Ethno-racial Diversity and Social Capital in English Schools Abstract Since the racial disturbances in the Northern English towns of Bradford, Oldham and Burnley, the belief has gained strength in educational circles that ethnically mixed schools contribute to inter-community trust and social cohesion. Several recent studies from the field of political science, however, have found that trust and participation are lower in ethnically diverse neighborhoods. This paper explores the relation between ethno-racial diversity and social capital in English schools using data from the IEA Civic Education Study. Controlling for various conditions at the individual and school class level, it finds no effect of diversity on generalized trust and ethnic tolerance. Diversity is positively related to participation in human rights and environmental organizations and participation in ethnic and religious associations, however. It further finds social capital, particularly its bridging form, to be a complex multidimensional phenomenon. Consequently, it proposes to decompose social capital into its constituent parts – trust, participation and tolerance. Key words: social capital, bonding, bridging, ethnic and racial diversity, schools 1 Introduction Desegregation has been a prime educational objective throughout the western world over the last forty years. Initially, the effort to create ethnically and racially mixed schools was motivated by indignation about racial inequality and discrimination produced by segregated school systems. The Civic Rights movement in the US came to epitomize this effort in its struggle to achieve civic equality, integration and emancipation for African Americans. The term ‘desegregation’ itself was very much part of and restricted to a social justice discourse. From the late 1990s, however, the objective of mixed schools also started to be embraced by scholars concerned about declining levels of community cohesion and growing ethnic and religious intolerance. These scholars felt that the prevailing practices of multiculturalism and recognition of minority cultures had only reinforced the isolation of ethnic and racial minorities and had led to more divisiveness. Herbert (2001), for instance, argued that faith schools serving the needs of particular ethnic groups constitute a kind of ‘educational apartheid’ segregating rather than integrating various ethnic communities. A string of events in the early 2000s – Nine-Eleven, the racial disturbances in the Northern English towns of Bradford, Oldham and Burnley, and the London and Madrid underground bombings – added great weight to the analysis of these scholars and led to the adoption of their views by various policy makers and government advisers. Thus, the Cantle Report, commissioned by the British Home Office, suggested that ethnically segregated schools have contributed to the disorders in Bradford, Oldham and Burnley and explicitly called for admissions policies maximizing the proportion of pupils 2 of the same cultural or ethnic background at 75 per cent in schools in multi-cultural areas (Cantle 2001). However, at a time when a consensus in education circles about the benefits of ethnic mixing seems to be growing, a tradition in the political science field pointing to the negative effects of diversity on social capital, trust and solidarity is also gaining strength. A recent important contribution in this tradition is the study on local level community cohesion in the United States by Putnam (2007). His main finding is that people in ethnically and racially diverse communities exhibit not only less out-group but also less in-group trust compared to people in homogenous neighborhoods. In other words, diversity has a “constricting” effect (p. 144), making people in diverse surroundings “hunker down – that is, to pull in like turtles” (p. 149). This raises the obvious question whether Putnam’s findings also apply to schools, like the neighbourhood a local level phenomenon. In other words, is it possible that schools with a mixed ethnic intake do not produce cross-cultural integration and bridging social capital (as hoped for by the Cantle report) but rather its opposite: isolation, disengagement and distrust? We will explore this question in the present paper by analyzing survey data collected among 14 year olds in England. We find, firstly, that social capital, particularly its bridging variety, is a complex phenomenon comprised of quite unrelated behavioral and attitudinal components. We further find that the effect of diversity on the attitudinal component (generalized trust and ethnic tolerance) is merely compositional and therefore spurious. Its effect on the behavioral component (participation in several organizations) is positive however, which supports the notion 3 that ethnic mixing contributes to inter-ethnic harmony (i.e. the key assumption of the Cantle report). We start by discussing the much used concept of social capital, the dependent variable of this paper. We then review the political science literature on relation between diversity and social capital and examine to what extent educational studies have addressed this issue. Subsequently, we discuss the data source, research design and methodology used for the analyses. Finally we present the results of these analyses. The conclusions rehearse the main findings and offer suggestions for further research. Social capital A term widely used in the social sciences to include notions of