Integration Regimes, Social Trust and Ethnocentrism

Transcription

1 Marc Hooghe* Tim Reeskens* Dietlind Stolle Ann Trappers* *. Catholic University Leuven (Belgium).. McGill University Montreal (Canada). The order of the authors is purely alphabetical, as all authors contributed equally to this paper. Corresponding authors: and DRAFT VERSION DO NOT QUOTE WITHOUT CHECKING WITH THE AUTHORS Integration Regimes, Social Trust and Ethnocentrism The Effects of Integration Policies on Social Cohesion in Europe Paper presented at the Joint Sessions of the European Consortium for Political Research Helsinki, May 7-12, 2007

2 1. Introduction Every year, about 2.75 million foreigners migrate to European countries (Salt 2005, 26). The European Commission has already indicated that economic considerations will necessitate the arrival of more immigrants, in order to compensate for shortages on the European labor market. These demographic trends and changes on the labor market all indicate that Western societies are increasingly becoming ethnically and culturally diverse (Cornelius & Rosenblum 2005; Hooghe et al. 2006a). In recent years, the impact of the migration on social cohesion has become a hotly debated topic in political science and society: the arrival of African refugees on Italian and Spanish coasts, the London and Madrid bombings, the riots in the suburbs of Paris, the controversy about the Mohammed Cartoons in Denmark and debates about the welfare state leave Europeans confronted with possible negative effects of global migration flows. According to theoretical assumptions, social cohesion is under pressure when a normative consensus among the members of society is missing or is being weakened (Durkheim 1973 [1893]). However, empirical research into the relation between ethnic diversity and social cohesion has led to diverging results. On the one hand, American research at the county or neighborhood level has found that ethnically fractionalized communities are less cohesive than homogeneous neighbourhoods (Alesina & La Ferrara 2002; Hero 2003). Also at an aggregated national level, a negative effect of diversity on social cohesion has been observed (Delhey & Newton 2005). When the shortcomings of this research regarding the geographic limitations and the treatment of the clustered nature of the data are taken into account, it is revealed, however, that the negative relationship between ethnic diversity and social cohesion is quite diffuse and needs to be qualified (Hooghe et al. 2006). Moreover, what is missing from this investigation on the effects of diversity on social cohesion is an analysis of how policies might be able to mitigate the seemingly negative consequences of diversity. The state, and state policies, are largely absent from the current research on the social effects of diversity. Countries confronted with large migration flow are forced to manage the ensuing ethnic and cultural diversity and to develop suitable policies in order to maintain social cohesion. Generally speaking, nations can choose between three options in their minority and integration policy, namely exclusion, assimilation and 1

3 integration (Cornelius et al. 2004), depending on the level of citizenship rights and support for the maintenance of their own distinct culture immigrants receive. Of course, various distinctions exist within these various policy frameworks as well. Based on the first wave of the European Social Survey (ESS 2002/3) and a self-collected dataset with information about naturalization law, citizenship status, minority voting rights and multicultural policies in the 21 countries of the ESS, we will use multi-level analysis to make assertions about the relationship between integration regimes and social cohesion in European societies. Our main research question is to examine whether different integration regimes have an impact on social cohesion indicators of that specific country. In a first chapter, we reflect on the literature with regard to integration, before presenting our data and methods. Finally, we will conclude and discuss future research necessary in the investigation of diversity and policy effects on social cohesion. 2. Theoretical overview Generalized trust is usually a reliable indicator to measure social cohesion (Stolle 2002). If many people have the feeling that most other people cannot be trusted, it will be more difficult for a community to pursue collective action efforts and to provide for collective goods. Recent research indicates that increasing racial and ethnic diversity in communities tends to reduce generalized trust. For ethnic minorities as well as majorities in Western democracies it is easier to develop trust if they live in homogeneous interaction contexts (Alesina and La Ferrara 2002; Leigh 2006; Soroka et al 2006). The current research on the relation between social diversity and social cohesion suffers from a number of shortcomings. First, most of the research thus far has been conducted in a US setting. It has to be remembered, however, that US society exhibits a very distinctive pattern of race relationships (Sniderman & Piazza 1993). Second, most of these studies focus exclusively on generalized trust as the main measurement of social cohesion, whereas other aspects of this construct remain untapped. In an increasingly diverse society, negative or even hostile feelings toward outsider groups might be just as detrimental to social cohesion as a lack of generalized trust. Third, by focusing just on one country, most studies fail to detect any policy influences. While most Western countries are confronted with the same 2

