Education Friendship Ethnocentrism Bremen · EDUCATION, FRIENDSHIP, AND ETHNOCENTRISM 5...

EDUCATION, FRIENDSHIP, AND ETHNOCENTRISM

1

The Impact of Education and Intergroup Friendship

on the Development of Ethnocentrism

A Latent Growth Curve Model Analysis of a Five Year Panel Study

among Belgian Late Adolescents

Cecil Meeusen, Marc Hooghe, and Ellen Quintelier (University of Leuven)

[email protected]

***ECPR Graduate Conference Bremen***


2

Abstract

In this paper, we investigate individual level changes in ethnocentrism during adolescence

and pre-adulthood. We use structural equation modeling for longitudinal designs on data from

the Belgian Political Panel Survey (BPPS, 2006-2011). In this panel, 2,428 Belgian

adolescents were questioned at three points in time: at the ages of 16, 18 and 21. Individual

change is analyzed by using Latent Growth Curve Modeling. Individual variability was

explained using two important predictors of ethnocentrism: education and intergroup

friendship. Adolescents in lower educational tracks have higher initial levels of ethnocentrism

and their levels of ethnocentrism continue to rise during the observation period. Adolescents

who change to lower education tracks between 2006 and 2008 increase more in ethnocentrism

than adolescents who stay in the same track. While intergroup friendship had an effect on

initial levels of ethnocentrism this contact did not have an effect on subsequent changes in the

level of ethnocentrism.

Keywords: ethnocentrism, education, adolescents, panel research, Belgium, contact

hypothesis


3

Introduction

Taking into account the trend toward higher levels of ethnic and cultural diversity in

Western societies, tolerance is becoming increasingly important as a social attitude (Harrell,

2010). The success of anti-immigrant parties in Europe in the last decades suggests that a

substantial number of citizens perceive immigration as a negative and possibly even

threatening phenomenon (Eatwell, 2000). In Belgium, e.g., voting for extreme-right parties is

principally determined by a negative attitude toward immigrants and migration policy in

general is one of the most important electoral motives (Deschouwer, Delwit, Hooghe, and

Walgrave, 2010). Ethnocentrism is not only a socially and politically salient attitude, it has

also been found to be persistent and already present at an early age (Sears and Funk, 1999).

Furthermore, ethnocentrism has been shown to be positively associated with discriminatory

behavior and other attitudes like homophobia (Dejaeghere, Hooghe, and Claes, 2012;

Hooghe, 2011). Given these social and political consequences of ethnocentrism, it becomes

all the more important to understand how this attitude is being shaped. In this paper we will

investigate the development of ethnocentrism during late adolescence and early adulthood.

Given the fact that the empirical research was conducted in a Western European context, we

will follow the lead of most European research and operationalize ethnocentrism as a negative

attitude toward immigration and diversity, as this is the most salient object of ethnocentrism

in a Western European context (Davidov, Meuleman, Billiet, and Schmidt, 2008)

Previous research has focused mainly on the evolution of ethnocentrism at the aggregate

level (e.g. Alwin and Krosnick, 1991; Coenders and Scheepers, 1998, 2008; Hooghe and

Wilkenfeld, 2008; Vollebergh, Iedema, and Raaijmakers, 2001). In this line of research, it is

assumed that aggregate levels of ethnocentrism can be influenced by media consumption, the

discourse of political parties or by trends in the level of ethnic diversity in society. This kind

of information, however, does not inform us about the individual level development of

ethnocentric attitudes. Investigating individual level trajectories during or just after the

impressionable years can provide us with information to arrive at a better understanding of the

causal mechanisms that lead to the development of ethnocentric attitudes. Methodologically,

we aim to extend previous work by using advanced structural equation methods for

longitudinal designs. A three wave panel design that measures adolescents’ level of

ethnocentrism at the ages of 16, 18 and 21 allows us to estimate individual change using a

Latent Growth Curve Model.


4

In this analysis we will focus on two important predictors of (the development of)

ethnocentrism: education and intergroup friendship (Dovidio, Geartner, and Kawakami, 2003;

Hello, Scheepers, Vermulst, and Gerris, 2004; Pettigrew, 1998; Savelkoul, Scheepers,

Tolsma, and Hagendoorn, 2011). Education is routinely found to be one of the most important

predictors of ethnocentrism (Coenders and Scheepers, 2003; Pettigrew and Tropp, 2011).

Highly educated individuals are more tolerant toward immigrants than are lowly educated

individuals (Hello et al., 2004). In the literature on the contact hypothesis, intergroup

friendship is considered to be a powerful tool to reduce ethnocentric prejudice (Allport, 1954;

Pettigrew, 1998). Intergroup friendship is expected to have an immediate effect on

ethnocentrism, but it is also assumed that the continued interaction with members of ethnic

minority groups further continues to reduce prejudice. Despite the fact that quite some of the

research stresses the role of education and of intergroup friendship in explaining trends in

ethnocentrism, it is striking to observe that most of the research is based on purely cross-

sectional observations. Ideally, proving the impact of both variables would require

longitudinal observations. That is exactly what we want to do in this paper, and we hope this

panel design will allow for a better understanding of the determinants of ethnocentrism as it

develops during late adolescence and early adulthood.

Literature and Hypotheses

Ethnocentrism can be defined as ‘a basic attitude expressing the belief that one’s own

ethnic group or one’s own culture is superior to other ethnic groups or cultures, and that one’s

cultural standards can be applied in a universal manner’ (Hooghe, 2008:11). Most of the

previous research suggests that ethnocentrism develops during adolescence (Alwin, Cohen,

and Newcomb, 1992; Hooghe and Wilkenfeld, 2008; Martinovic, van Tubergen, and Maas,

2009). Previous research has shown that changes in context variables like unemployment and

immigration levels can have an impact on aggregate levels of ethnocentrism (Coenders and

Scheepers, 1998; Semyonov, Raijman, and Gorodzeisky, 2006). We know far less, however,

about individual level determinants of changes in levels of ethnocentrism. In this paper, we

use two important variables to predict individual level attitude change over time: education

and intergroup friendship.

