British Educational Research Journal

Vol. 36, No. 3, June 2010, pp. 433–461

School regimes and education equity:

some insights based on PISA 2006
Miquel Àngel Alegre* and Gerard Ferrer
Universitat Autonòma de Barcelona, Spain
(Received
CBER_A_399091.sgm
British
10.1080/01411920902989193
0141-1926
Original
Taylor
02009
00
Dr
[email protected]
000002009
MiguelAlegre
&
and
Education
Article
Francis
9(print)/1469-3518
Francis
August Research
2008; revised
Journal
(online)
11 April 2009; accepted 22 April 2009)

This paper aims to analyze the effects of certain characteristics of the educational systems on the
social composition of schools. After accounting for significant effects of schools’ social composition
on student outcomes (this is confirmed on the basis of a multilevel analysis), we explore the
impacts of distinct components of what we name ‘school regimes’ on measures of school social
segregation (Hutchens indices) across countries and regions. The PISA 2006 database has been
used as the main source of information for such measures. Our analysis considers data for 32
OECD educational systems. Certain characteristics of school regimes are specially assessed: the
level of institutional differentiation existing in the educational career; the presence of private
schools in compulsory education; the level of school autonomy as regards the process of student
admission; and the models and criteria defining public regulation of school access processes.
Results of the regression analyses suggest that more market-oriented school regimes tend to in-
crease schools’ social segregation, whilst those characterised as more comprehensive and publicly
regulated tend to reduce it.

Introduction: the salience of the school composition effect

The main purpose of this paper is to examine how and to what extent certain charac-
teristics of educational systems influence school social segregation across countries
and regions. From our point of view, this object of analysis is relevant in itself. Never-
theless, its relevance would be considerably enhanced if a preliminary hypothesis
could first be substantiated. This hypothesis is the following: the social composition
of schools—and thus, the extent to which the distribution of different student groups
amongst schools is even or uneven—contributes significantly to the explanation of
inequalities amongst students’ learning opportunities.
Since the publication of the Coleman report (1966), research has continued to
corroborate the existence of significant effects of different school composition

*Corresponding author. Institut de Govern i Polítiques Públiques, Universitat Autònoma de

Barcelona, Campus UAB, Parc de Recerca, Mòdul A. 08193 Bellaterra, Spain. Email: miguelangel.
[email protected]

ISSN 0141-1926 (print)/ISSN 1469-3518 (online)/10/030433-29

© 2010 British Educational Research Association
DOI: 10.1080/01411920902989193
434 M. A. Alegre and G. Ferrer

variables on pupils’ outcomes (over and above the effects of the same variables at an
individual level). Some studies focus on the influence of the socio-economic profile
of classes and schools (Caldas & Bankston, 1997; Willms, 1986), others on the net
effects attributable to their ethnic composition (Hanushek et al., 2002; Dronkers &
Levels, 2007) and others on the aggregated and asymmetric impacts of the
ability composition (previous academic performance) in different school contexts
(Opdenakker & Van Damme, 2001; De Fraine et al., 2003). Risking over-
simplification, the main findings here show that academic results obtained by all
students tend to improve when they attend schools with a higher proportion of
students of a high socio-economic status, high ability students and native students.
Partly as a reaction to the research emphasis on school composition, at the end of
the 1980s the School Effectiveness Research (SER) paradigm emerged, its main
message being that schools, through their own effort, ‘can make a difference’ (and can
therefore also be effective in reducing inequalities between their pupils’ results) if they
base their everyday work on the development of a whole set of pedagogic and organ-
isational processes that have been shown to be positive in this sense (Scheerens &
Bosker, 1997; Teddlie & Reynolds, 2000).
Many criticisms have been made of the SER perspective and conclusions, mainly
by researchers who compare the net effects of both composition and process variables
on students’ outcomes. Indeed, while studies confirming that the effects of school
composition on students’ outcomes are far more significant than variables related to
different pedagogic and organisational processes have gained support (De Fraine
et al., 2003; Dumay & Dupriez, 2004), it has also been made clear that mutual inter-
actions exist between the two sets of variables (Baker et al., 2002; Dumay & Dupriez,
2007). Moreover, other qualitative studies have emphasised how the composition of
schools (on socio-economic, ability or ethnic terms) widens or narrows the margin for
the development of certain pedagogic and organisational processes and practices and
also influences the emergence of a more or less positive school climate (Thrupp,
1999; Lupton, 2005).

Theoretical background: school regimes and education equity

Institutional differentiation
Institutional differentiation is understood as a general mechanism for dividing
students into separate types of education with different curricula, different qualifica-
tions at the end of each track and different expectations with regard to the transition
to further education or work (Organisation for Economic Co-operation and Develop-
ment [OECD], 2007). Much of the literature that has sought to evaluate the influ-
ence of the educational system on the dynamics of school segregation or on the
possible accentuation of inequalities between pupils’ and schools’ performance has
focused its attention on the different levels of institutional differentiation and stratifi-
cation that these systems configure. For instance, various studies based on
Programme for International Student Assessment (PISA) data have noted a tendency
 School regimes and education equity 435

for less comprehensive educational systems to increase those achievement inequali-

ties that can be explained by the pupils’ socio-economic background (Duru-Bellat
et al., 2004; OECD, 2007). As regards the relationship between institutional
differentiation and school social segregation, Jenkins et al. (2008) reach the conclu-
sion that in countries with high levels of segregation, the bulk of segregation is
accounted for by unevenness in social background between the different tracks rather
than by differences within each track.
However, the level of formal comprehensiveness (or differentiation) of the education
system is not the only institutional factor that explains the existence of school situations
with varying levels of segregation. There are other contextual factors affecting the allo-
cation of students to schools: for example, the presence of private schools (more or less
publicly funded) (Dronkers & Robert, 2008), the level of diversification of the school
network (in terms of ownership, religious orientations, whether they are selective or
not, whether they are specialised or not etc.) (Gammarnikow & Green, 2003) and the
regulations framing parental choice and school choice procedures (West, 2006) etc.
Grosso modo, when it comes to assessing how these other contextual factors are set
up across countries, two general models tend to be compared: on the one hand, a
model based on the logics of the market, on the other, a model in which basic school
network operations are strongly subjected to public control and regulation.

The logics of quasi-market school developments

There are different paths that can lead school regimes to a ‘quasi-market’ articulation.
Three of the most widespread strategies have been: funding private schools (a strategy
followed in the Netherlands, Belgium, Ireland and Spain); developing school vouch-
ers programmes (different school vouchers schemes have been implemented in
several states of the USA, Chile, Sweden, Denmark, Colombia and New Zealand);
and diversifying publicly maintained schools (England and Wales).
Indeed, one of the main forces behind the tendency of a significant number of western
countries towards the introduction of quasi-market criteria in the configuration of their
public services (which include education) can be found in the fostering of a greater
freedom of choice in terms of the type and profile of the service to be received (Le
Grand & Bartlett, 1993; Ball, 2003). In relation to the issues of school choice/assign-
ment, quasi-market (or open enrolment) models are based on the premise of self-regu-
lation as they depend on the balance between supply of and demand for school places.
A number of quantitative and qualitative studies carried out in various countries and
contexts (Gewirtz et al., 1995; Waslander & Thrupp, 1995; Cullen et al., 2005; Burgess
et al., 2007) have concluded that this model tends to enhance the dynamics of school
socio-economic and ethnic segregation. The schools in greatest demand quickly run
out of places to offer, while those situated further down the ranking of excellence
(which generally accommodate the more disadvantaged pupils) still have a large
number of free places. These spaces are then taken up by pupils that may not have
satisfied the formal or informal admission criteria of the highly ranked schools or that
do not have the cultural and social capital required to be sufficiently competitive in
436 M. A. Alegre and G. Ferrer

the school selection market. Thus, in a market driven by student funding and open
enrolment schemes, schools that both lose pupil numbers and gain—as a cause and
as a consequence of the latter—greater proportions of disadvantaged students, tend
to fall into ‘cycles of decline’ (Whitty, 1997).
In contrast, other authors have advocated that open choice and open enrolment
schemes have led, in some contexts, to an increase in the levels of school segregation
just because market-oriented political reforms have been inadequate or even too
cautious (Chubb & Moe, 1990, 1992). A number of studies have pointed out that,
within school choice schemes, low-intake or minority students open their real margin
for schools choice as they benefit from the removal of specific barriers for choice, such
as ‘economic selection’ or ‘selection by mortgage’. According to this, open choice
schemes should lead towards a decrease in school segregation and, thus, towards
increasing the education opportunities for all students (Godwin & Kemerer, 2002;
Schütz et al., 2007).
Yet, other researchers reach the conclusion that the effects of school choice policies
and school competition on segregation and inequality are quite ambiguous (Fiske &
Ladd, 2000). For instance, Gorard et al. (2003), in their study of the evolution of
school segregation levels in England and Wales, argue: ‘choice policies do not appear
to have either the full benefits their advocates had hoped or the dangers of segregation
their opponents feared’ (p. 168).