4 phenomenon of increasing ethnic and cultural diversity, they have developed a wide array of policies as a response (Joppke 1999). Depending on the specific policy framework, we can expect that the potential negative effects of diversity might be overcome or augmented, in which case integration policies might work to mitigate or to amplify social tensions. Soysal, Castles, Koopmans, Hollifield and many others have suggested various typologies of what may be called "immigration regimes." We like to highlight here four dimensions of these regimes that might be of relevance in the debate about their effect on social cohesion. First, some countries have clearly opted for a politics of multiculturalism, allowing, and sometimes even encouraging minorities to develop their own distinct culture, instead of being forced to adapt to the mainstream original culture (Banting & Kymlicka 2003; Kymlicka 1995; Kymlicka & Norman 2000). Other countries are much more hostile to multicultural policies, expecting that newcomers will abandon at least part of their cultural heritage in order to integrate into the receiving society (Koopmans et al. 2005). It is a general expectation that multicultural policies reduce tensions between the traditional and the new groups in society, while at the same time, they address feelings of discrimination and oppression among minorities. Multiculturalism is usually considered as a fair and unbiased method to govern multicultural and diverse societies (Kymlicka 2001). The basic expectation is that in countries pursuing multicultural and group rights policies diversity can be handled successfully by effectively mitigating the alleged negative effects of diversity. Countries, in which the dominant group tries to impose its own culture on the newly arriving groups, might have more troubles with tensions and the exclusion of ethnic groups within society. However, critics of multicultural policy are more sceptical, and point out this might be an unequal policy which brings with it unfair treatment and privileging of some groups, and thus leads to cultural isolation of new ethnic groups, fewer majority-minority social interactions, deepening of social cleavages, indifferences or distrust (Barry 2001). Countries have also implemented different procedures for the naturalization process. In some countries, knowledge of the national language(s) and facts about the host society are required for naturalizations; and they might be accompanied by extended ceremonies. In other societies such prerequisites do not exist. We expect that the more integrative and activist naturalization procedures might actually help to create social cohesion between majority and minority members. Although they might be challenging criteria for foreigners, they do bring foreign citizens closer to the national culture into which they like to immigrate and therefore bridge 3

5 cultural and social distances. The lack of such integrative attempts might contribute to antagonisms between majority and minority populations. Beyond various levels of multi-culturalism and naturalization procedures, countries differ in their general openness toward immigrants in that they have implemented various levels of strictness when it comes to acquiring citizenship status. Whereas some countries make it a birth right, other countries have added several other conditions to citizenship requirements and do not guarantee citizenship to second or third generations, and potentially deny dual citizenship status as well. In Switzerland, e.g., citizenship status can only be acquired after a stay of twelve years, while the candidate furthermore has to prove that s/he is integrated in the Swiss community, and is accustomed to the Swiss way of life and practices. The latter requirements are not further specified in the legal text, and the local authorities of the canton in practice decide on whether the applicant is sufficiently integrated. Belgium, on the other hand, is at the other extreme end of the spectrum. Residents can acquire Belgian citizenship after three years on Belgian territory (two years for asylum seekers), and without further formal requirements. In an atmosphere of social tensions and conflicts between cultural and ethnic minorities and majorities in European societies, we expect that more open policies might lead to the amplification of ethnic and racial tensions; whereas restrictive policies might help to quiet ethnocentric and xenophobic elements in Western societies that face increasing diversity. One final immigration policy we like to highlight here are voting rights for foreign nationals. Voting rights work particularly well for establishing social cohesion because they are universal egalitarian policies, implemented to put other citizens at an equal footing with majority groups. This interpretation complements earlier research on institutional characteristics that seem to foster aspects of social capital and trust in our societies; this literature emphasizes that particularly impartial and fair institutions that are built on the principal of procedural justice are successful in building generalized trust and solidarity (Rothstein & Stolle forthcoming; Tyler 1998). This is perhaps in contrast to some of the multicultural policies which at least partially follow clearly the principle of preferential treatment with the overall goal of achieving equality. From a policy point of view, the most pressing question is to know whether these different integration regimes actually matter. In order to ascertain whether these policy dimensions 4

6 have an effect on the relation between diversity and social cohesion, research cannot limit itself to one single case, but should adopt a comparative approach (Citrin & Sides, forthcoming). In the subsequent analyses then the relationship between diversity, government policy and social cohesion will be tested more thoroughly than has been done so far. Our analysis is based on the first European Social Survey which was conducted in in 21 European countries. Because of the wide variety of integration regimes and policies in Western Europe with regard to social conditions, we can better test whether there is a relationship between diversity, policy and social cohesion than previous analyses which were based on just one country. The overview of integration policies was assembled using an expert survey throughout various European countries (Trappers et al. 2007). We proceed in two steps. First it is examined whether there is a relation between ethnic diversity and the levels of trust and ethnocentrism within these countries. In a second step, it is ascertained whether the integration policies that have been adopted by these countries relate in any way to outcomes of ethnocentrism or generalized trust, and whether any of these policies might mitigate the potentially effects of diversity. 3. Data and methods The data used in this paper are obtained from several sources. The European Social Survey (ESS) provides the individual-level data for the dependent and individual-level independent variables. The ESS survey was collected in 2002/2003 by means of face-to-face interviews in representative samples. 1 Because of uniform field work procedures and strict scientific norms, ESS data can be considered as reliable for cross-national research (Jowell et al. 2003). In this analysis, three different indicators for social cohesion are utilized as dependent variables: generalized trust, ethnocentrism and the assumptions about the presence of foreigners in the country. Figure 1 depicts the country specific means of generalized trust and ethnocentrism. Ethnocentrism is measured by three Likert-items, taping he respondent s opinion about the consequences of immigration in their country on the economy and cultural domains more specifically, and life in general. For more information on the scaling, see 1. Israel was excluded because of its very specific demographic profile and immigration pattern. 5