Education. One of the most consistent findings in this line of research is the negative

association between education and ethnocentrism: highly educated individuals are less


5

ethnocentric than lowly educated individuals (Pettigrew and Tropp, 2011). Educational

attainment is often seen as the most important predictor of tolerance toward ethnic minorities

(Coenders and Scheepers, 1998; Hello et al., 2004). In addition, Semyonov et al. (2006) found

this effect to be stable over time. It is argued that during adolescence, different socialization

agents have an impact on the formation of political predispositions (Niemi and Sobieszek,

1977). Nevertheless, Hello et al. (2004) found the influence of parents to be rather small

compared to the effect of education. Education is thus a very important cause of cultural

division in society (Hainmueller and Hiscox, 2007: 438): “The educational differences we can

observe between those individuals holding more pro- and anti-outsider views of the world

may be more of a symptom of the cultural divide between the two groups than they are a

cause.”

This effect of education among adolescents can be explained by invoking socialization

theory (Coenders and Scheepers, 2003). The socialization theory emphasizes the

developmental effect of education: Schooling introduces children and adolescents to civic

norms and values like tolerance (Hello et al., 2004). Higher educated individuals will

therefore be more successful in understanding the principles of tolerance and equality and

they will be able to generalize these principles to ethnic minorities. Furthermore, high

education levels are associated with a stronger cognitive capacity to learn about the out-group

(Kenworthy, Turner, Hewstone, and Voci, 2005).

The effect of education is so pervasive that we do not just expect it to have an effect on

the initial level of ethnocentrism, but also that it will continue to reduce levels of

ethnocentrism throughout the observation period as the adolescents are further socialized into

a culture that is congruent with the attitudes that prevail within the school system. Those with

negative education experiences, e.g., because they fail to pass exams, are expected to develop

more negative attitudes toward outsider groups. In their extensive research on the stability of

sociopolitical orientations, Alwin et al. (1992: 212-214) hypothesized that socioeconomic

variables contribute to the direction of attitude change. The results of the experiment with the

Bennington Women confirmed this hypothesis. Independent of their initial level of liberalism,

women with high educational achievements were more likely to continue to change their

attitudes in a liberal direction after college. Vollebergh et al. (2001) too found an additional

longitudinal ‘sleeper’ effect of education on ethnocentrism: Controlling for the effect at time

1, educational level was able to explain the variance in ethnocentrism at time 2.

Based on these findings, our first hypothesis is:


6

H1. (Changes in) Education level will have a negative impact both on the initial level of

ethnocentrism and on the trend during the observation period.

Intergroup friendship. The contact hypothesis suggests that intergroup contact reduces

feelings of prejudice toward the outgroup (Pettigrew, 1998). Allport (1954) distinguished four

conditions for intergroup contact to have a positive effect: equal group status within the

interaction context, common goals, intergroup cooperation and authority support. Later on,

researchers of the contact hypothesis added another condition: In order for intergroup contact

to be successful, the contact situation must have ‘friendship potential’ (Pettigrew, 1998). In

their research, Kenworthy et al. (2006: 286) confirm the importance of friendship potential:

“Cross-group friendship is thought to be one of the best predictors of better intergroup

attitudes because of its impact in terms of reducing anxiety and threat.” Research among

adolescents and adults has even shown a positive effect from extended friendship ties: Having

friends with intergroup friends by itself already reduces prejudice (Feddes, Rutland, and

Noack, 2009; Paolini, Hewstone, Cairns, and Voci, 2004). Pettigrew and Tropp (2005),

however, add a word of caution by suggesting that the conditions specified by Allport are

only facilitators of positive contact effects and not essential conditions.

Although frequently confirmed, the contact hypothesis has also been met with fierce

criticism. Relevant for research on intergroup friendship in adolescence is the critique that

cross-group friendships are confronted with obstacles. Adolescents are much more likely to

select same-race friends than cross-race friends. Moreover, cross-group friendships are more

difficult to sustain over time and they tend to decrease during the transition from childhood to

adolescence (Aboud, Mendelson, and Purdy, 2003; Feddes et al., 2009; Pettigrew and Tropp,

2011). So, although intergroup friendships might not be so common in adolescence, we do

expect strong negative effects on ethnocentrism for those adolescents that do have cross-

group friends.

Longitudinal research on the effects of intergroup friendship is scarce (Martinovic et al.,

2009; Pettigrew, 2008). An extensive longitudinal research project conducted by Levin, van

Laar, and Sidanius (2003) found longitudinal effects of intergroup contact. College students

with more outgroup and fewer ingroup friendships at the start of their college years, showed

less outgroup bias and anxiety at the end of college. Dhont et al. (in press) used a latent

change model to ascertain whether intergroup contact at an initial time point leads to a larger

subsequent decrease in prejudice over time. The assumption is that the interaction with

outgroup members continues to reduce prejudice over time as actors are exposed to

experiences and information that continues to erode their initial prejudice. Their results


7

confirmed the hypothesis that higher levels of initial intergroup contact were followed by

larger decreases in prejudice at a subsequent point in time and that an intensification of

intergroup contact was significantly and negatively related to prejudice. Nevertheless, the

analysis suffers from important limitations. The researchers used a very small sample (n=65)

of mostly female undergraduate students. As a consequence, generalizations are not

warranted. Moreover, the respondents were only questioned at two points in time, whereas a

latent change model requires panel data of at least three time points (Duncan, Duncan, and

Strycker, 2006: 17). Our second hypothesis therefore is:

H2. (Changes in the number of) Intergroup friendships will have a negative impact both

on the initial level of ethnocentrism and on the trend during the observation period.

While education and intergroup friendships serve as the main independent variables in

the analysis, self-evidently various control variables have to be included in the models.

Pettigrew and Tropp (2011) list some important covariates of ethnocentrism: Gender,

religious denomination, religiosity, social context, and socio-economic position. These

elements will therefore be included as control variables.

