
negative effects of segregation, adjusted for observable potential confounders.
It is clear that non-experimental research examining segregation effects typically cannot
exclude selection bias. Social segregation in education systems may generate disparities in
student achievement. However, achievement disparities may as well reflect preexisting
differences between students (i.e., differences not related to the exposure to socially segregated
schools). For instance, family characteristics such as social and economic resources contribute
to residential and school district choice and to children’s educational achievement, which
complicates the estimation of genuine segregation effects. Previous research from the United
States used measures of local government fragmentation prior to the observation period as
instruments for segregation, indicating that segregation does have a causal effect on inequalities
in educational attainment (Quillian, 2014). However, identifying robust instruments is difficult
(Owens, 2018). Here we use a comparative approach and standardized international student
assessment data to study whether social segregation within education systems moderates micro-
level associations between socioeconomic status and educational achievement under ceteris
paribus conditionswhen observable country-, school-, and individual-level determinants of
student achievement are taken into account. We argue that social segregation within education
systems contributes to social disparities in educational achievement by increasing inequalities
between disadvantaged and advantaged schools. Schools draw on social, economic, and cultural
resources of families informally, and we expect that an unequal distribution of such resources
will intensify disparities in learning environments and educational opportunities, ultimately
exacerbating social inequality in student achievement. In view of the challenges that potential
selection effects present, the results of our study provide empirical evidence consistent with,
but not definitively demonstrating, a causal association between social segregation in education
systems and social inequality in educational achievement.
4. Method
4.1. Data
The data are drawn from the 2012 wave of the Program for International Student Assessment
(PISA), a cross-national comparative survey that has analyzed 15 year olds’ achievement in
mathematics, science, and reading in a three-year cycle since 2000, with a special focus on one
of these subjects in each wave, which was here mathematics. PISA uses a stratified sampling
procedure and, in the first stage, schools with 15-year-old students are selected with a
probability proportional to the size of the school (primary sampling units). In the second stage,
students are selected at random within schools. The sample used here comprises 29 European
countries with 171,159 students (50.5% male) from 7,301 schools.
Table 1 summarizes the
Thirty-one European countries participated in the 2012 PISA wave. Liechtenstein was excluded owing to its
small sample size. Italy was excluded because it contained 6.2% of schools in which fewer than 20 students
participated in the survey, but analyses including Italy yield virtually identical results and lead to the same
conclusions. It should also be noted that schools are not necessarily comparable across all countries. This is
exemplified by the fact that, in some countries, schools were defined as administrative units that can consist of
several buildings. In others, individual buildings were defined as schools. Of the 29 countries included in our
number of students and schools for each country. The PISA final student weights are applied
so that the sample of each country reflects the total population of 15-year-old students within
each country (see OECD, 2009b, p. 47ff.). These weights are inversely proportional to the
probability of selecting a given student into the PISA sample, which considers the probability
of selecting the school within a country as well as the individual student within a school.
Table 1
Number of schools and students in the sample.
N schools
N students
Czech Republic
Great Britain
sample, 23 used individual schools as the primary sampling unit, whereas six used educational programs or tracks
within schools as the primary sampling units (BEL, HRV, HUN, NLD, ROU, SVN).
4.2. Measures
This section describes the variables used in this study. Table 2 displays the descriptive statistics
of these variables, pooled across countries; Table 3 displays the descriptive statistics of the
individual- and school-level variables for each country separately; Table 4 displays the
descriptive statistics of the dependent variable (5 plausible values) for each country.
Table 2
Descriptive statistics.
Note: N = 171,159. Descriptive statistics of binary and un-centered continuous variables. The continuous variables
were grand-mean centered for the analyses. (a) The reference category is “general academic program”. (b) As
opposed to public schools, private schools are funded by fees paid by parents (entirely if they are government-
independent, partially if they are government-dependent). (c) Gini coefficient of equivalized disposable income
(higher values of indicate greater inequality in disposable household income).
4.2.1. Dependent variable. The dependent variable is student achievement, estimated using the
PISA measurement of math proficiency. In PISA, math proficiency is conceptualized as an
individual’s capacity to formulate, interpret, and deploy mathematics in a variety of contexts,
which involves the application of important mathematical concepts, knowledge, and skills to
solve everyday problems (OECD, 2013). Although math proficiency constitutes only one aspect
of student achievement, it is considered as a particularly suitable subject for comparative
purposes across educational systems, in particular because several educational systems contain
large proportions of immigrant students whose language proficiency may vary considerably
(Levels, Dronkers, & Kraaykamp, 2008). Math proficiency is also used as a proxy for student
achievement to compare with findings from previous studies (Schlicht et al., 2010; Stadelmann-
Steffen, 2012). Math proficiency is estimated in the form of five plausible values, which
represent the range of abilities that a student can be expected to have, given the student’s
responses to the PISA test items (Wu, 2005). To determine population statistics, each plausible
value is first used separately in any analysis. Using Rubin’s rule (1987), the results of these
analyses are then averaged in order to produce the final statistics (OECD, 2009a). By employing
plausible values instead of raw estimates of student achievement, we minimize the effect of
measurement error bias in the outcome variable.
4.2.2. Independent variable. The independent variable is students’ socioeconomic status (SES),
measured using an index that considers parents’ occupational status (the international
socioeconomic index of occupational status, HISEI), parents’ educational level (number of
years in education according to the international standard classification of education, ISCED),
and home possessions (a construct consisting of items assessing family wealth, cultural
possessions, educational resources, and the number of books at home). In the PISA dataset, this
is known as the index of economic, social, and cultural status (ESCS). This index is comparable
across countries, as determined by similar scale reliabilities (Cronbach’s α) across countries, as
well as through principal component analyses, performed separately for each country,
indicating that across countries the three componentsparental occupational status, parental
education, and home possessionshad very similar loadings on the index of economic, social,
and cultural status, and thus correlated to a very similar degree with this index (OECD, 2014,
p. 352).
4.2.3. Central moderator variable. The key variable assumed to moderate the individual-level
relationship between SES and educational achievement is an index of social segregation within
national education systems (see Table 5). This index is estimated by means of intra-class
correlations of SES, using a multilevel modeling approach in line with previous studies (Ferrer-
Esteban, 2016; Goldstein & Noden, 2003; Mayer, 2002). The intra-class correlation (ICC)
measures the degree to which SES varies between, as opposed to within, schools. A high ICC
indicates high within-school similarity of students, meaning that students within a given school
are more similar in terms of SES to students within their school than to those in other schools.
The ICC can also be interpreted as the proportion of variance in SES that lies between schools.
Mathematically, it corresponds to the ratio of the school-level variance in SES to the total
variance in SES within a country. In order to partition the total variation in SES within a country
into two variance componentswithin schools and between schoolswe use an unconditional
multilevel regression model with SES as the outcome and with a random intercept at the student
level and a random intercept at the school level, performed separately for each country. This
model is specified as
SESij = β0j + εij
(eq. 1)
with β0j = α00 + μ0j
(eq. 2)
where, at the individual level, SESij is the socioeconomic status of student i in school j, β0j is
the mean SES in school j, and εij is the deviation of the SES of student i from the school mean,
or the residual error (eq. 1). At the school level, α00 is the grand mean, and μ0j is the deviation
of the mean SES of school j from the grand mean, or the residual error (eq. 2). The variances
of the residual errors εij and μ0j are assumed to have a normal distribution with a mean of zero
and to be mutually independent. They are denoted as 
and 
, respectively, and are also
referred to as variance components. As noted above, the ICC corresponds to the ratio of the
school-level variance in SES to the total variance in SES within a country. Thus, it is calculated
as ρ = 
/ (
+ 
), where 
is the school-level variance and 
is the individual-
level variance in SES.
The intra-class correlation of SES is a standard index of social segregation within
education systems (Agirdag, Van Avermaet, & van Houtte, 2013; Goldstein & Noden, 2003;
Modin, Karvonen, Rahkonen, & Östberg, 2015; Palardy, Rumberger, & Butler, 2015). There
are various other indices of segregation available (Duncan & Duncan, 1955; Gorard & Taylor,
2002; Hutchens, 2004; Jenkins et al., 2008; Reardon & Bischoff, 2011). However, typically,
and in contrast to the applied index, they are a function of observed proportions, such as poor
versus non-poor children in schools, and thus based on dichotomous measures (Leckie,
Pillinger, Jones, & Goldstein, 2012). The intra-class correlation relies on a continuous scale,
which captures the entire distribution of socioeconomic origin. This allows us to determine to
what extent students are socially dissimilar (segregated) between schools without determining
a cut-off value to differentiate students into broad socioeconomic categories. As any other index
of segregation, the intra-class correlation of SES provides an estimate of the unevenness in the
distribution of students across schools.
4.2.4. Alternative country-level influences on educational inequality. In addition to the index of
social segregation, the analysis includes the following country-level variables to evaluate their
effects on student achievement and whether they moderate the relationship between SES and
student achievement. As a measure of a country’s economic development, we consider the
gross domestic product (GDP) per capita in purchasing power standard (averaged across the
years 2003 through 2011, the period preceding the PISA assessment during which the students
attended compulsory school; data from Eurostat, 2017). As a measure of socioeconomic
inequality, we consider the level of income inequality within the population, notably the Gini
coefficient of equivalized disposable income (averaged across the years 2005 to 2012, given
the availability of data for this period; data from Eurostat, 2018).
To take into account the
amount of time that children spent in school annually in a given education system, we use the
annual taught time (in hours of 60 minutes), averaged across the compulsory education years
(20032011; data from Eurydice, 2013).
The preschool enrollment rate is estimated based on
data from the PISA database. It refers to the proportion of students who had been enrolled in
preschool (ISCED 0) for any given duration, in contrast to students who had never been
enrolled. Finally, we assess the public educational expenditure on compulsory education
(ISCED 14) as the percentage of the gross domestic product (GDP; averaged across the
period 20032011; data from Eurostat, 2017) (see Table 2). To avoid model overspecification,
we also perform models with only selected country-level variables, as explained in Section 6.
4.2.5. Control variables at the individual and school level. At the individual level, the analysis
controls for gender, immigrant status, language spoken at home, the school grade in which a
student is enrolled at the time of the assessment (school grade level), and whether a student
attended (0) a general academic program or (1) a pre-vocational or vocational program, because
Data for Serbia refer to 2013, owing to missing values for the preceding years.