7 Appendix I. Research has shown that this scale is measured in an equivalent way across the ESS-sample and, as a consequence, that it can be used in comparative analyses. Concerning the distribution of ethnocentrism in Europe, we see that the mean levels range from 3.37 to Countries with the lowest ethnocentrism levels are Luxembourg, Sweden, Finland and Switzerland: on a 0-10 scale, their country specific score is around 4 or lower. Countries with high levels of negative attitudes toward foreigners are Greece, together with the Eastern European countries Hungary, Slovenia, Portugal and Czech Republic, with scores of about 5.5 and higher. In this paper we also want to introduce a third, innovative measurement for hostile feelings toward outsider groups, i.e., the overestimation of the number of foreigners in the country. The ESS questionnaire included a question on the percentage of the population born outside the country. It is a well-known phenomenon that those who have hostile feelings toward minority groups tend to overestimate their presence within a community (Allport 1954). To assess this effect, we calculated the ratio between the estimation of the respondent and the actual percentage of the foreign-born population in that country (OECD figures), see more below. 6

8 Figure 1. Trust and Ethnocentrism in European countries DK NO FI SE IE CH NL GB AT DE LU BE ES FR CZ PT IT SI HU PL GR generalized trust ethnocentrism Note: The graph represents the mean scores per country on the generalized trust scale (0-10, in grey) and on the ethnocentrism scale (0-10, in black). The countries in the graph are Denmark (DK), Norway (NO), Finland (FI), Sweden (SE), Ireland (IE), Switzerland (CH), Netherlands (NL), United Kingdom (GB), Austria (AT), Germany (DE), Luxembourg (LU), Belgium (BE), Spain (ES), France (FR), Czech Republic (CZ), Portugal (PT), Italy (IT), Slovenia (SI), Hungary (HU), Poland (PL) and Greece (GR). The generalized trust scale was based on three items measuring the belief that most people can be trusted. For scaling information please see the Appendix I. Analysis of the crosscultural measurement equivalence of the generalized trust scale has revealed that there are some minor comparability problems (Reeskens & Hooghe, 2007), yet, we believe that the level of equivalence is sufficient enough to do comparative research. Figure 1 shows that generalized trust is particularly high in Scandinavian countries such as Denmark, Norway, Finland and Sweden. Trust levels are lowest in Poland and Greece. Overall, aggregate generalized trust ranges from 3.45 to 6.81 on a 0-10 scale. At the individual level, independent variables are included which in earlier research have proven to have an effect on trust and/or ethnocentrism. Research has shown, for example, that women, on average, are more trusting than men, while in most of the research, age too has a significant effect on trust and/or ethnocentrism (Putnam 2000; Stolle 2001; Uslaner 2002; Scheepers et al. 1992; Sniderman et al. 2003). Based on the literature, we can also expect the 7

9 highly educated, the employed, the financially secured, and those actively involved in churches to trust more and to have reduced feelings of ethnocentrism (Brehm & Rahn 1997). Since it is very difficult to arrive at a reliable and comparable measurement of income levels across 20 countries, we included financial satisfaction of the respondent as a proxy of socioeconomic position. 2 We also expect immigrants to be less trusting and to have lower feelings of ethnocentrism, of course, than majority citizens (Soroka et al. 2006). We also included an attitudinal scale on political efficacy because political powerlessness tends to have a negative effect on generalized trust, while it is positively related to ethnocentrism (Hooghe 2003). All scales and variables are documented in the Appendix I of this paper. At the country level, we have one control variable, namely the gross domestic product, taken from the Human Development Report statistics, because we expect a strong relationship between social cohesion indicators and GDP per capita (Delhey and Newton 2005). The stock of foreigners within a country, gathered from the OECD publications serves as our measurement of diversity; it was chosen from a test of the effect of 32 diversity measurements on social cohesion in the European context as the most promising candidate (Hooghe et al. 2006b). Since these statistics are assembled for numerous countries since the 1980s, they also allow the detection of changes over time. Although some questions have been raised about the degree of comparability of the OECD statistical sources (Lemaitre 2005), in practice they provide the most reliable source on migration to date (Hooghe et al. 2006a). Other measures at the country level are, of course, integration policy related variables. The first one is an index of immigrant multicultural policy, calculated by Banting and Kymlicka (2003). This Multiculturalism Policy-index is a simple scale, based on expert judgments on eight elements: (1) constitutional, legislative or parliamentary affirmation of multiculturalism (2) the explanation/celebration of multiculturalism in school curriculum; (3) the inclusion of ethnic representation/sensitivity in the mandate of public media or media licensing; (4) exemptions from dress-codes, Sunday-closing legislation, etc. (5) allowing dual citizenship; (6) the funding of ethnic group organizations or activities 2. In the last edition of the first wave of the European Social Survey (edition 6), this financial satisfaction measure of France was excluded from the general ESS dataset because of comparability issues. 8