Public regulation of school access

At the other end of the continuum we find school regimes rooted in public mecha-
nisms of control and intervention. As regards the regulation of school choice and
school access, here there is a wide range of possible measures: from the application of
quota policies (reservation of places for pupils with learning difficulties, for instance)
to the implementation of common prioritisation admission criteria for all public and
private dependent schools1 (proximity to the pupil’s home, siblings already enrolled
at the chosen school, links between primary and secondary schools etc).
A number of these policies have not been exempt from criticism, regarding both
their ideological assumptions and their real impacts. One of the policies that have
received greater attention has been the practice of basing the process of sorting
students into schools on the delimitation of school catchment areas. It is worth noting
that this type of regulation has been generically situated as the ‘natural’ opposite to
the principles of open enrolment and quasi-market procedures. According to various
studies, catchment areas, in practice, contribute to the reproduction of school segre-
gation as they overlap with social and ethnic residential segregation (Hoxby, 2000;
Gorard et al., 2003). Nevertheless, it has to be made clear that these arguments
consider the incidence of just one of the many possible modalities of zonification,
namely that which assigns each school to its immediate geographical catchment area.
By way of contrast, other studies find that levels of school social segregation are likely
to be reduced in cities where catchment areas are drawn to be internally diverse
(socio-economically and ethnically) and to function as enrolment areas for more than
 School regimes and education equity 437

one public school and more than one private dependent school (Benito & Gonzàlez,
2007).
To sum up, there is a wide and diverse range of possible components underpinning
what we call here ‘school regimes’. Some of these factors concern the basic articula-
tion (more or less comprehensive) of the educational career, others refer to the level
of marketisation of the school networks offering such programmes. This paper is
concerned with a list of variables related to both components and with the extent to
which they affect school social segregation across countries and regions.

Research questions
The purpose of this paper is to answer the following question: how and to what extent
do characteristics of different school regimes influence school social segregation?
More specifically, does the level of institutional differentiation play a significant role
in explaining levels of school social segregation that exist across countries and
regions? How and to what extent does the specific configuration of school networks
and school processes contribute to explaining these levels of school segregation? We
take as our point of departure the hypothesis that school regimes based on more
differentiated curricular frameworks and on clear market-oriented school networks
and processes will tend to show higher levels of social segregation than school regimes
with more comprehensive curricular frameworks within which school networks and
processes are more publicly controlled.
As previously mentioned, this object of analysis, although relevant in its own right,
would be of greater relevance if the following preliminary hypothesis could first be
substantiated: the social composition of schools contributes significantly to the
explanation of inequalities amongst student outcomes. Thus, the first part of our
study will examine how and to what extent the social composition of schools—and so
the realities of school social segregation that exist across countries and regions—
affects student achievement inequalities.

Methods
Data and sample
Our analysis considers data for 32 educational systems. The majority of these corre-
spond to national units (23 countries), whilst a few of them capture sub-national
school features (nine regions). This selection attempts to account for the reality of
‘school regimes’ with high levels of autonomy (even if they correspond to sub-
national political units) as regards the macro-variables we take into consideration.
Whatever the case, we are dealing here with western countries and regions belonging
to the Organisation for Economic Co-operation and Development (OECD).
The main source of the information used to construct the variables of our analyses
is the PISA 2006 database. As described below, we use PISA data to account for
measures of students’ and schools’ social background, schools’ management, schools’
438 M. A. Alegre and G. Ferrer

key proceedings and student outcomes (the scores obtained in the science test).2 As
to the case of sub-national units, PISA categorises two types of regional units: ‘adju-
dicated regions’ (where data adherence to the PISA sampling standards and interna-
tional comparability was internationally adjudicated) and ‘non-adjudicated regions
(where data adherence to the PISA sampling standards at sub-national levels was
assessed by the countries concerned). Two of our sub-national cases (Scotland and
Belgium-Flemish Community) are considered ‘adjudicated regions’, whilst the other
seven (Belgium-French Community, Belgium-German Community, Finland-
Finnish Speaking, Finland-Swedish Speaking, UK-England, UK-Northern Ireland
and UK-Wales) are ‘non-adjudicated regions’.

Variables and measures

Individual and school effects on student achievement. At this point, the dependent vari-
able will be the achievement scores obtained by students in the PISA 2006 science
test, which is the main competency domain assessed in this last PISA edition. A
school’s achievement score is calculated by aggregating the data concerning the
pupils that attend that school.
With respect to student background, we consider here a set of variables whose
effects on student outcomes and opportunities have been repeatedly demonstrated by
educational research: gender, immigrant status, mother tongue and the PISA index
of economic, social and cultural status (ESCS), where the latter is a proxy for the
social background of students.3
As to measures of schools’ social composition, we consider the proportion of
students in the lowest ESCS quartile and, separately, the proportion of students in
the highest ESCS quartile attending each school. This information comes from the
aggregation at the school level of the corresponding individual measures of ESCS.
With respect to ‘school process’ variables, we consider PISA variables and indices
that clearly correspond to others that have been identified as key process factors by
the SER. Thus, we will account for measures of school size, school management
(private or not), school resources (student–teacher ratio and teacher shortage), school
autonomy in education content and practices of ability grouping. In addition, we
consider variables that capture key elements of the concept of school ‘quasi-markets’
described above. Therefore we will focus on whether the school has a considerable
responsibility in approving students’ admission or not; whether its practices of
student admission give high priority to students’ academic records or not; whether it
competes with other schools for the same students or not; and whether it applies a set
of accountability procedures or not (see Table 1A of the Appendix).

Effects of school regimes on school social segregation. Here, the dependent variable is
the level of school social segregation in each of the countries and regions under
consideration. There are a number of distinct indices that are used for such a purpose
in comparative studies.4 In this study we use the Hutchens index, a square root index
 School regimes and education equity 439

based on the dissimilarity index parameters (Hutchens, 2001, 2004).5 It can be

expressed by the following formula:

S p 
∑  Pi  −
pi ri
H = ⋅
i =1   P R

Here, pi and ri are, respectively, the number of pupils in the selected group (for
instance, the number of students in the lowest quartile of socioeconomic status) and
the rest of the pupils in school i. P and R are the total number of students in each
group (selected group and the rest of pupils) in all the schools considered in the
analysis (for our purpose, the total number of schools in each educational system).
Possible values range from 0 (where in every school the proportion of the selected
group of students is exactly the same as its total proportion at a regional level) to 1
(maximum level of segregation). H may be interpreted as the sum of each school’s
shortfall from distributional evenness.
The PISA is a sample survey, which means that any measure of segregation estimated
using its data is subject to sampling variation. Nonetheless, as has been confirmed by
other studies, the number of schools per country sampled by the PISA survey is suffi-
cient to reduce bias to negligible levels (Jenkins et al., 2008). Moreover, in order to
reduce some upward bias attributable to the size of the pupil samples, our estimates
of H are calculated using the final students’ weights provided by the PISA dataset.
We use the Hutchens index to assess the evenness (and unevenness) of the distri-
bution between schools of the following student subgroups: the lower ESCS quartile,
the upper ESCS quartile, the 40th lowest ESCS percentile and the 40th highest
ESCS percentile (Table 2A of the Appendix).
As independent variables we include the contextual characteristics that form the
basis of what we refer to as school regimes (Table 3A of the Appendix). The notion
of ‘school regime’ that we develop here relies on two core components: the level of
differentiation or stratification in the educational career and the market configuration
of the school network. As regards the first component, we consider two variables
whose effects on student achievement inequalities have already been demonstrated in
a number of studies: (1) the age of first selection in the educational career; (2) the
number of tracks available to 15-year-olds.6
With regard to the second component of school regimes—the market configuration
of the school network—we assess the effects of the following variables on segregation:
the percentage of students attending private dependent schools and private indepen-
dent schools; the percentage of students attending schools that have considerable
responsibility in approving student admission; the percentage of students attending
schools that compete with other schools for the same students; and the percentage of
students attending schools where procedures of student admission take students’
academic reports as a prerequisite or give them high priority.
We include an independent variable that aims to describe the general frame or
model of parental choice among public secondary schools deployed in each country
440 M. A. Alegre and G. Ferrer