Data for Serbia are derived from OECD (2011), data for Switzerland from UNESCO (2011).
these variables have a direct effect on student achievement (Burger & Walk, 2016; Schlicht et
al., 2010). At the school level, the analysis controls for school type (public vs. private), the
proportion of first-generation immigrants in school, and school socioeconomic composition
(the aggregate SES of a school’s student population) in order to account for their hypothesized
effects on student achievement. All control variables are derived from the PISA database.
Table 3
Descriptive statistics for the individual- and school-level variables, by country.
Individual level
School level
Language at
home: same
as test
School grade
relative to
modal grade
status (SES)
School type:
private school
Proportion of
immigrants in
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
-0.51 (0.57)
0.11 (0.83)
0.04 (0.02)
-0.26 (0.21)
-0.42 (0.65)
0.18 (0.91)
0.05 (0.02)
-0.14 (0.29)
0.00 (0.33)
-0.23 (1.02)
0.04 (0.02)
-0.26 (0.21)
0.20 (0.34)
-0.35 (0.85)
0.04 (0.02)
-0.29 (0.19)
Czech Republic
0.41 (0.57)
0.06 (0.76)
0.05 (0.02)
-0.13 (0.30)
-0.18 (0.42)
0.28 (0.91)
0.05 (0.03)
-0.10 (0.30)
-0.21 (0.45)
0.15 (0.13)
0.04 (0.02)
-0.24 (0.22)
-0.19 (0.43)
0.35 (0.83)
0.05 (0.02)
-0.12 (0.29)
-0.27 (0.56)
-0.02 (0.80)
0.04 (0.02)
-0.22 (0.22)
0.27 (0.67)
0.19 (0.93)
0.04 (0.02)
-0.20 (0.26)
Great Britain
0.16 (0.40)
0.24 (0.81)
0.06 (0.04)
-0.02 (0.32)
-0.05 (0.27)
-0.05 (0.99)
0.04 (0.02)
-0.27 (0.20)
0.16 (0.54)
-0.20 (0.94)
0.04 (0.02)
-0.23 (0.22)
0.00 (0.00)
0.78 (0.81)
0.05 (0.02)
-0.30 (0.18)
0.46 (0.73)
0.13 (0.85)
0.04 (0.02)
-0.27 (0.20)
-0.12 (0.45)
-0.18 (0.87)
0.04 (0.02)
-0.25 (0.21)
0.05 (0.43)
-0.13 (0.91)
0.04 (0.02)
-0.22 (0.22)
0.27 (0.67)
0.08 (1.10)
0.06 (0.03)
-0.28 (0.15)
0.44 (0.57)
0.21 (0.78)
0.05 (0.02)
-0.28 (0.20)
0.00 (0.07)
0.47 (0.76)
0.05 (0.02)
-0.26 (0.21)
-0.04 (0.23)
-0.16 (0.92)
0.04 (0.02)
-0.28 (0.20)
-0.56 (0.76)
-0.48 (1.17)
0.05 (0.02)
-0.25 (0.21)
0.00 (0.32)
-0.46 (0.93)
0.04 (0.02)
-0.28 (0.19)
0.01 (0.16)
-0.30 (0.90)
0.04 (0.02)
-0.30 (0.19)
-0.46 (0.68)
-0.15 (0.92)
0.04 (0.02)
-0.22 (0.26)
0.02 (0.21)
-0.02 (0.85)
0.05 (0.02)
-0.10 (0.28)
-0.38 (0.64)
-0.11 (1.00)
0.07 (0.09)
-0.06 (0.37)
-0.02 (0.23)
0.29 (0.81)
0.04 (0.02)
-0.24 (0.22)
-0.02 (0.53)
0.11 (0.87)
0.05 (0.04)
-0.05 (0.30)
Note: SD = Standard deviation. Descriptive statistics for pooled data across countries reported in Table 2.
Table 4
Descriptive statistics for the dependent variable ‘student achievement’ (5 plausible values), by country.
Plausible values (PV)
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
507.71 (91.16)
507.53 (90.85)
507.84 (90.64)
507.76 (91.25)
508.05 (90.32)
520.38 (101.49)
519.52 (101.13)
519.61 (101.54)
519.92 (101.48)
519.25 (101.97)
442.16 (92.64)
442.38 (92.78)
442.95 (92.52)
442.70 (93.23)
442.16 (93.50)
469.83 (86.70)
469.98 (87.52)
470.10 (88.02)
469.74 (87.50)
470.21 (87.43)
Czech Republic
519.77 (96.69)
520.76 (96.86)
520.12 (96.90)
520.32 (96.90)
519.31 (97.03)
486.19 (86.59)
486.35 (85.71)
486.23 (86.10)
486.39 (86.70)
486.05 (86.25)
521.81 (80.26)
522.36 (81.18)
522.48 (81.01)
522.11 (79.82)
522.95 (81.21)
507.53 (89.86)
506.94 (89.75)
506.94 (89.34)
507.16 (89.45)
507.30 (89.56)
499.47 (96.98)
497.73 (96.58)
498.26 (96.64)
497.97 (96.32)
498.44 (96.58)
513.93 (96.74)
513.79 (96.26)
513.55 (96.94)
514.12 (96.66)
513.97 (96.39)
Great Britain
489.65 (91.12)
489.52 (91.22)
489.55 (91.11)
489.67 (91.30)
490.24 (91.11)
453.89 (87.57)
453.23 (87.97)
453.77 (87.13)
453.07 (87.85)
453.61 (87.81)
485.39 (91.34)
485.19 (91.38)
484.41 (91.27)
484.56 (91.01)
484.79 (90.64)
493.15 (92.38)
492.21 (91.13)
492.62 (91.31)
493.22 (91.61)
493.43 (92.84)
500.90 (84.51)
500.98 (84.20)
501.55 (84.51)
501.32 (84.85)
501.61 (84.78)
495.45 (80.97)
495.70 (80.95)
495.34 (80.31)
495.68 (80.66)
495.52 (81.32)
478.68 (88.70)
479.12 (89.33)
479.58 (88.74)
479.45 (88.82)
479.37 (89.11)
490.27 (95.33)
491.63 (95.67)
490.24 (95.31)
490.20 (96.05)
490.08 (95.05)
518.13 (92.60)
518.11 (92.56)
518.43 (92.58)
518.80 (92.27)
519.22 (92.18)
489.75 (89.84)
489.37 (89.45)
489.12 (90.07)
489.29 (89.68)
489.20 (90.41)
520.59 (91.15)
520.38 (90.75)
520.46 (91.15)
520.37 (91.21)
520.82 (91.23)
484.56 (93.93)
484.89 (93.97)
485.73 (93.99)
485.34 (93.76)
485.09 (93.41)
445.78 (80.36)
444.28 (80.46)
445.48 (80.67)
445.36 (80.60)
445.53 (80.41)
447.74 (89.43)
447.90 (90.18)
447.23 (89.75)
447.29 (89.70)
447.14 (89.99)
485.64 (102.24)
485.49 (101.42)
486.35 (101.85)
485.32 (100.73)
485.54 (102.06)
484.48 (89.50)
484.55 (89.88)
484.33 (89.95)
484.33 (90.28)
484.96 (90.15)
495.36 (88.43)
495.63 (88.67)
495.59 (88.35)
495.25 (88.43)
495.36 (88.50)
479.15 (90.66)
478.79 (91.53)
479.40 (91.32)
479.23 (91.58)
479.62 (90.89)
520.67 (92.62)
521.25 (92.73)
520.94 (92.99)
520.83 (92.80)
521.15 (92.85)
Note: SD = Standard deviation.
4.3. Analytic strategy
First, we apply bivariate analysis to assess the extent to which social segregation is related to
social inequality in student achievement at the macro level of European education systems. In
this analysis, we calculate a country-specific index of social inequality in achievement, which
corresponds to the coefficient of an ordinary least-squares (OLS) regression predicting math
achievement as a function of SES, while controlling for gender, home language, immigrant
background, and school grade (coefficients reported in Table 5). To obtain this index, we
perform an OLS regression (for each country separately), because we focus solely on
individual-level variables, whereas in further analyses we will perform multilevel regressions,
which also include additionalschool- and country-levelvariables. The PISA final student
weights are applied in this regression analysis, resulting in coefficients that are representative
for each country.
Second, we perform multilevel (linear mixed-effects) models to ascertain whether social
segregation moderates individual-level associations between SES and educational achievement.
Multilevel models take into account that the data are hierarchically clusteredstudents in
schools, and schools in countriesmeaning that the observations in the sample cannot be
considered as being independent (Snijders & Bosker, 2012). Standard (OLS) regression models
rely on the assumption of independence of the observations. With a hierarchical data structure,
this assumption is violated and hence the estimates of the standard errors of standard models
will be too small, which may lead to spuriously significant results (Hox, 2010). Multilevel
models allow for the simultaneous estimation of the direct effects of individual-, school-, and
country-level variables on student achievement, as well as to evaluate whether social
segregation within education systems strengthens the micro-level associations between social
origin and student achievement. The final model is represented as:
 
   
  
   
   
(eq. 3)
The educational achievement Y of a student i in school j in country k is estimated as a function
of the overall mean achievement across countries (β000), a vector of individual-level variables
(Xhijk to Xlijk) with their coefficients (βh to βl), a vector of school-level variables (Smjk to Sojk) with
their coefficients (αm to αo), and a vector of country-level variables (Cpk to Cuk) with their
coefficients (δp to δu). The model also includes a vector of cross-level interactions between the
individual-level variable ‘socioeconomic status’ and the country-level variables (Xlijk · Cvk),
with the respective coefficients (γv to γz). Furthermore, by including a random slope μ1jk ~ N(0,
1jk) on ‘socioeconomic status’ (Xlijk) at the school level, the model considers that the
association between socioeconomic status and student achievement differs between schools.
The random slope is determined by a fixed effect for the school average on socioeconomic
status and a random effect that defines the variance in the slopes between schools, as denoted
by the term (β010 + μ1jk) Xlijk, where β010 represents the slope on socioeconomic status (Xlijk) for
the average school and
1jk represents the between-school variance in this slope. Three random
terms are associated with the intercept and fixed effects, reflecting the remaining or residual
variation at the country level, ν0k ~ N(0,
0k), at the school level, μ0jk ~ N(0,
0jk), and at the
student level, ε0ijk ~ N(0,
0ijk). These random terms are assumed to have zero means given the
independent variables, to be drawn from normally distributed populations, and to be mutually
independent. The model allows for a correlation between the school-level variance in math
achievement (random intercept), ν0k, and the random slope on socioeconomic status at the
school level, μ1jk, thereby taking into account that any relationship between socioeconomic
status and student achievement may vary across schools. We use un-centered binary variables
and grand-mean centered continuous variables. There were no collinearity issues in the model,
with all variance inflation factors being below 2.39.