10 (7) the funding of bilingual education or mother-tongue instruction, and (8) affirmative action for disadvantaged immigrant groups. The second policy variable is a factor score of four indicators measuring the openness of a country towards citizens with respect to the years of residence for applying naturalization, the possibility of having dual citizenship, ius soli at birth second and third generation citizens and ius soli entitlement after birth for second generation citizens. The third policy variable is again a factor score generated by three indicators measuring the procedure before one can be naturalized: having knowledge of the language and of society, and attending a citizenship ceremony. Most of these data are obtained from comparative studies on naturalization law (Bauböck et al. s.d.; de Hart & van Oers 2005). The fourth variable measures the voting rights for non-eu nationals (Waldrauch 2005). In this, a distinction is made between countries which have granted these voting rights early in time, late in time and not at all. In Appendix II, the scale reduction statistics on these items can be found. The method used to model these data is a random intercept multilevel model. 3 There are various reasons for doing multilevel analysis in this cross-national dataset. First, the basic assumption of a multiple regression model is the one of independent observations. In doing comparative country research, citizens are not independent: Irish citizens have things in common that make them distinct from Swedish citizens, which implies that respondents are nested within their country backgrounds. Multilevel analysis copes with this assumption of independent measures. Second, whereas previous research has used aggregate measures in cross-national analyses, such methods neglect the true nested nature of the data and lead to wrong estimates of the standard errors. In our analysis, we will develop a random intercept model, meaning that we allow the intercepts between countries to vary Social research that tries to explain variance in a dependent variable by independent variables at two or more levels often uses a regular multivariate regression analysis, whether the data at micro-level is aggregated or not. However, this method is less appropriate in a multi-level context. Multivariate regression analysis requires some assumptions. One of these assumptions is the independence of the observations (Hox 2002, 5). This assumption is met at the micro-level: in all the countries studied, the data are obtained by a random sample. However, the countries were not randomly chosen but assigned on theoretical premises, i.e., the ESS-investigation. The violation of this assumption has consequences for the variance-distribution on both the individual and country levels, a fact which is adjusted in a multilevel regression analysis. In coping with this estimation problem, multilevel analysis therefore is the best statistical technique equipped to analyze micro and macro-level variables in their relationship to generalized trust and ethnocentrism. 4. More elaborated analyses sometimes use random slopes models. In these types of models, the slopes of the micro level variables are estimated. However, there is no theoretical reason for estimating random slopes models for explaining the impact of diversity and immigration policies on social cohesion that is we do not have any strong reason to assume that the effect of diversity and policies differs across our cases. 9

11 Yet, there are also downsides to using multilevel analysis. Simulation studies warn about estimation problems at the second level when using small samples, like for instance only 20 countries (Kreft 1996; Hox & Maas 2001; Meuleman, Reeskens & Billiet 2007). At the same time, other assumptions of the regression model must be met as well. For instance, the assumption of having no outliers can be problematic in research using diversity and integration regimes as predictors. On average, the mean stock of the foreign population within European countries lies between 0 and 10 percent. Yet, Switzerland and Luxembourg, with foreigner rates of 20 and almost 40 percent respectively, clearly stand out as can be seen in Figure 2. Because the stocks of the foreign population are important in our analysis, we need to exclude these two countries, so that our final estimation only includes 19 countries. Nonetheless, multilevel analysis is still the most appropriate technique to handle this independency problem and the correct estimation of standard errors. Figure 2. The leverage function of Switzerland and Luxembourg in the analysis of generalized trust regressed on the stocks of foreign population Note: The graph represents the plot of the stocks of foreign population (OECD) on generalized trust (ESS). Simulations have shown that the relation between the stocks and generalized trust depends on the presence of these two outlier countries in the data. For the country labels: check the note to Figure 1. For both dependent variables we proceed in the same manner. First we estimate a baseline model with the individual level variables and the control country level variable as well as diversity, namely the gross domestic product per capita and the stocks of foreign population. In a second step, we include the policy variables which measure aspects of immigration regimes. Our attention is focused on two issues: First, whether any policies might be able to mitigate the potentially negative effects of diversity; and second, which policies are most 10

12 positively associated with social cohesion in Europe. It has to be remembered, however, that these analyses are built on only 19 cases, and therefore not all variables at the country level could be included simultaneously. We therefore used only this slim baseline model for country level variables, and entered additional integration policy measures one by one, to ascertain the effects of every variable separately. 4. Generalized trust The first step in the investigation of the effects of immigration regimes on social cohesion is the analysis of generalized trust. Before performing a multilevel regression analysis, it needs to be estimated whether there is a sufficient level of variability in the dependent variables among the 19 countries. According to the baseline model, about percent of the variation in generalized trust can be explained by country-relevant information, thus a multilevel regression analysis is appropriate. 5 In Table 1 we provide the results of the multilevel regression model with the test of all relevant micro level variables together with gross domestic product per capita and the stock of foreigners within a country. As previous research has already established, a substantial part of the variation in generalized trust in Europe can be explained by the income level of the country (Delhey & Newton 2005). For this reason, we include this country characteristic into the baseline model which will be used below for the one by one test of the immigration policies. Simultaneously, the stock of the foreign population as a proxy for diversity is included into the model. 5. The intraclass correlation is calculated by dividing the level 2 (country) variance between the sum of the level 2 (country) variance and level 1 (individual) variance: ρ 2 u 0 2 u 0 =. σ σ + σ 2 e 11

