Jump to ContentJump to Main Navigation
Growing up in Diverse SocietiesThe Integration of the Children of Immigrants in England, Germany, the Netherlands, and Sweden$

Frank Kalter, Jan O. Jonsson, Frank van Tubergen, and Anthony Heath

Print publication date: 2018

Print ISBN-13: 9780197266373

Published to University Press Scholarship Online: May 2019

DOI: 10.5871/bacad/9780197266373.001.0001

Show Summary Details
Page of
date: 27 May 2020

Learning Together or Apart? Ethnic Segregation in Lower Secondary Schools

Learning Together or Apart? Ethnic Segregation in Lower Secondary Schools

Chapter:
(p.114) 5 Learning Together or Apart? Ethnic Segregation in Lower Secondary Schools
Source:
Growing up in Diverse Societies
Author(s):

Hanno Kruse

Frank Kalter

Publisher:
British Academy
DOI:10.5871/bacad/9780197266373.003.0005

Abstract and Keywords

Whether, or to what degree, minority students are able to learn together with majority peers in schools is among the important context factors for their integration paths. In this chapter we investigate the extent of ethnic segregation in lower secondary schools in the four CILS4EU countries. We demonstrate that there are vast differences in majority exposure at school, both across the four countries as well as across ethnic groups within each country. Further analyses suggest that these group differences may be due to at least three reasons: ethnic differences in residential segregation, in the allocation across different ability tracks as well as ethnically specific school choice preferences. Finally, we show that low levels of majority exposure at school may not always come with a disadvantaged learning environment: in Germany, the Netherlands and Sweden schools with low majority shares tend to hold fewer learning-related resources; the opposite seems to apply for schools in England.

Keywords:   ability tracking, ethnic minorities, integration, residential segregation, school segregation, school choice, native flight

5.1 Introduction

SCHOOLS ARE CERTAINLY AMONG the most important social contexts for adolescents, and it is here they spend a considerable amount of their time. Schools not only provide access to educational skills and certificates, they also represent the places where friendships are built, romantic relationships are started, and where basic values, attitudes, and cultural habits and preferences are shaped in daily interactions. It is thus needless to emphasise that certain aspects of the school environment can be decisive for the integration of children of immigrants in particular.

One of the most obvious of these characteristics is the ethnic composition of the student body within a school. The question whether children of immigrants are facing high shares of co-ethnics or rather high shares of the majority youth as their schoolmates can make a big difference to their integration paths. By definition, the ethnic composition among the students constitutes a key opportunity structure for inter-ethnic contact (Blau 1977; Joyner & Kao 2000; Moody 2001). As a consequence, it is thus also an important restriction to any behavioural or attitudinal aspect that is prone to influence processes within peer groups. Not least, there is also heavy dispute on whether the ethnic composition of a school or classroom has a direct impact on educational aspirations and the educational performance per se (Hanushek et al. 2002; Burgess et al. 2005; Armor & Duck 2007; Pong & Hao 2007; Sacerdote 2011).

In this chapter we are interested in the basic patterns of ethnic segregation in European schools and in potential basic causes behind these patterns. More precisely, this chapter raises three questions. First, we ask how strongly different ethnic minority groups in the four countries of the Children of Immigrants Longitudinal Survey in Four European Countries (CILS4EU) are exposed to majority students in their daily school lives. Secondly, we look for evidence concerning the main causes of ethnic minority group differences in this exposure to majority students, thereby concentrating on the role of ethnic residential (p.115) segregation and ability tracking in the school system. Finally, we inspect the attending circumstances of learning in a school environment with a relative low versus high share of majority students.

The structure of the chapter is as follows: we begin with a short discussion of the current state of knowledge concerning (group differences in) ethnic segregation in secondary schools in the four CILS4EU countries. Subsequently, we lay out three main causes of school segregation in further detail, thereby addressing the question of how each of the mechanisms might be responsible for differences in the extent of school segregation among ethnic groups. Having thus a clear idea about what the main mechanisms might be, we turn to our empirical cases of interest, that is, the four countries: England, Germany, the Netherlands and Sweden. Our analyses thereby proceed in three steps. First, we establish the extent of ethnic school segregation in terms of majority exposure at school for the different groups; secondly, we provide evidence for or against the different causes leading to (group differences in) school segregation; and finally, we investigate how school environments with high versus low majority exposure differ in terms of different learning-related resources. We end this chapter with some concluding remarks.

5.2 Ethnic Segregation in European Schools: Known Facts and Potential Causes

The central prerequisite for ethnic minority and majority students to be able to ‘learn together’ are ethnically integrated schools. Ethnic segregation across schools—an ethnically specific sorting process of students into administrative school units1—may often hinder this. Despite the striking significance of the topic, the current state of knowledge about the actual extent of ethnic school segregation in the four CILS4EU countries is still rather patchy. Two studies attempted a more systematic cross-country comparison (Gorard & Smith 2004; Schnepf 2006). Analysing the 2000 data from the Programme for International Student Assessment (PISA) and applying a common index, both studies find that the UK seems to show slightly higher levels of ethnic school segregation than the other three countries. Moreover, Gorard & Smith (2004) find that this segregation by ethnicity seems to be more pronounced than segregation by characteristics such as parental occupation or family wealth in the four countries. As one central shortcoming of the studies, however, the ethnicity measure simply distinguishes between those students who are native born versus those who are foreign born, thus providing a rather extreme, potentially exaggerated, image of ethnic segregation.

(p.116) Only few selected country-specific studies were able to apply wider concepts of ethnicity and to differentiate between different ethnic groups. Most research has been conducted in the English context, where it could be found that students of Pakistani background are among the most unevenly distributed groups across schools, closely followed by Bangladeshi students. Black Caribbean and black African students tend to be ethnically concentrated in schools as well, but on substantially lower levels (Burgess et al. 2005). Concerning the Netherlands, there is evidence for ethnic group differences in school segregation based on a nationally representative sample of 126 schools: it suggests that students of Moroccan and Surinamese backgrounds attend schools with the lowest majority shares, closely followed by students of Turkish and Antillean backgrounds (Sykes & Kuyper 2013). For the other countries little is known so far about which ethnic groups face the most ethnically concentrated schools.

In the following we identify three main causes for the emergence of ethnic segregation in schools—neighbourhood segregation, ability tracking and school choice preferences—and discuss how they can be responsible for differences in the extent of school segregation observable between the countries and the different ethnic groups.

5.2.1 The role of residential segregation

The most obvious reason behind ethnic school segregation is that it might simply be a by-product of ethnic residential segregation. The latter is a well-known and common phenomenon in Western European societies. Regardless of the country or region under investigation, previous studies have repeatedly shown that people of the same ethnic background tend to live in locally concentrated areas (Musterd 2005). Whereas the most extreme levels of uneven distributions over geographical spaces can clearly be found in the USA, the extent of residential segregation in the four CILS4EU countries is somewhat more moderate (Alba & Foner 2015). Among these, England has the most ethnically segregated regions and areas, whereas Germany is located on the other side of the spectrum with rather low levels of segregation (Musterd 2005; Alba & Foner 2015).