trust, cooperation and community cohesion is social capital. In fact, Putnam has been recognized as the key scholar promoting this concept ever since he published Making Democracies Work: Civic Traditions in Modern Italy (1993), in which he argued that social capital enhances the performance of democratic government. In his 2007 study it is therefore the relation between diversity and social capital that Putnam essentially examines, albeit with a strong focus on trust as one of its components. Defining the concept as “social networks and the associated norms of reciprocity and trustworthiness” (Putnam 2007: 137), he contends that social capital, like monetary and human capital, is a desirable asset in that it contributes to the health, happiness, civic mindedness and career opportunities of the individuals who possess it. Social capital is further said to have positive spillover effects 4 for outsiders: living in an area rich in social networks contributes to trust in neighbors and feelings of safety even if one does not join community life oneself. As we seek to replicate Putnam’s study for schools, we adopt his definition of the term and will operationalise the concept in accordance with this definition. Embracing Putnam’s definition of social capital does not mean that we apply the term without reservations. Two important objections can be raised. First, as the term incorporates several notions with some referring to attitudes (trust and norms) and others to behaviors (networks, association, cooperation), it can be criticized for being an incoherent concept comprising a number of disparate components. This criticism does not seem to be of great concern to Putnam, though, as he simply assumes social capital to refer to a coherent syndrome of attitudes and behaviors at the individual level: Our discussion of trends in social connectedness and civic engagement has tacitly assumed that all the forms of social capital that we have discussed are themselves coherently correlated across individuals. This is in fact true. Members of associations are much more likely than non-members to participate in politics, to spend time with neighbors, to express social trust and so on (Putnam 1995: 73). To be sure, there is research supporting this assumption for a number of western countries including the US and Great Britain (Brehm and Rahn 1997; Hall 1999). However, studies examining a much larger set of countries have found the components of social capital to be only weakly correlated across individuals (Newton and Norris 2000; Janmaat 2006). Janmaat (2006) moreover argues that the external distinctiveness of the concept can be called into question because the correlations among the social capital 5 components are not stronger than the correlations of these components with external conditions. From a policy perspective working with concepts that are not internally consistent is troublesome, as it may well mean that interventions aimed at fostering one particular component do not contribute to other components – or worse affect them negatively (see also Janmaat 2008; Tolsma et al 2008). In the empirical section we will therefore explore to what extent our indicators of social capital intercorrelate and form a distinctive syndrome. Secondly, it has been pointed out that social capital can take many forms, some of which are not commendable such as youth gangs, ethnic militias or terrorist organizations. Putnam (2007: 138) is the first to acknowledge this, calling for instance Al Qaeda “an excellent example of social capital, enabling its participants to accomplish goals they could not accomplish without that network”. Yet, he maintains that social capital by and large is a positive phenomenon leading to happier, healthier and better educated people and making democracy and the economy work better. Still, in our view this objection cannot be dismissed this easily, particularly because it relates to that other important issue in social capital theory: the distinction between bonding and bridging forms of social capital. The bonding variety refers to small, culturally homogenous communities characterized by ‘thick’ forms of trust, solidarity and cohesion, and high levels of exclusion. Bridging forms, by contrast, refer to broad outward-looking networks uniting people of various backgrounds who interact on the basis of mutual respect (Putnam 2000). The latter are usually seen in a more positive light as they are considered to contribute to tolerance, overall integration and societal cohesion (Granovetter 1978). The former on the other hand are looked upon with much more ambivalence: bonding 6 forms may help minority groups in finding their way in the wider society but they may also enhance ethnocentrism, oppose society and fuel intercommunity distrust and hostility. Examples of this divisive form of bonding social capital are abundant and not just restricted to the extreme groups mentioned above. It suffices to refer to the religious communities in Northern Ireland and in Lebanon and the ethnic communities in Bosnia and in Sri Lanka to show that even more moderate forms of exclusive bonding can have disastrous effects for society at large (Green et al 2006). Clearly, it was fear of the divisive variety of bonding social capital that has motivated the authors of the aforementioned Cantle report to advocate ethnic mixing. This makes the distinction between bonding and bridging social capital highly relevant for the current study. Is there a