or region. Resulting from the study of the official national and regional regulations
and policy statements, this variable consists of four categories: (1) Model of ‘forced
recruitment areas’, (2) Model of ‘unforced assignment areas’, (3) Model of ‘restricted
choice’ and (4) Model of ‘open choice’ (see Appendix).

Statistical procedures
To analyse the effects of school factors on student achievement a multilevel analysis
has been carried out, where independent variables are distributed at two levels:
students and schools. Variables included in multilevel analysis models have been
selected according to their significant effect as observed in preliminary multiple
regression models. Multilevel analyses are used to overcome the limitation of the
traditional regression approach, which violates the assumption of independence when
dealing with individuals clustered in a same group (schools, for instance). One of the
main interests here is to assess the changes in the between-school variance results
when increasing one percentage unit of both low and high status students in schools,
after accounting for the rest of the individual and school level variables.
The general equation of the developed multilevel analyses is expressed as follows:

Yij = α j + β1( X1 )ij + β2 ( X 2 )ij + L + β k ( X K )ij + ε ij

α j = γ 00 + U 0 j

where i refers to students (level 1) and j to schools (level 2); Yij is the outcome variable
(science scores); Xk refers to the explanatory variables; βk is the regression coefficient
for the variable k and ε accounts for the residual. αj is the intercept, which is divided
into a fixed part (γ00is the overall intercept, which is equal to the average of school
intercepts αj) and into a random part (U0j denotes school distance from the overall
intercept).
To account for the effects of school regimes’ characteristics on the Hutchens indi-
ces of segregation, the statistical procedure carried out consists of the operation of
multiple linear regressions, using different methods for adjusting the models (step-
wise, backward and forward). The multiple linear regression formula is expressed as
follows:

Yi = α + β1 X i 1 + L + β k X ik + ε i

where Y is the outcome variable, Xk refers to the explanatory variables, α is the inter-
cept, βk is the regression coefficient for the variable k, and ε accounts for the residual.

Results
Effects of school composition and school processes
Two ranges of multilevel models have been carried out in order to account for both
the proportion of students in schools belonging to the lower quartile of the ESCS
 School regimes and education equity 441

index and the proportion of students in the top quartile of the ESCS index. The two
sets of models, called ‘Low ESCS models’ and ‘High ESCS models’, are made up of
eight models each. All of them have been constructed to observe the change in
performance variance between schools, as explained by the school factors included
(Tables 1 and 2).
As shown in the Empty model (models 1.0 and 2.0), across all countries and
regions of our sample, the average proportion of variance attributed to school
variables is approximately 29% of the total variance. All of the factors that are
progressively included into the model are intended to explain a proportion of this
between-school variance.
As regards student level factors (models 1.1 and 2.1), students’ socio-economic
background (ESCS), being female, being native and speaking the language of the
PISA test (which is the language of instruction of each country), explain 34% of
between-school variance.
Once individual effects have been controlled for, models 1.2 and 2.2 account for
the effects of the proportion of low and the proportion of high ESCS students in
schools, respectively. Not surprisingly, we clearly observe that the higher the propor-
tion of lower ESCS students at schools, the lower the average performance of
students. In contrast, an increase in high ESCS students in schools is associated with
a better performance average. In sum, the proportion of low and high ESCS students
in schools explains 14 and 13% of between-school variance, respectively. These
proportions increase (reaching 20 and 19%, respectively) when school process factors
are not included in the models.
Conversely, all the other variables included in the analysis (which we refer to as
‘school process’ variables) only explain between 4 and 5% of the between-school vari-
ance attributed to school variables (models 1.3 to 1.7, and 2.3 to 2.7). This propor-
tion increases (up to 10.5%) when school composition factors are not included in the
models (models 1.8 and 2.8).
The school process factors which are more consistently related with students’
performance are: school size; average number of students per teacher; ability group-
ing for all subjects within school; school posting achievement data publicly; school
with high academic selectivity in student admission; and school-level index of teacher
shortage. The significance of these variables does not fluctuate significantly after
accounting for other student- and school-level factors.
In contrast, other school process factors are not so consistent in their association
with performance variance. As explained above, both competition amongst schools
and private management lose their significance when accounting for the social compo-
sition of schools. With regard to school autonomy in educational content, this variable
does not show any significant effect across all the models. With respect to schools’
responsibility for student admissions, our results take two distinct directions. When
the effect of the proportion of low ESCS students at schools is detracted (together with
the effects of all the other process variables under consideration: model 1.7), the positive
impact of schools being responsible for student admissions on student performance
remains significant. Conversely, when the effect of the proportion of high ESCS
 Table 1. Multilevel analysis. Low ESCS models

Fixed effects 1. Multilevel low ESCS models

Student-level effects 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8

Intercept γ00 504.73 468.47 505.31 494.38 495.01 488.43 470.59 470.20 421.62
(0.76) (0.94) (1.19) (1.62) (1.63) (2.35) (2.71) (2.71) (2.65)
1. Index of ESCS 26.42*** 25.10*** 25.05*** 25.05*** 25.03*** 24.94*** 24.92*** 26.09***
(0.24) (0.24) (0.24) (0.24) (0.24) (0.24) (0.24) (0.24)
2. Gender (being female) −5.89*** −6.01*** −6.01*** −6.02*** −6.13*** −6.16*** −6.16*** −6.07***
(0.39) (0.39) (0.39) (0.39) (0.39) (0.39) (0.39) (0.39)
3. Immigration (being 18.80*** 18.71*** 19.00*** 18.99*** 18.79*** 18.96*** 18.97*** 19.15***
442 M. A. Alegre and G. Ferrer

native) (0.72) (0.71) (0.71) (0.71) (0.72) (0.72) (0.72) (0.72)

4. Language (speaking test 23.41*** 23.39*** 23.40*** 23.37*** 23.02*** 22.96*** 22.93*** 22.93***
language) (0.77) (0.76) (0.76) (0.76) (0.77) (0.77) (0.77) (0.77)
School-level effects 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
5. Percentage of low ESCS −1.45*** −1.38*** −1.37*** −1.37*** −1.30*** −1.29***
students in schools (0.03) (0.03) (0.03) (0.03) (0.03) (0.03)
6. School size 1.25*** 1.26*** 1.24*** 0.81*** 0.73*** 1.50***
(0.12) (0.12) (0.12) (0.13) (0.13) (0.14)
7. School management: 0.85 1.27 0.92 −0.50 0.12 11.36***
Schools privately (1.59) (1.60) (1.61) (1.58) (1.58) (1.72)
managed
8. School autonomy index 0.55 0.10 −0.01 −0.24 −0.49
in educational (0.58) (0.59) (0.58) (0.58) (0.64)
content
9. Ability grouping (ability −7.85*** −8.01*** −8.74*** −8.56*** −12.5***
grouping for all (1.66) (1.65) (1.62) (1.62) (1.78)
subjects)
10. School with high level 1.89 1.11 0.87 6.23*
of competition (1.34) (1.31) (1.31) (1.44)
 Table 1. (Continued)

