Inclusive attitudes towards immigrants: a … · Web viewGiven that this belief is generally...

53
Do Ethnically Mixed Classrooms Promote Inclusive Attitudes Towards Immigrants Everywhere? A Study Among Native Adolescents in 14 Western Countries Abstract: This paper examines attitudes among 14-year-old native students in 14 Western countries to assess how out- group size, as measured by the proportion of first- and second-generation migrant children in a class, is related to inclusive views on immigrants. It develops three competing hypotheses: (1) higher proportions of immigrants contribute to inclusive views everywhere; (2) higher proportions have a negative effect on inclusive views everywhere; (3) the effect of out-group size depends on the ratio of first- to second- generation migrant children: the higher this ratio, the weaker the effect. It discovers that out-group size is positively related to inclusive views on immigrants in countries where second-generation outnumber first-generation migrant children (i.e. the old immigration states), and that there is no significant link with such views in countries where the reverse is the case (i.e. the new immigration states). The same regularity applies at the classroom level: in classes with more second than first generation students, out-group size enhances inclusive views while it shows no relationship to such views in classes with more first than second generation students. The results thus support the third hypothesis. The non-relation in contexts with many first generation students may well be a temporary phenomenon, however. Once immigrant communities have become more settled and integrated in the destination countries, positive effects of ethnic mixing could well emerge everywhere. Key words: inter-group contact, inclusive attitudes, classroom ethnic composition, contact theory, conflict theory, ICCS study Introduction

Transcript of Inclusive attitudes towards immigrants: a … · Web viewGiven that this belief is generally...

Do Ethnically Mixed Classrooms Promote Inclusive Attitudes Towards Immigrants

Everywhere? A Study Among Native Adolescents in 14 Western Countries

Abstract: This paper examines attitudes among 14-year-old native students in 14 Western countries to assess how out-group size, as measured by the proportion of first- and second-generation migrant children in a class, is related to inclusive views on immigrants. It develops three competing hypotheses: (1) higher proportions of immigrants contribute to inclusive views everywhere; (2) higher proportions have a negative effect on inclusive views everywhere; (3) the effect of out-group size depends on the ratio of first- to second-generation migrant children: the higher this ratio, the weaker the effect. It discovers that out-group size is positively related to inclusive views on immigrants in countries where second-generation outnumber first-generation migrant children (i.e. the old immigration states), and that there is no significant link with such views in countries where the reverse is the case (i.e. the new immigration states). The same regularity applies at the classroom level: in classes with more second than first generation students, out-group size enhances inclusive views while it shows no relationship to such views in classes with more first than second generation students. The results thus support the third hypothesis. The non-relation in contexts with many first generation students may well be a temporary phenomenon, however. Once immigrant communities have become more settled and integrated in the destination countries, positive effects of ethnic mixing could well emerge everywhere.

Key words: inter-group contact, inclusive attitudes, classroom ethnic composition, contact theory, conflict theory, ICCS study

Introduction

As Western societies have become increasingly diverse owing to immigration and cross-

border mobility, the question of how welcoming the native population is to newcomers has

become increasingly relevant. If the native population has exclusionary attitudes, the

integration of immigrants into the receiving society will be severely compromised, with

possible negative consequences for overall social cohesion. Governments call on schools to

cultivate the opposite of such attitudes among youngsters, i.e. tolerance and accommodating

views. One particular way in which schools are thought to promote such values is by having

an ethnically mixed intake. The idea that mixing contributes to tolerance is based on the

assumptions that (1) a culturally diverse environment provides more opportunities for cross-

cultural interaction (Blau, 1974) and that (2) under a given set of conditions such interaction

enhances intercultural understanding and mitigates unfounded negative opinions of the ethnic

other (Allport, 1954).

Remarkably, to my knowledge, despite the obvious social and political relevance of

this question for immigration countries, no study has explored whether the effect of mixed

schooling on attitudes towards immigrants is the same across many different national

contexts. Pettigrew and Tropp (2006) did address the topic of universality in their

comprehensive meta-analysis of no fewer than 700 studies from a wide variety of national

contexts. They found that interethnic contact was positively related to favourable out-group

attitudes in 90% of the studies they reviewed. However, one cannot simply assume on the

basis of this study that ethnically mixed schools will therefore also promote more tolerant

attitudes to immigrants everywhere, because the studies included in Pettigrew and Tropp’s

review assessed only the effect of actual intergroup contact (i.e. not that of mixed settings)

and mainly concerned interracial (Black–White) attitudes.

Most of the studies investigating the relationship between mixed schooling and out-

group attitudes have in fact been single-country studies. Interestingly, only two of

these, Dollase (2001) and Van Geel and Vedder (2010), found school

diversity to be positively related to favourable out-group attitudes. The

studies by Wagner et al (1989), Kokkonen et al (2010), Dejaeghere et al

(2012) and Bekhuis et al (2013) found no relationship between classroom

diversity and out-group attitudes, while Vervoort et al (2011) found that

diversity enhanced negative out-group attitudes. It would be premature to

conclude on the basis of this short review that the effect thus varies

across countries, since the single-country studies have used different

measures of school ethnic composition and interethnic attitudes. Two

studies have used similar measures cross-nationally. However, one of

these (Barber et al, 2010) has assessed the relationship only on the basis

of pooled international data and thus ignored the possibility of country-

specific effects, while the second study carried out separate analyses on

only three countries (Janmaat, 2012). The last-named study found

classroom diversity to be positively related to tolerant attitudes to

immigrants in Sweden and Germany but to show no significant link with

this outcome in England.

This article will assess the issue of universality more profoundly by

exploring how classroom ethnic composition is related to inclusive

attitudes to immigrants among native 13- and 14-year-old pupils in each

of 14 Western countries. These countries have all experienced

immigration recently and/or longer ago, which means the disposition of

the native population towards immigrants is relevant in all of them.

Inclusive attitudes refer to the belief that immigrants should have the same rights,

duties and opportunities as all other citizens, irrespective of their country of origin and ethno-

racial background. Given that this belief is generally considered to be a key civic virtue

underpinning liberal democracy (Putnam, 1993), enabling immigrants to participate in the

host society, its relevance can hardly be overstated. I consider inclusive beliefs about

immigrants to be a subset of positive out-group attitudes in the same way that tolerant

attitudes are. Although inclusive beliefs have been found to be strongly related to ethnic

tolerance (e.g. Scheepers et al, 2002), the two concepts need to be distinguished conceptually

as the notion of accepting the ethnic other as equal is a crucial component of inclusive beliefs

while it does not necessarily form part of ethnic tolerance. Examining these beliefs among

13- and 14-year-olds is important because they are developed during adolescence and become

relatively stable thereafter (Sears, 1990). The classroom is the context to focus on since this is

the environment in which adolescents spend a lot if not most of their time.