The potential causes behind ethnic neighbourhood segregation are thus at least indirect reasons for the emergence of ethnic school segregation, to the extent that students are allocated to schools based on geographical or neighbourhood principles (such as closeness to school). Respective research typically discusses and finds support for mechanisms that, roughly, are either based on ethnically specific restrictions or on ethnically specific preferences concerning actors’ residential choices. Most causes have at some point been identified as relevant in at least one of the four countries (e.g. Gramberg 1998; Noreisch 2007; Stevens 2007; Bolt et al. 2008; Auspurg et al. 2011; Stevens et al. 2011; Glikman & Semyonov 2012).

On the side of restriction-related mechanisms, the main starting point is basically of socioeconomic nature (Massey & Denton 1993). Most obviously, ethnic inequalities in income and wealth may lead to a situation where more affluent (p.117) ethnic groups can afford to live in better neighbourhoods, whereas less well-off groups are more constrained in their residential choices, especially so upon their arrival in the receiving country. But there are also a number of more indirect reasons for restriction-driven neighbourhood segregation. Firstly, some ethnic groups rely more heavily on publicly subsidised housing (i.e. social housing) than others. Depending on country-specific idiosyncrasies—such as the extent and local concentration of the publicly funded housing stock—social housing may thus intensify neighbourhood segregation, for example in the Netherlands (Alba & Foner 2015). Secondly, ethnically specific sorting into workplaces and/or locally clustered industries (e.g. Glitz 2014) may be responsible as well, given that living close to work is an important factor driving residential choices. Thirdly, information on vacant housing often stems from informal networks that tend to be ethnically homogeneous (McPherson et al. 2001), resulting in bounded choice alternatives. Fourthly, and finally, discrimination in the housing or credit market may exist, which might penalise or even prevent certain ethnic groups from becoming home owners or from moving into specific (majority-dominated) neighbourhoods.2

Concerning preference-related mechanisms, three main arguments exist. The first is based on residential preferences that only indirectly relate to actors’ ethnicity. For example, ethnic differences in the family or age structure of the households would yield diverging residential preferences, given that young families usually aim for a different neighbourhood infrastructure than, for example, single-headed households. The second argument is based on residential preferences that are explicitly related to actors’ ethnicity. It suggests that immigrants aim for local support from ethnic networks—potentially based on family and kin—especially so upon arrival in the receiving country. Consequently, they settle close to other (same-ethnic) members of their origin group. Finally, the third, most prominent preference-driven mechanism follows the seminal work of Schelling (1978). According to his famous ‘self-forming neighbourhood model’, ethnic residential segregation may arise as an unintended consequence of people’s aim to satisfy their moderate preferences for co-ethnic neighbours.

Whatever the causes behind ethnic residential segregation, the reason why it will lead to the emergence of ethnic school segregation is straightforward: students tend to attend schools that are located close to where they live (Karsten et al. 2003; Burgess et al. 2015). This may be due to legal obligations following from predefined school placement areas. But even in countries like those in CILS4EU, where secondary school choices are not strictly geographically bound, home-to-school distances may nevertheless be minimised for reasons of convenience or due to concerted school choices of neighbouring parents. Obviously, such (p.118) school choices result in ethnic school compositions that reflect those of the local neighbourhoods to a considerable degree.

Given that the degree of ethnic residential segregation varies between countries, and given that in the same country some ethnic groups live in more segregated neighbourhoods than others, this might be a first plausible reason behind country and group differences in school segregation. In terms of country differences, the level of ethnic residential segregation in German, Dutch and Swedish cities seems to be lower than in many English cities (Musterd 2005), which would be in line with the mentioned tendencies in overall ethnic school segregation. Moreover, similar to what has been observed in terms of ethnic differences in school segregation in England, minority members of Pakistani and Bangladeshi background live in the most segregated neighbourhoods, closely followed by Caribbeans and black Africans (Iceland et al. 2011; Simpson 2012). In the Netherlands, those of Turkish and Moroccan background are most unevenly distributed across local areas (Musterd & Ostendorf 2009). From this perspective, the extent to which ethnic groups are segregated in schools seems to align with their extent of neighbourhood segregation to which they are subject. Based on the finding that Eastern Europeans in Germany live in less segregated neighbourhoods than Turks in Germany (Sager 2012; Glitz 2014) we would therefore expect the latter group to also be subject to higher levels of school segregation. Similarly, people of Turkish background in Sweden should attend much more segregated schools than those of Finnish background (Andersson 1998; 2007).

However, it seems to be a common finding that the extent of ethnic school segregation is usually even more extreme than that of residential segregation in the respective regions (Burgess et al. 2005; Johnston et al. 2006). This suggests that school segregation—and related country and group differences—is not solely a by-product of residential segregation, but that there must be further important mechanisms amplifying or complementing the ethnic specific sorting into schools. We lay out the two most important ones in the following: ability tracking and school choice preferences.

5.2.2 Ability tracking

Institutional rules concerning student allocations over different school types may be decisive in the emergence of ethnic school segregation. Arguably one of the most influential features of a school system is the extent to which students’ abilities play a role in what school they attend. Secondary schooling in the four CILS4EU countries differs substantially in this regard (EC/EACEA/Eurydice 2015). Sweden and England pursue a rather integrated policy. Both countries deploy one overarching school type only (at least in the first nine/ten years of schooling). This school type is attended by all students, regardless of their abilities. In contrast, the Netherlands and Germany rely on a rather strict version of ability tracking: contingent upon their performance at the end of primary school, students may or may not be eligible to enter a high-track school, eventually setting the stage for (p.119) higher tertiary education. In the Netherlands, students usually split at the age of 12 into one of several general school types. In simplified terms, they generally have the choice between ‘VMBO’, ‘HAVO’ and ‘VWO’, with only the latter qualifying for subsequent university education (EC/EACEA/Eurydice 2015). Often, however, schools combine several of these tracks under one roof, thus making transitions between them more feasible. In Germany, institutional arrangements vary somewhat across the different federal states. Generally, there are three classical school types, namely ‘Hauptschule’, ‘Realschule’ and ‘Gymnasium’, with the latter qualifying for university education. In addition, there are two different types of comprehensive schools (‘Integrierte Gesamtschule’, ‘Schule mit mehreren Bildungsgängen’) combining the three different ability tracks. The school transition age is also state-specific and varies between 10 and 12 (EC/EACEA/Eurydice 2015).

Allocating students into schools according to their abilities may lead to increased levels of ethnic segregation in schools. It is well known that school achievement correlates with origin group membership in Germany and the Netherlands, mainly due to differences in socioeconomic background (Kristen & Granato 2007; van de Werfhorst & van Tubergen 2007). Thus, minority and majority students are not equally likely to attend high-track schools in both countries; majority members tend to be overrepresented in high-track schools, immigrants’ children in schools of a lower track. Therefore, we should observe the level of ethnic school segregation exceeding that of neighbourhood segregation in these two countries.

But not all ethnic groups are equally likely to attend low-track schools. Those groups with a lower average socioeconomic background (for example, young people of Turkish background in Germany, see Kristen & Granato 2007; Dollmann 2010) are more likely to do so than ethnic groups whose average socioeconomic background is closer to that of the majority (for example, young people of Polish background in Germany, Kristen & Granato 2007; Dollmann 2010). From this perspective, ability tracking (in combination with ethnic differences in high-track attendance rates) may also be responsible for observed group differences in ethnic school segregation.