tension between bonding and bridging social capital as the Cantle report implicitly implies, and are schools with a mixed ethnic intake therefore undermining bonding forms but fostering bridging forms? Or is diversity affecting both forms of social capital, as Putnam’s study shows? Or is diversity connected in completely unexpected ways to bonding and bridging forms of social capital? These are the key questions we will address in the empirical section. The attentive reader might ask why we focus on Putnam’s conception of social capital when there is literature relating the concept specifically to youngsters, which is after all the target group of this study. Coleman (1988), for instance, argues that social capital as manifested in the family and in the immediate environment of children, i.e. micro forms of social capital, are particularly important for the creation of human capital and school performance of youngsters. In his view, a crucial property of this environment is the degree of intergenerational closure of networks of parents and children: the better 7 parents know the parents of their children’s friends, the more likely they are to develop mutually agreed norms and sanctions regulating their children’s behavior and contributing to their academic achievements. Coleman’s conception of social capital, however, is incapable of addressing issues of inter-community trust, solidarity and selforganisation, which is the key concern of the Cantle report. Indeed, it is entirely imaginable that multi-ethnic communities display both high levels of intergenerational closure within each group (i.e. high levels of bonding) and high levels of inter-group hostility (i.e. low levels of bridging) simultaneously. For this reason we focus on Putnam’s bonding and bridging forms of social capital and disregard scholars who understand social capital as an individual or micro-level property (e.g. Bourdieu (1986) who considers social capital to be a resource possessed and exploited by individuals). Another reason to embrace Putnam’s conception of social capital is the growing body of research claiming that its key components, trust, tolerance and participation, are shaped in one’s formative years and will change little thereafter (Stolle and Hooghe 2004; Flanagan and Sherrod 1998). It is thus highly relevant to explore the conditions affecting the acquisition of these attitudes and behaviors by adolescents since this is precisely the age group in which these outcomes take definite shape. The diversity perspective At its core the diversity perspective argues that cultural similarity enhances trust, solidarity and cooperation. While some scholars seem to accept this claim as a given fact 8 (e.g. Alesina and La Ferrara (2002: 208), who argue that people have “a natural aversion to heterogeneity”), others have sought to explain the relation theoretically. Conflict theorists for instance assert that it is the competition over scarce resources that fuels intragroup solidarity and inter-group hostility. The more often people encounter members of ethnic out-groups in their daily lives and the larger the size of these out-groups, the tighter their in-group bonding and stronger their fear of ethnic others (Bobo 1999; Quinlan 1995). From a slightly different viewpoint, evolutionary theorists have also pointed to the mechanism of group competition. Thus Salter (2004) contends that the current propensity of people to bond with ethnic kin and be distrustful of ethnic others is the result of a long historical process in which clans and tribes with internal mutual support schemes have outperformed groups lacking these support systems in the group struggle for survival. In this perspective, multiethnic societies will continue to be troubled by ethnic conflict, ethnic nepotism and faulty welfare systems. However the diversity argument is elaborated theoretically, its advocates have marshaled an impressive amount of research evidence in support of their claims. Most of this research relates to the United States and Canada, where a series of studies have found a negative relation between ethnic or racial heterogeneity on the one hand and trust, cooperation and solidarity on the other at the city or state level (e.g. Alesina, Baqir and Easterly 1997; Hero and Tolbert 1996; Luttmer 2001; Soroka, Johnston and Banting 2004). Combining tract level data from the census and public opinion data from the General Social Survey, Luttmer, for instance, discovered that white support for welfare spending diminishes as the proportion of black recipients of welfare in the tract population increases. This result indicates that racial heterogeneity seems to be 9 particularly harmful for social solidarity if racial cleavages coincide with social inequalities. Focusing on Canada, Soroka, Johnston and Banting (2004) established that interpersonal trust diminishes as the proportion of visible minorities in census tracts increases, and that trust in turn is positively linked to pro-welfare attitudes. The aforementioned study by Putnam (2007) is the latest in this string of studies. Controlling for a number of individual and contextual conditions, he found that diversity does not only reduce both in- and out-group trust, but is also showing a negative link with other civic attitudes such as cooperation, altruism, political efficacy and confidence in local institutions. Putnam believes though that this negative impact of diversity is likely to be only temporary. In the long