Fixed effects 1. Multilevel low ESCS models

School-level effects 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8

11. Responsibility student 6.71*** 4.16** 4.01* 4.16*

admission (principal, (1.71) (1.68) (1.68) (1.85)
teachers, gov. board)
12. School with high 22.96*** 22.92*** 29.49***
academic selectivity (1.62) (1.62) (1.79)
of admittance
13. School average number 1.48*** 1.43*** 1.70***
of students per (0.14) (0.14) (0.16)
teacher
14. School-level index of −1.53** −1.57** −4.75***
teacher shortage (0.59) (0.59) (0.64)
15. Accountability: school 4.38*** 5.56***
posting achievement (1.17) (1.30)
data
Random effects 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
Within-schools variance 7186.60 6644.75 6637.29 6636.18 6635.92 6598.67 6597.31 6597.25 6603.11
(23.33) (21.57) (21.52) (21.51) (21.51) (21.43) (21.42) (21.42) (21.46)
Between-schools variance 2974.02 1959.62 1362.82 1336.96 1331.91 1320.18 1234.56 1230.89 1648.42
(63.68) (44.97) (33.30) (32.66) (32.54) (32.40) (30.73) (30.66) (38.82)
Var. prop. attributed to 29.27
school (rho, ρ) (%)
Within-schools var. prop. 7.54 7.64 7.66 7.66 8.18 8.20 8.20 8.12
explained (%)
Between-schools var. prop. 34.11 54.18 55.05 55.22 55.61 58.49 58.61 44.57
explained (%)

Note: *** p < .001; ** p < .01; * p < .05; ESCS = economic, social and cultural status.
School regimes and education equity 443
 Table 2. Multilevel analysis. High ESCS models

Fixed effects 2. Multilevel high ESCS models

Student-level effects 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8

Intercept γ00 504.73 468.47 435.47 427.16 428.20 423.04 406.71 406.52 421.62
(0.76) (0.94) (1.15) (1.37) (1.39) (2.10) (2.46) (2.46) (2.65)
1. Index of ESCS 26.42*** 25.08*** 25.02*** 25.02*** 25.01*** 24.91*** 24.90*** 26.09***
(0.24) (0.24) (0.24) (0.24) (0.24) (0.24) (0.24) (0.24)
2. Gender (being female) −5.89*** −6.03*** −6.03*** −6.04*** −6.15*** −6.18*** −6.18*** −6.07***
(0.39) (0.39) (0.39) (0.39) (0.39) (0.39) (0.39) (0.39)
3. Immigration (being native) 18.80*** 18.91*** 19.21*** 19.21*** 19.01*** 19.18*** 19.20*** 19.15***
444 M. A. Alegre and G. Ferrer

(0.72) (0.71) (0.71) (0.71) (0.72) (0.72) (0.72) (0.72)

4. Language (speaking test 23.41*** 23.60*** 23.60*** 23.57*** 23.22*** 23.14*** 23.12*** 22.93***
language) (0.77) (0.76) (0.76) (0.76) (0.77) (0.77) (0.77) (0.77)
School-level effects 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8
5. Percentage of high ESCS 1.34*** 1.29*** 1.28*** 1.27*** 1.21*** 1.20***
students in schools (0.03) (0.03) (0.03) (0.03) (0.03) (0.03)
6. School size 1.31*** 1.33*** 1.31*** .79*** 0.72*** 1.50***
(0.12) (0.12) (0.12) (0.13) (0.13) (0.14)
7. School management: −2.32 −1.82 −2.21 −3.39* −2.84∧ 11.36***
Schools privately (1.61) (1.63) (1.64) (1.60) (1.61) (1.72)
managed
8. School autonomy index in 0.39 0.01 −0.12 −0.32 −0.49
educational content (0.58) (0.59) (0.58) (0.58) (0.64)
9. Ability grouping (ability −7.73*** −7.92*** −8.50*** −8.36*** −12.5***
grouping for all subjects) (1.66) (1.66) (1.63) (1.62) (1.78)
10. School with high level of 2.40∧ 1.62 1.42 6.23***
competition (1.35) (1.32) (1.32) (1.44)
 Table 2. (Continued)

Fixed effects 2. Multilevel high ESCS models

School-level effects 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8

11. Responsibility student 5.07** 2.71 2.58 4.16*

admission (principal, (1.72) (1.69) (1.69) (1.85)
teachers, gov. board)
12. School with high academic 20.82*** 20.81*** 29.49***
selectivity of admittance (1.64) (1.64) (1.79)
13. School average number of 1.70*** 1.66*** 1.70***
students per teacher (0.14) (0.14) (0.16)
14. School-level index of −1.23* −1.27* −4.75***
teacher shortage (0.59) (0.59) (0.64)
15. Accountability: school 3.71** 5.56***
posting achievement data (1.18) (1.30)
Random effects 2.0 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8
Within-schools variance 7186.60 6644.75 6638.22 6636.93 6636.70 6599.56 6598.85 6598.85 6603.11
(23.33) (21.57) (21.53) (21.52) (21.52) (21.44) (21.43) (21.43) (21.46)
Between-schools variance 2974.02 1959.62 1388.05 1358.38 1353.47 1342.12 1251.69 1248.69 1648.42
(63.68) (44.97) (33.89) (33.15) (33.04) (32.91) (31.22) (31.17) (38.82)
Var. prop. attributed to school 29.27
(rho, ρ) (%)
Within-schools var. prop. 7.54 7.63 7.65 7.65 8.17 8.18 8.18 8.12
explained (%)
Between-schools var. prop. 34.11 53.33 54.33 54.49 54.87 57.91 58.01 44.57
explained (%)

Note: *** p < .001; ** p < .01; * p < .05; ESCS = economic, social and cultural status.
School regimes and education equity 445
446 M. A. Alegre and G. Ferrer

students is accounted for (next to the effects of other process variables: model 2.7),
schools’ responsibility for admissions has no significant impact on variability in scores.

School regimes and school social segregation

Before assessing how and to what extent certain features of school regimes contribute
to explaining levels of school social segregation, we first show a general overview of
social segregation values across the countries and regions under consideration.
As shown in Figure 1, the school regimes with lower levels of school segregation
with regard to the distribution of students in the lower ESCS quartile correspond to
the Scandinavian countries (Finland, Norway, Sweden, Iceland and Denmark)
together with Wales and the German Community of Belgium. All their Hutchens values
are < 0.10 (in the lower 40% of the observed values). At the other end of the spectrum,
Hungary, Greece, France, the French Community of Belgium, Germany, Portugal
and Spain show the higher levels of school segregation as regards low ESCS students
(all their Hutchens indices are > 0.15; in the higher 40% of the observed values).
Figure 1. Hutchens Segregation Indices (25th, 40th, 60th, and 75th percentiles of PISA ESCS index) across countries and regions














 ' * '" "   ' * '" " 

         
 ' " "   ' " " 
&( 0  " &( 0  "
+ * &(  '
* '  + *
%  ' &( +   %  '
/   %  '
&( +   %  ' /  

-.    * ' 

&(  '  '

%  ' -.   

+ * ,  ' + * ,  '

+   '"    " 

# ' !" 

*$  ' &( ) '

!"  # '

&( ) ' %  '

#$    
 '
*$  '
& '   "
% 
% 
 
   " 
& '   "
#$    
    
!" 
+   '"
    
    
 
 
  
  
 
!" 
    

 
 



' "   )# 2    
' " *  )# 1     ' "   )# 3   
' " *  )#    

Note: In both graphs, countries have been ordered respectively according to segregation value Source: Prepared by the authors on the basis of OECD PISA database 2006
calculated with the ESCS 25th and 75th percentile

Figure 1. Hutchens Segregation Indices (25th, 40th, 60th, and 75th percentiles of PISA ESCS
index) across countries and regions
 School regimes and education equity 447

When considering the measures of school segregation and accounting for the distri-
bution of students in the upper ESCS quartile, the results are only slightly dissimilar.
Finland, the German Community of Belgium, Wales and Scotland present the lower
Hutchens values (< 0.10), whilst Hungary, France, Greece, Austria, Portugal, Spain,
Slovak Republic, The Netherlands, the French Community of Belgium, the USA and
Germany show the higher ones (> 0.15).