5.2.3 School choice preferences and native flight

Even in the absence of explicit tracking according to students’ abilities, schools may differ in many other ways. Formally, they can be public or private schools, potentially even of a religious denomination. Further, schools may differ in their curricula (for example, international schools teaching in a foreign language) or in the extent to which they provide extracurricular activities. From this perspective it seems plausible that there are schools which may be more attractive for some ethnic groups than for others.

Aside from these rather formal differences, schools may also vary in their reputation; whether they are generally perceived as being a ‘good’ or a ‘bad’ school. Reputational differences are often closely linked to the socioeconomic (p.120) and ethnic compositions of the student bodies. Research since the 1970s, mainly from the USA, suggests the existence of so-called ‘white flight’: schools with low majority proportions are seen as learning-unfriendly environments and are in consequence avoided by students (and their parents)—especially from the majority population (Coleman et al. 1975; Giles 1976; Clotfelter 2001; for the Netherlands, see Karsten et al. 2003). Along these lines we refer to the same phenomenon in the European context as ‘native flight’, given that we do not investigate racial but ethnic differences (see also Betts & Fairlie 2003; Rangvid 2010). Empirically it is challenging to distinguish explicit native flight—that is, discriminatory ethnic avoidance behaviour—from preferences related to more formal characteristics of the schools that simply correlate with ethnic school compositions. Studies that convincingly provide evidence for the existence of (one of) the two mechanisms therefore barely exist. Nevertheless, we can assert that both tendencies may ultimately lead to an increase in ethnic school segregation.

Furthermore, school choice preferences may be ethnically specific. For example, ethnic groups of Muslim denomination may be attracted to Islamic faith schools whereas ethnic groups with a Catholic background will not (for the Netherlands, see Denessen et al. 2005). Moreover, native flight may be more extreme in schools dominated by ethnic groups with a lower socioeconomic background (for example, Turkish-dominated schools in Germany versus Polish- and Russian-dominated schools in Germany). Ethnic school segregation may, in this regard, also be the result of group differences in school preferences.

5.3 Analyses

5.3.1 School segregation comparing groups and countries

Having laid out the main causes of school segregation—residential segregation, ability tracking, and school preferences—and why the extent of school segregation may be ethnically specific, we will now investigate to what degree ethnic minority groups and the majority learn apart or together, thereby relying on the CILS4EU data. To do so, we take the students in our sample as our units of analysis and look at the share of majority members among their schoolmates. To approximate this share, we estimate the proportion of majority members among the student respondents in a sampled school.3 The resulting variable is an exposure measure of segregation (Massey & Denton 1988) and can be interpreted straightforwardly as being the probability that a randomly picked student from one’s grade is a majority member. According to the elaborations in Chapter 3, majority students are defined as having two native-born parents. The estimate for the share of majority is calculated from the sample data, not counting the respondent him or herself.

(p.121) Figure 5.1 compares this individual-level measure of majority exposure across ethnic groups and countries. The majority exposure of the majority group itself is given in brackets after each country name. In England we find, for example, that the probability that a randomly picked schoolmate of a majority student is also a majority member is on average 0.79. In other words, English majority members are strongly exposed to other majority members in their daily school lives. The other three countries provide a somewhat similar impression, with Germany and Sweden showing the lowest majority exposure values for majority members of 0.76 and 0.75, while the Netherlands has the highest value of 0.86. Of course, these values are to a large degree determined by the overall proportion of minority students in the country sample, which—applying design weights—are 0.73 for England, 0.71 for Germany, 0.83 for the Netherlands and 0.70 for Sweden (see Table 3.1 in Chapter 3). The fact that the average exposure measures for the majority students are larger than these overall proportions of majority students in the sample already points to the fact that the majority–minority distribution across schools is not equal, but that there is a noticeable degree of ethnic school segregation in all four countries.

The four graphs in Figure 5.1 visualise how much the average in-school majority exposure of different ethnic groups differs from the reference group value as indicated by the grey reference line. For example, for a student with Pakistani background in England the probability of a schoolmate being a majority member is only 0.36, resulting in a difference of 0.42, as depicted by the square in the respective row in Figure 5.1.4 The same logic and procedure is applied for the other ethnic groups and in the other three countries.

Among all groups we distinguish in England, Pakistani students have the weakest exposure to majority students and are in these operational terms most segregated, followed by those of Indian background. This corresponds well with prior findings assessing ethnic school segregation for earlier periods, based on different, aggregated data and measures (Burgess et al. 2005).5 All other groups differ significantly from the majority reference category as well, though on a somewhat less pronounced level. In the other three countries, all ethnic minority groups are also clearly less exposed to the majority than the majority itself, the differences fail to reach significance on the chosen level of 95% only for the residual categories ‘other’ in Germany and the Netherlands, Eastern Europe in the Netherlands and Finland in Sweden. However, there is considerable variation in the extent of the exposure between different ethnic minority groups. Among the groups with particularly low majority exposure in schools are—not mentioning the broader summarising categories—Serbs and Turks in Germany, Moroccans and Turks (p.122)

Learning Together or Apart? Ethnic Segregation in Lower Secondary Schools

Figure 5.1 Majority exposure at school in the survey countries

Note: Design weighted, accounting for clustering. Abbreviations in order of appearance: NWS EUR = North, West, South Europe; OTH = Other; MENA+ = North Africa and Middle East plus Afghanistan and Pakistan; S-AFR = Sub-Saharan Africa; E EUR = Eastern Europe; ASIA = South, South-East and East Asia; CAR = Caribbean; IND = India; PAK = Pakistan; RUS = Russia; POL = Poland; ITA = Italy; TUR = Turkey; SER = Serbia; SUR = Suriname; MOR = Morocco; FIN = Finland; KOS = Kosovo; B&H = Bosnia & Herzegovina; IRA = Iraq.

(p.123) in the Netherlands, and students with Iraqi, Turkish, Bosnian-Herzegovinian or Kosovar background in Sweden.

The four graphs in Figure 5.1 raise the obvious question whether and how the strength of ethnic school segregation can also be compared across countries. The visual impression is that the differences are in general somewhat more pronounced in England and Sweden than in Germany and the Netherlands. However, due to the different ethnic group sizes such a comparison is not straightforward. One possible simple indicator could be the difference in majority exposure between the majority and all ethnic minority students in total. In England this would yield –0.22 (= 0.57–0.79), in Germany –0.15 (= 0.61–0.76), –0.13 (= 0.73–0.86) in the Netherlands, and –0.17 (= 0.58–0.75) in Sweden—thus confirming the visual impression. Another potential measure that also takes the differences between ethnic groups into account would be the explained variance of the country-specific regression models underlying the four graphs in Figure 5.1. Given that all four models solely include all group dummies as independent variables, the explained variance therefore indicates to what extent ethnicity may already determine a respondent’s majority exposure at school.6 The respective R2-values are 0.26 in England, 0.18 in Germany, 0.17 in the Netherlands and 0.24 in Sweden (see last rows of model M1 in Tables A5.1–A5.4 in the Appendix).