The following section develops three hypotheses on the generalisability of the effect

of mixed classes. Subsequently, the data source and variables of interest are explained. I close

by testing the hypotheses and discussing the main findings.

Classroom Ethnic Mix and Inclusive Views on Immigrants: Three Competing

Hypotheses

As this study primarily aims to assess whether the relationship between ethnically mixed

classrooms and inclusive attitudes is the same across different national contexts, it is

important to discuss why one might expect this relationship to be similar or to vary cross-

nationally. One reason for expecting it to be similar is the startling similarity of school

education across the Western world in terms of structure and content (Benavot et al, 1991).

Practically everywhere, school education is characterised by groups of 20–30 youngsters of

the same age being taught by a qualified adult in a confined area (classroom and school). This

commonality of conditions makes it plausible to propose that the effect of classroom ethnic

composition observed in one country can be generalised to other countries.

The next question is thus what the direction of this effect will be: do ethnically mixed

classrooms promote or undermine inclusive attitudes towards immigrants? The literature on

interethnic relations does not offer a conclusive answer. On the one hand, it leads one to

conjecture that mixed classrooms do indeed promote inclusive attitudes. This conjecture is

based on a number of theoretical propositions linking the explanatory condition to the

outcome in a stepwise fashion. The first two of these propositions have already been

mentioned. One concerns the idea that as a social context, particularly a well-bounded micro

one such as a school class, becomes more mixed, so the probability of intergroup contact

increases because of the proximity of members of the out-group (Blau, 1974). The second is

the main tenet of contact theory, which holds that as intergroup contact increases, so the in-

group becomes less prejudiced and develops more favourable attitudes to the out-group

(Allport, 1954). Third, it may be proposed that the micro setting of the classroom usually

meets the conditions that Allport considered vital for intergroup contact to have such positive

effects. These conditions are (1) equal status of the groups engaging in the contact, (2)

common goals, (3) intergroup cooperation, and (4) institutional support and monitoring. In a

classroom there is indeed equality of status among students (at least formally), the contact

often includes activities involving common goals and cooperation (such as team sports), and

the setting is closely supervised by a teacher, who is not likely to allow intolerant behaviour

and remarks (Kokkonen et al, 2010; Janmaat, 2012). Even if not all these conditions are met

in a classroom, intergroup contact is still likely to yield positive intergroup attitudes, as there

is increasing agreement nowadays that Allport’s conditions should be seen as facilitating

factors rather than as necessary conditions (Schlueter and Scheepers, 2010). Fourth, it has

been argued that interethnic contact fosters favourable attitudes not only towards those

members of the out-group directly involved in the interaction but also towards the entire out-

group, and the aforementioned meta-analysis of Pettigrew and Tropp (2006) has indeed found

evidence for this.

On the other hand, one may surmise that mixed classes foster exclusionary attitudes

towards immigrants. This is based on the assumption from conflict (or ethnic competition)

theory that the larger the proportion of the out-group in a given population is, the more the

dominant group will feel threatened in its privileged position, culturally and economically,

and consequently the more exclusionary and hostile it will become towards the out-group

(Blalock, 1967; Quillian, 1995; Schneider, 2008). Most of the research supporting this theory

examines the effect of ethnic composition at the level of countries (e.g. Quillian, 1995;

Schneider, 2008; Scheepers et al, 2002), i.e. macro social contexts in which different groups

can easily live separate lives with little cross-cultural interaction. That such interaction plays

a crucial mediating role has been demonstrated by Semyonov and Glikman (2008). They

found that, once levels of positive interethnic contact were taken into account, people living

in highly diverse neighbourhoods expressed more negative views on immigrants than people

living in homogeneous neighbourhoods. In other words, whether mixed settings foster or

diminish positive out-group attitudes depends on the level of actual intergroup contact: in

settings with little contact the outcome predicted by conflict theory occurs; in settings with

many contacts the outcome anticipated by contact theory emerges (Sturgis et al, 2011).

These findings and reflections lead to the expectation that the outcome predicted by

conflict theory would not occur in the classroom setting, since intercultural contact would

seem hard to avoid in this micro environment. Yet one cannot assume that intercultural

contact is a given in mixed classrooms, as Vermeij et al (2009) found that out-group contacts

did not increase (and in-group contacts did not reduce) as the class became more diverse in

terms of out-group size. Moreover, other studies of educational settings found that the

outcome predicted by conflict theory did occur. Researching student attitudes in the

Netherlands, Vervoort et al (2011), for instance, found that, in classrooms with high

proportions of minority students, ethnic-majority respondents reported more negative out-

group attitudes. In America, Forman (2003) found a curvilinear effect in that racial

prejudice among White students first declined and then rose again to the point of

surpassing initial levels of prejudice as the percentage of Black students in the class

increased. Negative outcomes of diversity have also been found for behaviour. Tolsma et

al (2013), for instance, found both intra- and inter-ethnic bullying to rise with

increasing levels of classroom ethnic diversity.

Combining the proposition on universal effects with the conjectures on the direction

of the effect, I arrive at the following contrasting hypotheses: (1) higher proportions of

immigrant students in class foster inclusive attitudes to immigrants among native students in

all national contexts; (2) higher proportions of immigrant students in class diminish inclusive

attitudes to immigrants among native students in all national contexts.

Yet the effect of mixed classes is unlikely to be only a function of educational

conditions and out-group size. It can be proposed that the composition of the population with

an immigrant background (with regard to history of settlement) also influences this effect.