Characteristics of school regimes and levels of school social segregation: correlations

A correlation analysis has been carried out in order to confirm the association between
school regime factors and the segregation index, as well as to gain an understanding of
the sense and strength of relationships between explanatory factors themselves.
As shown in Table 3, the age at which education systems operate their first selection
in the educational career is the only contextual factor negatively correlated with school
segregation indices (association variation of 21% with the Hutchens index for students
in the lower ESCS quartile and of 25% with the Hutchens index for students in the upper
ESCS quartile). Not surprisingly, the number of school tracks available to 15-year-olds
is also correlated with segregation, since both early selection and the number of tracks
available to 15-year-olds are highly correlated (association variation of 13.5% with the
Hutchens lower ESCS students and 17% with the Hutchens upper ESCS students).
All the other contextual variables show statistically significant positive coefficients of
correlation with indices of school segregation: students in private independent schools
(19 and 15%), students in schools competing with other schools (17 and 13%), students
in schools with considerable responsibility for admission (20 and 15%) and the students
in schools which consider students’ academic records as an important criterion for
approving admission (20 and 25%). The only school regime factor not correlated with
segregation is the proportion of students enrolled at private dependent schools.
As regards the correlations between school regime factors, the variables that
demonstrate a higher level of association are those referring to the institutional differ-
entiation each educational system operates. Indeed, the first age of selection and the
number of tracks available to 15-year-olds are associated in 71% of their variation.
Moreover, highly differentiated systems (with regard to these variables) tend to have
higher proportions of students at selective schools and at schools with autonomy for
approving student admissions.
Underpinning these correlations is a positive association between, on the one hand,
the proportion of students at schools with a considerable responsibility for admissions
and, on the other hand, the proportion of students at schools approving admissions
on the basis of academic records (association of 27%) or the number of students at
schools in competition with other schools for the same studentship (32%). The
proportion of students at private dependent schools shows less consistent correlations
with the proportion of students at schools competing with other schools and with the
number of school tracks available to 15-year-olds, whilst the presence of private
independent schools does not appear associated with any of the contextual variables
under consideration.
 Table 3. Characteristics of school regimes and segregation indices: correlations

schools
choice

selection

years-olds
b. Number of
c. Proportion
of students

a. First age of
e. Proportion of
g. Proportion of

school tracks
d. Proportion of
for admission
with academic

selection in the
available to 15-
h. Process parental

education system
students in schools

students in private
students in schools

independent schools

in private dependent
f. Proportion of student
in schools responsible

competing for students

1. First age of selection in Coef. corr. −0.845** −0.206 0.261 −0.169 −0.377* −0.806** −0.308
the education system P-value 0.000 0.265 0.156 0.363 0.037 0.000 0.086
448 M. A. Alegre and G. Ferrer

2. Number of school tracks Coef. corr. −0.845** 0.390* −0.185 0.265 0.456** 0.640** 0.503**
available to 15-years- P-value 0.000 0.030 0.318 0.149 0.010 0.000 0.003
olds
3. Proportion of students in Coef. corr. −0.206 0.390* −0.054 0.373* 0.152 0.157 0.463**
private dependent P-value 0.265 0.030 0.772 0.039 0.413 0.399 0.009
schools
4. Proportion of students in Coef. corr. 0.261 −0.185 −0.054 0.214 −0.151 −0.246 −0.103
private independent P-value 0.156 0.318 0.772 0.248 0.417 0.182 0.582
schools
5. Proportion of students in Coef. corr. −0.169 0.265 0.373* 0.214 0.562** 0.234 0.578**
schools competing for P-value 0.363 0.149 0.039 0.248 0.001 0.206 0.001
students
6. Proportion of students in Coef. corr. −0.377* 0.456** 0.152 −0.151 0.562** 0.517** 0.416*
schools responsible for P-value 0.037 0.010 0.413 0.417 0.001 0.003 0.020
admission
 Table 3. (Continued)

choice

selection

years-olds
b. Number of
c. Proportion
of students

a. First age of
e. Proportion of
g. Proportion of

school tracks
d. Proportion of
for admission
with academic

selection in the
available to 15-
h. Process parental

education system
students in private
students in schools
students in schools

independent schools

in private dependent
f. Proportion of student
in schools responsible

competing for students

7. Proportion of students in Coef. corr. −0.806** 0.640** 0.157 −0.246 0.234 0.517** 0.286
schools with academic P-value 0.000 0.000 0.399 0.182 0.206 0.003 0.119
selection
Coef. corr. −0.308 0.503** 0.463** −0.103 0.578** 0.416* 0.286
8. Process parental choice P-value 0.086 0.003 0.009 0.582 0.001 0.020 0.119

Students below the ESCS Coef. corr. −0.458** 0.368* 0.030 0.432* 0.413* 0.444* 0.447* 0.106
25th percentile P-value 0.008 0.038 0.872 0.015 0.021 0.012 0.012 0.563
Students below the ESCS Coef. corr. −0.449** 0.377* 0.067 0.426* 0.443* 0.447* 0.444* 0.111
40th percentile P-value 0.010 0.033 0.720 0.017 0.013 0.012 0.012 0.544
Students above the ESCS Coef. corr. −0.452** 0.402* 0.086 0.428* 0.409* 0.409* 0.469** 0.138
60th percentile P-value 0.009 0.023 0.645 0.016 0.022 0.022 0.008 0.453
Students above the ESCS Coef. corr. −0.497** 0.410* 0.090 0.392* 0.359* 0.389* 0.502** 0.099

Hutchens
Segregation Index
75th percentile (High) P-value 0.004 0.020 0.630 0.029 0.047 0.031 0.004 0.592

Note: ** values statistically significant at the 1% (p < 0.01); * values statistically significant at the 5% (p < 0.05).
School regimes and education equity 449
450 M. A. Alegre and G. Ferrer

Characteristics of school regimes and levels of school social segregation: regression models.
Regression models have been utilised to assess how and to what extent school regime
variables are associated with variance in school segregation across countries and
regions. As shown in the correlation analysis described above, these contextual vari-
ables are: first age of selection in the education system; proportion of students in
private independent schools; proportion of students in schools competing with other
schools in the same area; proportion of students in schools with responsibility for
student admissions; and proportion of students in schools that consider academic
records in student admissions.
Four ranges of models have been constructed (see Table 4), called ‘Low ESCS’
(where the Hutchens index for students in the lower ESCS quartile is the dependent
variable), ‘Mid-low’ (where the dependent variable is the Hutchens index for students
below the ESCS 40th percentile), ‘Mid-high’ (where the dependent variable is the
Hutchens index for students above the ESCS 60th percentile) and ‘High’ (where the
Hutchens index for students in the upper ESCS quartile is the dependent variable).
Models have been included according to their strength in explaining variance in
segregation indexes (> 50% explained).
Models 1.1, 2.1, 3.1 and 4.1 account for the effect of the first age of selection and
the proportion of students enrolled in private independent schools on segregation in
schools. The proportion of students at private dependent schools is not included in
this basic model as it shows no significant relationship with school segregation.
Results show that comprehensive systems tend to have lower levels of school segrega-
tion: an increase in the first age of selection in education systems is significantly
related to a decrease in segregation values. On the other hand, these models show that
an increase in the proportion of students enrolled at private independent schools is
associated with an increase in the national/regional levels of segregation of low and
mid-low ESCS students (models 1.1 and 2.1) and of mid-high and high ESCS
students (models 3.1 and 4.1). Overall, these two variables—first age of selection in
the education system and the proportion of students in private independent schools—
contribute to explaining 53.3, 52.6, 53.5 and 56.5% of the variance in segregation
indices in models 1.1, 2.1, 3.1 and 4.1, respectively.
When adding into the models the proportion of students in schools competing with
other schools in the same area, no changes are observed with regard to the explana-
tion of variance in segregation indices. This applies to the case of all the Hutchens
indices under consideration (models 1.2, 2.2, 3.2 and 4.2). Indeed, this variable—
students in competing schools—loses its individual explanatory weight after account-
ing for variables such as age of first selection or students in private independent
schools.
In models 1.3, 2.3, 3.3 and 4.3, the proportion of students in schools that report
having a considerable responsibility for student admissions is added. Resulting from
the addition of this variable, a substantial increase is observed (of 5.3 to 9.7 points)
in the explanation of school segregation of low ESCS students (model 1.3), mid-low
ESCS students (model 2.3) and mid-high ESCS students (model 3.3). In contrast,
this same variable does not have any significant effect on the levels of school
 Table 4. Characteristics of school regimes and segregation indices: regression models