5.3.2 The impact of neighbourhood segregation and ability tracking

As outlined, the most obvious reason for ethnic school segregation is that it might simply be a consequence of ethnic residential segregation. In our study we have information of the ethnic density in residential areas of respondents in form of subjective reports. Among other questions, we asked the students how many of the people living in their neighbourhood are majority members; they could answer in five categories which were coded with the numbers 1 to 5, ranging from ‘almost none’ (1) over ‘few’ (2), ‘about half’ (3), ‘many’ (4) to ‘almost all’ (5). In Figure 5.2 we use this information to assess the exposure of ethnic minority students to the majority in their neighbourhoods. For reasons of simplicity, we assume the scale of the five response categories to be metric, thus allowing us to apply ordinary least squares (OLS) regression.7 The estimates thus represent the average number of categories from which the ethnic minority groups differ from the majority on the outlined five-point scale.8

We find that members of almost all ethnic minority groups—with few exceptions in the Netherlands and one exception in Germany—feel significantly less exposed to majority members in their neighbourhoods than majority members (p.124) themselves. When we look at how strongly the neighbourhood exposure to the majority differs for the different groups in the four survey countries, we see that the patterns mirror those found for within-school exposure in Figure 5.1 to a considerable degree, though not perfectly. To illustrate this, we kept the vertical order of the ethnic groups in each graph identical to that in Figure 5.1. Accordingly, the groups with the lowest within-school exposure in all four countries are at the bottom of each graph—Pakistanis in England, Serbs in Germany, Moroccans in the Netherlands, Iraqis in Sweden. Figure 5.2 shows that they also report the lowest levels of majority exposure in their neighbourhoods. Pakistanis’ responses in England, for example, are on average 1.5 categories lower than those of the majority students. A similarly clear relation between the relative degrees of school and neighbourhood segregation holds for almost all other ethnic minority groups in all four countries. Deviations in the rank order are most apparent in Germany, which suggests that there might be interesting additional reasons for ethnic school segregation besides ethnic concentration in neighbourhoods and that these might be more influential in Germany than in the other countries.

What degree of the ethnic differences in ethnic school segregation is due to differences in ethnic neighbourhood segregation? A straightforward answer to this central question can be given with our data by looking at the impact of neighbourhood majority exposure when including it as an independent variable into the regression models on ethnic school segregation. Figure 5.3 shows the respective results.9 As in Figure 5.1, it visualises how much the average in-school majority exposure differs from the line for the reference group. The circles in Figure 5.3 represent the relative group effects on school majority exposure when controlling for the subjectively experienced neighbourhood exposure in a multivariate regression model. The values indicated by the squares are identical to those in Figure 5.1. The difference between the squares and circles is thus the part of school segregation that can be explained by (subjectively experienced) neighbourhood segregation.

As expected, group differences in school segregation decline substantially for almost all groups in all four countries. For example, whereas the level of in-school majority exposure of Pakistani students in England is –0.42 smaller than that of majority students (square in respective line), this difference decreases to –0.29 (circle in respective line) assuming that both groups are faced with the same neighbourhood composition. Similar patterns are observable for most other ethnic groups in England and also the other countries. Overall, these results suggest that residential segregation contributes largely to the emergence of school segregation. According to the amount of variance explained by the models this seems specifically to be the case in England and Sweden.10 Intriguingly, however, significant differences remain for most groups in the four destination countries even net of their majority exposure in the neighbourhoods, except for a few groups in Germany and the Netherlands and for Finns in Sweden. This means that while of crucial (p.125)

Learning Together or Apart? Ethnic Segregation in Lower Secondary Schools

Figure 5.2 Perceived majority exposure in the neighbourhood in the survey countries

Note: Design weighted, accounting for clustering. Abbreviations in order of appearance: NWS EUR = North, West, South Europe; OTH = Other; MENA+ = North Africa and Middle East plus Afghanistan and Pakistan; S-AFR = Sub-Saharan Africa; E EUR = Eastern Europe; ASIA = South, South-East and East Asia; CAR = Caribbean; IND = India; PAK = Pakistan; RUS = Russia; POL = Poland; ITA = Italy; TUR = Turkey; SER = Serbia; SUR = Suriname; MOR = Morocco; FIN = Finland; KOS = Kosovo; B&H = Bosnia & Herzegovina; IRA = Iraq.

(p.126)

Learning Together or Apart? Ethnic Segregation in Lower Secondary Schools

Figure 5.3 Role of neighbourhood segregation for ethnic school segregation in the survey countries

Note: Design weighted, accounting for clustering. Abbreviations in order of appearance: NWS EUR = North, West, South Europe; OTH = Other; MENA+ = North Africa and Middle East plus Afghanistan and Pakistan; S-AFR = Sub-Saharan Africa; E EUR = Eastern Europe; ASIA = South, South-East and East Asia; CAR = Caribbean; IND = India; PAK = Pakistan; RUS = Russia; POL = Poland; ITA = Italy; TUR = Turkey; SER = Serbia; SUR = Suriname; MOR = Morocco; FIN = Finland; KOS = Kosovo; B&H = Bosnia & Herzegovina; IRA = Iraq.

(p.127) importance, neighbourhood segregation is not the entire reason for school segregation in our survey countries.

A certain weakness in these analyses lies in the fact that the majority exposure in the neighbourhood is measured by a subjective assessment only, which might not perfectly match the facts. The contribution of neighbourhood segregation to the explanation of school segregation might thus be somewhat under- or overrated. Fortunately, in two of the four survey countries, Germany and the Netherlands, we were able to match objective information on the respondents’ neighbourhood composition to our data.11 It turns out that these objective measures modestly correlate with the perceived majority exposure as reported by the respondents. In Germany, the subjective and objective measure align more closely than in the Netherlands (Germany: r = 0.51, the Netherlands: r = 0.39), which may have to do with the fact that the objective measures in the two countries slightly differ in terms of their geographical scales.

Figure 5.4 shows how the objective measure compares to the subjective one when including it as an independent variable into the model explaining ethnic school segregation.12 The squares and circles represent the same values as in Figure 5.3, that is, the gross group effects and the group effect net of subjective majority neighbourhood segregation. The triangles now show the respective values when using the objective measure instead. We find that in both Germany and the Netherlands the objective indicator explains a little more of the ethnic school segregation when looking at the groups with the larger differences. That means that the subjective measure somewhat underestimates the impact of neighbourhood segregation in these cases. For the groups with the smaller differences the opposite tends to be the case. Nevertheless, the deviations between the effects resulting from the objective and subjective measure are only small, and nowhere significant. Most importantly, the basic conclusion stays unchanged: neighbourhood exposure—at least as conceptualised here—explains a lot, but by far not all group differences in ethnic school segregation that we observe.

As discussed, in both Germany and the Netherlands students at the age of 14, that is, our sample population, have already been tracked into different school types according to their abilities, while this is not the case in the more integrated school systems of England and Sweden. Based on data from the former two countries, we can thus study whether ability tracking might be responsible for the group differences in within-school majority exposure. To do so, we estimated a further model specification controlling not only for the objective majority exposure in the neighbourhood of the respondent but also including additional school type dummies capturing whether the students attend a low-, intermediate- or high-track school. The remaining group effects after including these controls are (p.128)

Learning Together or Apart? Ethnic Segregation in Lower Secondary Schools

Figure 5.4 Role of neighbourhood segregation and ability tracking in Germany and the Netherlands

Note: Design weighted, accounting for clustering. Abbreviations in order of appearance: OTH = Other; NWS EUR = North, West, South Europe; ASIA = South, South-East and East Asia; E EUR = Eastern Europe; RUS = Russia; POL = Poland; ITA = Italy; MENA+ = North Africa and Middle East plus Afghanistan and Pakistan; TUR = Turkey; S-AFR = Sub-Saharan Africa; SER = Serbia; CAR = Caribbean; SUR = Suriname; MOR = Morocco.

indicated by the diamonds in Figure 5.4. We find that the ethnic differences in school segregation notably decrease in Germany. In the Netherlands, the change in ethnic differences remains rather small compared to the results from the previous model. From this perspective, it seems as if ability tracking does play an important role for the emergence of school segregation in Germany, but to a much lesser degree in the Netherlands.