In conclusion, our aim is to exploit variation in the level of social segregation within
education systems across countries to describe systematic patterns of covariation between social
segregation and educational inequality, when observable potential confounders are considered.
Our models do not differentiate statistically between the effects of the systematic underlying
processes that lead to segregated schools (such as the intertwined residential and school choice
decisions of families and schools’ decisions regarding which students to admit), and the effects
of exposure to segregated schools (cf., Leckie et al., 2012). We argue that student achievement
and educational inequality are shaped by both such underlying processes and exposure to
segregated schools. Accordingly, the estimates of the models reported hereafter may be
interpreted as estimates of the combined influence of selection into schools and treatment,
or exposure to socially homogeneous or heterogeneous student populations, in these schools,
under ceteris paribus conditions.
Given the cross-sectional nature of PISA, there is no direct measure of prior student achievement in the dataset.
Following prior research using PISA data, we include school grade at assessment as a rough (in fact, the only
available) proxy for prior student performance, presuming that 15 year olds who were enrolled in lower grades at
5. Results
Table 5 displays the index of social segregation within education systems and the index of social
inequality in achievement for each country.
The index of social segregation varies between 0.090 and 0.456. Greater values indicate
that students of a given socioeconomic-status group were found to a greater degree in distinct
schools and thus isolated from students of a different socioeconomic-status group. Levels of
social segregation were relatively low in Norway, Finland, and Sweden, whereas they were
considerably higher, for instance, in Slovakia, Hungary, and Bulgaria.
The index of social inequality in achievement is a measure of the relationship between
SES and the level of student achievement. It ranges from 15.60 in Spain to 51.67 in the Czech
Republic. That is, on average across European countries, a one-unit increase in SES was related
to a 31.48-point better achievement, or roughly a 0.31 standard-deviation improvement in
achievement. The variation in the indices of social inequality in achievement between countries
contributes to discussion on the degree to which student achievement is a result of inherited
ability, providing the basis for a predisposition to learning, and/or of socialization and
environmental influences (e.g., Nielsen, 2006). The degree of social inequality in achievement
varied considerably across countries, which implies that any predisposition derived from the
family context or otherwise cannot be the sole determinant of educational achievement.
Because country-specific differences in educational inequality cannot be ascribed to any genetic
or social inheritance, the macro environment seemed to play a decisive role in shaping this
the time of the assessment had performed worse in previous years (Chiu, 2010; Lee, Zuze, & Ross, 2005). We
acknowledge the limitations of our approach in Section 6.
Table 5
Index of social segregation within the education system, index of social inequality in
educational achievement.
Index of social
Index of social inequality in
achievement (SE)
Czech Republic
Great Britain
Note: Information about the indices in section 4.2.
The correlation between the index of social segregation within education systems and the index
of social inequality in achievement (r(27) = 0.372, p < .05) provides an estimate of the extent
to which social segregation in education systems was related to social gradients in student
achievement at the aggregate level of European education systems (see also Fig. 1). The
moderate positive relationship identified here supports theory in respect to social class
inequalities in education being more pronounced in those education systems where
socioeconomically diverse students are less evenly distributed across schools.
Figure 1. Scatterplot of the index of social segregation and the index of social
inequality in achievement. Abbreviations: AUT: Austria, BEL: Belgium, BGR:
Bulgaria, CHE: Switzerland, CZE: Czech Republic, DEU: Germany, DNK:
Denmark, ESP: Spain, EST: Estonia, FIN: Finland, FRA: France, GBR: Great
Britain, GRC: Greece, HRV: Croatia, HUN: Hungary, IRL: Ireland, ISL: Iceland,
LTU: Lithuania, LUX: Luxembourg, LVA: Latvia, NLD: Netherlands, NOR:
Norway, POL: Poland, PRT: Portugal, ROU: Romania, SRB: Serbia, SVK:
Slovakia, SVN: Slovenia, SWE: Sweden.
However, the bivariate relationship between social segregation within education systems and
social inequality in achievement does not allow us to gauge whether social segregation
moderates social inequality in educational achievement. Thus, we also estimated a series of
multilevel models to determine whether the strength of the relationship between SES and
educational achievement at the individual level varies across education systems that exhibit
different levels of social segregation, when potential alternative influences are considered at the
individual, school, and country levels.
The dependent variable had complete data for all of the students in the sample; however
three individual-level covariates contained missing valuesschool grade (0.4%), immigrant
status (2.6%), and SES (1.9%). Assuming that the probability of a missing value on these
variables was not conditional on unobserved values of these variables, given the observed
values (Rabe-Hesketh & Skrondal, 2008; Rubin, 1976), we performed multilevel analyses using
full maximum likelihood estimation, which is widely considered to provide robust estimations
if the assumed model is accurate. Furthermore, this allowed us to compare the goodness of fit
of several models through likelihood ratio tests (Muthén & Shedden, 1999; Schafer & Graham,
2002). In robustness analyses, we also replaced missing data with imputed data, computing
maximum likelihood estimates through the expectation-maximization algorithm, which allows
for estimation of parameters in a probabilistic model (Do & Batzoglou, 2008). These additional
analyses confirmed the conclusions that we draw from the analyses presented hereafter.
Table 6 summarizes the results of four increasingly complex multilevel models. The
unconditional (or null) modelwith only intercepts at the individual, school, and country level,
and student achievement as the outcomereveals that 5.6% of the variance in student
achievement was at the country level, whereas 12.7% was at the school-within-country level.
However, variance components between schools and countries can only be reasonably
interpreted when school grade level is considered, given that school grade level explains a large
proportion of the variance in student achievement. Thus, in a quasi-unconditional model (not
shown), which included school grade as the only predictor, we found that 9.9% of the variance
in student achievement was at the country level, whereas 15.3% was at the school-within-
country level. This result implies that student achievement scores varied systematically not only
at the individual level, but also between schools and countries.
Table 6. Multilevel models predicting student achievement.
Model 0
Model 1
Model 2
Model 3
Fixed effects
Individual level
First-generation immigrant
Language spoken at home: same as test language
School grade relative to modal grade
Pre-vocational or vocational program
Socioeconomic status (SES)
School level
School type: private school
Proportion of first-generation immigrants in school
School socioeconomic composition
Country level
Gross domestic product (GDP) per capita
Income inequality: Gini coefficient
Annual taught time in compulsory education
Preschool enrollment rate
Educational expenditure (as % of the GDP)
Social segregation within the education system
Cross-level interactions
SES x GDP per capita
SES x Income inequality
SES x Annual taught time
SES x Preschool enrollment rate
SES x Educational expenditure
SES x Social segregation in the education system
Random effects
Individual-level variance (SD)
School-level variance (SD)
Country-level variance (SD)
Random slope on SES at the school level (SD)
Correlation between the school-level variance (random intercept)
and the random slope on SES
Note: Unstandardized coefficients with standard errors (SE) are reported for the fixed effects. Variances with standard deviations (SD) are reported for the random effects. Maximum-likelihood
estimation was used. The significance of the coefficient estimates of the fixed effects was determined using Wald tests. To partition the variance in student achievement into three components
(at the individual, school and country level), model 0 was calculated with unweighted data. Models 1 to 3 were calculated with weighted data, as explained in section 4.1.
*** p < .001 (two-tailed tests).
In Model 1, we added the individual- and school-level predictors. The results of this model
corroborate findings of earlier studies regarding the statistically significant relationships
between student-level characteristics and educational achievement (Levels et al., 2008; Schlicht
et al., 2010). On average, male students outperformed female students, while immigrant
students underperformed. Moreover, students whose home language corresponded to the PISA
test language, and students who were enrolled in higher grades at school, outperformed their
peers who spoke a foreign language at home and were enrolled in lower school grades,
respectively. The relationship between students’ socioeconomic status and their educational
achievement was positive and highly significant. We modeled between-school variation in the
relationship between socioeconomic status and educational achievement by adding a random
slope on socioeconomic status at the school level. That is, because the relationship between
socioeconomic status and educational achievement varied across schools, we allowed the slope
on socioeconomic status to vary across schools. Including this random slope improved the
model fit significantly, as indicated by a likelihood ratio test based on a comparison of the log-
likelihoods of a model without a random slope and a model with a random slope, χ2 (2, N =
171,159) = 1481.5, p < .001 (see Raudenbush & Bryk, 2002, on likelihood ratio tests to compare
the fit of nested models based on model deviance statistics). At the school level, school type
(public vs. private) and the proportion of first-generation immigrants in school were not
significantly related to student achievement, whereas school socioeconomic composition was.
On average, each one-unit increase in school socioeconomic composition was associated with
a 15.67-point improvement in student achievement, controlling for student socioeconomic
status at the individual level and the other covariates in Model 1. This corresponds roughly to
an increase in achievement of a 0.16 standard deviation. Thus, the average difference in
achievement between students attending the most socioeconomically disadvantaged schools
and students attending the most advantaged schools was approximately 36.97 points, or
approximately a 0.37 standard deviation. Figure 2 further illustrates this relationship by
revealing that the school average achievement level was higher in those schools with more
privileged student populations. This finding does not allow for the conclusion that school
socioeconomic composition necessarily caused an improvement in student achievement, given
that PISA did not assess student ability prior to school entry. The higher achievement levels of
schools that draw a majority of their population from more privileged backgrounds could be a
consequence of greater student ability or of peer effects or a combination of both. Hence, Figure
2 does not provide evidence of a school composition effect (see also Pokropek, 2015), but it
does provide descriptive evidence of a positive association between school socioeconomic
composition and student achievement levels. Note that the box plots are based on pooled data
from all countries included in the study (countries with a greater number of participating schools
contribute more data points to the analysis; each school has equal weight).
Figure 2. Box plots of the distribution of school average achievement across schools
with varying socioeconomic compositions, divided into quintiles. The horizontal line
within the boxes shows the median. The box edges represent the 1st and the 3rd
quartile. The end of the upper whisker equals (Q3 + 1.5 * IQR), the end of the lower
whisker equals (Q1 1.5 * IQR). Observations outside the whiskers are plotted as
In Model 2, we added all of the country-level variables. This model shows that none of these
variables had a statistically significant direct effect on student achievement. This includes a
non-significant main effect of social segregation within the education system, suggesting that
the level of social segregation within an education system was not significantly related to the
average level of student achievement in a country.