1. Low ESCS models

Dependent variable: Low ESCS Seg. Index 1.1 1.2 1.3 1.4 1.5
Constant 0.284 0.284 0.137 −0.019 0.137
Explanatory variables: ‘School regimes’
First age of selection in the education system −0.013** −0.013** −0.011** −0.011**
Students in private independent schools 0.012** 0.012** 0.013** 0.012** 0.013**
Students in schools competing for students excl.
Students in schools responsible for admission 0.001** 0.001* 0.001**
Students in schools with academic selection 0.001** excl.
Explained variance of Low Segregation Index 53.3 53.3 62.8 53.7 62.8
2. Mid-low ESCS models
Dependent variable: Mid-low ESCS Seg. Index 2.1 2.2 2.3 2.4 2.5
Constant 0.246 0.246 0.120 −0.013 0.120
Explanatory variables: ‘School regimes’
First age of selection in the education system −0.011** −0.011** −0.009** −0.009**
Students in private independent schools 0.010** 0.010** 0.011** 0.010** 0.011**
Students in schools competing for students excl.
Students in schools responsible for admission 0.001** 0.001* 0.001**
Students in schools with academic selection 0.001** excl.
Explained variance of Mid-low Segregation Index 52.6 52.6 62.3 52.8 62.3
School regimes and education equity 451
 Table 4. (Continued)
3. Mid-high ESCS models
Dependent variable: Mid-high ESCS Seg. Index 3.1 3.2 3.3 3.4 3.5
Constant 0.244 0.244 0.149 0.069 0.149
Explanatory variables: ‘School regimes’
First age of selection in the education system −0.011** −0.011** −0.009** −0.009**
Students in private independent schools 0.010** 0.010** 0.010** 0.010** 0.010**
Students in schools competing for students excl.
Students in schools responsible for admission 0.001* excl. 0.001*
Students in schools with academic selection 0.001** excl.
Explained variance of Mid-high Segregation Index 53.5 53.5 58.8 50.1 58.8
452 M. A. Alegre and G. Ferrer

4. High ESCS models

Dependent variable: High ESCS Seg. Index 4.1 4.2 4.3 4.4 4.5
Constant 0.308 0.308 0.308 0.083 0.308
Explanatory variables: ‘School regimes’
First age of selection in the education system −0.014** −0.014** −0.014** −0.014**
Students in private independent schools 0.012** 0.012** 0.012** 0.011** 0.012**
Students in schools competing for students excl.
Students in schools responsible for admission excl. excl. excl.
Students in schools with academic selection 0.001** excl.
Explained variance of High Segregation Index 56.5 56.5 56.5 50.2 56.5

Note: ** values statistically significant at the 1% (p < 0,01); * values statistically significant at the 5% (p < 0,05); ESCS = economic, social and cultural status.
 School regimes and education equity 453

segregation of students in the upper ESCS quartile (model 4.3). In other words, the
patterns of the distribution of the most privileged students are not affected by the
extent to which school regimes control the amount of responsibility schools have for
student admissions.
Models 1.4, 2.4, 3.4 and 4.4 allow a comparison of the effects on school segrega-
tion associated with the proportion of students in private independent schools; the
proportion of students at schools with considerable responsibility for student admis-
sion; and the proportion of students at schools that select students on the basis of
their academic record. Overall, these three variables have a significant and positive
effect on the levels of school segregation of low and mid-low ESCS students (models
1.4 and 2.4). However, in models 3.4 and 4.4, the proportion of pupils at schools
responsible for admissions loses its significance, suggesting that the distribution of
more privileged groups depends more specifically on the extent to which selection by
ability is applied than on schools’ general level of autonomy for approving admission.
Models 1.5, 2.5, 3.5 and 4.5 add the age of first selection in the educational career
to these variables. In this case, whilst school responsibility for student admissions is
still significant, the effect of academic selectivity loses its relevance with respect to the
indices of segregation of low, mid-low and mid-high ESCS students. This is mainly
due to the high level of variation shared by the age of first selection and the extent to
which school regimes apply academic selection (65%). Concerning the segregation of
high ESCS students, in model 4.5 the loss of significance of academic selection is
accompanied by the exclusion of the variable of responsibility for admissions (in line
with the results from model 4.3).
To conclude, the school regime factors that show more significant effects on school
segregation across countries/regions are related to levels of institutional differentia-
tion, the presence of private independent schools and the margin given to school
choice. With regard to the latter, the most consistent factor in explaining segregation
variance is the national/regional proportion of students at schools with a considerable
responsibility for student admissions. School selection practices based on ability show
a significant explanatory effect on segregation when not accounting for the age at
which school regimes operate their first selection into educational careers.

Conclusions and discussion

Effects of school composition and school processes
As previously mentioned, if we can corroborate the significance of the effects of
school composition on student outcomes, we can then substantiate the relevance of
assessing how and to what extent certain contextual factors (components of school
regimes) influence school social segregation across countries and regions. At this
initial stage, we have reached the expected conclusion (an argument that has been
widely confirmed in a number of other studies): the socio-economic composition of
the student body significantly affects students’ performance (over and above the
contribution of the same variable in the model at an individual level). Moreover, this
454 M. A. Alegre and G. Ferrer

factor contributes to explaining a large proportion of the variance in results between

schools, which is attributable to school factor effects.
Thus, for example, one can observe how each increase of the proportion of low
ESCS (lower quartile) students attending a school is accompanied by a significant
decrease in the achievement outcomes of all students, whereas the opposite effect
occurs when we consider an increase of the proportion of high ESCS (upper quartile)
students. At the same time, confirming the conclusions reached by De Fraine et al.
(2003) and Dumay and Dupriez (2004), it is worth noting that the significance of
variables relating to schools’ social composition remain stable when other school
variables referring to school institutional frameworks and pedagogical processes are
included in the models. Indeed, variables of schools’ social composition tend to annul
or reduce the significance of an important set of these latter variables (e.g. schools
being privately managed, schools competing with other schools for the same student-
ship or schools being responsible for admissions).

On the effect of institutional differentiation

With respect to the effects of the basic configuration of the academic career on segre-
gation, our results are in line with the conclusions of comparative studies that show
that more stratified educational systems tend to generate more inequalities in achieve-
ment results obtained by students and schools (Duru-Bellat et al., 2004; OECD,
2007). Our findings suggest that these inequalities are fundamentally associated with
the high levels of school segregation that this stratification tends to provoke. Of the
variables used to capture the level of educational systems’ institutional differentiation,
one remains particularly significant in all of our models: the age of first selection in
the educational career. This evidence confirms the hypothesis that school regimes
based on early selection features generate higher levels of school social segregation,
basically profiting from the possibility of concentrating low ability students (with a
high probability of also having a low ESCS) in schools or academic tracks of less
academic value.
This conclusion provides relevant insights for the policy debate centred on the
equity and quality of school systems. Basic curricular configuration seems to be
responsible for significant effects on school segregation (and thus on learning oppor-
tunities) and, as shown in other comparative studies, the extent to which students are
selected early and sorted into different institutional tracks does not appear to be
significantly associated with the national averages of student performance (OECD,
2007).

On the effects of public/private stakeholders

At this point, our results enter into discussion with comparative studies that conclude
that, at a national level, a greater proportion of private schools is not significantly
associated with larger socio-economic disparities in schooling outcomes (OECD,
2007), nor with school segregation measures of certain countries (Jenkins et al.,
 School regimes and education equity 455

2008). In all probability these discrepancies are due to the distinction we operate
between private dependent schools and private independent schools. This distinction
allows us to generate findings that probably remain hidden when both types of private
school are grouped in the same category. In short, our analyses show no significant
effects of the national/regional proportion of private dependent schools on school
segregation. In contrast, the national/regional proportion of private independent
schools always appears as a significant factor: as this proportion increases, so does the
national/regional level of school segregation.
Hence, the presence of private independent schools is a key factor when accounting
for school segregation, showing a net effect that goes beyond the impacts on school
segregation of both schools’ level of responsibility in student admissions and the level
of stratification of the academic career. Further analysis focused on the effects of
private schools on student learning inequalities should consider the distinction
between private dependent and independent schools and examine in detail the sort of
selection practices (formal and informal) that the two types operate.