In general we find that ethnic neighbourhood segregation and ability tracking alone are not able to account for all the observed ethnic differences in school segregation, but that considerable differences in within-school majority exposure remain for most of the ethnic minority groups in the four survey countries. Given that some of the remaining group differences are still of substantial size it also seems unlikely that this is due to measurement issues only. This means that additional mechanisms might play an important role as well. As discussed, school preferences—especially in the form of avoidance behaviour by the majority (‘native flight’)—are a frequently suggested possibility. The results might thus be read as indirect evidence for this phenomenon to be of empirical importance in the four countries.

(p.129) 5.3.3 Majority exposure in school and learning-related school characteristics

The revealed group differences in majority exposure in school gave rather concrete insight to what degree majority and minority students learn together or apart. Further, we now have important empirical hints on the potential causes underlying ethnic group differences in the daily exposure to majority peers. In a final step, we now want to inspect what a higher or lower majority exposure means in terms of the more general learning environment. Put somewhat more technically, we investigate to what extent learning-related school characteristics may correlate with majority exposure in school in the four countries. Such correlated school characteristics seem rather likely: in most Western European countries socioeconomic disparities exist between the majority population and ethnic minority groups, with the latter usually being in disadvantaged positions. Given that a higher socioeconomic background usually implies better access to various kinds of resources (economic, social etc.), it seems likely that a higher majority exposure in school also implies a better-equipped, more conducive learning environment.

Figure 5.5 shows how parents’ average socioeconomic status (SES) (in terms of their highest International Socio-Economic Index of Occupational Status (ISEI) score, see Ganzeboom et al. 1992) relates to the schools’ majority proportions. Unsurprisingly, the ethnic composition of a school is closely associated with its socioeconomic composition in Germany, the Netherlands and Sweden, the bivariate relation being significantly positive; the higher the majority proportion in a school the higher its average parental ISEI score. However, this is much less the case in England. Here, we observe an inverse u-shaped relation, as the grey trend line suggests.13 The highest average parental ISEI scores can thus be found in schools with moderate majority proportions. What makes the English case so peculiar is that its largest minority group (i.e. Indian) does not have a lower average SES than the majority population but rather a higher one (see Chapter 2). Concerning the presence of learning-related resources in schools with high and low majority proportions we should therefore expect to see different patterns in England to those in the other three countries.

Among the most important learning conditions provided by the school context is the general cognitive ability level of the students. One powerful feature of the CILS4EU data is that it entails an objective and internationally comparable measure for students’ cognitive abilities. It is based on a written ten-minute test that all students in the survey were asked to take. The test consisted of 27 graphical problems that could either be solved correctly or not, resulting in a maximally attainable score of 27 (Weiß 2006). Given that the test was solely based on graphics and was free of any language (except for the instructions that (p.130)

Learning Together or Apart? Ethnic Segregation in Lower Secondary Schools

Figure 5.5 Majority proportion and average parental occupational status (ISEI scores) in schools in the survey countries

Note: Not weighted; * p<.05.

had been translated in accordance with the test publishers), its comparability across countries could be guaranteed (CILS4EU 2016). The first line in Table 5.1 reports to what extent average test scores correlate with majority presence in schools across the four countries. In Germany and Sweden, the correlation is significantly positive, thus showing that schools with higher majority proportions tend to have significantly higher average ability levels. Or, putting it the other way round, attending a school with a higher minority proportion means being surrounded by schoolmates with lower cognitive abilities on average. In the other two countries, however, this does not hold true. In England, we even observe an (insignificant) negative correlation.

Table 5.1. Correlates of a high majority proportion in schools across survey countries

England

Germany

Netherlands

Sweden

Cognitive abilities

−0.10

0.39*

0.19

0.30*

Language proficiency

0.08

0.54*

0.30*

0.56*

Single parenthood

−0.10

−0.17*

−0.48*

−0.35*

Cash margin available

−0.09

0.10

0.16

0.07

Delinquency

−0.07

−0.20*

−0.15

0.00

Note: Design weighted, correlation coefficients at school level;

(*) p < 0.05.

(p.131) Turning to further school-level characteristics, similar country-specific patterns emerge. In Germany and Sweden, the schools’ majority proportion is strongly associated with average proficiency in the destination-country language, which was measured, among others, by country-specific synonym tests in the CILS4EU study (see also Chapter 9). In the Netherlands, this association exists as well, but on a more moderate level. Again, the only exception is England where the correlation is zero. This may be due to the fact that the mother tongue of many minority groups in England is also English. But again, it may also have to do with minority groups in England being not necessarily disadvantaged in terms of their SES (see Chapter 2). This is also further reflected in the fact that England is the only country where a school’s majority proportion is not associated with the prevalence of single parenthood (see also Chapter 6). Being able to provide a certain amount of cash money (see also Chapter 4) is not more or less common in majority-dominated schools in all four countries. The same also holds, finally, for delinquency (see also Chapter 13) in terms of deliberately damaging things that belong to others, theft, or carrying a knife or weapon, except for Germany, where schools with high majority proportions show significantly lower levels.

To summarise, the patterns observable concerning the correlation between majority exposure in school and learning-related school characteristics seem to be rather complex and country-specific. In Germany, attending a school with high minority shares is clearly associated with being faced with a learning-unfriendly environment. In Sweden, this seems to be the case as well, except that there is indication that ideational aspirations are higher. The Netherlands provides an overall mixed image. Finally, in England, the opposite tendency seems to apply: the higher the minority share at school, the more learning-conducive the school context.

5.4 Discussion

In this chapter we studied to what extent majority and minority adolescents learn together in integrated school environments in the four CILS4EU countries. The extent to which different ethnic minority groups are exposed to majority students in their schools gave us a very differentiated answer. Though almost all ethnic minority groups in all four countries face significantly fewer majority students among their schoolmates than the majority students themselves, for some groups this difference is rather small, for others rather large. Among the most segregated origin groups were Pakistanis in England, Serbs in Germany, Moroccans in the Netherlands and Iraqis in Sweden. However, even these most segregated groups are far from being fully separated from the majority. By the way, and worth emphasising, the analyses also tell, as the flipside of the coin, an interesting story about the exposure of majority students to minority students: majority students, and this is an overarching finding for all four countries, are the ones with the (p.132) least exposure to other ethnic groups, thus are the group with the highest risk of ‘learning apart’.

There is strong indication that, as was expected, ethnic segregation in neighbourhoods is the most important factor driving the emergence of segregation in schools. Patterns of ethnic differences in residential segregation are rather similar to those of school segregation, and the majority exposure in the neighbourhood is able to account for a substantial part of the group differences in exposure to majority peers in school. This holds for all four survey countries, most strongly in Sweden and England. In the Netherlands and in Germany we were able to analyse the impact of a second central potential cause leading to school segregation: ability tracking. While, surprisingly, the empirical explanatory power in the Netherlands turned out to be only marginal, the results clearly support the view that this is an important factor in Germany. In spite of all evidence for the validity of both mechanisms, overall results suggest that residential segregation and the existence of ability tracking alone cannot account fully for the observed group differences in the four countries. It therefore seems likely that students’ (or parents’, respectively) school choice preferences may play a role for the emergence of ethnic school segregation as well. Whether this is actually the case, however, could not be directly investigated here.