13 Table 1. Multilevel model for generalized trust Fixed part Intercept Gender Age Educational level Employment status Financial satisfaction Religious involvement Immigrant status Political powerlessness GDP per capita (in thousands) Stock of foreign population (in percentages) Random part Level 1. Individuals (residual variance) Level 2. Countries (intercept variance) R² at individual level 6 R² at country level # individuals # countries Female Ref: Male Secondary education Higher education Ref: Lower education Unemployed Other employment status Ref: Employed Modal member Marginal member Ref: Core member Born abroad Ref: Majority * p < 0.10; ** p < 0.05; *** p < 0.01; **** p < Parameter 3.400**** 0.170**** 0.001** 0.084*** 0.268**** *** 0.094**** 0.286**** **** **** ** **** 0.104**** ** 2.964**** 0.197*** 21.41% 64.95% 34, t-value The explained variance at the first level is defined as the proportional reduction in the mean squared prediction error (Snijders & Bosker 1999, 102): variance, 2 σ e R 2 2 [ σ u + σ ]( ) 0 e A 2 [ σ + σ ]( B) u e 2, where refers to the level 2 0 to the level 1 variance, (A) to the fact that these information is obtained from the estimated model and (B) from the null (empty) model. The explained variance at the second level can be defined as the proportional reduction in mean squared prediction error for the predicted group-specific value of the dependent 2 2 [ σ ( / )]( ) 2 u + σ 0 e G A variable (Snijders & Bosker 1999, 103): R2 1, where G refers to the number of groups. 2 2 / G ( B [ σ + ( σ )] ) u 0 e σ u0 12

14 Table 1 also confirms previous individual-level research on generalized trust: all theoretically relevant variables have a significant effect on the trust scale. With regard to gender, women are more trustful than men. Older people are more trusting than younger ones. The socioeconomical status has powerful effects as well: education, employment status and financial satisfaction are positively related to generalized trust. Concerning religious involvement, we see that the modal and marginal members are less trusting than the core members, namely these people who go at least once a week to a religious service. On average, people who are born abroad are less trustful than European majority citizens. Political powerlessness shows a strong negative relation: people who feel that they cannot influence politics are in general also less trustful. With respect to the two macro variables, the strong relation of gross domestic product with generalized trust is confirmed: the wealthier the nation, the higher the levels of trust of its citizens. Homogeneous societies with a lower proportion of foreigners have higher levels of generalized trust (Delhey & Newton 2005). In crude figures, the results of the multilevel regression analysis suggest that an increase of five percent on national diversity levels relates to a decrease of 0.6 on the generalized trust scale. Thus, generally, fostering trust in heterogeneous societies seems to be difficult, confirming some of the earlier results on diversity and trust. The question now is whether select immigration policies are able to mitigate these negative effects of diversity. These results are presented in Table 2, in which we add policy variables to the Baseline Model one by one. The first tested diversity regime variable, the Banting and Kymlicka immigrant multicultural policies index, shows contrary to our expectations a negative relation with generalized trust: the higher the level of multicultural policies, the lower the levels of trust, although the relation is not statistically significant. The same negative relation can be found for the openness of countries towards immigrants and trust: as expected, countries which are more restrictive towards new people have higher trust levels among their citizens. Yet, again here too the relation is not significant. Having a strict procedure for acquiring naturalization, such as knowing the national language and society and attending a citizenship ceremony, has also no statistically significant relation with trust. Yet, once again it seems that countries with these strict procedures show higher levels of generalized trust. Only the last set of tested variables, namely voting rights, show a highly significant relation: there is more generalized trust in countries where voting rights for non-eu nationals were granted rather early in time, for instance in the 1980s, compared with countries which granted voting rights for non-eu citizens late and compared to those that do not grant foreign voting rights at all. 13

15 What is remarkable is that the early voting right variable cancels out the negative and significant effect of the stocks of foreign population on generalized trust, which is not the case for any other policy variable. This means that the establishment of voting rights at an early stage mitigates the negative effect of diversity. Table 2. Multilevel macro model for generalized trust Model I Model II Model III Model IV param t-value Param t-value param t-value param t-value Fixed part GDP/cap 0.09*** *** *** ** 1.97 Stocks -0.13** ** * IMMCP Openness Procedure Late voting Early voting Ref: no vote Random part L1. Individuals L2. Countries R² individual R² country # individ # countries 2.87*** 0.23*** 22.99% 59.86% 28, * p < 0.10; ** p < 0.05; *** p < *** 0.24*** 23.28% 58.77% 26, *** 0.23*** 21.56% 58.91% 26, *** 2.96*** 0.13*** 23.02% 71.37% 34, To visualize the relationship between voting rights for non-eu citizens and generalized trust in Europe more clearly, we have presented the graphical plot of the bivariate relation between these two variables (Figure 3). The graph indicates that countries which have adopted voting rights rather early in time (Denmark, Norway, Finland, Sweden, Ireland and the Netherlands) have significant higher levels of generalized trust compared with countries which have adopted voting rights for non-eu citizens rather late (United Kingdom, Belgium, Hungary and Slovenia), or have not adopted voting rights for foreigners at all (Austria, Germany, Spain, France, Czech Republic, Italy, Portugal, Poland and Greece). Granting voting rights to non-nationals apparently is a very important symbolic recognition of shared community 14