run, “successful immigrant societies create new forms of social solidarity and dampen the negative effects of diversity by constructing new, more encompassing identities” (ibid. pp. 138,139) He explicitly notes that ethnic and racial identities are dynamic, socially constructed phenomena and that therefore the substance of diversity can change with some cleavages becoming less and others more salient over time. Research highlighting a negative link between diversity and trust is counterbalanced however by studies finding no effect for diversity. In her study of formal and informal social capital in 839 British neighbourhoods, Letki (2008), for instance, found that the socio-economic status of a neighbourhood was the most important driver of social capital, while its ethno-racial composition hardly mattered. Oliver and Wong (2003) even found a positive relation between community heterogeneity and favorable attitudes towards out-groups. People living in homogenous areas turned out to have stronger ethnic stereotypes and to be more suspicious of ethnic others. Tolsma et al 10 (2008) observe that ethnic diversity can be related in completely different ways to various components of social cohesion. In their study of 503 neighbourhoods and 245 municipalities in the Netherlands, they found that ethnic diversity is not conducive to volunteering and the frequency of contacts with neighbours, but contributes positively to ethnic tolerance and shows no relation with generalized trust. Summarizing this contradictory evidence, we can conclude that the nature of the link between diversity and social capital seems to vary across different components of social capital and different countries and to depend on the number and kind of control variables included in the analysis. The review thus only adds weight to the concern raised above that social capital encompasses a collection of highly disparate notions. Remarkably, the education literature seems not to have picked up on the explosion of political science studies suggesting a negative link between diversity and social capital. Traditionally, educationalists have been more concerned with the effects of ethnic and racial segregation on student performance (e.g. Orfield 1978; Rumberger and Willms 1992). Yet, there is a distinct tradition within the field of education examining the consequences of (de)segregation for inter-racial friendships, inter-cultural understanding and comfort in dealing with multi-cultural settings. Recent studies in the US by Frankenberg et al (2003) and Holme et al (2005), for instance, found that the experience of racially mixed schools left graduates with a better understanding of different cultures and an “increased sense of comfort in interracial settings” (ibid p. 14). Research by Ellison and Powers (1994) and Sigelman et al (1996), moreover, shows that the tolerant attitudes and interracial friendships developed in racially integrated schools persist into adulthood. Holme et al (2005) further claim that the daily experience of interracial 11 schooling is much more effective in this regard than multicultural curricula or student exchange programs. Bruegel (2006), investigating inter-ethnic frienships among pupils of 12 primary schools in London and Birmingham, reaches similar conclusions. In her view, “the day-to-day contact between children [of different ethnic background, XX] has far more chance of breaking down barriers between communities, than school twinning and sporting encounters” (ibid. p. 2). Given her positive appraisal of racially mixed schools, she is skeptical of policies promoting school choice as these might have the unintended effect of promoting segregation. Other research in the UK focusing on community relations in Northern Ireland has argued that integrated (i.e. mixed faith) schools “impact positively on identity, outgroup attitudes, forgiveness and reconciliation” (McGlynn et al, 2004: 1). All these studies can thus be said to strongly refute the aforementioned conflict perspective: daily contact between people of different ethnic or racial background does not contribute to interethnic hostility but to its opposite: inter-ethnic understanding and trust, in short to bridging social capital Yet, not all educational research points in this direction. In fact, a thorough review of the literature on the effects of school desegregation observed that the extant research was strikingly inconsistent in its findings (Schofield 2001). While many studies indeed point to the positive effects of ethnic mixing for out-group attitudes, others suggest the opposite or claim there is no effect, and again others argue that the effects are different for majority and minority groups. Our analyses will allow us to explore all these possibilities for both bonding and bridging social capital. 