Data and Methods

Data

We use the data from the Belgian Political Panel Survey, 2006-2011 (BPPS) (Hooghe,

Havermans, Quintelier, and Dassonneville, 2011). This panel survey was designed to collect

data about political values, attitudes and participation among 16, 18 and 21 year old

adolescents from the two major language communities (French and Dutch) in Belgium. In

order to collect the data for this panel, a two step design was used. First, a random school

sample was drawn, that was representative with regard to province and school type (public or

private education).1 Subsequently, students of the selected schools were questioned by the

researchers during class hours. The first wave of the BPPS was conducted in 2006 (average

age 16). The response analysis in 2006 indicated that the survey was representative for school

type, education track, gender, and language group. Two additional waves were conducted in

2008 (average age 18) and 2011 (average age 21). In total 109 schools participated in de

survey, with an average of 22 pupils per school. Respondents that changed schools between

2006 and 2008, were contacted in 2008 by mail. As the 2011 respondents all left school, they


8

were contacted through mail and internet survey. There were no mixed-mode effects on the

reports of ethnocentrism in the survey. Like in most panel surveys, the BPPS survey had to

deal with panel attrition (small overrepresentation of female respondents and respondents of

the Dutch language group), but the analysis showed that the third wave of the panel still can

be considered as representative for the 21 year old population of Belgium. Weighing factors

were introduced for language and gender, but they all remained below 1.40 (Hooghe et al.,

2011).

In total, 3,025 respondents participated in all three waves of the panel. It has to be

remembered, however, that we are mainly interested in detecting a negative feeling toward

outside groups. Self-evidently, ethnocentric prejudice can be found both among the majority

and minority groups of the population. Measuring ethnocentric prejudice among ethnic

minorities, however, requires different measurement scales because the object of prejudice

will be different among these minorities (Meeus, Duriez, Vanbeselaere, Phalet, and Kuppens,

2009). In line with previous analyses of the same dataset, we therefore limited ourselves to

those respondents that unequivocally can be considered as belonging to the majority group

within Belgium (Dejaeghere, Hooghe, and Claes, 2012). We do so by limiting the analysis to

the respondents whose parents both held Belgian citizenship status at the moment the

respondent was born. This led to a loss of almost 600 respondents, resulting in a final sample

of 2,428 respondents belonging to the ethnic majority group.

Measurements

Ethnocentrism. Ethnocentrism is operationalized using a four-point Likert-scale of three

opinion items expressing a negative view on the presence of immigrants (Davidov et al.,

2008). This is in line with most of the research on ethnocentrism in Europe, where prejudice

is directed mainly toward (descendents of) immigrants (Cunningham, Nezlek, and Banaji,

2004). Respondents were asked to give an opinion on the following questions: “It is better if

all share the same customs and traditions”, “If a country wants to reduce tension it should stop

immigration” and “Immigration is a danger to our way of life.”

Because we use multiple indicators to measure ethnocentrism at three points in time, it is

necessary to test for factorial invariance. Longitudinal factorial invariance ascertains whether

the relation between the latent variable (i.e. ethnocentrism) and the indicators is equal over

time, i.e. if the latent variable measures the same latent concept over time (Widaman, Ferrer,

and Conger, 2010). According to the fit statistics of the strong factorial invariance model

(Factor loadings and intercepts of corresponding items fixed to be equal over time), we can


9

conclude that ethnocentrism indeed measures the same concept across time (χ²=35.4; df=25;

χ²/df=1.4; RMSEA=.013; CFI=.99). The analysis also demonstrates that the loadings of the

items are very strong in all three measurements. Subsequent analyses therefore are based on

this measurement model (see Appendix A.1). For a more elaborated description of the

measurement model we refer to Appendix A.2.

Education. Since all respondents for this panel survey were recruited at the age of 16

during high school, there is no variation with regard to level of finished education. The

Belgian education system, however, is characterized by an intensive form of tracking,

differentiating pupils in high status and low status tracks (Van Houtte, 2004). Respondents,

therefore, can be sorted on whether they follow a general academic, a technical or a

vocational track. Furthermore, if pupils perform badly at exams, there is a practice that they

are demoted to a ‘lower’ track, and between 2006 and 2008, this happened with 356

respondents. The reverse change, students rising to a ‘higher’ track, did not occur. In order to

operationalize education as closely as possible, we therefore included information about the

education track of the respondent. Next to education track we include another predictor of

education: Educational goal. Previous research indicates that this expectation can serve as a

powerful operationalization of the academic orientation of pupils (Andrew and Hauser, 2011).

Both during the first and during the second wave of the panel study, respondents were asked

about their most likely future education level: ‘I will not finish secondary education’, ‘I will

finish secondary education’, ‘I will follow higher education’, and ‘I will get a university

degree’. This question proved to be quite predictive: Among the respondents who at age 16

assumed that they would go for higher education, 84 per cent was indeed enrolled in higher

education at the age of 21. Among those who predicted they would only finish high school or

even drop out, this was just 25 per cent. The correlation between the 2006 and the 2008

answers was .647.

Intergroup friendship. Intergroup friendship can be measured in multiple ways

(Pettigrew and Tropp, 2011). Most studies use the number or proportion of outgroup friends a

respondent reports (e.g. Binder, Brown, and Zagefka, 2009; Paolini et al., 2004; Savelkoul et

al., 2011). A meta-analysis of Davies et al. (2011) indicated that this kind of straightforward

measure is a reliable operationalization of intergroup friendship. Although it has to be

acknowledged that this is only a crude single item question, this meta-analysis clearly

suggests that this is a reliable indicator for intergroup contacts and friendship. Similarly, we

measure intergroup friendship with the question “How many of your close friends have a

different ethnic or racial background than yours?” (1 = none: 2 = almost nobody; 3 = a few; 4


10

= half; 5 = many; 6 = almost all; 7 = all of them). This question too proved to be rather stable

over time, with a correlation of .415 between the first and the second observation, and again

.415 between the second and the third observation.

Control variables. In our model, we include gender (0 = male; 1 = female), education

level of the parents (factor of highest educational attainment of mother and father)2, language

group (0 = Dutch; 1 = French), religious denomination (not religious, Christian/Catholic and

other religions) and religious practice (frequency of attending religious ceremonies in 2006,

scale from 1 to 5) as control variables at the individual level. Because of the multi-level

sampling procedure (pupils within schools), we also control for the ethnic composition of the

school, as it is plausible that this will affect the likelihood of intergroup friendships. Ethnic

composition at school is measured as the percentage of students born in a non-European

country or with a mother or father born in a non-European country. Following the Canadian

work on the definition of visible minorities (Claes, Hooghe, and Stolle, 2009), we defined the

ingroup as all European respondents, thus including the neighbouring countries of Belgium.

Children originating from non-European countries are regarded as ‘visible’ minorities. The

average percentage of visible minorities at school was 10.1% (SE=14.45).