Finally, our analyses suggest that the learning conditions students face at school vary with the school’s majority proportion. Higher majority proportions are associated with more favourable socioeconomic compositions in Germany, the Netherlands and Sweden, and thus with a number of context characteristics that may be decisive for students’ further life chances. Interestingly, however, this is not the case in England. This is mainly due to the fact that minority members of Asian and especially Indian background in England are on average of higher socioeconomic background than the majority population. In the English case it would therefore be false to assume that attending a school with a high minority proportion would automatically imply being faced with a less favourable learning environment.

Our main intent in this chapter was to provide a descriptive perspective on the most important social environment of immigrant youths in the CILS4EU countries, namely their school (and neighbourhood) contexts. Exploiting one central strength of CILS4EU we tried to keep the analyses as comparative as possible. As so often, this comes at a cost: for example, to guarantee an internationally comparable perspective for all four survey countries we relied on respondents’ subjective reports of their neighbourhood compositions instead of more objective measures. Even though we could show for two countries that our results would not change substantially when using a more objective measure, this is nevertheless a shortcoming of our analyses to take into account. Another point to keep in mind is that all analyses rest on linear and thus theoretically simplified models. We deemed this parsimonious modelling approach to be appropriate for our descriptive purposes. A serious investigation of the causes of school segregation, however, would have to account for a complex, most likely (p.133) non-linear process of sorting into schools. It should therefore be stressed that the evidence we provide concerning the causes leading to school segregation are mere indications and do by no means serve as definite proof for or against their existence.

Our findings suggest that there may be young people with an immigrant background in the CILS4EU countries who indeed have little contact with the majority in their everyday school environments. Pakistani students in England, for example, have on average only 36% majority schoolmates. However, other groups are exposed to over 70% majority members in their schools, such as the Northern/Western/Southern European group throughout all four countries. As these examples suggest, ‘learning apart’ from the majority is a group-specific phenomenon and seems to be rather an exception than the rule. Further chapters of this book digging deeper into the cogs and wheels of young immigrants’ integration processes will have to account for this fact when trying to explain why some students seem to be more integrated than others in Western European societies.

References

Bibliography references:

Alba, R. & Foner, N. (2015), Strangers no More: Immigration and the Challenges of Integration in North America and Western Europe (Princeton NJ, University Press).

Andersson, R. (1998), ‘Socio-Spatial Dynamics: Ethnic Divisions of Mobility and Housing in Post-Palme Sweden, Urban Studies, 35(3): 397–428.

Andersson, R. (2007), ‘Ethnic Residential Segregation and Integration Processes in Sweden’, in K. Schönwälder (ed.), Residential Segregation and the Integration of Immigrants: Britain, the Netherlands and Sweden (Berlin, WZB Wissenschaftszentrum Berlin für Sozialforschung), 61–90.

Armor, D.J. & Duck, S. (2007), ‘The Effect of Black Peers on Black Test Scores’, Working Paper, Arkansas, Education Working Paper Archive.

Auspurg, K., Hinz, T. & Schmid, L. (2011), ‘Contexts and Conditions of Ethnic Discrimination: Evidence from a Field Experiment in German Housing Markets’, Working Paper, Konstanz, Konstanz University.

Betts, J.R. & Fairlie, R.W. (2003), ‘Does Immigration Induce “Native Flight” from Public Schools into Private Schools?’, Journal of Public Economics, 87(5): 987–1012.

Blau, P.M. (1977), Inequality and Heterogeneity: A Primitive Theory of Social Structure, 7th edn (New York, Free Press).

Bolt, G., van Kempen, R. & van Ham, M. (2008), ‘Minority Ethnic Groups in the Dutch Housing Market: Spatial Segregation, Relocation Dynamics and Housing Policy’, Urban Studies, 45(7): 1359–84.

Burgess, S., Wilson, D. & Lupton, R. (2005), ‘Parallel Lives? Ethnic Segregation in Schools and Neighbourhoods’, Urban Studies, 42(7): 1027–56.

Burgess, S., Greaves, E., Vignoles, A. & Wilson, D. (2015), ‘What Parents Want: School Preferences and School Choice’, The Economic Journal, 125(587): 1262–89.

CILS4EU (Children of Immigrants Longitudinal Survey in Four European Countries) (2016), Children of Immigrants Longitudinal Survey in Four European Countries: Technical Report: Wave 1–2010/2011, v1.2.0 (Mannheim, Mannheim University).

(p.134) Cleveland, W.S. (1979), ‘Robust Locally Weighted Regression and Smoothing Scatterplots’, Journal of the American Statistical Association, 74(368): 829–36.

Clotfelter, C.T. (2001), ‘Are Whites Still Fleeing? Racial Patterns and Enrollment Shifts in Urban Public Schools, 1987–1996’, Journal of Policy Analysis and Management, 20(2): 199–221.

Coleman, J.S., Kelly, S. & Moore, J. (1975), ‘Trends in School Desegregation, 1968–1973’, Research Report, Washington DC, The Urban Institute.

Denessen, E., Driessena, G. & Sleegers, P. (2005), ‘Segregation by Choice? A Study of Group-Specific Reasons for School Choice’, Journal of Education Policy, 20(3): 347–68.

Dollmann, J. (2010), Türkischstämmige Kinder am ersten Bildungsübergang (Wiesbaden, VS Verlag für Sozialwissenschaften).

EC/EACEA/Eurydice (European Commission/Education, Audiovisual and Culture Executive Agency/Eurydice) (2015), The Structure of the European Education Systems 2015/16: Schematic Diagrams: Eurydice Facts and Figures (Luxembourg, Publications Office of the European Union).

Ganzeboom, H.B., De Graaf, P.M. & Treiman, D.J. (1992), ‘A Standard International Socio-Economic Index of Occupational Status’, Social Science Research, 21(1): 1–56.

Giles, M.W. (1976), ‘Determinants of Resegregation: Compliance/Rejection Behavior and Policy Alternatives’, Research Report, Washington DC, The Urban Institute.

Glikman, A. & Semyonov, M. (2012), ‘Ethnic Origin and Residential Attainment of Immigrants in European Countries’, City & Community, 11(2): 198–219.

Glitz, A. (2014), ‘Ethnic Segregation in Germany’, Labour Economics, 29: 28–40.

Gorard, S. & Smith, E. (2004), ‘An International Comparison of Equity in Education Systems’, Comparative Education, 40(1): 15–28.

Gramberg, P. (1998), ‘School Segregation: The Case of Amsterdam’, Urban Studies, 35(3): 547–64.

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, Working Paper No. 8741, Cambridge, National Bureau of Economic Research.

Iceland, J., Mateos, P. & Sharp, G. (2011), ‘Ethnic Residential Segregation by Nativity in Great Britain and the United States’, Journal of Urban Affairs, 33(4): 409–29.

Johnston, R., Burgess, S., Wilson, D. & Harris, R. (2006), ‘School and Residential Ethnic Segregation: An Analysis of Variations across England’s Local Education Authorities’, Regional Studies, 40(9): 973–90.