16 membership. How one should model the causal relation between voting rights and generalized trust, however, is a matter to be dealt with in future analysis. Figure 3. Voting rights for non-eu citizens and generalized trust in Europe Note: The graph represents classification of countries on their voting rights for non-eu citizens (Waldrauch 2005) on generalized trust (ESS). For the country labels: check the note to Figure Ethnocentrism In addition to the investigation of diversity effects on generalized trust, the second step is an in-depth analysis on ethnocentrism as a dependent variable. The null model of the multilevel regression analysis shows that 8.64 percent of the variability in ethnocentrism can be explained by information at the country level. Despite the fact this intraclass correlation is remarkably lower compared with generalized trust, we believe that the nested data structure cannot be neglected and, once again, multilevel analysis seems to be the appropriate technique to investigate both micro- and macro-level effects on ethnocentrism, although overall we 15

17 expect country factors to have less influence in the ethnocentrism models compared to the trust models. First, we analyze the micro-level variables together with the theoretically relevant control variable GDP per capita and again the stocks of foreign population. In a later phase, we analyze the policy variables. As Table 3 below shows, the micro-level variables confirm in most of the cases the theoretical discussion. Men seem to have more negative feelings about immigrants compared to women. Older people are more negative towards outsider groups compared to the younger ones. With respect to the indicators for socio-economic status, namely educational level, employment status and financial satisfaction, all three show significant relations with a confidence interval of 99 percent: the higher educated are less hostile towards immigrants, just as the people who are satisfied with their financial situation; respondents who are unemployed on the other hand are significantly less open to foreigners. Religious orientation also is related to ethnocentrism: modal and marginal members show higher levels of prejudice. Respondents, who where not born in the country, show lower ethnocentrism levels compared with the majority population. People with high levels of political powerlessness indicate more negative feelings toward outsiders. The test of GDP per capita and the actual stocks of the foreign population did not reveal any significant results: citizens in wealthy countries and homogeneous countries are not significantly more positively oriented towards foreigners than less those in developed and more heterogeneous countries. The null effect of diversity on feelings of ethnocentrism is of high interest in the current heated debate, where scholars usually focus on the negative effects of diversity. 16

18 Table 3. Multilevel model for ethnocentrism Fixed part Intercept Gender Age Educational level Employment status Financial satisfaction Religious involvement Immigrant status Political powerlessness GDP per capita (in thousands) Stock of foreign population (in percentages) Random part Level 1. Individuals (residual variance) Level 2. Countries (intercept variance) R² at individual level R² at country level Female Ref: Male Secondary education Higher education Ref: Lower education Unemployed Other employment status Ref: Employed Modal member Marginal member Ref: Core member Born abroad Ref: Majority # individuals # countries * p < 0.10; ** p < 0.05; *** p < 0.01; **** p < Parameter *** 0.009**** **** **** 0.102** **** **** 0.539**** **** *** 14.43% 26.21% 33, t-value The diversity regime measures reveal some significant results on attitudinal ethnocentrism. First, once again the Banting and Kymlicka immigrant multicultural policies score has a positive effect: the higher the level of multicultural policies, the higher the levels of prejudice towards immigrants; yet, the effect is not statistically significant. The same story appears again for the openness of the country towards immigrants: more open countries have higher levels of ethnocentrism compared with countries which are more restrictive, but again this effect is not significant. Citizens in countries, which have an extended procedure for acquiring naturalization, show higher levels of foreigner hostility. Yet, this effect is once again not 17

19 significant. And most importantly, again voting rights for non-eu citizens, show a relationship with ethnocentrism. However, we can observe now that countries that granted voting rights late have the highest ethnocentrism figures. Citizens in countries in which voting rights were granted early on do have the lowest ethnocentrism levels, but this result is nonsignificant compared to countries where there are no voting rights for non-nationals at all. Table 4. Multilevel macro model for ethnocentrism Model I Model II Model III Model IV param t-value param t-value Param t-value param t-value Fixed part GDP/cap Foreign Stocks Multiculturalism Openness Naturalization Procedure Late voting Early voting Ref: no vote Random L1. Individ. L2. Countries R² individual R² country 3.13*** 0.24*** 13.19% 22.06% *** 0.24*** 12.36% 21.00% *** 0.21*** 12.44% 25.66% * *** 0.19*** 15.34% 32.97% # individ 27,910 26,420 25,891 33,918 # countries * p < 0.10; ** p < 0.05; *** p < Once again, taking a rather loose significance level of 90% into account, voting rights seem to matter in explaining ethnocentrism levels within Europe. But while we could observe a linear relation with trust levels, the effect on ethnocentrism seems rather curvilinear (Figure 4). Countries which have only recently granted voting rights to non-eu citizens differ significantly from countries which do not grant voting rights at all: there are higher ethnocentrism levels among their population. Again, early voting rights are associated with the best results or lowest levels in ethnocentrism, but this is not significant. Overall the same 18