Methods

To measure individual trajectories of ethnocentrism over time we use a Latent Growth

Curve Model (LGCM) with multiple indicators (Duncan et al., 2006; Meredith and Tisak,

1990; Schlueter, Davidov, and Schmidt, 2007). Figure 1 illustrates the logic of an LGCM:

individual growth for ethnocentrism is a function of a latent intercept and a latent slope. The

latent intercept reflects the average initial value at the start of the longitudinal change process.

The latent slope indicates the average individual change rate over time. LGC models have the

advantage that it is possible to study predictors of individual change over time. This way,

variant growth processes between subgroups can be examined. In other words, LGCM

investigates whether, and to what extent, predictors are able to explain variance in the

different trajectories. In this paper, we will evaluate the effect of educational goal and

intergroup friendship on changes in the level of ethnocentrism (see Figure 1).

To answer the research questions we will structure the analysis as follows: First, we will

estimate the stability coefficients of ethnocentrism at three points in time using an

autoregressive model. Second, we will study the individual change in ethnocentrism. We start

the estimation with an unconditional growth model. This model focusses on the factor

loadings of the intercept and slope. Next, we elaborate the model by adding predictor

variables (the conditional model). The final development model controls for the effects of


11

ethnic composition at school. As a consequence, we add a third level variable to the model

(Level 1 = Time; Level 2 = Individual; Level 3 = School) (Muthén, 1997). Parameters are

calculated using Robust Maximum Likelihood estimations. All models are estimated with

Mplus.

FIGURE 1 AROUND HERE

Results

Descriptives

First, we look at the overall change in ethnocentrism between the three measurement

points. Table 1 displays the mean scores of ethnocentrism (4-point sum scale). We notice that

the mean level of ethnocentrism is approximately equal in 2006 (2.47) and 2008 (2.45). At the

age of 21 (2011), young adults reach a significantly higher average level of ethnocentrism

(2.58). The current data do not allow us to distinguish age and period effects. Nevertheless,

comparing the development of ethnocentrism in the BPPS with the development found in the

European Social Survey (ESS, 2012), provides a first indication of period effects (Figure 2).3

The aim of the current article, however, is not to explain these overall trends, but rather to

explain the differences of the trends of specific groups of respondents, so we will not pursue

an explanation for this general trend further.

TABLE 1 AROUND HERE


Intergroup friendship is approximately equal between 2006 and 2008 and increases

significantly between 2008 and 2011 (mean of 1.68 in 2006; 1.69 in 2008; 1.99 in 2011). In

2006 63.2 per cent of the adolescents follows general education, and in 2008 this percentage

dropped to 54.5 per cent (Appendix A.3). The percentage of respondents in technical

education increases between 2006 and 2008 (24.8 per cent to 31.4 per cent). The proportion of

respondents in vocational education stays approximately the same (12.0 per cent in 2006 and

10.8 per cent in 2008). Between 2006 and 2008 12.1 per cent of the respondents were

demoted to a lower education track. Furthermore, we can observe that at the ages of 16 and


12

18, adolescents have ambitious goals with regard to their educational career. More than 80 per

cent of the respondents plans on following higher education after finishing high school.

Second, we look at the correlation between ethnocentrism at the three measurement

points. We found high positive correlations (.642 between 2006 and 2008, .652 between 2008

and 2011, and .533 between 2006 and 2011), providing a first indication of the stability of

ethnocentrism during adolescence. To account for the measurement error of the model, we

estimated the stability coefficients making use of an autoregressive model. The stability

model showed a good fit to the data (χ²=67; df=26; χ²/df=2.58; RMSEA=.026; CFI=.998).

The standardized stability coefficients between the two time intervals reveal only moderate

stability (β=.667 between 2006 and 2008; β=.676 between 2008 and 2011, p<.001) (Schlueter

et al., 2007).4

Individual trajectories

Subsequently, our goal is to identify the determinants of individual differences in the

evolution of ethnocentrism during the observation period. First, we fit the unconditional linear

growth model, i.e. a model without predictors (Model I in Table 2). To specify the linear

growth model, the factor loadings of the intercept are constrained to 1 and the factor loadings

of the slope to 0, 2, and 5 (time-lags between the measurement points, see Figure 1). Table 2

shows that the linear growth model fits the data well (χ²=57; df=26; χ²/df=2.19;

RMSEA=.023; CFI=.996). The mean of the latent intercept represents the level of

ethnocentrism of adolescents at the age of 16 (Time 1). The slope is significantly different

from zero and positive; this indicates a growth in ethnocentrism (.183, p<.001). The variances

of the latent intercept (non-standardized estimate=3.938, p<.001) and slope (non-standardized

estimate=0.147, p=.001) indicate reasonable variability around the average intercept and

slope, i.e. inter-individual differences in the change trajectories (Byrne, 2012). The significant

variances provide justification for the addition of variables to explain the variation in

individual trajectories. The covariance between the latent intercept and slope is negative (-

.268, p=.012), meaning that respondents with high initial levels of ethnocentrism, experience

less growth over the five-year period. In other words, the difference between the levels of

ethnocentrism in 2006 levels off in 2008 and 2011.

In order to explain these differences in the development of ethnocentrism, we add our

three variables of interest, education track, educational goal and intergroup friendship, to the

model. We tested a model interpreting education track, educational goal and intergroup

friendship in 2006 as time-invariant variables (χ²=218; df=113; χ²/df=1.93; RMSEA=.018;

CFI=.990). Controlling for language group, religious denomination and practice, education


13

level of the parents and gender, education track and intergroup friendship have strong effects

on the latent intercept (Model II). Respondents in general education have a lower initial level

of ethnocentrism than the respondents in technical education (-.122, p<.001).5 The more

outgroup friends, the lower the initial level of ethnocentrism (-.174, p<.001). These findings

partly confirm our hypotheses. Further, girls are less ethnocentric than boys, adolescents with

highly educated parents are less ethnocentric than adolescents with lowly educated parents6

and non-religious adolescents are a to a small extent less ethnocentric than Catholic

adolescents. Language group, educational goal and religious practice have no significant

effect on the initial level of ethnocentrism.

Our main interest, however, is not the initial level of ethnocentrism but the evolution of

this attitude. We notice a significant effect of education track on the slope (-.160, p<.001).