More specifically, one may hypothesise that in countries where first-generation migrants

(greatly) outnumber second-generation ones (that is, countries that have only recently

attracted many immigrants), the link between mixed classes and inclusive views is weaker

than in countries where the opposite applies (Hypothesis 3a). This conjecture is based on

the assumption that the second generation is generally better integrated. Having been

raised in the country of destination second generation children have had many more

opportunities to learn the destination country language, to familiarize themselves with

that country’s culture and customs, and to interact with members of native majority

than first generation children (Estevez and Sampaio 2013). Research has demonstrated

that second generation children indeed have a better command of the destination

country language (Tubergen and Kalmijn 2005; OECD 2009) and have higher levels of

friendships with native majority children (Esser 2013). Others have found destination

language proficiency and interethnic friendships (or contacts with the native majority

more broadly) to be closely correlated (Wagner and Machleit 1986; Matinovic et al

2009), suggesting that destination language proficiency is crucial for contacts with the

majority or that the two are mutually reinforcing. Thus, the process predicted by

contact theory, i.e. that of interethnic contact reducing prejudice, is most likely to occur

in diverse contexts with many second generation children. This is all the more likely in

view of the finding that intense and high-quality forms of intergroup interaction, such

as interethnic friendships, are especially conducive to more favourable out-group

attitudes (Kao and Joyner, 2004). We may add that the same phenomenon should also

be visible at the classroom level, i.e. the more second generation relative to first

generation students there are in the classroom, the stronger the link between mixed classes

and inclusiveness (Hypothesis 3b), since micro-environments such as classrooms vary in

immigrant composition as well.

In short, while the first two hypotheses propose similar effects across countries,

the third expects the effect of mixed classes to depend on the characteristics of the

immigrant population both at the country and classroom level. I am aware that the

effect of diversity on inclusive attitudes may be shaped by other factors as well, such as

the political climate of the country of destination and the distance between the culture

of the majority and those of immigrant groups. Space constraints and data limitations

prevented me from exploring these factors in greater detail. Should the cross-national

pattern of findings not correspond to any of those predicted by the aforementioned

hypotheses, then such factors are likely to have played a role.

Data, Variables and Methods

Data source

For the current study I use data from the 2009 International Civic and Citizenship Education

Study (ICCS). The ICCS study collected data on the civic attitudes and competences of Grade

8 students (ages 13 and 14) in 38 countries worldwide. In each participating country between

1,700 and 4,500 students from 66 to 196 schools were surveyed in a nationally representative

sample. Following a two-stage stratified cluster design, the study first randomly selected the

schools on the basis of probability proportional to size and then selected one whole classroom

in each school (Schulz et al, 2010). All the students in this classroom were surveyed.

I selected the samples from 14 Western states as data for the current study. The choice

was restricted to Western states because these states have recently and/or historically

experienced immigration.1 The selection further only includes states where first- and second-

generation migrant children together numbered at least 2.5% of the respondents.2 Table 1 lists

the 14 selected countries and shows the proportions of native, first- and second-generation

migrant children as well as the total number of respondents in each country. The percentage of

migrant children (of first and second generation combined) ranges cross-nationally between

27.4 (Switzerland) and 7.1 (Italy). The countries also vary considerably in the relative

weighting of the first- and second-generation students, which makes the selection of countries

ideal for testing Hypothesis 3. Some countries have four times as many first- as second-

generation students (Cyprus, Italy, Spain and Ireland), reflecting the sudden and recent shift

from emigration to immigration countries that these states have experienced. Others have

almost or more than twice as many second- as first-generation students (Switzerland, Austria,

Sweden and the Netherlands), indicating that they have attracted immigrants for some time.

Table 1 about here

Dependent variable – inclusive views on immigrants

1 I decided not to include post-communist states in the selection of countries due to the specific migration histories of these countries. Most of the second generation migrant children in these countries have parents who migrated within the borders of the same state (that is when the state they are now residing in formed part of a larger state), which means they cannot be considered international migrants. 2 This cut-off point was established because in near-to homogenous countries the number of classrooms with higher proportions of immigrant children is insufficient to allow one to make reliable statements on the effect of immigrant proportion.

To measure inclusive attitudes to immigrants I devised a scale similar to the one developed by

the ICCS experts. It is based on the same five items as the one in the ICCS database and

shows the same high levels of internal coherence (with Cronbach’s alphas ranging

between .74 and .89 cross-nationally). The items are as follows:

‘How much do you agree or disagree with the following statements about immigrants?’

Immigrants should have the opportunity to continue speaking their own language;

Immigrant children should have the same opportunities for education that other

children in the country have;

Immigrants who live in a country for several years should have the opportunity to vote

in its elections;

Immigrants should have the opportunity to maintain their own customs and lifestyle;

Immigrants should have all the same rights that everyone else in the country has.

< 1– strongly disagree; 2 – disagree; 3 – agree; 4 – strongly agree >

The scale was constructed by taking the mean of the responses to the five items. As a result,

the scale has values ranging between 1 (maximum exclusionary views) and 4 (maximum

inclusive views), with a midpoint of 2.5 denoting neutral attitudes to immigrants. The

advantage of this scale is that it enables one to interpret the values in both a relative and an

absolute sense, while the ICCS scale, which has values standardised to an international mean

of 50, allows for only a relative interpretation. The scale is an individual-level variable.

Independent variables

The main variable of interest is immigrant proportion, a class-level variable representing the

share of first- and second-generation migrant children in the class. The key decision in the

construction of this variable was whether to include the second generation in the out-group.

This was not an easy decision to make, as the ICCS did not ask respondents whom they

considered to be ‘immigrants’ when they posed the aforementioned five questions.

Consequently, it was not possible to determine whether native students included only first-

generation migrant children (i.e. immigrants in the proper sense of the word) or also second-

generation migrants in their understanding of ‘immigrants’. This is a handicap because one

should ideally know where the in-group draws the boundary between itself and the out-group

before developing a variable reflecting out-group size.

The widespread use of denigrating terms such as Pakis, Les Beurres and Ausländer,

which stigmatise immigrants and their offspring, suggests that second and possibly third

generations are still seen as ‘immigrants’ by many members of the native majority. Others,

however, might well understand the concept of immigrants in the official way, i.e. as

referring only to first-generation migrants. From the perspective of the second generation,

too, it is not clear where the boundary lies, given the propensity of many second-generation

children to develop hybrid identities combining feelings of attachment to their country of

birth with those towards the country of origin of their parents (Berry et al, 2006; Faas, 2009).