20 conclusion holds for both generalized trust and ethnocentrism: voting rights seem to correlate in one way or another with ethnocentrism. Figure 4. Voting rights for non-eu citizens and ethnocentrism in Europe Note: The graph represents classification of countries on their voting rights for non-eu citizens (Waldrauch 2005) on ethnocentrism (ESS). For the country labels: check the note to Figure 1. As we mentioned already, in this paper we also want to experiment with a new measurement for ethnocentrism, i.e., the extent to which respondents overestimate the presence of foreigners in their own country. Our assumption here is that those who are hostile toward the presence of foreigner will display a tendency to overestimate the presence of foreigner in their own community. First, Figure 5 makes clear that there is indeed a considerable amount of variation with regard to this variable. While the respondents in Luxembourg, Switzerland and Denmark do a pretty good job in estimating the presence of foreigners in their own country, the inhabitants of Slovenia and Hungary tend to overestimate this presence quite strongly. 19

21 Figure 5. Ratio between guessed and actual presence of foreigners in European countries LU CH DK AT DE IE BE NO GR FI CZ SE ES IT PT FR GB NL SI HU Note: The graph represents the deviation ratio between the guessed and actual stocks of foreign population. For the country labels: see Figure 1. Poland is removed from this analysis because of its outlier status. In our effort to explain these differences we proceed in the same manner. Again, education and gender prove to be strong variables to explain the overestimation of the presence of foreigners. Somewhat surprisingly, respondents who themselves have been born abroad, also tend to overestimate the presence of foreigners in their host country. We should also note that higher GDP per capita reduces this cognitive form of ethnocentrism, and diversity in itself tends to be beneficial here. People in diverse societies make better guesses about the proportion of foreigners than those in homogeneous countries. 20

22 Table 5. Multilevel model for overestimation ratiog Fixed part Intercept Gender Age Educational level Employment status Financial satisfaction Religious involvement Immigrant status Political powerlessness Female Ref: Male Secondary education Higher education Ref: Lower education Unemployed Other employment status Ref: Employed Modal member Marginal member Ref: Core member Born abroad Ref: Majority GDP per capita (in thousands) Stock of foreign population (in percentages) Random part Level 1. Individuals (residual variance) Level 2. Countries (intercept variance) R² at individual level R² at country level # individuals # countries Parameter 7.325*** 0.809**** **** **** **** 0.400*** 0.230**** **** 0.199*** 0.216*** 0.517**** 0.499**** * *** **** *** 16.72% 47.84% 27, * p < 0.10; ** p < 0.05; *** p < 0.01; **** p < Poland was removed from this analysis. t-value Subsequently, we add the policy level variables to our analysis, which yields some unexpected results. Respondents living in a country with a strong policy of multiculturalism, or with relatively open integration regimes tend to overestimate the presence of foreigners, and the same is true for the establishment of early voting rights. Here we seem to observe the downside of multicultural and open integration policies: these policies seem to create the impression among the population that the country is being flooded by the arrival of new ethnic minorities. Most importantly though voting rights matter once more: just as we observed with regard to the more conventional ethnocentrism measurement, wrong 21

23 estimations on the stocks of foreigners are significantly higher in countries that adopted voting rights for non-nationals late. For those countries which granted voting rights to foreigners already at an early stage, there is not a significant effect when comparing to countries with no voting rights. In short, the effect is curvilinear once more. Table 6. Multilevel macro model for overestimation ratio Model I Model II Model III Model IV param t-value param t-value param t-value param t-value Fixed part GDP/cap -0.05** Foreign Stocks -0.32*** *** *** *** Multiculturalism 0.42*** 5.11 Openness 0.73*** 3.60 Naturalization Procedure Late voting Early voting Ref: no vote Random L1. Individ. L2. Countries R² individual R² country 9.27*** 0.25*** 57.41% 86.73% *** 0.38*** 57.21% 84.33% *** 0.66*** 55.21% 79.91% *** *** 1.30*** 21.70% 66.35% # individ 24,522 23,557 22,614 27,571 # countries * p < 0.10; ** p < 0.05; *** p < Conclusion and discussion With a comparative data set we were able to perform a simple test whether specific policies targeted at the integration and inclusion of foreign citizens in any way relate to social cohesion in European societies. There are two noteworthy results that push the analysis of social cohesion further. First, we found that those respondents, who live in countries in which foreigners are given the most encompassing voting rights and where these rights have been established at an early stage of rising immigration, were most trusting; they exhibited the lowest levels of ethnocentrism and also indicated the best guess of the proportion of their 22