Adolescents in general education have a lower rate of change (rate of increase) than

adolescents in technical and vocational education (Figure 2). In contrast to hypothesis 2, the

association between the latent slope and intergroup friendship is not significant. Having

outgroup friends at the age of 16 has no effect on the change rate of ethnocentrism. Looking

at the control variables, we notice a significant effect of language group on the slope (.106,

p=.003). Respondents living in the French-speaking part of Belgium experience a steeper rise

in the level of ethnocentrism than respondents in the Dutch-speaking part of Belgium. We

also notice a small effect of gender on the slope: Although girls have a lower initial level of

ethnocentrism, they increase more in ethnocentrism between 2006 and 2011 than boys (.076,

p=.016).


In Model III of Table 2 we include variables measuring the difference in education track,

educational goal, and intergroup friendship between 2006 and 2008 (χ²=269; df=140;

χ²/df=1.92; RMSEA=.018; CFI=.987). This allows us to investigate the effect of changes in

education level and intergroup friendship. An increase in educational goal has a small

negative effect on the growth in ethnocentrism (-.079, p=.023). Adolescents with increased

educational goals in 2008 have a lower rate of increase in ethnocentrism than adolescents with

decreased educational goals in 2008. Yet, the strongest effect of education is found for

students who changed tracks between 2006 and 2008 (.101, p=.001): Adolescents that were

demoted to a lower track have a higher rate of increase in ethnocentrism than adolescents that

stayed in the same track. Changes with regard to intergroup friendships have no significant

longitudinal effect on ethnocentrism.


14

In Model IV we control for the ethnic composition of the school (χ²=74; df=20;

χ²/df=3.70; RMSEA=.038; CFI=.963).7 The covariance between the intercept and slope at the

between-school level is not significant. Ethnic composition of the school has no effect on the

intercept, nor on the slope. Moreover, controlling for the school-level has no big influence on

the estimates of the within-school variables. Only for the slope, the effects of educational goal

and difference in intergroup friendship become significant (p<0.05). We can conclude

therefore that while the ethnic composition of the school does not have an effect on the

evolution of ethnocentrism, the contact hypothesis is mainly confirmed with regard to the

effect of intergroup friendship.

TABLE 2 AROUND HERE

Discussion

In this paper, we attempted to extend previous work on the evolution of ethnocentrism by

analyzing a five year panel survey among Belgian late adolescents and young adults. More

specifically, we used the Belgian Political Panel Survey data for the period 2006-2011 and

this allowed us to track the development of ethnocentrism among 2,428 respondents between

the ages of 16 and 21.

Support for hypothesis 1 was strong and unequivocal: We found the education track of

the respondent to be a strong predictor both of ethnocentrism in 2006 (initial level) and of the

evolution of ethnocentrism over the five year period. At the age of 16, we can observe already

a substantial and significant ‘tolerance gap’ between adolescents in high and those in low

education tracks. Education furthermore affects ethnocentrism in a number of ways. Not only

does education track have a direct effect on the level of ethnocentrism, there also seems to be

a kind of ‘sleeper’ effect, determining the subsequent slope of ethnocentrism. Furthermore,

pupils that experienced a demotion to a lower education track even grow more strongly

intolerant of the presence of ethnic minorities. The fact that education also has this

longitudinal effect, suggests that in the higher status tracks the Belgian school system is in

fact quite successful in socializing adolescents into a more tolerant value pattern.

Hypothesis 2 on the impact of intergroup friendship was only partly confirmed. While

friendship with members of outsider groups had a negative impact on the initial level of

ethnocentrism, it did not have a significant effect on the evolution of ethnocentrism during the

observation period. To drive that point home: Adolescents that expanded their intergroup

friendship network between 2006 and 2008 did not follow a different developmental


15

trajectory with regard to ethnocentrism if we compare them to their counterparts that lost

some of their cross-ethnic friendships. We do acknowledge that, in line with the contact

theory, intergroup contacts can be important. The effect of intergroup contact on the initial

level of ethnocentrism was even stronger than the effect of education track. This might imply

that the effect of intergroup friendships is concentrated mainly among a younger age group

and decreases with age. To ascertain whether this really is the case, future observation using

this panel might be very useful.

This research has some limitations. Our panel data only cover three measurement points

over a five year period (the minimum required for LGCM). With more measurement points

the estimations of our growth trajectories would be more precise and reliable (Byrne, 2012:

343). Moreover, more measurement points allow for greater flexibility in model fitting and

testing. Another limitation can be found in the applicability of the contact theory. One of the

most frequent critiques of the contact hypothesis deals with the direction of causality

(Pettigrew, 1998; Savelkoul et al., 2011). Selection bias may be an underestimated factor of

influence: Prejudiced actors most likely will avoid contact with persons from a different

ethnic background. If that would be the case, the mechanism of intergroup friendship would

be reversed. Research on the causality of intergroup contact is inconclusive. Some authors

have found the effect of intergroup contact on prejudice to be larger than the effect of

prejudice on intergroup contact (e.g. Brown, Eller, Leeds, and Stace, 2007). Others, however,

found similar effects (e.g. Binder et al., 2009). Here too, we can only repeat that future

measurement points in this ongoing panel study should enable us to develop a better

understanding of the precise causal mechanism.

In future research it seems important to elaborate on the aforementioned shortcomings.

The current results indicate that ethnocentrism is developed as a result of school and social

experiences during adolescence. As a consequence it is important to investigate its formation

and evolution at younger ages. On the other hand, we only studied in depth the effects of two

predictors, education and intergroup friendship. Other important predictors like

authoritarianism can be added in future research.

In sum, we found significant variability in the individual development of ethnocentrism

during adolescence and young adulthood. An important predictor of the variability is

education track. Adolescents in low education tracks have higher initial levels of

ethnocentrism and their ethnocentrism increases at a higher rate compared to adolescents in

higher education tracks. Intergroup friendship has no significant effects on rate of change, but

is an important predictor for initial levels of ethnocentrism. While our results do provide


16

partial support for the contact hypothesis in this regard, the most important determinant of

changes in ethnocentrism clearly is education and educational stratification.


17

References

Aboud, F., Mendelson, M., and Purdy, K. (2003). Cross-race peer relations and friendship

quality. International Journal of Behavioral Development, 27(2), 165-173.

Allport, G. W. (1954). The nature of prejudice. New York: Doubleday.