On the effects of parental choice (parents choosing schools)

One of the leitmotifs of contemporary educational reforms has been widening the
margin of school choice available to families (Le Grand & Bartlett, 1993; Gewirtz
et al., 1995). Beyond the policy debate, however, it is worth pointing out that compar-
ative educational research has not formulated conclusive arguments about the effects
various policies which aim to widen parental choice have on education equity (and
inequality). Indeed, is not uncommon to find different studies reaching contradictory
conclusions in this respect (Gorard et al., 2003).
As far as our analysis is concerned, it is important to note that the variables used to
account for different levels of ‘parental choice’ are inevitably limited. Firstly, the use
of the national/regional proportion of students at schools competing with other
schools for the same studentship (from the PISA dataset) does not guarantee an
analysis of a representative sample of schools with regard to this variable. Secondly,
the information gathered beyond the PISA dataset in order to categorize the general
models of parental choice deployed in each country or region (forced/unforced
assignment area, restricted choice or open choice) are unable to capture the diversity
of local regulations and policies that are being implemented within each country and
region.
Nevertheless, our results do allow us to suggest the existence of relevant tendencies.
For instance, the national/regional proportion of students at schools competing with
other schools for the same students appears as a significant variable (positively asso-
ciated with the level of school segregation) in a regression model that only combines
it with the variable of age at first selection. However, once the proportion of students
at private independent schools is included, its significance is annulled. This suggests
that the margin available for parental choice does not have relevant effects on school
segregation after the characteristics of the basic configuration of the school network
are controlled for. In short, our results do not support arguments that suggest that
456 M. A. Alegre and G. Ferrer

market oriented reforms will be accompanied by benefits in school equity, nor the
opposite statement. An open choice model, although formally extending the range of
school options available to families, does not in itself lead to greater heterogeneity
across schools’ social composition; nor does such effect appear associated with more
publicly controlled schemes of parental choice.

On the effects of school choice (schools ‘selecting’ students)

By contrast, the margin of autonomy schools have with regard to approving student
admission has been shown to have significant effects on segregation. Here it is neces-
sary to be cautious. The variable, constructed from the PISA dataset and used to
describe this margin of autonomy, takes into account the proportion of students in
schools where either the principal/teachers, the school’s governing board or both have
a considerable amount of responsibility in approving student admissions. Here, a
representative sample of schools with regard to this variable is not assured.
Nonetheless, the effect of this variable appears solid and stable across the majority
of the models: school regimes with a higher proportion of students at schools consid-
erably responsible for student admissions tend to show higher levels of social
segregation. Such an effect remains significant when the rest of the variables account-
ing for the basic configuration of the school network are controlled for and this
contributes considerably to explaining the variation in levels of school social segrega-
tion. The exception here is the level of segregation of high ESCS students. In this
case, the extent to which school regimes give schools considerable responsibility for
admission do not have any significant effect. In other words, the uneven distribution
of the most privileged students is not being corrected even in school regimes where
the school’s autonomy for approving student admissions is more restricted.
It is worth highlighting that the aforementioned variable used as a reference for
school choice does not make any particular specification regarding which criteria are
being applied by schools for student admissions. At this point, we should turn to the
discussion of the extent to which school selection practices specifically based on
students’ abilities (academic record) tend to increase school social segregation.
Indeed, there are a number of studies concluding the significance of such an effect
(West et al., 2004; Jenkins et al., 2008). Our analysis corroborates these conclusions
only in part.
We have taken the national/regional proportion of students at schools that give high
consideration to student records when deciding on admissions as a variable of refer-
ence. This variable is positively associated with national/regional levels of school
segregation, as well as highly correlated with the level of stratification existing in the
educational career. When the age of first selection is not included in models account-
ing for the segregation of low and mid-low ESCS students, the proportion of students
at schools that select by ability does not reduce the significance of the general level of
responsibility for admissions that school regimes give to schools. This finding
suggests that the very fact of schools being responsible for student admissions affects
the distribution of less privileged pupils beyond the specific admittance criteria
 School regimes and education equity 457

schools report to be applying. On the contrary, the proportion of students at schools

selecting on ability annuls the significance of the proportion of students at schools
responsible for admissions when the segregation of the most privileged pupils is
considered.
To sum up, it seems clear that, more than the margin given to families to select
between different schools (parental choice), what fundamentally matters when it
comes to assessing school segregation is the margin given to schools to intervene in
student admissions. When considered with regard to the policy debate, the implica-
tions of this conclusion are relevant and have already been underlined in other studies
(West, 2006): in order to generate the contextual conditions that could lead to an
increase in social heterogeneity within schools (and, consequently, to higher rates of
equity in the academic outcomes obtained by students and schools), more public
control of student admission processes applied by schools with a high level of admis-
sions autonomy is needed.

Notes
1. Private dependent schools are here considered as schools that are privately managed but receive
public funding, whilst private independent schools are privately managed and not publicly
funded.
2. See OECD (2007) for a description of the PISA 2006 programme, as well as for a description
of the national and sub-national sample numbers.
3. The PISA ESCS index is composed of individual measures for parental occupational status,
family level of education and home possessions. See the PISA 2003 technical report (OECD,
2005) for a detailed explanation of its construction.
4. For an interesting discussion of the use of school segregation indices, see Allen and Vignoles
(2007) and Gorard (2007).
5. For an overview of the advantages of the Hutchens index in comparison to other segregation
indices, see Jenkins et al. (2008).
6. Tracks are defined here as parallel educational itineraries that divide students into different
school settings (not necessarily into different institutions) and that may lead to different
credentials with distinct academic values.

References
Allen, R. & Vignoles, A. (2007) What should an index of school segregation measure? Oxford
Review of Education, 33(5), 643–668.
Baker, D.P., Goesling, B. & Letendre, G.K. (2002) Socioeconomic status, school quality and
national economic development: a cross-national analysis of the ‘Heyneman-Loxley Effect’ on
mathematics and science achievement, Comparative Education Review, 46(3), 291–312.
Ball, S. J. (2003) Class strategies and the education market (London, RoutledgeFalmer).
Benito, R. & Gonzàlez, I. (2007) Processos de segregació escolar a Catalunya (Barcelona,
Mediterrània).
Burgess, S., McConnell, B., Propper, C. & Wilson, D. (2007) The impact of school choice on sort-
ing by ability and socioeconomic factors in English secondary education, in: L. Wößmann &
P. E. Peterson (Eds) Schools and the equal opportunity problem (Cambridge, MA, MIT Press).
Caldas, S. J. & Bankston, C. (1997) Effect of school population socioeconomic status on
individual academic achievement, Journal of Education Research, 90(5), 269–277.
458 M. A. Alegre and G. Ferrer