To get some idea of where the boundary between the in- and the out-group is drawn, I

assessed the mean values of inclusive attitudes for the native group and the two immigrant

groups (see Table 2). The results show that both the first- and the second-generation migrants

are significantly more inclusive in their views on immigrants than the native students across

the board (only in two countries is the difference from native students not significant for

second-generation ones). From the perspective of first-generation students this makes good

sense. Being immigrants themselves, they would in essence have consented to being placed

in a subordinate position relative to the native group if they had expressed exclusionary views

on immigrants. More revealing is that the second generation also hold more accommodating

views. In 10 of the 14 states their mean values are closer to those of the first generation than

to those of the native group, and in eight states their inclusive views are equal to or even

surpass those of the first generation. This suggests that not only the first- but also the second-

generation students have identified (or at least shown sympathy) with the group of people –

immigrants – they were asked to state their views on. It gives us a clue that the boundary

between the in- and the out-group lies between the native group on the one hand and the first

and second generation on the other. I therefore decided to include second-generation migrant

children in the out-group when constructing the immigrant proportion variable.

Table 2 about here

As some research has found non-linear effects (e.g. Dollase, 2001; Forman, 2003), the

variable was transformed into a categorical variable to enable an assessment of the effect of

various levels of immigrant share vis-à-vis the reference category (100% native students).

The other categories are low levels (0.1–15%), medium levels (15.1–50%) and high levels

(50.1–99%) of immigrant share. I created these categories with unequal range because

immigrant proportion is skewed towards the homogeneous end. Table 3, which presents the

distribution of students across the four categories, clearly shows this. For example, despite

their narrow ranges, the categories of 100%-native and low-level-immigrant-share constitute

the largest in terms of number of students enrolled in two and ten countries respectively. Only

in the two countries with the highest proportion of immigrants (Switzerland and New

Zealand) is the medium-level category the largest group.

Table 3 about here

The other independent variable of interest is immigrant composition. This country-level

variable represents the ratio of the number of first- to the number of second-generation

migrant students. It has a minimum of .47 (Switzerland: the first-generation group is twice as

small as the second generation) and a maximum of 6.93 (Ireland: the first-generation group is

seven times the size of the second generation).

I used five control variables at the individual and the classroom level to assess

whether the link between immigrant proportion and inclusive views is genuine and not a

reflection of other influences. The individual-level variables included gender, social

background, cultural resources and civic knowledge. Social background was measured by

means of the highest education level obtained by one of the parents. Cultural resources were

tapped with the number of books at home. Civic knowledge was captured with a ready-made

composite measure in the ICCS database (the variable NWLCIV). This measure is based on

79 items testing student knowledge in four areas: civic society, civic principles, civic

participation and civic identities (Schulz et al, 2011). The four individual-level control

variables have been shown to be strongly related to a wide range of civic attitudes including

views on immigrants (see, for instance, Scheepers et al 1992; Galston, 2001; Schultz et al,

2010). The classroom-level control variable is classroom social status, which is the class

average of social background. Many studies, particularly those examining neighbourhood

characteristics, have highlighted the independent effect that the social status of the

environment has on a range of civic outcomes (e.g. Letki, 2008; Oliver and Mandelberg,

2000).

Method of analysis

Given the nested structure of the data – with individuals nested in classes, and classes

nested in countries (considering the one class per school sample) - and the fact that the

independent variables are pitched at different levels, I used multilevel analysis (MLA) to

explore the relationships between immigrant proportion and inclusive views on immigrants.

When data is nested, MLA provides for a more accurate estimation of the effects of

contextual variables than conventional multiple regression techniques (Snijders and Bosker,

1999). The mixed-methods option in SPSS was used to construct two-level random

intercept models (individuals and classes) and a three-level random slope model with

interaction effects (individuals, classes and countries). These models were based on native

students only. All independent variables in the MLA models were grand-mean-centred.

Results

Table 4 shows how immigrant proportion is linked to inclusive views on immigrants in each

of the 14 countries (based on 14 country-specific two-level multilevel models). These

relationships have all been controlled for by the variables listed in the notes below the table.

The countries have been ranked in descending order according to immigrant composition

(Column 8). While significant positive relations emerge in seven countries, significant

negative links can be observed in just one country (Italy). There is thus much more evidence

for mixed classrooms fostering inclusive views than for their doing the opposite. The medium

category of immigrant proportion shows a significant positive link with the outcome in five

countries. In other words, in these countries the native students in classrooms with medium

proportions of immigrants are significantly more inclusive in their attitudes than the native

students in all-native classrooms (the reference category). Interestingly, these countries form

quite a diverse group in terms of demographics, political traditions and migrant integration

policies, including states such as, England, Switzerland and Sweden. In eight countries the

link is not significant, while in only one country is it significantly negative. High proportions

of immigrants (see Column 6) also appear to be conducive to inclusive views, judging by the

highly significant positive relations in five countries. The category is unrelated to the

outcome in nine countries. In contrast, small proportions of immigrants do not seem to make

much of a difference, as this category demonstrates only a (barely) significant positive

relationship in two countries.

Table 4 about here

These results are clearly not consistent with the second hypothesis, which expects higher

shares of immigrants to have a negative effect on inclusive attitudes. Not only do positive

relationships far outnumber negative ones, but it is actually at higher shares of immigrants

(the categories ‘medium’ and ‘high’) that immigrant proportion is related positively to

inclusive attitudes. The first hypothesis, which anticipated that immigrant proportion would

produce a positive effect across the board, is supported regarding the direction of the effect

but not regarding its universality, given the non-relationship in half of the countries. Looking

more closely at this group of countries where no or a negative relationship is found, it appears

that it includes all those in which there are more first- than second-generation immigrant

students (Ireland, Spain, Greece, Cyprus, Italy and New Zealand – see Column 8). This

lends strong support to Hypothesis 3a, which predicted that the effect of immigrant

proportion would be weaker in contexts where the former outnumber the latter.

This hypothesis is also confirmed by a three-level random slope multilevel

analysis on the pooled data including immigrant composition as a country level

variable, immigrant proportion (as a continuous measure) at the classroom level3 and

an interaction term for immigrant composition and immigrant proportion (see Table 5).

The main effect of immigrant proportion is positive and significant (b = 0.031; p =

0.032), while the interaction term shows a significant negative relationship (b = -0.016; p

= 0.022). This means that the more first relative to second generation migrant children

there are in a country, the weaker is the positive effect of immigrant proportion on

inclusive attitudes, which is consistent with Hypothesis 3a. Interestingly, the cross-

national variation in the effect of immigrant proportion is not significant (b = 0.0016; p

= 0.091), while it is significant in a model without the interaction term (b = 0.0026; p =

0.043) (not shown in table). Hence, the main reason why the effect of immigrant

proportion differs across countries is immigrant composition. Once immigrant

composition is taken into account, countries no longer show significant differences in

the effect of immigrant proportion on inclusive attitudes, indicating that this effect is

not strongly influenced by other conditions which vary across countries such as political

climate, cultural distance with the native majority, etc.