Alwin, D. F., Cohen, R. L., and Newcomb, T. M. (1992). Political attitudes over the life span.

The Bennington women after fifty years. Madison: University of Wisconsin Press.

Alwin, D. F. and Krosnick, J. A. (1991). Aging, cohorts, and the stability of sociopolitical

orientations over the life span. American Journal of Sociology, 97, 169-195.

Andrew, M. and Hauser, R. (2011). Adoption? Adaption? Evaluating the formation of

education expectations. Social Forces, 90(2), 497-520.

Binder, J., Brown, R., Zagefka, H., Funke, F., Kessler, T., Mummendey, A., Maquil, A.,

Demoulin, S., and Leyens, J. P. (2009). Does contact reduce prejudice or does prejudice

reduce contact? Journal of Personality and Social Psychology, 96(4), 843-856.

Brown, R., Eller, A., Leeds, S., and Stace, K. (2007). Intergroup contact and intergroup

attitudes: A longitudinal study. European Journal of Social Psychology, 37, 692-703.

Byrne, B. M. (2012). Structural Equation Modeling With Mplus. New York: Routledge.

Coenders, M. and Scheepers, P. (1998). Support for ethnic discrimination in the Netherlands

1976-1993: Effects of period, cohort, and individual characteristics. European

Sociological Review, 14(4), 405-422.

Coenders, M. and Scheepers, P. (2003). The effect of education on nationalism and ethnic

exclusionism. Political Psychology, 24(2), 313-343.

Coenders, M. and Scheepers, P. (2008). Changes in resistance to the social integration of

foreigners in Germany, 1980-2000. Journal of Ethnic and Migration Studies, 34(1), 1-26.

Claes, E., Hooghe, M., and Stolle, D. (2009). The Political Socialization of Young People in

Canada: Is There a Differential Effect of Citizenship Education on Visible Minorities?

Canadian Journal of Political Science, 42(3), 613-636.

Cunningham, W. A., Nezlek, J. B., and Banaji, M. R. (2004). Implicit and explicit

ethnocentrism: Revisiting the ideologies of prejudice. Personality and Social Psychology

Bulletin, 30(10), 1332-1346.

Davidov, E., Meuleman, B., Billiet, J., and Schmidt, P. (2008). Values and support for

immigration. A cross-country comparison. European Sociological Review, 24, 583-599.


18

Davies, K., Tropp, L. R., Aron, A., Pettigrew, T. F., and Wright, S. C. (2011). Cross-group

friendships and intergroup attitudes: A meta-analytic review. Personality and Social

Psychology Review, 15(4), 332-351.

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(1), 108-117.

Deschouwer, K., Delwit, P., Hooghe, M., and Walgrave, S. (Eds.) (2010). Les voix du peuple.

Brussels: Éditions de l’Université de Bruxelles.

Dhont, K., Van Hiel, A., De Bolle, M., and Roets, A. (in press). Longitudinal intergroup

contact effects on prejudice using self- and observer-reports. British Journal of Social

Psychology. DOI: 10.1111/j.2044-8309.2011.02039.x.

Dovidio, J. F., Gaertner, S. L., and Kawakami, K. (2003). Intergroup contact: The past,

present, and the future. Group Processes and Intergroup Relations, 6(1), 5-21.

Duncan, T. E., Duncan, S. C., and Strycker, L. A. (2006). Latent variable growth curve

modeling. Concepts, issues, and applications. Mahwah: Lawrence Erlbaum.

Eatwell, R. (2000). The rebirth of the ‘extreme right’ in Western Europe? Parliamentary

Affairs, 53(3), 407-425.

ESS (2012). European Social Survey. Retrieved from www.europeansocialsurvey.org.

Feddes, A. R., Rutland, A., and Noack, P. (2008). Direct and indirect friendship effects on

minority and majority children’s interethnic attitudes. Child Development, 80(2), 377-

390.

Hainmueller, J. and Hiscox, M. J. (2007). Educated preferences: Explaining attitudes toward

immigration in Europe. International Organization, 61, 399-442.

Harrel, A. (2010). Political tolerance, racist speech, and the influence of social networks.

Social Science Quarterly, 92(3), 724-740.

Hello, E., Scheepers, P., Vermulst, A., and Gerris, J. R. M. (2004). Association between

educational attainment and ethnic distance in young adults. Socialization by schools or

parents? Acta Sociologica, 47, 253-275.

Hooghe, M. (2008). Ethnocentrism. In: W. A. Darity (Ed.), International encyclopedia of the

social sciences. Volume 3: Ethnic – Inequality, Gender (pp.11-12). Detroit: Gale.

Hooghe, M. (2011). The impact of gendered friendship patterns on the prevalence of

homophobia among Belgian late adolescents. Archives of Sexual Behavior, 40, 543-550.

Hooghe, M., Havermans, N., Quintelier, E., and Dassonneville, R. (2011). Belgian political

panel survey (BPPS), 2006-2011. Technical report. Leuven: KU Leuven.


19

Hooghe, M. and Wilkenfeld, B. (2008). The stability of political attitudes and behaviors

across adolescence and early adulthood. Journal of Youth and Adolescence, 37, 155-167.

Hooper, D., Coughlan, J., and Mullen, M. R. (2008). Structural equation modeling:

Guidelines for determining model fit. Electronic Journal of Business Research Methods,

6(1), 53-60.

Kenworthy, J. B., Turner, R. N., Hewstone, M., and Voci, A. (2005). Intergroup contact:

When does it work, and why? In J. F. Dovidio, P. Glick, & L. A. Rudman (Eds.), On the

nature of prejudice. Fifty years after Allport (pp. 278-292). Malden: Blackwell.

Levin, S., van Laar, C., and Sidanius, J. (2003). The effects of ingroup and outgroup

friendships on ethnic attitudes in college: A longitudinal study. Group Processes and

Intergroup Relations, 6(1), 76-92.

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(3),

303-318.

Meeus, J., Duriez, B., Vanbeselaere, N., Phalet, K., and Kuppens, P. (2009). Examining

dispositional and situational effects on outgroup attitudes. European Journal of

Personality, 23, 307-328.

Meredith, W. and Tisak, J. (1990). Latent curve analysis. Psychometrika, 55(1), 107-122.