In Model 3, we further included the cross-level interactions between SES and the
country-level variables. Although our main focus here is on the interaction between SES and
social segregation, we briefly summarize the findings regarding the other interactions in a first
step, because all of these interactions were statistically significant. They indicate that the
association between SES and student achievement was weaker in countries with a higher GDP
and a longer annual taught time; however, this association increased with income inequality,
preschool enrollment rates, and educational expenditure. The main finding of Model 3 was that
social segregation moderated educational inequality in that the effect of SES on student
achievement was stronger in countries with higher levels of social segregation within the
education system, even when the alternative system-level influences were considered. To assess
the contribution of the moderating effect of social segregation, we performed a likelihood ratio
test, comparing the log-likelihoods of a model that included the cross-level interaction between
socioeconomic status and social segregation, and a model without this interaction term. This
test indicated that adding the cross-level interaction to the model significantly improved the
model fit, χ2 (2, N = 171,159) = 21.7, p < .001. Figure 3 illustrates the interaction between SES
and social segregation, showing how the marginal effect of SES on student achievement
changed as the degree of social segregation within education systems did, when all of the other
variables in the model were kept constant (cf. Preacher, Curran, & Bauer, 2006). The black line
indicates that, on average, a one-unit increase in SES was associated with an increase in student
achievement of approximately 29 points in the least segregated education systems (Norway and
Finland), and of approximately 40 points in the most segregated system (Bulgaria). Expressed
in standard deviation units, an increase in SES by one standard deviation was associated with,
approximately, a 0.29 standard-deviation increase in student achievement in the least
segregated systems, and a 0.40 standard-deviation increase in student achievement in the most
segregated systems. The 95% confidence interval shows that the statistical uncertainty
associated with the coefficients increased slightly as the degree of social segregation within
education systems grew, which can be explained by the smaller number of countries that
exhibited a comparatively high degree of social segregation.
Figure 3. Change in the marginal effect of SES on student achievement as the
degree of social segregation within education systems increases.
The individual-, school-, and country-level variances shown in Table 6 represent the effects of
any unobserved covariates at the respective levels. In line with previous empirical studies and
theory (Dronkers, 2010; Schlicht et al., 2010), the unexplained individual-level variance
remained larger than the unexplained variances at the school and country levels. The weak
positive correlation between the school-level variance (random intercept) and the random slope
on SES at the school level (r = 0.14) indicates that the association between SES and student
achievement was slightly stronger in schools with higher levels of average student achievement;
however, differences between schools were negligible given the weak correlation.
We performed robustness tests to check for omitted variable and overspecification bias,
and for variation in the results when using subsets of countries in the analysis. First, we assessed
the sensitivity of the results to changes in model specification by entering additional, potentially
confounding, country-level covariates: (1) an indicator of whether countries used centrally
administered examinations to test student performance, (2) the proportion of schools that used
assessments in order to compare students with national performance, (3) the variance in parental
education attainment (as an inequality measure), and (4) an index of vocational specificity of
the education system (dual system)using data from PISA, Eurostat (2015), and Bol and Van
de Werfhorst (2013). These variables were not included in the main models because of the
unavailability of data for some of the sampled countries. Second, we ran several models in
which we removed country-level variables because theoretically we might risk overspecifying
our model by including six country-level variables simultaneously, although statistically we did
not identify any multicollinearity issues. Third, we performed a type of cross-validation by
replicating the analysis based on reduced datasets, sequentially excluding (1) one country, or
(2) random pairs of countries (50 combinations), or (3) countries with comparatively strongly
decentralized education policies (Austria, Belgium, Germany, Hungary, Switzerland; cf.,
Schlicht et al., 2010) from each replication. All of these additional tests corroborated the results
reported here and lead to the same conclusions.
6. Discussion
Complementing prior research on correlates of socio-spatial separation of students, this study
assessed to what extent social segregation occurred within education systems in Europe, and it
examined patterns of covariation between social segregation within education systems and
social inequality in educational achievement, using a cross-national comparative design that
considered observable potential confounders at the individual, school, and country levels.
The findings indicate that schools were segregated along socioeconomic lines across
European countries, albeit to varying degrees. Although the extent of social segregation was
comparatively small in Scandinavian countries, it was substantially greater in some Central and
Eastern European countries. For instance, it was approximately five times greater in Bulgaria
than in Norway. Moreover, findings suggest that social segregation within education systems
was related to social inequality in student achievementthe higher the level of social
segregation within an education system, the stronger the aggregate-level relationship between
SES and student achievement in a country. However, social gradients in student achievement
could be the result of inequalities within society at large, or of the economic and education
policy context, rather than the consequence of social segregation within the education system.
We therefore examined whether social segregation in education systems moderates social
inequality in student achievement when such country-level influences are considered (see
Appendix B for a discussion of how the alternative country-level influences moderated
educational inequality).
We found that, ceteris paribus, the effect of SES on student achievement was
significantly stronger in education systems with higher levels of social segregation, suggesting
that social segregation within education systems may contribute to the intergenerational
transmission of educational (dis)advantage. This finding is in line with research from the United
States revealing that spatial inequalities created by social segregation increase achievement
gaps between advantaged and disadvantaged students (Owens, 2018). Moreover, the finding
casts doubt on the view that the consequences of school segregation are “at the limit of our
detectability” (Gorard, 2006, p. 87). Rather, the present investigation of nationally
representative samples in a cross-national design measurably points toward the fact that social
segregation may amplify inequality in educational outcomes. However, the moderating effect
of social segregation on educational inequality was relatively modest. In the least socially
segregated education systems, a one standard-deviation (SD) increase in SES was associated
with an increase in student achievement by approximately 0.29 SDs, whereas in the most
segregated systems it was associated with an increase in achievement of roughly 0.40 SDs. By
way of comparison, this 0.11-SD difference was somewhat smaller than the 0.18-SD difference
in the effect of SES on achievement that was attributable to variations in the annual taught
timein those education systems with the least time spent on teaching per year, a 1-SD increase
in SES was associated with an increase in student achievement by 0.37 SDs, whereas in those
systems with the most time spent on teaching per year, a 1-SD increase in SES was associated
with an increase in achievement by 0.19 SDs, keeping all other covariates constant. However,
the 0.11-SD difference attributable to social segregation was larger than the roughly 0.05-SD
difference that was attributable to variations in economic development (GDP). It was also larger
than the 0.05-SD difference attributable to variations in income inequality (Gini) and the 0.06-
SD difference attributable to variations in preschool enrollment rates; finally, it was comparable
in magnitude with the 0.10-SD difference ascribable to variations in educational expenditure.
Multiple sensitivity analyses confirmed the robustness of our findings (see Section 5).
However, it must be acknowledged that estimates of segregation indices based on sample
surveys tend to be biased upward because they capture both the uneven distribution of students
across schools that results from actual segregation processes (i.e., the systematic underlying
processes of segregation such as school choice decisions and residential choices of families)
and the uneven distribution of students across schools that arises as a result of randomness.
Even if students were allocated to schools completely at random, we would measure some
unevenness in the distribution of diverse students across schools simply as a consequence of
random allocation (Leckie et al., 2012; Ransom, 2000). Consequently, differences in the index
of social segregation between countries are in part the result of sampling variability and must
therefore be interpreted with caution. However, an index of segregation that would measure
deviations from randomness, rather than deviations from evenness in the distribution of students
across schools (Allen, Burgess, Davidson, & Windmeijer, 2015), would lead to a highly similar
ranking of countries by school segregation, given that PISA sampled approximately the same
number of students per school across countries (we have removed Italy from our analyses
because it contained a relatively large proportion of schools in which fewer than 20 students
participated in the survey). We recognize that for countries with a large average school size the
index of social segregation between schools is expected to be smaller because larger schools
will lead to smaller socioeconomic differences between schools, and greater differences within
However, any measure of segregation that is based on a sample survey, such as PISA,
is subject to sampling variation, and previous research has demonstrated that the number of
schools sampled per country in the PISA survey is sufficiently large to minimize bias to
negligible levels (Jenkins et al., 2008).
It should not be disregarded that social segregation within education systems may be a
consequence of residential segregation, for instance, where particular schools are in more
affluent catchment areas, while others are found in districts with a high level of social housing
(Croxford & Paterson, 2006; Dupriez & Dumay, 2006; Ferrer-Esteban, 2016). Moreover, the
reputation of the school may well give rise to residential segregation, with property markets
responding to demand from families (Gorard, 2000; Kane, Riegg, & Staiger, 2006; Leech &
Campos, 2001). Thus, the relationship between school social segregation and residential
segregation may be theoretically conceived of as a reciprocal relationship of mutual
determination between school and housing markets (Taylor & Gorard, 2001). There are
currently no standardized cross-national data that could be cross-referenced at a European level
with the data on the schools from the PISA survey. Our research therefore cannot distinguish
between residential and school segregation. The essential question that it does address,
however, is whether the clustering of children along social background linesas observed in
education systemsstrengthens the relationship between social origin and educational
achievement, with the results indicating that this is the case.
Finally, and as previously mentioned, it is important to be aware that effects of social
segregation between schools on social inequality in achievement may be mediated by school
characteristics, such as academic entry requirements or overall ability levels (e.g., Harris &
Williams, 2012; Liu, Van Damme, Gielen, & Van Den Noortgate, 2015). Experimental
Analyses in which the countries with the largest average school sizes were excluded (LUX, NLD, ROU, GBR;
Eurydice, 2012) provided evidence of a slightly stronger interaction effect between socioeconomic status and
social segregation within education systems than was the case with the interaction effect presented here.
longitudinal studies would provide the opportunity to examine causal mediation effects.
However, such studies may pose significant ethical challenges and are therefore not necessarily
feasible. With this in mind, the value of the PISA data for cross-national comparative analyses
is substantial. Parallel data extending across European countries are rare. Thus, the standardized
international assessments provide unique data for analyzing educational inequalities, where
otherwise only smaller and non-representative samples were available (Hanushek &
Wössmann, 2014). These large-scale assessments allow for exploring variation that exists only
across countries. Even if the degree of social segregation may vary across areas within
countries, variation between European countries is considerable and therefore particularly
worthy of investigation (Fig. 1 and Tab. 5). In conclusion, and in the absence of longitudinal or
experimental data, the current study provides robust descriptive evidence in support of theory
that social segregation within European education systems is detrimental to equity in education.