Table 5 about here

3 This continuous measure is the percentage of children in the class who are first or second generation migrants (see also the footnote under Table 5).

Does the same regularity apply to the micro-level of the classroom? In other words, is

diversity more effective in enhancing native students’ inclusive attitudes in classrooms

where second generation outnumber first generation migrant children (Hypothesis 3b)?

I investigated this question by splitting the pooled sample in two groups – one in which

1st generation outnumbered 2nd generation children in the classroom and one in which

the reverse was the case – and by conducting a two-level MLA on each of these groups

(see Table 6).4 We see that immigrant proportion, neither as a continuous measure nor

as a categorical one, shows any relation to inclusive attitudes in the group of classrooms

where 1st generation children outnumber 2nd generation ones (Models 1 and 2). In

contrast, it does show a strong positive link with inclusiveness in the classrooms where

there are more 2nd than 1st generation children (Models 3 and 4). What is more, the

strongest link can be seen at the highest level of out-group size (i.e. at 50.1 – 99%

immigrant children in the classroom) (Model 4). These results are in agreement with

Hypothesis 3a as they show that immigrant composition also makes a difference at the

micro-level in terms of modifying the effect of immigrant proportion on inclusive

attitudes.

Table 6 about here

Limitations

Some remarks and qualifications need to be made with regard to the current study. First, I

could not explore some of the causal mechanisms presumed to drive the effect of mixed

4 I opted for this approach rather than a single MLA including classroom immigrant composition and classroom immigrant proportion as main effects and an interaction term for both because the latter approach caused multicollinearity problems between the two main effects.

classes on out-group attitudes since ICCS does not have items tapping interethnic contact,

proficiency in the language of the destination country or perceived threat. Yet I would

argue that this drawback has not been a major impediment, as the primary objective of this

study was to assess the cross-national generalisability of the effect of mixed classes, not to

explore the causal reasoning linking mixed classes to (positive or negative) out-group

attitudes. A second and more serious point concerned the difficulty of demarcating the in-

and the out-group, which has been identified as a problem by others as well (e.g. Van Geel

and Vedder, 2010; Vervoort et al, 2011). Both contact and ethnic competition theory assume

that clear in- and out-groups can be identified and that these groups constitute mutually

exclusive categories. However, we know from the literature on majority and minority ethnic

identities that the process of identity formation among young people is characterised by

hybridity, complementarity, nestedness and contextuality (Tizard and Phoenix, 2002; Faas,

2009). This substantially complicates the notion of ‘intercultural’ contact and requires us to

take the salience and degree of belonging to particular identities into account, which should

be a task for future research.

Finally, as with all research relying on cross-sectional data, the analyses of this study

could prove only association, not the direction of causality. This is potentially a serious

problem as one could imagine that mixed settings are a consequence rather than a cause of

inclusive views on immigrants. They are a consequence if children with parents who are

more open-minded about immigrants in the first place self-select into mixed classes, or

(which is more likely) if parents with anti-immigrant attitudes enrol their children in all-

native classrooms. Although a selection bias can be partly eliminated by controlling for

individual background variables (as I have done), and is likely to be minimal in societies with

restricted school choice (Kokkonen et al, 2010), one cannot rule out the possibility that some

self-selection has occurred.

Discussion and conclusion

In the current age of austerity, which is likely to enhance in-group favouritism among native

groups (Hoskins et al, 2012), it becomes all the more important to assess what schools can do

to promote inclusion and tolerance. On the whole, the results of this study are welcome news

for the advocates of desegregation, as they suggest that ethnically mixed schools are well

positioned to promote inclusive out-group attitudes among native students. This positive

effect of ethnic mixing may well be powerful enough to override the influence of context-

specific factors, as the current study observed such positive effects in seven countries varying

widely in population size, political traditions and ethno-linguistic composition. True, no

effect of ethnic mixing was found in half of the countries examined. Yet, aside from Flanders

and New Zealand, these countries are all new immigration states which have never

experienced large inflows of migrants before. The immigrant children in these countries

are mostly of the first generation and are thus unlikely to have developed a good

command of the destination country language and/or to have established many contacts

and friendships with native majority youth yet. These circumstances are likely to

hamper the ability of diverse settings to foster inclusive out-group attitudes among

native majority children. Although I could not assess the steps in this causal reasoning,

as the data source used does not have suitable measures of intergroup contact and

language proficiency, the finding that out-group size contributes to inclusiveness in

settings where second generation children outnumber first generation ones (both at the

country and classroom level) does point in this direction. This finding, moreover,

suggests that the non-relationship between out-group size and inclusive attitudes in the

new immigration countries is only a temporary phenomenon: once immigrants become

settled and produce offspring, this non-relationship is likely to turn into a positive one.

In old immigration states with established immigrant communities the positive effect of

mixed schooling on inclusive attitudes could therefore well be universal.

The results of this study are clearly not in agreement with the hypothesis based on

conflict theory, which assumed out-group size to be negatively related to favourable out-

group attitudes. Aside from Italy, no country showed this relationship. In a sense these

findings are not wholly unexpected, since conflict theory is mostly supported by studies

examining the effects of ethnic diversity in macro social environments where intergroup

contact need not occur. By contrast, in the very micro setting of classrooms, the social

context of the current study, interethnic contact can hardly be avoided, and it should therefore

not surprise us that the findings are more consistent with contact theory.

Another strong indication that the mechanism proposed by conflict theory does not

apply (i.e. that of out-group size triggering a sense of threat among the in-group, which in

turn triggers hostile attitudes towards the out-group) is that positive effects of mixed settings

were found at medium and high levels of immigrant proportion in the classroom. Apparently,

native students forming only a small majority or even a minority in a class were not

intimidated by the presence of immigrant peers, or at least not to the extent that it would

prevent them from expressing more inclusive views on immigrants than native students in all-

native classrooms. Interestingly, this finding also suggests that the problem in making ethnic

mixing ‘work’ is not too much out-group presence but too little. Possibly, in view of the

tendency of homophily to shape interaction (Blau, 1974; Moody, 2001), native students in

low-diversity classrooms can still avoid contact with immigrant peers while this is no longer

possible at higher levels of immigrant presence. Low heterogeneity levels have been found to

be ineffective by others. Dollase (2001), for instance, found that the incidence of hostile

attitudes towards foreigners among native German students dropped significantly only when

proportions of foreign students in schools exceeded 10%.