Muthén, B. (1997). ‘Latent Variable Growth Modeling with Multilevel Data’. In Maia

Berkane (ed.), Latent Variable Modeling and Applications to Causality (pp. 149-161),

New York: Springer.

Niemi, R. G. and Sobieszek, B. I. (1977). Political socialization. Annual Review of Sociology,

3, 209-233.

Paolini, S., Hewstone, M., Cairns, E., and Voci, A. (2004). Effects of direct and indirect

cross-group friendships on judgement of Catholics and Protestants in Northern Ireland.

Personality and Social Psychology Bulletin, 30(6), 770-786.

Pettigrew, T. F. (1998). Intergroup contact theory. Annual Review of Psychology, 49, 65-85.

Pettigrew, T. F. (2007). Future directions for intergroup contact theory and research.

International Journal of Intercultural Relations, 32, 187-199.

Pettigrew, T. F. and Tropp, L. R. (2005). Allport’s intergroup contact hypothesis: Its history

and influence. In J. F. Dovidio, P. Glick, & L. A. Rudman (Eds.), On the nature of

prejudice. Fifty years after Allport (pp. 262-277). Malden: Blackwell.

Pettigrew, T. F. and Tropp, L. R. (2011). When groups meet. The dynamics of intergroup

contact. New York: Psychology Press.


20

Savelkoul, M., Scheepers, P., Tolsma, J., and Hagendoorn, L. (2011). Anti-Muslim attitudes

in The Netherlands: Tests of contradictory hypothesis derived from ethnic competition

theory and intergroup contact theory. European Sociological Review, 27(6), 741-758.

Schlueter, E., Davidov, E., and Schmidt, P. (2007). Applying autoregressive cross-lagged and

latent growth curve models to a three-wave panel study. In K. van Montfort, J. Oud, and

A. Satorra (Eds.), Longitudinal models in the behavioral and related sciences (pp. 315-

336). Mahwah: Lawrence Erlbaum.

Sears, D. O. and Funk, C. L. (1999). Evidence of the long-term persistence of adults’ political

predispositions. Journal of Politics, 61(1), 1-28.

Semyonov, M., Raijman, R., and Gorodzeisky, A. (2006). The rise of anti-foreigner sentiment

in European societies, 1988-2000. American Sociological Review, 71(3), 426-449.

Van Houtte, M. (2004). Tacking effects on school achievement. A quantitative explanation in

terms of the academic culture of school staff. American Journal of Education, 110(4),

354-388.

Vollebergh, W. A. M., Iedema, J., and Raaijmakers, Q. A. W. (2001). Intergenerational

transmission and the formation of cultural orientations in adolescence and young

adulthood. Journal of Marriage and Family, 63, 1185-1198.

Widaman, K. F., Ferrer, E., and Conger, R. D. (2010). Factorial invariance within longitudinal

structural equation models. Child Development Perspectives, 4(1), 10-18.


21

Endnotes

1. In Belgium there are three types of schools: Schools that are funded by the community

(public education), schools that are funded by provinces and municipalities (local public

education), and private (i.e., mostly Catholic) schools. 2. Factor based on two items: “What is the highest educational degree of your mother”

and “What is the highest educational degree of your father.” Principal Component Analysis

indicated factor loadings of .741 for both items. The factor explains up to 77 per cent of the

variance between the items. 3. Ethnocentrism in the European Social Survey was measured with a 10-point Likert

scale of the following items:

1. Immigration is bad or good for the economy

2. The country’s cultural life is undermined or enriched by immigrants

3. Immigration makes a country a worse or better place to live. 4. The stability coefficients are the same over time. The Sartorra-Bentler scale χ²-

difference test indicates a significant difference between the unrestricted and the restricted

model (equality constraint of stability coefficients) (TRd=4.82; df=1; p=.028). 5. Respondents in general education do also differ significantly from the respondents in

vocational education, both for intercept and slope effects. 6. Differentiating between the effects of the education level of the mother and the father

provided the same results (Education level mother: β intercept=.057 (p<.1); β slope=-.067

(p>0.1); Education level father: β intercept=-.090 (p<.01); β slope=.001 (p>.1). 7. Because of the increased complexity, the three-level LGC is not based on the factorial

invariance model. The residual variance of ethnocentrism2006 and ethnocentrism2011 was

negative, therefore we fixed the residual variances at 0.


22

1 1

1 0 3 5

Education Intergroup friendship Intercept Slope

Ethnocentrism Ethnocentrism Ethnocentrism 2006 2008 2011

Figure 1. Two-level Latent Growth Curve Model for intra-individual variability of ethnocentrism

explained by education level and intergroup contact. Intercept loadings constrained to 1; Slope

loadings constrained to 0, 2, and 5.


23

Figure 2. Development of ethnocentrism in the European Social Survey 2002-2010

(Belgium). Source: ESS 2002-2010.

4.6

4.7

4.8

4.9

5.0

5.1

5.2

2002 2004 2006 2008 2010 2011


24

Figure 3. Developmental trajectories of adolescents in different education tracks. Entries are

sample averages of ethnocentrism for respondents in general, technical, and vocational

education in 2006. Source: BPPS 2006-2011.

2.00

2.50

3.00

2006 2008 2011

General education Technical education

Vocational education


25

Table 1

Mean and standard errors (SE) for ethnocentrism, intergroup friendship and educational

goal, proportion of respondents in education track

Time 1 (2006) Time 2 (2008) Time 3 (2011) T-Tests

Mean (SE)

N Mean (SE)

N Mean (SE)

N Time 1-Time 2

Time 2 - Time 3

Time 1 - Time 3

Ethnocentrism 2.47 (0.694)

2,226 2.45 (0.746)

2,349 2.58 (0.740)

2,357 1.134 -8.677 ***

-6.839 ***

Intergroup friendship

1.68 (0.986)

2,406 1.69 (0.987)

2,408 1.99 (1.040)

2,394 -0.788 -12.881 ***

-13.212 ***

General education

63.2 % 1,913 54.5 % 1,650 n.a. n.a.

Technical track

24.8 % 749 31.4 % 949 n.a. n.a.

Vocational training

12.0 % 363 10.8 % 326 n.a. n.a.

Educational goal

3.15 (0.740)

2,385 3.16 (0.737)

2,385 n.a.

-0.033 n.a.