Joyner, K. & Kao, G. (2000), ‘School Racial Composition and Adolescent Racial Homophily’, Social Science Quarterly, 81(3): 810–25.

Karsten, S., Ledoux, G., Roeleveld, J., Felix, C. & Elshof, D. (2003), ‘School Choice and Ethnic Segregation’, Educational Policy, 17(4): 452–77.

Kristen, C. & Granato, N. (2007), ‘The Educational Attainment of the Second Generation in Germany Social Origins and Ethnic Inequality’, Ethnicities, 7(3): 343–66.

Massey, D.S. & Denton, N.A. (1988), ‘The Dimensions of Residential Segregation’, Social Forces, 67(2): 281–315.

Massey, D.S. & Denton, N.A. (1993), American Apartheid: Segregation and the Making of the Underclass, 8th edn (Cambridge MA, Harvard University Press).

McPherson, M., Smith-Lovin, L. & Cook, J. M. (2001), ‘Birds of a Feather: Homophily in Social Networks’, Annual Review of Sociology, 27: 415–44.

(p.135) Moody, J. (2001), ‘Race, School Integration, and Friendship Segregation in America’, American Journal of Sociology, 107(3): 679–716.

Musterd, S. (2005), ‘Social and Ethnic Segregation in Europe: Levels, Causes, and Effects’, Journal of Urban Affairs, 27(3): 331–48.

Musterd, S. & Ostendorf, W. (2009), ‘Residential Segregation and Integration in the Netherlands’, Journal of Ethnic and Migration Studies, 35(9): 1515–32.

Noreisch, K. (2007), ‘School Catchment Area Evasion: The Case of Berlin, Germany’, Journal of Education Policy, 22(1): 69–90.

Pong, S.L. & Hao, L. (2007), ‘Neighborhood and School Factors in the School Performance of Immigrants’ Children’, International Migration Review, 41(1): 206–41.

Rangvid, B.S. (2010), ‘School Choice, Universal Vouchers and Native Flight from Local Schools’, European Sociological Review, 26(3): 319–35.

Sacerdote, B. (2011), ‘Peer Effects in Education: How Might They Work, How Big are They and How Much Do We Know Thus Far?’, in E.A. Hanushek, S.J. Machin & L. Woessmann (eds), Handbook of the Economics of Education, 3rd edn (Amsterdam, North-Holland), 249–77.

Sager, L. (2012), ‘Residential Segregation and Socioeconomic Neighbourhood Sorting: Evidence at the Micro-Neighbourhood Level for Migrant Groups in Germany’, Urban Studies, 49(12): 2617–32.

Schelling, T.C. (1978), Micromotives and Macrobehavior (New York, Norton).

Schnepf, S.V. (2006), ‘How Different are Immigrants? A Cross-Country and Cross-Survey Analysis of Educational Achievement’, in C.A. Parsons & T.M. Smeeding (eds), Immigration and the Transformation of Europe (Cambridge, Cambridge University Press), 200–34.

Simpson, L. (2012), More Segregation or More Mixing? (Manchester, Centre for Dynamics of Ethnicity).

Stevens, P.A. (2007), ‘Researching Race/Ethnicity and Educational Inequality in English Secondary Schools: A Critical Review of the Research Literature between 1980 and 2005’, Review of Educational Research, 77(2): 147–85.

Stevens, P.A., Clycq, N., Timmerman, C. & van Houtte, M. (2011), ‘Researching Race/Ethnicity and Educational Inequality in the Netherlands: A Critical Review of the Research Literature between 1980 and 2008’, British Educational Research Journal, 37(1): 5–43.

Sykes, B. & Kuyper, H. (2013), ‘School Segregation and the Secondary-School Achievements of Youth in the Netherlands’, Journal of Ethnic and Migration Studies, 39(10): 1699–716.

Van de Werfhorst, H.G. & van Tubergen, F. (2007), ‘Ethnicity, Schooling, and Merit in the Netherlands’, Ethnicities, 7(3): 416–44.

Weiß, R.H. (2006), Grundintelligenztest CFT 20-R (Göttingen, Hogrefe).

Table A5.1. Multivariate analysis (OLS regression) of majority exposure at school: England

Model 1

Model 2

Origin groups (ref.: majority)

   North, West, South Europe

−0.091***

−0.057***

(0.021)

(0.018)

   Eastern Europe

−0.226***

−0.159***

(0.042)

(0.036)

   Caribbean

−0.298***

−0.215***

(0.042)

(0.035)

   MENA+

−0.216***

−0.170***

(0.041)

(0.037)

   Pakistan

−0.424***

−0.289***

(0.043)

(0.035)

   Sub-Saharan Africa

−0.222***

−0.167***

(0.039)

(0.029)

   Asia

−0.240***

−0.166***

(0.025)

(0.03)

   India

−0.313***

−0.218***

(0.063)

(0.045)

   Other

−0.193***

−0.153***

(0.04)

(0.035)

Majority in neighbourhood (subj., ref.: almost none)

   Few

−0.008

(0.036)

   About half

0.081*

(0.04)

   Many

0.226***

(0.042)

   Almost all

0.268***

(0.045)

Intercept

0.787***

0.560***

(0.015)

(0.039)

No. of obs. (students)

4,315

4,202

   R2

0.26

0.38

Note: Design weighted, accounting for clustering; standard errors in parentheses;

(*) p < 0.05,

** p < 0.01,

(***) p < 0.001. MENA+: Middle East and North Africa plus Afghanistan. (p.137)

Table A5.2. Multivariate analysis (OLS regression) of majority exposure at school: Germany

Model 1

Model 2

Model 3

Model 4

Origin groups (ref.: majority)

   North, West, South Europe

−0.060**

−0.033

−0.021

−0.011

(0.022)

(0.022)

(0.018)

(0.014)

   Italy

−0.165***

−0.133***

−0.082**

−0.048

(0.042)

(0.038)

(0.03)

(0.026)

   Eastern Europe

−0.071*

−0.030

−0.028

−0.018

(0.028)

(0.023)

(0.02)

(0.015)

   Poland

−0.110***

−0.073***

−0.079***

−0.064**

(0.025)

(0.021)

(0.022)

(0.021)

   Russia

−0.107***

−0.052

−0.061**

−0.031

(0.029)

(0.028)

(0.023)

(0.024)

   Serbia

−0.293***

−0.222***

−0.165***

−0.096***

(0.037)

(0.03)

(0.026)

(0.022)

   MENA+

−0.226***

−0.168***

−0.131***

−0.112***

(0.031)

(0.029)

(0.027)

(0.027)

   Turkey

−0.240***

−0.174***

−0.112***

−0.080***

(0.024)

(0.021)

(0.019)

(0.019)

   Sub-Saharan Africa

−0.252***

−0.187***

−0.112***

−0.089***

(0.035)

(0.033)

(0.028)

(0.023)

   Asia

−0.071*

−0.048

−0.020

−0.023

(0.031)

(0.028)

(0.023)

(0.021)

   Other

−0.036

−0.022

−0.024

0.001

(0.028)

(0.022)

(0.025)

(0.03)

Majority in neighbourhood (subj., ref.: almost none)

   Few

0.069*

(0.03)

   About half

0.131***

(0.031)

   Many

0.204***

(0.028)

   Almost all

0.267***

(0.029)

Majority in neighbourhood

0.015***

0.013***

(obj., in %) School type (ref.: Lower sec.)