In sum, this study suggests that policy makers should consider ethnic mixing as a

strategy to promote more inclusive out-group attitudes among the native majority. They

should not expect immediate results, however, in schools with many first-generation

immigrant students. Neither should they expect a minimal presence of immigrant students in

class to do the job.

References

Allport, G. (1954) The Nature of Prejudice, Reading, MA: Addison-Wesley.

Barber, C., Torney-Purta, J. and Fennelly, K. (2010). Adolescents’ Attitudes Toward

Immigrants’ Rights and Nationalism in 25 Countries. Unpublished paper available at

http://www.iea-irc.org/index.php?id=cived_and_iccs, accessed 20 May 2012.

Bekhuis, H., Ruiter, S. and Coenders, M. (2013). Xenophobia among Youngsters: The Effect

of Interethnic Contact, European Sociological Review, 29, 229–242.

Benavot, A., Cha, Y.-K., Kamens, D., Meyer, J.W., and Wong, S.Y. (1991). Knowledge for

the Masses: world models and national curricula, 1920–1986. American Sociological

Review, 56, 85–100.

Berry, J.W., Phinney, J.S., Sam, D.L. and Vedder, P. (2006). Immigrant Youth in Cultural

Transition: Acculturation, Identity and Adaptation Across National Contexts. New York:

Lawrence Erlbaum.

Blalock, H. M. (1967) Toward a theory of minority-group relations, New York: John Wiley

& Sons.

Blau, P.M. (1974). Parameters of Social Structure. American Sociological Review, 39, 615–

35.

Dejaeghere, Y., Hooghe, M. and Claes, E. (2012). Do Ethnically Diverse Schools Reduce

Ethnocentrism? A Two-Year Panel Study Among Majority Group Late Adolescents in

Belgian Schools. International Journal of Intercultural Relations, 36, 108–117.

Dollase, R. (2001). Die Multikulturelle Schulklasse – oder: Wann ist der Ausländeranteil zu

hoch? Zeitschrift für Politische Psychologie, 9, 113–26.

Duncan, G.J. and Raudenbusch, S.W. (1999) Assessing the effects of context in studies of

child and youth development, Educational Psychologist, 34, 29–41.

Esser, H. (2013). The Integration of Second Generation Immigrants in Germany: An

Explanation of Cultural Differences. In: R.S. Sigel and M. Hoskin (eds.), Education for

Democratic Citizenship: A Challenge for Multi-Ethnic Societies. New York: Lawrence

Erlbaum, 45-70.

Esteves, A. and Sampaio, D. (2013). Language Proficiency among Immigrants and the

Establishment of Interethnic Relations: A Comparative Analysis of Bilbao, Lisbon and

Rotterdam, Finisterra, XLVIII, 96, 65-88.

Faas, D. (2009) Reconsidering Identity: The Ethnic and Political Dimensions of Hybridity

Among Majority and Turkish Youth in Germany and England, British Journal of

Sociology, 60, (2), 299–320.

Forman, T. (2003). From Pet to Threat? Minority Concentration, School Racial Context and

White Youths’ Racial Attitudes. Dept of Sociology, University of Illinois at Chicago.

Unpublished manuscript.

Galston, W. (2001). ‘Political Knowledge, Political Engagement and Civic Education,’

Annual Review of Political Science, 4, 217–234.

Hoskins, B., Kerr, D., Abs, H. J., Janmaat, J. G., Morrison, J., Ridley, R., and Sizmur, J.

(2012). Participatory Citizenship in the European Union: Analytic Report. Available at:

http://ec.europa.eu/citizenship/pdf/report2_analytic_report.pdf

Janmaat, J.G. (2012). The Effect of Classroom Diversity on Tolerance and Participation in

England, Sweden and Germany, Journal of Ethnic and Migration Studies, 38, 21–39.

Kao, G., and Joyner, K. (2004). Do race and ethnicity matter among friends? Activities

among interracial, interethnic, and intraethnic adolescent friends. The Sociological

Quarterly, 45, 557–573.

Kokkonen, A. Esaiasson, P. Gilljam, M. (2010). Ethnic Diversity and Democratic

Citizenship: Evidence from a Social Laboratory, Scandinavian Political Studies, 33, 331–

335.

Letki, N. (2008) Does diversity erode social cohesion? Social capital and race in British

neighbourhoods, Political Studies, 21, 109–123.

Martinovic, B., van Tubergen, F. and Maas, I. (2009). Dynamics of Interethnic Contact: A

Panel Study of Immigrants in the Netherlands, European Sociological Review, 25, 303-

18.

Moody, J. (2001). Race, School Integration, and Friendship Segregation in America.

American Journal of Sociology, 107, 679–716.

OECD (2009). PISA 2009 Results: Overcoming Social Background; Equity in Learning

Opportunities and Outcomes; Volume II. Paris: OECD.

Oliver, J.E. and Mandelberg, T. (2000) Reconsidering the environmental determinants of

white racial attitudes, American Journal of Political Science, 44, 574–89.

Pettigrew, T.F. (1998). Intergroup Contact Theory. Annual Review of Psychology, 49, 65–85.

Pettigrew, T.F. and Tropp, L.R. (2006). A Meta-Analytic Test of Intergroup Contact Theory,

Journal of Personality and Social Psychology, 90, 751–783.

Putnam, R. (1993) Making Democracy Work: Civic Traditions in Modern Italy. Princeton,

NJ: Princeton University Press.

Quillian, L. (1995) Prejudice as a Response to Perceived Group Threat: Population

Composition and Anti-Immigrant and Racial Prejudice in Europe, American Sociological

Review, 60, 586–611.

Scheepers, P., Felling, A., & Peters, J. (1992). Anomie, Authoritarianism and Ethnocentrism:

Update of a Classic Theme and an Empirical Test. Politics and the Individual, 2, 43-60.

Scheepers, P. Gijsberts, M. and Coenders, M. (2002). Ethnic Exclusionism in European

Countries: Public Opposition to Civil Rights for Legal Migrants as a Response to

Perceived Ethnic Threat, European Sociological Review, 18, 17–34.