7. Conclusion
Extending research on the geography of opportunity (Logan, Minca, & Adar, 2012), this cross-
national comparative study shows that the degree of social segregation within education
systems varied considerably across European countries. It also highlights a relationship at the
system level between social segregation and the degree of social inequality in student
achievement. The average level of student achievement in a country was not affected by the
level of social segregation within the education system. However, the effects of social origin
on student achievement were stronger in more socially segregated systems, although the
respective differences between systems were relatively small. These findings provide new
evidence of the potentially damaging effect of a socio-spatial separation of students, indicating
that socioeconomic segregation in European education systems may contribute to some extent
to the perpetuation of educational and, by extension, social disadvantage from one generation
to the next.
Research ethics
This study involves analysis of publicly available, de-identified data. Any information
presented here is such that study participants cannot be identified.
Conflicts of interest
Appendix A
The scatter plots in Figure A.1 and Figure A.2 illustrate that the degree of social segregation in
European education systems is to some extent related to the number of years that children spent
in a tracked regime (Fig. A.1) and to the number of tracks that are implemented in a given
system (Fig. A.2). However, there is also considerable variation in the degree of social
segregation among education systems that use similar or even identical tracking regimes. Note
that the data on the tracking regimes are derived from Eurydice (2010) which provides official
information about the structure of European education systems. The Organization for Economic
Co-operation and Development (OECD, 2013) reports data that differ slightly for certain
countries (e.g., 4 tracks in Germany, 3 tracks in Hungary). Analyses based on OECD data
confirm the results presented in this article and lead to the same conclusions.
Figure A1.
Figure A2.
Appendix B
All cross-level interactions in our model were statistically significant, indicating that all
country-level variables moderated the individual-level relationship between socioeconomic
status and student achievement (i.e., educational inequality). First, levels of educational
inequality were lower in countries with a higher GDP. This supports the theory that social
inequality in educational outcomes declines with economic development (Marks, 2009),
suggesting that in the context of economic growth a transition occurs “from ascriptive rules of
social mobility to mobility patterns based on personal achievements and meritocratic ideas”
(van Doorn et al., 2011, p. 97). Second, income inequality was positively associated with the
degree of educational inequality, which ties in with recent findings from a study of selected
OECD countries (Chmielewski & Reardon, 2016). Although policy documents emphasize the
role of education in “breaking the link between socioeconomic background and life prospects”
(OECD, 2012, p. 18), school systems in Europe seem to face significant challenges in breaking
this link. Instead, they might even reproduce and exacerbate pre-existing family income-related
inequality between children (cf., Downey & Condron, 2016). Third, educational inequality was
weaker in countries with a longer annual taught time. This corroborates prior research whereby
a more intense schooling may diminish socioeconomic differentials in educational outcomes
(Schlicht, Stadelmann-Steffen, & Freitag, 2010). Fourth, the preschool enrollment rate was
positively associated with the degree of educational inequality. This unexpected finding might
be explained by socioeconomic differentials in the duration of preschool attendance across
Europe, with children of higher-SES families attending preschool for longer periods of time
(authors, 2016). As a consequence, a higher preschool enrollment rate may increase, rather than
decrease, social inequality in educational outcomes. Fifth, a higher level of educational
inequality was observed in countries with greater educational expenditure. This result
challenges the view that an increase in public spending on education might prevent the
occurrence of educational inequality. Instead, it suggests that public expenditure on education
may benefit in particular socioeconomically advantaged students who are potentially better able
to capitalize on public education. Finally, the degree of educational inequality was stronger in
countries with a higher level of social segregation in the education system, as explained in detail
in the article.
Agirdag, O., Van Avermaet, P., & van Houtte, M. (2013). School segregation and math
achievement: A mixed-method study on the role of self-fulfilling prophecies. Teachers
College Record, 115(3), 150.
Alegre, M. A., & Ferrer, G. (2010). School regimes and education equity: Some insights
based on PISA 2006. British Educational Research Journal, 36(3), 433461.
Allen, R., Burgess, S., Davidson, R., & Windmeijer, F. (2015). More reliable inference for the
dissimilarity index of segregation. The Econometrics Journal, 18(1), 4066.
Ammermüller, A. (2005). Educational opportunities and the role of institutions (Discussion
Paper No. 0544). Mannheim: ZEW. Retrieved from
Belfi, B., Goos, M., Pinxten, M., Verhaeghe, J. P., Gielen, S., Fraine, B., & van Damme, J.
(2014). Inequality in language achievement growth? An investigation into the impact
of pupil socio-ethnic background and school socio-ethnic composition. British
Educational Research Journal, 40(5), 820846.
Benito, R., Alegre, M. À., & Gonzàlez-Balletbò, I. (2014). School segregation and its effects
on educational equality and efficiency in 16 OECD comprehensive school systems.
Comparative Education Review, 58(1), 104134.
Bernelius, V., & Vaattovaara, M. (2016). Choice and segregation in the ‘most egalitarian’
schools: Cumulative decline in urban schools and neighbourhoods of Helsinki,
Finland. Urban studies, 53(15), 31553171.
Böhlmark, A., Holmlund, H., & Lindahl, M. (2016). Parental choice, neighbourhood
segregation or cream skimming? An analysis of school segregation after a generalized
choice reform.Journal of Population Economics, 29(4), 11551190.
Bol, T., & Van de Werfhorst, H. G. (2013). Educational systems and the trade-off between
labor market allocation and equality of educational opportunity. Comparative
Education Review, 57(2), 285308.
Borman, G. D., & Dowling, M. (2010). Schools and inequality: A multilevel analysis of
Coleman’s equality of educational opportunity data. Teachers College Record, 112,
Brunello, G., & Checchi, D. (2007). Does school tracking affect equality of opportunity? New
international evidence. Economic Policy, 22(52), 781861.
Burger, K. (2010). How does early childhood care and education affect cognitive
development? An international review of the effects of early interventions for children
from different social backgrounds. Early Childhood Research Quarterly, 25(2), 140
Burger, K. (2013). Early childhood care and education and equality of opportunity.
Theoretical and empirical perspectives on social challenges. Wiesbaden, London:
Burger, K. (2015). Effective early childhood care and education: Successful approaches and
didactic strategies for fostering child development. European Early Childhood
Education Research Journal, 23(5), 743-760.
Burger, K. (2016a). Intergenerational transmission of education in Europe: Do more
comprehensive education systems reduce social gradients in student achievement?
Research in Social Stratification and Mobility, 44, 54-67.
Burger, K. (2016b). A transdisciplinary approach to research on early childhood education.
GAIA Ecological Perspectives for Science and Society, 25(3), 197-200.
Burger, K. & Walk, M. (2016). Can children break the cycle of disadvantage? Structure and
agency in the transmission of educational advantage across generations. Social
Psychology of Education, 19(4), 695-713.
Campbell, M., Haveman, R., Sandefur, G., & Wolfe, B. (2005). Economic inequality and
educational attainment across a generation. Focus, 23(3), 1115.
Castellano, K. E., Rabe-Hesketh, S., & Skrondal, A. (2014). Composition, context, and
endogeneity in school and teacher comparisons. Journal of Educational and
Behavioral Statistics, 39(5), 333367.
Cebolla-Boado, H., Radl, J., & Salazar, L. (2017). Preschool education as the great equalizer?
A cross-country study into the sources of inequality in reading competence. Acta
Sociologica, 60(1), 4160.
Chiu, M. M. (2010). Effects of inequality, family and school on mathematical achievement:
Country and student differences. Social Forces, 88(4), 16451676.
Chiu, M. M. (2015). Family inequality, school inequalities, and mathematics achievement in
65 countries: Microeconomic mechanisms of rent seeking and diminishing marginal
returns. Teachers College Record, 117(1), 132.
Chmielewski, A. K. (2014). An international comparison of achievement inequality in within-
and between-school tracking systems. American Journal of Education, 120(3), 293
Chmielewski, A. K., & Reardon, S. F. (2016). Patterns of cross-national variation in the
association between income and academic achievement. AERA Open, 2(3), 127.
Crosnoe, R., & Muller, C. (2014). Family socioeconomic status, peers, and the path to
college. Social Problems, 61(4), 602624.
Croxford, L., & Paterson, L. (2006). Trends in social class segregation between schools in
England, Wales and Scotland since 1984. Research Papers in Education, 21(4), 381
Davis-Kean, P. E. (2005). The influence of parent education and family income on child
achievement: The indirect role of parental expectations and the home environment.
Journal of Family Psychology, 19(2), 294304.
De Fraja, G., & Martinez Mora, F. (2012). The desegregating effect of school tracking (CEPR
Discussion Paper No. 9204). Retrieved from
Do, C. B., & Batzoglou, S. (2008). What is the expectation maximization algorithm? Nature
Biotechnology, 26(8), 897899.
Downey, D. B., & Condron, D. J. (2016). Fifty years since the Coleman report: Rethinking
the relationship between schools and inequality. Sociology of Education, 89(3), 207
Driessen, G. (2002). School composition and achievement in primary education: A large-scale
multilevel approach. Studies in Educational Evaluation, 28, 347368.
Dronkers, J. (2010). Features of educational systems as factors in the creation of unequal
educational outcomes. In J. Dronkers (Ed.), Quality and inequality of education.
Cross-national perspectives (pp. 299327). Dordrecht, New York: Springer.
Dumay, X., & Dupriez, V. (2008). Does the school composition effect matter? Evidence from
Belgian data. British Journal of Educational Studies, 56(4), 440477.
Duncan, O. D., & Duncan, B. (1955). A methodological analysis of segregation indices.
American Sociological Review, 20(2), 210217.
Dupriez, V., & Dumay, X. (2006). Inequalities in school systems: Effect of school structure or
of society structure? Comparative Education, 42(2), 243260.
Esser, H., & Relikowski, I. (2015). Is ability tracking (really) responsible for educational
inequalities in achievement? A comparison between the country states Bavaria and
Hesse in Germany (No. IZA DP No. 9082). Bonn: IZA.
Eurostat. (2015). Inequality of income distribution - Eurostat. Retrieved from
Eurostat. (2017). Expenditure on education as % of GDP or public expenditure. Brussels:
European Commission. Retrieved from
Eurostat. (2018). Gini coefficient of equivalised disposable income. Brussels: European
Commission. Retrieved from
Eurydice. (2010). The structure of the European education systems 2010/11. Brussels:
Eurydice. (2012). Key data on education in Europe. Brussels: European Commission.