Note. Average scores on ethnocentrism (1-4 scale), Intergroup friendship (1-4 scale) and Educational goal (1-4 scale). Percentage of respondents in education track. Paired Samples T-Test score. Source: BPPS, 2006-2011. Sign. for T-tests: *** p<.001


26

Table 2

MLR Estimates of the Latent Growth Curve Model of Ethnocentrism

Effect Model I Model II Model III Model IV

WITHIN-School

Intercept Fixed at zero Fixed at zero Fixed at zero

Slope .070*** .055 .082

Var. intercept 3.938*** 2.897*** 2.857*** .227***

Var. slope .147*** .112*** .108*** .008***

Covar. intercept and slope -.204* -.180** -.180** -.012**

Intercept on

Language group (1=French) -.033 -.030 -.055

Gender (1=female) -.178*** -.176*** -.196***

Religious denomination

Christian/Catholic (=ref.)

Not religious -.065* -.064* -.057

Other religion -.005 -.012 -.044

Religious practice -.006 -.009 -.009

Education level parents -.154*** -.154*** -.118***

Educational goal time 1 -.031 -.030 -.050

Education track time 1

General education -.122*** -.125*** -.104*

Technical education (=ref.)

Vocational education .030 .031 .027

Intergroup friendship time 1 -.174*** -.168*** -.182***

Slope on

Language group (1=French) .106** .099** .149***

Gender (1=female) .076* .080* .106***

Religious denomination

Christian/Catholic (=ref.)

Not religious .059 .061* .064

Other religion -.022 -.010 .002

Religious practice .022 .036 .046

Education level parents -.057 -.038 -.025

Educational goal time 1 -.056 -.081 .138*

Diff.educgoal time 1̠ time 2 -.079* .098*

Education track time 1

General education -.160*** -.161*** -.210***

Technical education (=ref.)

Vocational education .015 .022 .019

Changed track (1=Yes) .101*** .099**


27

Intergroup friendship time 1 .054 .020 .006

Diff. intergr. friends. time 1̠ time 2

-.041 .088*

BETWEEN-School

Intercept 3.345***

Slope 0.056*

Var. intercept .024*

Var. slope .000

Covar. Intercept and slope -.002

Intercept on

Ethnic composition .131

Slope on

Ethnic composition .526

χ² 57*** 218*** 269*** 74***

Df 26 113 140 20

χ²/df 2.19 1.93 1.92 3.70

RMSEA .023 .018 .018 .038

CFI .996 .990 .987 .963

Note. Entries for covariates are standardized Robust Maximum Likelihood Estimates (MLR). Entries for intercept and slope variance and covariance are unstandardized MLR estimates. Model I: LGCM without covariates; Models II and III: LGCM with time-invariant covariates; Model IV: Three-level LGCM with time-invariant within- and between-school covariates. χ²/df statistic accounts for the impact of large sample sizes (Hooper, Coughlan, and Mullen, 2008). RMSEA = Root Mean Square Error of Approximation; CFI = Comparative Fit Index. *** p<.001; ** p<.01; *p<.05.


28

.587

.826

.779

.621

.888

.816

.649

.886

.197 .224

.206

.642

.533

.652

.133 .093

.104

.828

Appendix A.1

Results of Longitudinal Strong Factorial Invariance for Ethnocentrism

Ethno1(06)

Ethno2(06) Ethnocentrism 2006

Ethno3(06)

Ethno1(08)


Ethno3(08)

Ethno1(11)


Ethno3(11)

Χ²=35.4; df=25; X²/df=1.4; RMSEA=.013; CFI=.999 Note. Entries are standardized Maximum Likelihood Estimates of measurement model with longitudinal strong factorial invariance (Ethno3(11) freely estimated). All estimates have p<.001. Subsequent analyses are based this measurement model. BPPS 2006-2011, n=2,428.


29

Appendix A.2

Longitudinal Invariance of a Measurement Model of Ethnocentrism

χ² Df ∆χ² ∆df RMSEA [90% CI] CFI

Configural invariance 30.7 18 .017 [.005, .027] .999

Equal factor loadings 32.8 22 2.1 4 .014 [.000, .024] .999

Equal intercepts 45.6 26 12.8*** 4 .019 [.010, .028] .999

Equal intercepts (one indicator intercept free at Time 3)

35.4 25 2.6 3 .013 [.000, .022] .999

Note. Entries are Maximum Likelihood Estimates. ∆χ² = nested χ² difference; RMSEA = Root Mean Square Error of Approximation; 90% CI = 90% Confidence Interval for RMSEA; CFI = Comparative Fit Index. The configural model (i.e. same pattern of fixed and free factor loadings across time) fits the data well (χ²=30.7; df=18; RMSEA=.017; CFI=.99). Next, in the weak factorial invariance model, factor loadings were constrained to be equal across time. The χ²-difference test between the configural and weak factorial invariance model was not significant: the constraints did not reduce the model fit. The χ²-difference between the weak and strong factorial invariance (i.e. equal intercepts across time) model was significant (p<0.001): the intercepts are variant across time. In a final model, the intercept of the third item measured at Time 3 was freed. The result is a good model fit of the strong factorial invariance model with one varying indicator (χ²=35.4; df=25; RMSEA=.013; CFI=.99). Subsequent analyses are based on the strong factorial invariance model. Source: BPPS 2006-2011, n=2,428. *** p<.001.


30

Appendix A.3

Frequencies for language group, gender, religion and educational level of the mother and

father

Language (N=2,428) N % Dutch 1687 69.5% French 741 30.5% Gender (N=2,428) Male 1184 48.8% Female 1244 51.2% Religious denomination (N=2,406) Christian/Catholic 1779 74% Not religious 580 24% Other religion 47 2% Religious practice (N=2,425) Never 1075 44.3% Few times a year 1200 49.5% Few times a month 86 3.5% Once a week 60 2.5% More than once a week 4 0.2% Educational level mother (N=2,272) Lower secondary 275 12.1% Higher secondary 833 36.5% Higher education 875 38.3% University 289 13.1% Educational level father (N=2,248) Lower secondary 326 14.4% Higher secondary 829 36.9% Higher education 616 27.4% University 477 21.2%

Date post:	04-Mar-2020
Category:	Documents
Upload:	others
View:	2 times
Download:	0 times

Education Friendship Ethnocentrism Bremen · EDUCATION, FRIENDSHIP, AND ETHNOCENTRISM 5...

Documents