Schlueter, E. and Scheepers, P. (2010). The Relationship between Outgroup Size and Anti-

Outgroup Attitudes: A Theoretical Synthesis and Empirical Test of Group Threat- and

Intergroup Contact Theory, Social Science Research, 39, 285–295.

Schneider, S. L. (2008). Anti-Immigrant Attitudes in Europe: Outgroup Size and Perceived

Ethnic Threat, European Sociological Review, 24, (1), 53–67.

Schulz, W., J. Ainly, J. Fraillon, D. Kerr and B. Losito (2010). ICCS 2009 International

Report: The Civic Knowledge, Attitudes and Engagement Among Lower Secondary School

Students in 38 Countries. Amsterdam: IEA.

Schulz, W., Ainly, J. and Fraillon, J. (eds) (2011) ICCS Technical Report. Amsterdam: IEA.

Sears, D. O. (1990). Whither Political Socialization Research? The Question of Persistence.

In: Ichilov, O. (Ed.), Political Socialization, Citizenship Education and Democracy. New

York: Teachers College Press, 69–97.

Semyonov, M. and Glikman, A. (2009). Ethnic Residential Segregation, Social Contacts, and

Anti-Minority Attitudes in European Societies, European Sociological Review, 25, 693–

708.

Snijders, T. and Bosker, R.J. (1999). Multilevel analysis: An introduction to basic and

advanced multilevel modelling, London: Sage Publications.

Sturgis, P., Brunton-Smith, I., Read, S. and Allum, N. (2011). ‘Does Ethnic Diversity Erode

Trust? Putnam’s “Hunkering Down” Thesis Reconsidered’, British Journal of Political

Science, 41, 57–82.

Tizard, B. and Phoenix, A. (2002). Black, White or Mixed Race? Race and Racism in the

Lives of Young People of Mixed Parentage, London: Routledge.

Tolsma, J., van Deurzen, I. and Veenstra, R. (2013). Who is Bullying Whom in Ethnically

Diverse Primary Schools? Social Networks, 35, 51-61.

Torney-Purta, J. (2002) The school’s role in developing civic engagement: a study of

adolescents in twenty-eight countries, Applied Developmental Science, 6, 203–212.

Van Geel, M., Vedder, P. (2010). Multicultural Attitudes among Adolescents: The Role of

Ethnic Diversity in the Classroom, Group Processes & Intergroup Relations, 14, 549–

558.

Van Tubergen, F. and Kalmijn, M. (2005). Destination-Language Proficiency in Cross-

National Perspective: A Study of Immigrant Groups in Nine Western Countries,

American Journal of Sociology, 110, 1412-57.

Vermeij, L., Van Duijn, M.A.J., and Baerveldt, C. (2009). Ethnic Segregation in Context:

Social Discrimination among Native Dutch Pupils and Their Ethnic Minority Classmates.

Social Networks, 31, 230–9.

Vervoort, M. H.M., Scholte, R.H.J., Scheepers, P.L.H. (2011). Ethnic Composition of School

Classes, Majority–Minority Friendships, and Adolescents’ Intergroup Attitudes in the

Netherlands, Journal of Adolescence, 34, 257–267.

Wagner, U. and Machleit, U. (1986). Gastarbeiter in the Federal Republic of Germany:

Contact between Germans and Migrant Populations. In: Hewstone, M. (Eds.). Contact

and Conflict in Intergroup Encounters. Oxford: Blackwell, 59–78.

Wagner, U., Hewstone, M. and Machleit U. (1989). Contact and Prejudice between Germans

and Turks. Human Relations, 42, 561–74.

Table 1. The composition of the student population by immigration background

Native Second- First- Total students generation generation number (%) migrant migrant of

studentsa studentsa students (%) (%) (N)

Switzerland 72.6 18.6 8.8 2,924New Zealand 73.4 9.2 17.4 3,979Austria 79.1 14.0 6.9 3,385Sweden 83.0 11.3 5.7 3,464England 85.9 8.6 5.5 2,916Ireland 88.0 1.5 10.5 3,355Spain 88.2 2.3 9.5 3,309Norway 88.7 6.6 4.7 3,013Netherlands 89.2 7.1 3.7 1,964Greece 89.5 3.7 6.8 3,153Flanders (Belgium) 90.2 5.4 4.4 2,968Denmark 91.5 5.4 3.1 4,508Italy 92.8 1.5 5.6 3,366Cyprus 92.9 1.0 6.1 3,194a First-generation migrant students are children who have been born in the country of origin. Second-generation immigrant students are children who have been born in the receiving country but whose parents (both of them) were born in the country of origin.

Table 2. Inclusive views on immigrants by immigration background

Native Second- First- students generation generation

(mean) migrant migrant students students (mean) (mean)

Greece 3.2 3.4** 3.4***Sweden 3.1 3.7*** 3.7***Norway 3.1 3.5*** 3.5***Spain 3.1 3.3* 3.5***New Zealand 3.1 3.4*** 3.5***Ireland 3.1 3.2 3.5***Cyprus 3.1 3.1 3.3***Italy 3.0 3.3* 3.5***Denmark 3.0 3.4*** 3.4***Switzerland 3.0 3.3*** 3.4***Austria 2.9 3.4*** 3.3***Flanders (Belgium) 2.9 3.3*** 3.3***England 2.8 3.3*** 3.3***NetherlandsTotal

2.83.0

3.3***3.4***

3.3***3.4***

* P < 0.05; ** P < 0.01; *** P < 0.001. The significance levels concern the difference of the mean of each of the two immigrant groups from that of the native students.

Note: The scale ranges from 1 [minimum inclusive views] to 4 [maximum inclusive views]. The value 2.5 is the midpoint representing neither exclusionary nor inclusive views.

Table 3. Distribution of students across the four categories of immigrant proportion (by percentage)

100% native

Low(0.1–15%Immigrant)

Medium(15.1–50%Immigrant)

High(50.1–99%Immigrant)

Switzerland 8.6 22.4 59.8 9.1New Zealand 15.6 27.0 46.0 11.4Austria 11.4 49.9 30.4 8.3Sweden 25.7 38.6 23.5 12.2England 29.6 41.3 21.7 7.4Ireland 18.1 56.1 23.9 1.9Spain 28.7 41.3 27.8 2.2Norway 29.3 50.7 16.4 3.6Netherlands 19.8 55.4 20.6 4.1Greece 20.4 53.6 25.7 0.3Flanders (B) 40.1 34.9 21.0 4.1Denmark 33.1 46.2 17.1 3.6Italy 42.2 41.9 15.2 0.6Cyprus 21.7 61.6 16.0 0.7

Table 4. The relation between classroom immigrant proportion and inclusive views on immigrants: the results of 14 country-specific multilevel analyses.