Eurydice. (2013). Recommended annual taught time in full-time compulsory education in
Europe 2012/13. Brussels: EACEA.
Fekjær, S. N., & Birkelund, G. E. (2007). Does the ethnic composition of upper secondary
schools influence educational achievement and attainment? A multilevel analysis of
the Norwegian case. European Sociological Review 23(3), 30923.
Felouzis, G., & Charmillot, S. (2013). School tracking and educational inequality: A
comparison of 12 education systems in Switzerland. Comparative Education, 49(2),
Ferreira, F. H. G., & Gignoux, J. (2014). The measurement of educational inequality:
Achievement and opportunity. The World Bank Economic Review, 28(2), 210246.
Ferrer-Esteban, G. (2016). Trade-off between effectiveness and equity? An analysis of social
sorting between classrooms and between schools. Comparative Education Review,
60(1), 151183.
Goldstein, H., & Noden, P. (2003). Modelling social segregation. Oxford Review of
Education, 29(2), 225237.
Gorard, S. (2000). Education and social justice. The changing composition of schools and its
implications. Cardiff: University of Wales Press.
Gorard, S. (2006). Is there a school mix effect? Educational Review, 58(1), 8794.
Gorard, S., & Taylor, C. (2002). What is segregation? A comparison of measures in terms of
“strong” and “weak” compositional invariance. Sociology, 36(4), 875895.
Gustafsson, J.-E., Nilsen, T., & Yang Hansen, K. (2018). School characteristics moderating
the relation between student socio-economic status and mathematics achievement in
grade 8. Evidence from 50 countries in TIMSS 2011. Studies in Educational
Evaluation, 57, 1630.
Hanushek, E. A., & Wössmann, L. (2014). Institutional structures of the education system and
student achievement: A review of cross-country economic research. In R. Strietholt,
W. Bos, J.-E. Gustafsson, & M. Rosén (Eds.), Educational policy evaluation through
international comparative assessments (pp. 145176). Münster: Waxmann.
Hanushek, Eric A., Kain, J. F., Markman, J. M., & Rivkin, S. G. (2003). Does peer ability
affect student achievement? Journal of Applied Econometrics, 18(5), 527544.
Harris, D., & Williams, J. (2012). The association of classroom interactions, year group and
social class. British Educational Research Journal, 38(3), 373397.
Heyneman, S. P., & Loxley, W. A. (1983). The effect of primary-school quality on academic
achievement across twenty-nine high- and low-income countries. American Journal of
Sociology, 88(6), 11621194.
Holtmann, A. C. (2016). Excellence through equality of opportunity Increasing the social
inclusiveness of education systems benefits disadvantaged students without harming
advantaged students. In H.-P. Blossfeld, S. Buchholz, J. Skopek, & M. Triventi (Eds.),
Models of secondary education and social inequality: An international comparison
(pp. 6176). Cheltenham: Edward Elgar Publishing.
Hox, J. (2010). Multilevel analysis: Techniques and applications (2nd ed.). New York:
Hutchens, R. (2004). One measure of segregation. International Economic Review, 45(2),
Jackson, M., Erikson, R., Goldthorpe, J. H., & Yaish, M. (2007). Primary and secondary
effects in class differentials in educational attainment: The transition to A-Level
courses in England and Wales. Acta Sociologica, 50(3), 211229.
Jenkins, S. P., Micklewright, J., & Schnepf, S. V. (2008). Social segregation in secondary
schools: How does England compare with other countries? Oxford Review of
Education, 34(1), 2137.
Jerrim, J., & Macmillan, L. (2015). Income Inequality, Intergenerational Mobility, and the
Great Gatsby Curve: Is Education the Key? Social Forces, 94(2), 505533.
Kane, T. J., Riegg, S. K., & Staiger, D. O. (2006). School quality, neighborhoods, and
housing prices. American Law and Economics Review, 8(2), 183212.
Kariya, T., & Rosenbaum, J. E. (1999). Bright flight: Unintended consequences of detracking
policy in Japan. American Journal of Education, 107(3), 210230.
Kearney, M. S., & Levine, P. B. (2014). Income inequality, social mobility, and the decision
to drop out of high school (Working Paper No. 20195). National Bureau of Economic
Lareau, A., & Weininger, E. B. (2003). Cultural capital in educational research: A critical
assessment. Theory and Society, 32(5/6), 567606.
Lauen, D. L., & Gaddis, S. M. (2013). Exposure to classroom poverty and test score
achievement: Contextual effects or selection? American Journal of Sociology, 118(4),
Lavy, V., Paserman, M. D., & Schlosser, A. (2012). Inside the black box of ability peer
effects: Evidence from variation in the proportion of low achievers in the classroom.
The Economic Journal, 122(559), 208237.
Leckie, G., Pillinger, R., Jones, K., & Goldstein, H. (2012). Multilevel modeling of social
segregation. Journal of Educational and Behavioral Statistics, 37(1), 330.
Lee, J.-S., & Bowen, N. K. (2006). Parent involvement, cultural capital, and the achievement
gap among elementary school children. American Educational Research Journal,
43(2), 193218.
Lee, V. E., & Burkam, D. T. (2002). Inequality at the starting gate: Social background
differences in achievement as children begin school. Washington, D.C.: Economic
Policy Institute.
Lee, V. E., Zuze, T. L., & Ross, K. N. (2005). School effectiveness in 14 sub-Saharan African
countries: Links with 6th graders’ reading achievement. Studies in Educational
Evaluation, 31(23), 207246.
Leech, D., & Campos, E. (2001). Is comprehensive education really free? A case study of the
effects of secondary school admissions policies of house prices in one local area
(Warwick Economic Research Paper No. 581). Coventry: University of Warwick.
Levels, M., Dronkers, J., & Kraaykamp, G. (2008). Immigrant children’s educational
achievement in western countries: Origin, destination, and community effects on
mathematical performance. American Sociological Review, 73(5), 835853.
Liu, H., Van Damme, J., Gielen, S., & Van Den Noortgate, W. (2015). School processes
mediate school compositional effects: model specification and estimation. British
Educational Research Journal, 41(3), 423447.
Lockheed, M., Prokic-Bruer, T., & Shadrova, A. (2015). The experience of middle-income
countries participating in PISA 2000-2015. Paris: Organisation for Economic Co-
operation and Development.
Logan, J. R., Minca, E., & Adar, S. (2012). The geography of inequality: Why separate means
unequal in American public schools. Sociology of Education, 85(3), 287301.
Lucas, S. R. (2001). Effectively maintained inequality: Education transitions, track mobility,
and social background effects. American Journal of Sociology, 106(6), 16421690.
Maaz, K., Trautwein, U., Lüdtke, O., & Baumert, J. (2008). Educational transitions and
differential learning environments: How explicit between-school tracking contributes
to social inequality in educational outcomes. Child Development Perspectives, 2(2),
Marcinczak, S., Musterd, S., van Ham, M., & Tammaru, T. (2016). Inequality and rising
levels of socio-economic segregation: Lessons from a pan-European comparative
study. In T. Tammaru, S. Marcinczak, M. van Ham, & S. Musterd (Eds.), East meets
West: New perspectives on socio-economic segregation in European capital cities (pp.
358382). London: Routledge.
Marks, G. N. (2009). Modernization theory and changes over time in the reproduction of
socioeconomic inequalities in Australia. Social Forces, 88(2), 917944.
Mayer, S. E. (2002). How economic segregation affects children’s educational attainment.
Social Forces, 81(1), 153176.
McPherson, A., & Willms, J. D. (1987). Equalisation and improvement: Some effects of
comprehensive reorganisation in Scotland. Sociology, 21(4), 509539.
Modin, B., Karvonen, S., Rahkonen, O., & Östberg, V. (2015). School performance, school
segregation, and stress-related symptoms: Comparing Helsinki and Stockholm. School
Effectiveness and School Improvement, 26(3), 467486.
Musterd, S., Marcińczak, S., van Ham, M., & Tammaru, T. (2017). Socioeconomic
segregation in European capital cities. Increasing separation between poor and rich.
Urban Geography, 38(7), 10621083.
Muthén, B., & Shedden, K. (1999). Finite mixture modeling with mixture outcomes using the
EM algorithm. Biometrics, 55(2), 463469.
Neuenschwander, M. P., Vida, M., Garrett, J. L., & Eccles, J. S. (2007). Parents’ expectations
and students’ achievement in two western nations. International Journal of Behavioral
Development, 31(6), 594602.
Nielsen, F. (2006). Achievement and ascription in educational attainment: Genetic and
environmental influences on adolescent schooling. Social Forces, 85(1), 193216.
OECD. (2009a). PISA 2006 technical report. Paris: OECD Publishing.
OECD. (2009b). PISA data analysis manual. Paris: OECD Publishing.
OECD. (2011). Education at a glance 2011. Paris: OECD Publishing.
OECD. (2012). Equity and quality in education: Supporting disadvantaged students and
schools. Paris: OECD Publishing.
OECD. (2013). PISA 2012 assessment and analytical framework. Paris: OECD Publishing.
OECD. (2014). PISA 2012 technical report. Paris: OECD Publishing.
Opdenakker, M.-C., & van Damme, J. (2007). Do school context, student composition and
school leadership affect school practice and outcomes in secondary education? British
Educational Research Journal, 33(2), 179206.
Owens, A. (2018). Income segregation between school districts and inequality in students’
achievement. Sociology of Education, 91(1), 127.
Palardy, G. J. (2013). High school socioeconomic segregation and student attainment.
American Educational Research Journal, 50(4), 714754.
Palardy, G. J., Rumberger, R. W., & Butler, T. (2015). The effect of high school
socioeconomic, racial, and linguistic segregation on academic performance and school
behaviors. Teachers College Record, 117(12), 152.
Pépin, L. (2011). Education in the Lisbon Strategy: Assessment and prospects. European
Journal of Education, 46(1), 2535.
Pfeffer, F. T. (2008). Persistent inequality in educational attainment and its institutional
context. European Sociological Review, 24(5), 543565.
Pfeffer, F. T. (2015). Equality and quality in education. A comparative study of 19 countries.
Social Science Research, 51, 350368.
Pokropek, A. (2015). Phantom effects in multilevel compositional analysis problems and
solutions. Sociological Methods & Research, 44(4), 677705.
Preacher, K. J., Patrick J. Curran, & Bauer, D. J. (2006). Computational tools for probing
interactions in multiple linear regression, multilevel modeling, and latent curve
analysis. Journal of Educational and Behavioral Statistics, 31(4), 437448.