(0.001)

(0.001)

   School comb. sev. tracks

0.247***

(0.04)

   Intermediate secondary

0.168***

(0.03)

   Comprehensive

0.183***

(0.031)

    (p.138) Upper secondary

0.181***

(0.024)

   School for special needs

0.119**

(0.042)

   Rudolf Steiner

0.272***

(0.021)

Intercept

0.755***

0.528***

−0.641***

−0.651***

(0.015)

(0.028)

(0.084)

(0.085)

No. of obs. (students)

5,013

4,970

4,587

4,587

R2

0.18

0.27

0.37

0.49

Note: Design weighted, accounting for clustering; standard errors in parentheses;

(*) p < 0.05,

(**) p < 0.01,

(***) p < 0.001;

MENA+: Middle East and North Africa (except Turkey) plus Afghanistan and Pakistan.

Table A5.3. Multivariate analysis (OLS regression) of majority exposure at school: Netherlands

Model 1

Model 2

Model 3

Model 4

Origin groups (ref.: majority)

   North, West, South Europe

−0.051*

−0.040

−0.039

−0.032

(0.024)

(0.026)

(0.021)

(0.017)

   Eastern Europe

−0.052

−0.045

−0.019

−0.016

(0.032)

(0.027)

(0.031)

(0.028)

   Caribbean

−0.133***

−0.109***

−0.068

−0.064

(0.032)

(0.031)

(0.035)

(0.034)

   Suriname

−0.161***

−0.135**

−0.077***

−0.077***

(0.047)

(0.042)

(0.023)

(0.023)

   MENA+

−0.085**

−0.061*

−0.046*

−0.050**

(0.031)

(0.025)

(0.02)

(0.019)

   Morocco

−0.319***

−0.278***

−0.202***

−0.204***

(0.047)

(0.044)

(0.034)

(0.035)

   Turkey

−0.239***

−0.206***

−0.136***

−0.132***

(0.038)

(0.036)

(0.026)

(0.026)

   Sub-Saharan Africa

−0.147***

−0.128***

−0.103***

−0.105***

(0.035)

(0.034)

(0.023)

(0.023)

   Asia

−0.077***

−0.068**

−0.044*

−0.038*

(0.023)

(0.021)

(0.018)

(0.016)

   Other

−0.040

−0.036

−0.003

−0.002

(0.035)

(0.034)

(0.034)

(0.029)

Majority in neighbourhood (subj., ref.: almost none)

   Few

−0.038*

(0.017)

   About half

−0.037

(0.021)

    (p.139) Many

0.007

(0.02)

   Almost all

0.050***

(0.013)

Majority in neighbourhood

0.006***

0.006***

(obj., in %) School type (ref.: Lower sec.)

(0.001)

(0.001)

   School comb. sev. tracks

0.044

(0.044)

   Intermediate secondary

0.055

(0.034)

   Comprehensive

0.063

(0.035)

   Upper secondary

0.009

(0.042)

Intercept

0.855***

0.828***

0.309***

0.272***

(0.014)

(0.02)

(0.064)

(0.067)

No. of obs. (students)

4,363

4,323

4,362

4,362

R2

0.17

0.21

0.40

0.41

Note: Design weighted, accounting for clustering; standard errors in parentheses;

(*) p < 0.05,

(**) p < 0.01,

(***) p < 0.001;

MENA+: Middle East and North Africa (except Turkey and Morocco) plus Afghanistan and Pakistan.

Table A5.4. Multivariate analysis (OLS regression) of majority exposure at school: Sweden

Model 1

Model 2

Origin groups (ref.: majority)

   North, West, South Europe

−0.064***

−0.054***

(0.015)

(0.015)

   Finland

−0.028

−0.002

(0.019)

(0.019)

   Eastern Europe

−0.177***

−0.145***

(0.029)

(0.027)

   Bosnia & Herzegovina

−0.302***

−0.253***

(0.031)

(0.027)

   Kosovo

−0.223***

−0.185***

(0.039)

(0.033)

   MENA+

−0.284***

−0.241***

(0.025)

(0.022)

    (p.140) Iraq

−0.347***

−0.273***

(0.034)

(0.033)

   Turkey

−0.314***

−0.267***

(0.046)

(0.042)

   Sub–Saharan Africa

−0.330***

−0.262***

(0.038)

(0.036)

   Asia

−0.117***

−0.091***

(0.023)

(0.021)

   Other

−0.089***

−0.056**

(0.021)

(0.019)

Majority in neighbourhood (subj., ref.: almost none)

   Few

−0.102***

(0.021)

   About half

−0.063**

(0.024)

   Many

0.041*

(0.02)

   Almost all

0.078***

(0.02)

Intercept

0.752***

0.706***

(0.01)

(0.021)

No. of obs. (students)

5,025

4,856

R2

0.24

0.30

Note: Design weighted, accounting for clustering; standard errors in parentheses;

(*) p < 0.05,

(**) p < 0.01,

(***) p < 0.001;

MENA+: Middle East and North Africa (except Iraq and Turkey) plus Afghanistan and Pakistan.

Notes:

(1) We refer to ethnic segregation across administrative school units rather than across school locations throughout the chapter in order to account for constellations where separate administrative school units—often resembling different school tracks—are located within the same school building complex (mostly prevalent in the Netherlands).

(2) Discriminatory behaviour in the housing or credit market is based on ethnically specific preferences usually on the side of the majority. It is thus a matter of perspective whether to define discrimination as a restriction for the residential choices of the ethnic minority groups or, alternatively, as a (majority) preference-related mechanism.

(3) In each sampled school, participation in the survey was restricted to two classrooms only, chosen at random.

(4) All analyses underlying Figure 5.1 are based on the linear regression model M1, which can be found in Tables A5.1–A5.4 in the Appendix. We use the standard error of this estimate to draw a line representing the 95% confidence interval around the squares. All analyses are design weighted and the estimates of the standard errors account for clustering in schools.

(5) Note that in other studies Bangladeshi students often turn out to be even slightly more strongly segregated than Pakistani, but they could not be distinguished separately here given the number of cases in the sample (see Chapter 3).

(6) One has to be careful, however, as this measure is dependent on the broadness of the grouping and thus not ‘compositional invariant’ (Massey & Denton 1988: 283–7), meaning that if we combine groups into broader categories the R2-value naturally gets smaller.

(7) Alternative specifications that account for the ordinal nature of the measure (i.e. ordered logistic regression) provide substantially identical results.

(8) Regression models underlying the estimates in Figure 5.2 are not shown here.

(9) Figure 5.3 is based on models M1 and M2 in Tables A5.1–A5.4 in the Appendix.

(10) The respective R2-values can be found in the last rows of Tables A5.1–A5.4 in the Appendix.

(11) Whereas the objective measure of the majority share in respondents’ neighbourhoods is based on information from federal statistics in the Netherlands, it relies on name-based classification by the private geo-marketing company Microm in Germany.

(12) Figure 5.4 is based on models M1, M2, M3 and M4 in Tables A5.2 and A5.3 in the Appendix.

(13) The trend line is computed using locally weighted polynomial regression, better known as the locally weighted scatter-plot smoother (LOWESS) procedure (Cleveland 1979).