Low Medium High Immigrant (0.1–15% immigrants) (15.1–50% immigrants) (50.1–99% immigrants) composition

(ratio 1st / 2nd b t ratio b t ratio b t ratio generation)

Switzerland 0.027 0.98 0.052* 2.33 0.050** 2.87 0.47Austria 0.045 1.78 0.045 1.93 0.067*** 3.64 0.49Sweden 0.032 1.50 0.083*** 3.87 0.063*** 3.73 0.50Netherlands 0.065* 2.45 0.076** 2.74 0.032 1.05 0.52Denmark 0.026* 2.13 0.066*** 4.45 0.065*** 3.62 0.57England 0.003 0.11 0.053* 2.12 0.001 0.41 0.64Norway 0.001 0.38 0.039 1.82 0.064** 2.59 0.71Flanders (B) -0.018 -1.08 0.035 1.85 0.039 1.63 0.81Greece -0.010 -0.57 -0.015 -0.84 -0.061 -0.56 1.84New Zealand -0.021 -1.07 0.001 0.04 -0.001 -0.44 1.89Italy -0.041* -2.10 -0.084*** -3.57 -0.099 -1.79 3.73Spain -0.020 -0.99 -0.025 -1.30 0.001 0.41 4.13Cyprus -0.016 -0.85 -0.017 -0.76 -0.016 -0.31 6.10Ireland 0.016 0.93 0.018 1.04 -0.024 -0.93 6.93

* P < 0.05; ** P < 0.01; *** P < 0.001.

NB: Each of the 14 models represents a two-level random intercept multilevel linear regression analysis including gender, social background, cultural resources, civic knowledge and classroom social status as control variables. The table only shows the coefficients of the categories of low, medium and high proportions of immigrants in the classroom. The reference category is 100% native students. The full results of these analyses can be obtained from the author on request.

Table 5. The relation between classroom immigrant proportion and inclusive views on immigrants: the results of a three-level multilevel analysis on the pooled data.

b t ratio Degrees of freedom

Intercept 3.010*** 103.14 12

Immigrant composition (L3) a 0.023 1.70 12Immigrant proportion (L2 - random effect) b 0.031* 2.50 10Classroom social status (L2) a 0.009 1.67 2391Gender (1-boy; 2-girl) (L1) a 0.190*** 29.57 34151Cultural resources (L1) 0.034*** 9.41 34400Social background (L1) 0.023*** 6.00 33301Civic knowledge (L1) 0.011*** 31.84 33350Immigrant composition x immigrant proportion b -0.016* -2.64 12

Cross-national variance in the intercept 0.0113* 2.38Cross-national variance in the slope of immigrant proportion

0.0016 1.69

Cross-classroom variance in the intercept 0.0172*** 14.24

* P < 0.05; ** P < 0.01; *** P < 0.001.

a L3 = country level variable; L2 = classroom level variable; L1 = individual level variable.

b Allowing the slope of immigrant proportion to vary across countries means that the classroom level variable of immigrant proportion and the interaction effect with immigrant composition are treated as country-level variables and are thus based on the degrees of freedom (N) associated with that level. As the number of observations is so small at this level, I decided to use the continuous measure of immigrant proportion rather than the categorical one (as using the latter would result in too many interaction terms in relation to the degrees of freedom). Immigrant proportion and the interaction effect turn out to be statistically significant even when based on such a small number of observations.

Note: An analysis of the empty model (not shown in the table) reveals that the classroom level captures 7.4% of the total variance in inclusive attitudes. According to Duncan and Raudenbusch’s (1999) rule of thumb, this proportion represents a medium to large effect size, indicating that there is every reason to examine the impact of classroom-level variables such as immigrant proportion. The country level captures 3.3% of the total variance. The model shown in the table explains 8.8%, 41.5% and 16.5% of the variance at the individual, classroom and country level, respectively.

Table 6. The link between immigrant proportion and inclusive views on immigrants in two groups of classrooms differing in immigrant composition (results of four two-level multilevel analyses).

Classrooms where first- Classrooms where second- generation students outnumber generation students outnumber second-generation ones (Models 1 & 2) first-generation ones (Models 3 & 4)

Model 1 Model 2 Model 3 Model 4

b t ratio b t ratio b t ratio b t ratioIntercept 3.049*** 383.16 3.048*** 377.29 3.003*** 307.41 3.006*** 303.57

Immigrant proportion (continuous) (L2) 0.014 1.31 0.059*** 5.81Immigrant proportion (categorical) (L2) Low (0.1–15% immigrants) (ref cat) a

Medium (15.1–50% immigrants) 0.008 1.07 0.028** 3.30 High (50.1–99% immigrants) 0.011 1.11 0.040*** 4.79Classroom social status (L2) 0.023** 2.74 0.023** 2.72 0.044*** 4.16 0.042*** 3.90Gender (1-boy; 2-girl) (L1) 0.182*** 17.12 0.182*** 17.12 0.194*** 15.41 0.194*** 15.42Cultural resources (L1) 0.039*** 6.79 0.039*** 6.78 0.029*** 4.06 0.029*** 4.07Social background (L1) 0.014* 2.32 0.014* 2.33 0.032*** 4.04 0.033*** 4.07Civic knowledge (L1) 0.011*** 19.58 0.011*** 19.59 0.011*** 15.84 0.011*** 15.86

Level 1 explained variance (%) b 7.0 7.0 8.1 8.1Level 2 explained variance (%) b 31.2 31.2 35.4 34.4N Level 1 (individuals) a 12787 12787 9055 9055N Level 2 (classrooms) a 960 960 862 862

* P < 0.05; ** P < 0.01; *** P < 0.001.a The number of observations at the individual and classroom level is reduced because of the exclusion of the all-native classrooms from the models. For this reason the reference category of the categorical measure of immigrant proportion was changed to ‘low’.b Analysis of the empty model shows that the classroom level captures 8.5% and 9.8% of the total variance in inclusive attitudes in Models 1&2 and Models 3&4, respectively.

32