Quillian, L. (2014). Does segregation create winners and losers? Residential segregation and
inequality in educational attainment. Social Problems, 61(3), 402426.
Rabe-Hesketh, S., & Skrondal, A. (2008). Multilevel and longitudinal modeling using Stata
(2nd ed.). College Station, TX: Stata Press.
Ransom, M. R. (2000). Sampling distributions of segregation indexes. Sociological Methods
& Research, 28(4), 454475.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods (2nd ed.). Thousand Oaks: SAGE.
Raudenbush, S. W., & Willms, J. D. (1995). The estimation of school effects. Journal of
Educational and Behavioral Statistics, 20(4), 307335.
Reardon, S. F. (2011). The widening academic achievement gap between the rich and the
poor: New evidence and possible explanations. In G. J. Duncan & R. J. Murnane
(Eds.), Whither Opportunity? Rising Inequality, Schools, and Children’s Life Chances
(pp. 91116). New York: Russell Sage Foundation.
Reardon, S. F., & Bischoff, K. (2011). Income inequality and income segregation. American
Journal of Sociology, 116(4), 10921153.
Reardon, S. F., & Owens, A. (2014). 60 years after Brown: Trends and consequences of
school segregation. Annual Review of Sociology, 40(1), 199218.
Rubin, D. B. (1976). Inference and missing data. Biometrika, 63(3), 581592.
Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. New York: John Wiley
& Sons.
Rumberger, R. W., & Palardy, G. P. (2005). Does segregation still matter? The impact of
student composition on academic achievement in high school. Teachers College
Record, 107(9), 19992045.
Saporito, S., & Sohoni, D. (2007). Mapping educational inequality: Concentrations of poverty
among poor and minority students in public schools. Social Forces, 85(3), 12271253.
Schafer, J. L., & Graham, J. W. (2002). Missing data: our view of the state of the art.
Psychological Methods, 7(2), 147177.
Schlicht, R., Stadelmann-Steffen, I., & Freitag, M. (2010). Educational inequality in the EU:
The effectiveness of the national education policy. European Union Politics, 11(1),
Schmidt, M. G. (2004). Die öffentlichen und privaten Bildungsausgaben Deutschlands im
internationalen Vergleich. Zeitschrift für Staats- und Europawissenschaften, 2(1), 7
Schütz, G., Ursprung, H. W., & Wössmann, L. (2008). Education policy and equality of
opportunity. Kyklos, 61(2), 279308.
Snijders, T. A. B., & Bosker, R. J. (2012). Multilevel analysis: An introduction to basic and
advanced multilevel modeling (2nd ed). Los Angeles: Sage.
Sortkær, B. (2018). Feedback for everybody? Variations in students’ perception of
feedback. In Nordic Council of Ministers (Ed.), Northern Lights on TIMSS and PISA
2018. (pp. 161-182). Copenhagen: Nordic Council of Ministers.
Stadelmann-Steffen, I. (2012). Education policy and educational inequality - Evidence from
the Swiss laboratory. European Sociological Review, 28(3), 379393.
Strand, S. (2010). Do some schools narrow the gap? Differential school effectiveness by
ethnicity, gender, poverty, and prior achievement. School Effectiveness and School
Improvement, 21(3), 289314.
Taylor, C., & Gorard, S. (2001). The role of residence in school segregation: Placing the
impact of parental choice in perspective. Environment and Planning A, 33(10), 1829
Televantou, I., Marsh, H. W., Kyriakides, L., Nagengast, B., Fletcher, J., & Malmberg, L.-E.
(2015). Phantom effects in school composition research: Consequences of failure to
control biases due to measurement error in traditional multilevel models. School
Effectiveness and School Improvement, 26(1), 75101.
UNESCO. (2011). World data on education [IBE/2011/CP/WDE/RB]. Geneva: IBE.
van Doorn, M., Pop, I., & Wolbers, M. H. J. (2011). Intergenerational transmission of
education across European countries and cohorts. European Societies, 13(1), 93117.
Van Ewijk, R., & Sleegers, P. (2010). The effect of peer socioeconomic status on student
achievement: A meta-analysis. Educational Research Review, 5(2), 134150.
Waldinger, F. (2006). Does tracking affect the importance of family background on students’
test scores? Unpublished manuscript, LSE. Retrieved from
Wu, M. (2005). The role of plausible values in large-scale assessments. Studies in
Educational Evaluation, 31(23), 114128.
Yaish, M., & Andersen, R. (2012). Social mobility in 20 modern societies: The role of
economic and political context. Social Science Research, 41(3), 527538.
Yang Hansen, K., & Gustafsson, J.-E. (2016). Causes of educational segregation in Sweden
school choice or residential segregation. Educational Research and Evaluation, 22(1
2), 2344.
Yang Hansen, K., & Gustafsson, J.-E. (in press). Identifying the key source of deteriorating
educational equity in Sweden between 1998 and 2014. International Journal of
Educational Research,
Yang Hansen, K., Gustafsson, J.-E., & Rosén, M. (2014). School performance differences and
policy variations in Finland, Norway and Sweden. In K. Yang Hansen, J.-E.
Gustafsson, M. Rosén, S. Sulkunen, K. Nissinen, P. Kupari, … A. Hole (Eds.),
Northern Lights on TIMSS and PIRLS 2011. Differences and similarities in the Nordic
countries (pp. 2548). Copenhagen: Nordic Council of Ministers.
Yang Hansen, K., Rosén, M., & Gustafsson, J.-E. (2011). Changes in the multi‐level effects
of socio‐economic status on reading achievement in Sweden in 1991 and 2001.
Scandinavian Journal of Educational Research, 55(2), 197211.
... It may also result as a consequence of selective (i. e., tracked) education systems, ability grouping, or other reasons, such as neighborhood segregation or increases in school choice and school marketization (Burger, 2019;Strello et al., 2021;Strietholt et al., 2019). Because policies which change the composition of students are also likely to change the composition of teachers (Jackson, 2009), controlling for socioeconomic student segregation between schools should capture unobserved policy changes which are correlated with teacher sorting. ...
... We compute the intra-class correlation coefficient (ICC) of student SES, using a multi-level modelling approach well-established in previous work (i.e. Burger, 2019;Goldstein & Noden, 2003;Raudenbush & Bryk, 2002). The ICC denotes the proportion of variance at the student level which can be attributed to the clustering of schools (or the ratio of school-level SES variance to total SES variance in a country). ...
... The analysis finds consistent evidence that social segregation is more detrimental for mathematics achievement inequity than sorting across both dimensions. This is in line with other research on social segregation and student outcomes (Hanushek & Woessmann, 2006;Burger, 2019;Strello et al., 2021;Teltemann & Schunk, 2019;Sacerdote, 2011). Education systems have thus far largely failed to adequately address social segregation (Gutiérrez et al., 2019), and our analysis provides yet more evidence that it is harming the most vulnerable students. ...
Full-text available
Recent and older studies have reported either a persistence or a widening of the socioeconomic achievement gap-the difference in performance between students in top and bottom socioeconomic groups. Using a panel data technique with country fixed effects for 32 education systems and six waves of data from the Trends in International Mathematics and Science Study, we examine whether the sorting of teachers by specialization level in mathematics education and novice status across students of different socioeconomic backgrounds exacerbates mathematics achievement inequity despite the presence of a time-varying control for socioeconomic school segregation. We find modest evidence that sorting by mathematics education is associated with achievement inequity, but no evidence supporting the importance of sorting based on teacher experience. Socioeconomic school segregation, on the other hand, clearly and persistently exacerbates achievement inequity. The results have policy implications regarding the effective distribution of educational resources.
... In some cases, the racialized condition, for instance, can be a more powerful disadvantage, toward the struggle for equity, within the same economic and social strata. Indeed, belonging to a low socioeconomic level, just like being born to a minority group, being an immigrant or of immigrant descent, may, among other factors, become a predictor of lower educational achievement and higher school dropout (Sirin, 2005;Gonçalves and França, 2008;Coimbra and Fontaine, 2015;García and Weiss, 2017;Burger, 2019). These intricate roots make it more challenging to evaluate the foundations of inequalities and generate a sustainable change to this phenomenon. ...
Full-text available
Despite continuous eorts, the educational achievement gap is still, in most societies, a significant obstacle to ensuring more equity and social justice. Much of this inequality derives from belonging to historically discriminated groups. Indeed, coming from a lower socioeconomic status (SES), of an immigrant, or descendant situation, being Black, Hispanic, Gypsy, or any other racialized condition, still strongly influences academic attainment, school dropout and career choices. However, many innovative strategies and policies have been implemented to minimize this bias. This investigation proposes to gather, assess, and analyze these most recent interventions and perceive which of these present a better level of e�cacy. Using the PRISMA guidelines, this Systematic Review of Literature yielded 27 studies that fit the inclusion criteria. The analysis considered the level of e�cacy, intervention method and scope. Results show that targeted strategies, such as working on reading abilities and school subjects’ focused interventions are more eective in improving minorities’ and lower SES students’ attainment. Other beneficial initiatives include whole-school, state and community-based projects, innovative pedagogies, and, finally, programs that deal with the psycho-social consequences of racism and discrimination, e.g., the internalization of negative perceptions and expectations. Overall, there is a strong need to develop mixed-method and longitudinal designs that will further our knowledge about what type of measure works, while considering a situated and contextual perspective, instead of a one-size-fits-all approach.
... The considerable research on both between-and within-school stratification and its effect on student inequity allow us to make some relatively straightforward predictions about how it will relate to teacher sorting (Burger, 2019;Parker et al., 2016;Strello, Strietholt, Steinmann, & Siepmann, 2021). Both between-school tracking (academic vs. vocational) and within-school tracking (ability grouping between classes) has clearly and consistently been shown to increase performance gaps between low-and high-SES students as well as other forms of social inequity (Strello, Strietholt, Steinmann, & Siepmann, 2021). ...
... Almost all education systems allocate students to distinct educational tracks that differ in terms of academic requirements. This purposive clustering of students in classes within schools or in different types of schools is known as tracking, but it has also been referred to as ability grouping, sorting, or differentiation (Becker et al., 2012;Burger, 2016Burger, , 2019aChiu et al., 2017;Dockx et al., 2019). Advocates argue that tracking allows for effective and appropriately paced instruction that is ideally adapted to students' skills and needs, enabling maximum learning for all students. ...