Chubb, J. & Moe, T. (1990) Politics, markets and America’s schools (Washington, Brookings
Institution Press).
Chubb, J. & Moe, T. (1992) A lesson in school reform from Great Britain (Washington, Brookings
Institution Press).
Coleman, J. S. et al. (1966) Equality of educational opportunity (Washington, Department of Health,
Education and Welfare, Office of Education).
Cullen, J. B., Jacob, B. A. & Levitt, S. D. (2005) The impact of school choice on student outcomes:
an analysis of the Chicago public schools, Journal of Public Economics, 89(5–6), 729–760.
De Fraine, B., Van Damme, J., Van Landeghem, G., Opdenakker, M. C. & Onghena, P. (2003)
The effect of schools and classes on language achievement, British Educational Research
Journal, 29(6), 841–859.
Dronkers, J. & Levels, M. (2007) Do school segregation and school resources explain region-of-
origin differences in the mathematics achievement of immigrant students? Educational
Research and Evaluation, 13, 435–462.
Dronkers, J. & Robert, P. (2008) School choice in the light of the effectiveness: differences of various
types of public and private schools in 19 OECD countries, Journal of School Choice, 2, 260–301.
Dumay, X. & Dupriez, V. (2004) Effet établissement: effet de processus et/ou effet de
composition? [School effects: process effect and/or composition effect?] Les Cahiers de
Recherche en Education et Formation, 36. Available online at http://www.uclouvain.be/cps/ucl/
doc/girsef/documents/036cahier.2.pdf
Dumay, X. & Dupriez, V. (2008) Does the school composition effect matter? Evidence from
Belgian data, British Journal of Educational Studies, 56(4), 440–447.
Duru-Bellat, M., Mons, N. & Suchaut, B. (2004) Caractéristiques des systèmes éducatifs et
compétences des jeunes de 15 ans. L’éclairage des comparaisons entre pays, Les Cahiers de l’IREDU
(Bourgogne, IREDU, CNRS-Université de Bourgogne).
Fiske, E. & Ladd, H. (2000) When schools compete: a cautionary tale (Washington, Brookings
Institution Press).
Gammarnikow, E. & Green, A. (2003) Social justice, identity formation and social capital: school
diversification policy under New Labour, in: C. Vincent (Ed.) Social justice, education and
identity (London, Routledge).
Gewirtz, S., Ball, S. J. & Bowe, R. (1995) Markets, choice and equity in education (Buckingham,
Open University Press).
Godwin, R. K. & Kemerer, F. R. (2002) School choice tradeoffs: liberty, equity and diversity (Austin,
University of Texas Press).
Gorard, S. (2007) What does an index of school segregation measure? A commentary on Allen and
Vignoles, Oxford Review of Education, 33(5), 669–677.
Gorard, S., Taylor, C. & Fitz, J. (2003) Schools, markets, and choice policies (London,
RoutledgeFalmer).
Hanushek, E. A., Kain, J. F. & Rivkin, S. G. (2002) New evidence about Brown v. Board of Education:
the complex effects of school racial composition on achievement, NBER working papers (Cambridge,
MA, National Bureau of Economic Research).
Hoxby, C. M. (2000) Does competition among public schools benefit students and taxpayers?
American Economic Review, 90(5), 1209–1238.
Hutchens, R. (2001) Numerical measures of segregation: desirable properties and their implica-
tions, Mathematical Social Sciences, 42, 13–29.
Hutchens, R. (2004) One measure of segregation, International Economic Review, 45(2), 555–578.
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), 21–37.
Le Grand, J. & Bartlett, W. (Eds) (1993) Quasi-markets and social policy (Basingstoke, Macmillan
Press).
Lupton, R. (2005) Social justice and school improvement: improving the quality of schooling in
the poorest neighbourhoods, British Educational Research Journal, 31(5), 589–604.
 School regimes and education equity 459

Organisation for Economic Co-operation and Development (OECD) (2005) PISA 2003 technical
report (Paris, OECD).
Organisation for Economic Co-operation and Development (OECD) (2007) PISA 2006: science
competencies for tomorrow’s world. Vol. 1: Analysis (Paris, OECD).
Opdenakker, M. C. & Van Damme, J. (2001) Relationship between school composition and char-
acteristics of school process and their effect on mathematics achievement, British Educational
Research Journal, 27(4), 407–432.
Scheerens, J. & Bosker, R. J. (1997) The foundation of educational effectiveness (Oxford, Pergamon).
Schütz, G., West, M. R. & Wößmann, L. (2007) School accountability, autonomy, choice and the equity
of student achievement: international evidence from PISA 2003. OECD Education Working
Papers, 14 (Paris, OECD).
Teddlie, C. & Reynolds, D. (2000) The international handbook of school effectiveness research
(London, Falmer Press).
Thrupp, M. (1999) Schools making a difference? Let’s be realistic (Buckingham, Open University
Press).
Waslander, S. & Thrupp, M. (1995) Choice, competition and segregation: an empirical analysis of
a New Zealand secondary school market, 1990–1993, Journal of Education Policy, 10(1), 1–26.
West, A. (2006) School choice, equity and social justice: the case for more control, British Journal
of Educational Studies, 54(1), 15–33.
West, A., Hind, A. & Pennell, H. (2004) School admissions and ‘selection’ in comprehensive
schools: policy and practice, Oxford Review of Education, 30(3), 347–369.
Whitty, G. (1997) Creating quasi-markets in education: a review of recent research on parental
choice and school autonomy in three countries, Review of Research in Education,(22), 3–47.
Willms, J. D. (1986) Social class segregation and its relationship to pupils’ examination results in
Scotland, American Sociological Review, 51(2), 224–241.
460 M. A. Alegre and G. Ferrer

Appendix 1. Models and variables

Table 1A. Explanatory variables: school composition and school process factors

Description Name Variable type Codification

Students’ level
1. Index of economic, social and cultural status escs continuous –
2. Gender (being female) female dummy 0 = male
1 = female
3. Immigration (being native) native dummy 0 = immigrant
1 = native
4. Language (speaking test language) samelang dummy 0 = other language
1 = test language
Schools’ level
5. Proportion of students Low ESCS students escs_low continuous –
according to ESCS in schools
High ESCS students escs_high continuous –
in schools
6. School size xschsize continuous –
7. School management: schools being privately xprivman dummy 0 = public
managed 1 = private
8. School autonomy index in educational content xfacc continuous –
9. Ability grouping (school with ability grouping for xabgr dummy 0 = no
all subjects within school) 1 = yes
10. School with high level of competition xschlcomp dummy 0 = no
1 = yes
11. Responsibility student admission (principal or admprbo dummy 0 = no
teachers/school governing board) 1 = yes
12. School with high academic selectivity of school xhisele dummy 0 = no
admittance 1 = yes
13. School resources: school average number of xstratio continuous –
students per teacher (student-teacher ratio)
14. School resources: school-level index of teacher xtcshort continuous –
shortage
15. Accountability: school posting achievement xacc4 dummy 0 = no
data publicly 1 = yes

Table 2A. Dependent variable: Hutchens school segregation indices

Social distribution of students: Hutchens Segregation Index

School segregation of students below the ESCS 25th percentile HutchensLow

School segregation of students below the ESCS 40th percentile HutchensMidlow
School segregation of students above the ESCS 60th percentile HutchensMidhigh
School segregation of students above the ESCS 75th percentile HutchensHigh
 School regimes and education equity 461

Table 3A. Explanatory variables: school regime factors

Institutional differentiation

First age of selection in the education system firstsel

Number of school tracks available to 15-year-olds tracks15
School providers
Proportion of students in private dependent schools privdep
Proportion of students in private independent schools privindep
School autonomy related to students’ admission
Proportion of students in schools responsible for students’ admission admprbo
Proportion of students in schools considering academic record in admission schoolselabil
Regulation models of school access
Policies of parental choice choicepro
Proportion of students in schools competing for students schcompet

General models of parental choice (among public secondary schools)

1. Model of ‘forced recruitment areas’, which includes those countries and regions
whose local authorities oblige students to enrol at the public school that is located
in their area of residence. The only way to avoid the assigned local school is to
request a place in a private school that is not subject to such a regulation.
2. Model of ‘unforced assignment areas’. As in the previous case, local authorities do
establish the map of catchment areas corresponding to each public school. In the
countries and regions that mainly follow this procedure, pupils are allocated ‘by
default’ to their local nearby school. However, families are permitted to opt for
publicly maintained schools that are outside their area of residence, or even in
another local authority.
3. Model of ‘restricted choice’. Under these schemes families are given a wide
margin of freedom to choose the public school to which they would like to send
their children. In most cases, parents are asked to fill in an application form select-
ing their chosen schools in order of preference. The allocation criteria used by
local admission authorities are implemented only in the case of over-subscribed
schools in order to give priority to the enrolment of those students who live in their
catchment zone and who have placed these schools as their first choice.
4. Model of ‘open choice’. In this case, families are free to choose any of the public
schools in their local authority (or even beyond it). Here, no public regulations
intervene to control the school choices in the hands of parents, nor the admission
criteria to be implemented in the event of over-subscription.