24 foreign populations in their country. Thus countries who adopted voting rights early do have the best social cohesion across all indicators, and in the case of trust this effect is significant. In addition, other policy initiatives such as open immigration regimes or more encompassing multi-cultural policies influence aspects of ethnocentrism but not generalized trust. The influence, however is not always as we expected. Whereas more restrictive immigration regimes are better in fostering low levels of ethnocentrism than open regimes; we also found contrary to our expectations that multi-cultural policies seem to lead to the impression that there are more foreigners in the country than the actual figures show. Still, with these limitations in mind, the findings are highly interesting for the policy community, as they do suggest that policies have the potential of ameliorating effects on the tensions in increasingly diverse societies. However, before drawing any such causal conclusions, more research needs to investigate this finding. At the moment, we do not have any evidence for the causal direction of this relationship. It is certainly plausible that high trust societies attend to the rights of minorities earlier and in a more encompassing way, because generally high-trust societies are more oriented towards equality in the first place. At the same time, we find it just as plausible that policies that are egalitarian and universal in character and targeted at integrating minorities better into every-day political and social life (as in the case of voting rights) have beneficial effects on breaking down tensions and distrust between different societal groups. The causal mechanisms here would be based on the idea that these policies indeed have symbolic as well as actual effects when it comes to the status of foreigners. At the symbolic level, granting the right to vote certainly reveals that foreigners are indeed fully recognized and equal members of society. At the same time, voting rights enhance the status of immigrants in societies, and perhaps even positively influence their trust levels, their participation in the social-political life and as a result bring them in better contact with majority populations. The adoption of voting rights therefore seems to be a step in the right direction, especially if the goal is to foster generalized trust as an indicator of social cohesion. Although this empirical relationship stands, the results regarding the negative effects of late voting rights on ethnocentrism still set limits for policy proposals. It is important to know though that multicultural politics and recognition of group rights, however, might also have a detrimental effect, as they might lead to the perception that societies are quite diverse, even when this is not the case in reality. 23

25 The final verdict though about diversity, diversity policy and social cohesion will have to wait until we proceed to the next stage of our research project when we will try to better understand the extent to which the gap between immigrant and majority populations is closed in countries where we find active policies for the integration of immigrants. Second, when excluding outliers form the analysis, in Europe too it was confirmed that the influx of foreign citizens had significant detrimental effects on social cohesion. However, the inclusion of select policy variables diminished these negative effects. In the case of voting rights the policy effect cancels the negative effects of diversity on generalized trust. This finding in itself is important for the discussion about the negative effects of diversity particularly within North America. It is also plausible that our results are not directly comparable to many North American studies as we examine the effect of diversity at a much higher level of aggregation, namely the country (as opposed to smaller units such as neighbourhood or census tracts). Nevertheless it seems clear that the conclusion about ethnic diversity exerting negative effects on social capital that tends to dominate the US literature might have been drawn too early, and should at the very least consider the beneficial effects of immigration policy. 24

26 References Alesina, A. & E. La Ferrara (2002). Who Trusts Others? Journal of Public Economics, 85, Banting, K. & W. Kymlicka (2003). Do Multiculturalism Policies Erode the Welfare State? Paper presented at the meeting of the International Sociological Association, Toronto, August 21-24, Barry, Brian (2001). Culture and Equality. An Egalitarian Critique of Multiculturalism. Cambridge: Harvard University Press. Baubock, R. et al. (s.d.). Migration and Citizenship: Legal Status, Rights and Political Participation. State of the Art Report for IMISCOE cluster B3. Amsterdam: International Migration, Integration & Social Cohesion. Brehm, J. & W. Rahn (1997). Individual-Level Evidence for the Causes and Consequences of Social Capital. American Journal of Political Science, 41(3), Citrin, J. and J. Sides. Forthcoming. "European Opinion about Immigration: The Role of Identities, Interests, and Information." British Journal of Political Science. Cornelius, et al. (2004). Controlling Immigration: A Global Perspective, Second Edition. Stanford: Stanford University Press. Cornelius, W. & M. Rosenblum (2005). Immigration and Politics. Annual Review of Political Science, 8, De Hart, B. & R. van Oers (2005). International Trends in Nationality Law: A Comparison. Nijmegen: Radboud University of Nijmegen. Delhey, J. & K. Newton (2005). Predicting Cross-National Levels of Social Trust. Global Pattern or Nordic Exceptionalism? European Sociological Review, 21(4), Durkheim, E. (1973 [1893]). De la Division du Travail Social. Paris: PUF. Hero, R. (2003). Social Capital and Racial Inequality. Perspectives on Politics, 1, Hooghe, Marc (2003). Value Congruence and Convergence Within Voluntary Associations: Ethnocentrism in Belgian Organizations, Political Behavior, 25(2), Hooghe, M., et al. (2006a). Migration to European Countries. A Structural Explanation of Patterns, Leuven: Centre for Citizenship and Democracy. Hooghe, M. et al. (2006b). Ethnic Diversity, Trust and Ethnocentrism in Europe. A Multilevel Analysis of 21 European Countries. Paper Presentated at the 103 Conference of the American Political Science Association, Philadelphia, August 31-September 3, Hox, J. (2002). Multilevel Analysis. Techniques and Applications. Mahwah: Erlbaum. 25