... International large-scale assessments appeared as principal data sources, and authors frequently employed secondary data analysis. Herein, PISA, the Programme for International student Assessment by the OECD, figured prominently (e.g., Burger 2019;Chiu 2015;Julia 2016;Özdemir 2016;Santibañez and Fagioli 2016;Zhu 2018bZhu , 2018a, and so did TIMSS, the Trends in International ...
This thesis is an exploration of questions pertaining to the legitimacy of International Education Agendas (IEAs) with two main contributions. Firstly, I provide a conceptually sound and empirically viable approach to assess IEAs’ legitimacy. And secondly, I investigate what can be learnt about the legitimacy of IEAs when this approach is put into practice. I thereby focus on a sociological understanding of legitimacy which is fundamentally concerned with how global governance arrangements are perceived by an identified audience. To define IEAs I introduce the distinction between the formal core of an IEA, such as UN resolutions, and the formal interpretive frames, such as global monitoring reports and thematic indicators frameworks. Specifically, my research questions focus on the legitimacy beliefs held by academics towards equity in education, as an ideational feature of the current IEA, the Education 2030 – SDG 4 agenda. I argue in my thesis that academics can be considered an audience of particular relevance to IEAs, and likewise, equity in education arguably constitutes a critical ideational feature of SDG 4. Furthermore, the present thesis recognizes the importance of ideational competition in global governance dynamics and identifies IEAs as one of the loci where this ideational competition happens. In response to gaps in the existing literature, a heuristic conceptual framework for the analysis of legitimacy beliefs towards IEAs is introduced. Building on this conceptual framework, a blended reading research design is employed to investigate the congruence between, on the one hand, academics’ perceptions of the conceptualization of equity in education embedded in IEA, and, on the other hand, those conceptualizations that are found in the scholarship on equity in education. Eventually, three specific dimensions of textual data were considered, namely, representation, structure, and knowledge, in order to allow for a comprehensive understanding of the scholarship on equity in education. Findings point to the divergence between conceptualizations of equity in education that emerge from the formal interpretive frames of the present IEA, and those that are embedded in the initial wording of the UN resolution, as well as the most prominent conceptualizations of equity in education in the scholarship. The conceptualization that emerge from the formal interpretative frames of the SDG 4 agenda is identified as principally concerned with an achievement gap perspective, whereas a substantial portion of the most recent active scholars have engaged with critical pedagogy and social justice perspectives on equity in education. Empirically, there exists a lack of congruence between equity in education as conceptualized in the SDG 4 agenda and the structure of the scholarship on the issue. A situation that did not exist under the previous international education agenda. This in turn indicates that there exist grounds for academics to withhold legitimacy beliefs towards the SDG 4 agenda, potentially eroding its authority and putting its success at risk. Furthermore, the methodological design that was employed in the present study demonstrates the potential to combine scope and depth by integrating computational approaches with qualitative analyses in positivistic approaches to studying global governance in education.
... In 2015, ensuring inclusive and equitable quality education was included as one of the United Nation's Sustainable Development Goals. However, the uneven distribution of educational resources remains a huge obstacle to achieving this goal [3][4][5]. As the basic unit that carries educational resources, schools have an important impact on the optimization of the settlement networks and the improvement of house prices, etc. [6,7]. ...
Full-text available
China’s education has developed rapidly in recent years, but the issue of educational equality still exists. Currently, there are few studies on educational resources, and their spatial pattern and fairness remain unclear. Thus, this study selected the point of interest data and spatial analysis methods to depict the spatial pattern of educational resources (containing the number of teachers, students, facilities, etc.). Then, we evaluated the equity of educational resources (including the number of schools and school teachers) in terms of geographic and population distribution by combining statistical yearbook data with two indices (the index of dissimilarity and agglomera-tion degree) to promote healthy urban development. The results show the following. (1) Educa-tional resources have a multicenter spatial structure of “dual cores and multiple sub-centers.” The Moran index reflects a weak positive spatial correlation between educational resources. (2) The index of dissimilarity is between 0.02 and 0.21, which shows that the allocation of resources is relatively balanced. Regarding internal units, obvious differences exist in the agglomeration degree and equilibrium of educational resources.
In the academic year of 2004–2005, the Spanish region of Madrid began to implement a bilingual educational programme (MBP hereinafter) in state schools. One of the objectives of this pro-gramme was to make the study of a foreign language (English) accessible to students from economically disadvantaged families who cannot afford private foreign language classes. Our study aims to evaluate whether students from a disadvantaged socio- economic background really do have the same probability of parti-cipating in the MBP as their more privileged peers. The analysis use the PISA 2015 database which corresponds to the representative sample of the Community of Madrid in Spain, with added adminis-trative information supplied by the Madrid Regional Ministry of Education concerning the identification of bilingual and non- bilingual schools. Using these data, we estimate a logit model directed at identifying which factors explain the choice by students of whether to attend a bilingual state school. The results obtained reveal that the probability of attending a bilingual school is higher for students belonging to socio-economically and culturally better- off households. This suggests that the MBP could be fostering segregation within the state education sector in Madrid.
Europe is experiencing heightened public attention toward anti-immigration policy reforms and restrictions. Despite the potential importance of these policy changes, we do not know whether these policies influence how immigrant children perceive their futures in their host countries. Employing secondary data analysis of the Program for International Student Assessment and the Migrant Integration Policy Index data, I show that a decrease in policy support for immigrant integration is associated with a decrease in how good of a job immigrant children expect to have when they are adults. Since students’ occupational expectations influence their eventual status attainment, this article shows that a decrease in pro-integration policies has important implications for the integration of immigrants into their host countries and for their life trajectories.
Full-text available
Educational equity was found deteriorating in Sweden from the late 1980s onwards. However, it was not clear what the main sources to this trend are. The present study was therefore to investigate the development of educational equity at different levels of the Swedish educational system to identify possible sources to the change. Students’ school grades and family educational background for the cohorts leaving compulsory school between 1998 and 2014 were analysed in three-level hierarchical models. Intensified segregation with respect to contextual composition and academic outcomes across different schools was found to be the main contribution to the declining of educational equity. The finding was discussed in the light of differentiated learning opportunity, resulted of the recent school reforms in Sweden.
Full-text available
Research has shown that parents tend to pass educational advantage or disadvantage on to their children. However, little is known about the extent to which the intergenerational transmission of education involves children’s agency. In this study we drew from two traditions in sociological and social psychological theorizing—the theory of cultural and social reproduction and the theory of human agency—to examine whether agency influences children’s educational performance, and if so, whether this influence can be observed among children across social classes. We used data from the Spanish sample of the Program for International Student Assessment (N = 25,003 15-year-olds). Results indicate that the level of child agency was weakly positively related to social class, that child agency impacted on a child’s educational performance, and that the positive effect of agency on educational performance did not vary by social class. This suggests that strategies to enhance disadvantaged children’s agency may prove useful in reducing social gradients in educational performance. More generally, our findings may ignite a debate about the role that social structure and human agency play in shaping social inequality and mobility.
Full-text available
There is much debate on early childhood education in research, politics, and society at large. Furthermore, early childhood education research and policy have become increasingly complex and multidimensional. Nonetheless, transdisciplinary research frameworks are still scarce. This report gives insights into a project that used a transdisciplinary approach, outlining its conceptualization and its potential for the definition and analysis of issues that transcend traditional boundaries between academic disciplines, academic and nonacademic knowledge production, and different domains of society.
The Coleman report, published 12 years after the Brown decision, confirmed that widespread school segregation in the United States created inequality of educational opportunity. This study examines whether racial and socioeconomic segregation, which is on the rise in the United States, is still contributing to the achievement differences among students. The study used data from the National Education Longitudinal Survey of 1988 to estimate multilevel models of achievement growth between Grades 8 and 12 in mathematics, science, reading, and history for a sample of 14,217 students attending a representative sample of 913 U.S. high schools. The study found that the average socioeconomic level of students’ schools had as much impact on their achievement growth as their own socioeconomic status, net of other background factors. Moreover, school socioeconomic status had as much impact on advantaged as on disadvantaged students, and almost as much impact on Whites as on Blacks, raising questions about the likely impact of widespread integration. The impact of socioeconomic composition was explained by four school characteristics: teacher expectations, the amount of homework that students do, the number of rigorous courses that students take, and students’ feelings about safety. The results suggest that schools serving mostly lower-income students tend to be organized and operated differently than those serving more-affluent students, transcending other school-level differences such as public or private, large or small. This article then addresses the question of whether such school characteristics can be changed by policies to reform schools and funding systems versus policies to desegregate schools.
Early childhood care and education has become a key issue in social science as well as in politics. In many countries, different actors increasingly use early childhood programs to tackle a variety of societal challenges. The studies collected in this book contribute theoretical and empirical dimensions to a body of research that has neglected a number of questions to date. They analyze (1) the effects of early childhood care and education on the development of children from different social backgrounds, (2) sociocultural disparities in the use of childcare services, (3) the history of childcare in France and the United States of America since the creation of the first formal daycare facilities, and (4) the interplay of desire for linguistic proficiency, acquisition of language, and educational processes in early childhood (this study is published in German). By examining different phenomena using various methodologies, these studies add pedagogical, sociological, and historical perspectives to the scholarly discourse on early childhood care and education. Contents · Equality of opportunity in view of social inequalities · New contributions to early childhood care and education research · Effects of early childhood care and education programs on intellectual development · History of childcare in France and in the United States · Desire, language, and education in early childhood · Multiple methodical approaches · Implications for practice and research Target groups · Researchers and students in educational science, sociology, and history Author Dr. Kaspar Burger is lecturer and researcher in education and developmental psychology at the University Institute Kurt Bösch in Sion, Switzerland.
Large achievement gaps exist between high- and low-income students and between black and white students. This article explores one explanation for such gaps: income segregation between school districts, which creates inequality in the economic and social resources available in advantaged and disadvantaged students’ school contexts. Drawing on national data, I find that the income achievement gap is larger in highly segregated metropolitan areas. This is due mainly to high-income students performing better, rather than low-income children performing worse, in more-segregated places. Income segregation between districts also contributes to the racial achievement gap, largely because white students perform better in more economically segregated places. Descriptive portraits of the school districts of high- and low-income students show that income segregation creates affluent districts for high-income students while changing the contexts of low-income students negligibly. Considering income and race jointly, I find that only high-income white families live in the affluent districts created by income segregation; black families with identically high incomes live in districts more similar to those of low-income white families. My results demonstrate that the spatial inequalities created by income segregation between school districts contribute to achievement gaps between advantaged and disadvantaged students, with implications for future research and policy.