12 Data, indicators and method We explore the relationships between diversity and social capital by analyzing data of the IEA Civic Education Study (Cived) (Torney-Purta et al 2001). This study consisted of a large scale survey conducted in April 1999 among a sample of 90,000 14-year-olds in 28 countries worldwide. To this day, the Cived study has not enjoyed the same level of popularity as other large international surveys addressing civic values, such as the World Values Survey, the ISSP and the Eurobarometer. This is somewhat surprising given the quality of the data. Not only are the national samples much larger in the Cived study (around 3000 respondents in each country), the non-response is also significantly lower than in the other surveys. One of the advantages is that respondents of immigrant origin are represented to a sufficient degree (the share of these minorities ranges between seven to twenty per cent of the national samples of a number of West-European states). Given the nested character of the national samples, with one class being selected in 120-200 sampled schools in each participating country, the Cived study further allows researchers to explore both contextual effects (such as diversity) and individual-level factors. We selected the national sample of England, which is composed of 3043 students selected in 128 schools (i.e. classes). Social capital 13 Ideally, the bonding and bridging forms of social capital are explored with measures reflecting in- and out-group trust, solidarity and cooperation. Unfortunately, apart from two items on participation in an ethnicity or religion-based organizations, the Cived questionnaire does not have questions capturing the ethnic in/out-group notion. It does however include items covering the different scales at which the bonding and bridging forms can be said to operate. While bonding social capital is clearly a micro-level phenomenon reflecting close contacts with family, friends, classmates and ethnic kin, bridging social capital refers to the often infrequent ties that bind people of different backgrounds together at various levels, micro, meso and macro. The latter moreover can be said to include notions of tolerance and respect for people who are culturally or ideologically different. In line with these definitions, we devised one measure reflecting the bonding form and three measures reflecting the bridging form of social capital: Bonding - Bridging Participation in ethnic and religious - Participation in human rights and organizations environmental organizations - Generalized trust - Ethnic tolerance The measure for bonding social capital was tapped with the item “have you participated in the following organizations? (1) a cultural association based on ethnicity; (2) an organization sponsored by a religious group”. Based on the answers to this item, we 14 created a binary variable with values 0 [no participation] and 1 [participated in one or more organizations]. We did likewise for participation in human rights and environmental organizations, which we considered to reflect bridging social capital. Although this measure does not capture the in/out-group dimension, it does nicely represent the broad, outward-looking forms of association said to be typical of the bridging variety. Moreover, according to Hooghe (2003) an organization’s ideology is more important than the ethnic or social make up of its member base for the promulgation of values like tolerance and intercultural understanding. Analyzing survey data of the adult population in Flanders, Belgium, he found that membership of environmental and human rights organizations was particularly negatively related with ethnocentrism (i.e. the counterpart of tolerance), controlling for many background variables including education. Generalized trust was tapped with an item asking respondents how much of the time they trusted the people “who live in this country” <categories: never – only some of the time – most of the time – always>. Because of the explicit reference to the national scale in this item we considered it to reflect bridging social capital. The bridging nature of this item is moreover supported by Uslaner (2000: 575), who found generalized trust to be closely linked to trust in strangers. We further used a ready-made scale in the Cived database as a measure of ethnic tolerance, our third indicator of bridging social capital. The scale clusters the following five items on immigrants and is internally consistent with an alpha reliability of .82 (Torney-Purta et al 2001, 208): (1) Immigrants should have the opportunity to keep their own language; 15 (2) Immigrants’ children should have the same opportunities for education that other children in the country have; (3) Immigrants who live in a country for several years should have the opportunity to vote in elections; (4) Immigrants should have the opportunity to keep their own customs and lifestyle; (5) Immigrants should have all the same rights that everyone else in a country has. Categories: <strongly disagree – disagree – agree – strongly agree> These items, it must be noted, seem to reflect primarily accommodating views on immigrants rather than ethnic tolerance. However, Ford (2008) showed that such views are strongly linked to ethnic tolerance (as measured, in his research, by attitudes on social contact with ethnic minorities at work or in the family). Pettigrew (1997) moreover found these views to correspond in the same way as ethnic tolerance to a number of explanatory conditions including intergroup contact. Theoretically it is also highly plausible that the two are intimately related. Expressing disagreement with the five statements implies privileging the native majority over immigrants, a mindset which intuitively goes together with ethnocentrism and prejudice – the very antonyms of ethnic tolerance. For these reasons we assumed the five-item scale to be a good proxy of ethnic tolerance. The higher the values on this scale, the more the respondent agrees with the five statements and the more tolerant we consider him/her to be. 16 Independent variables We collected the answers to the question “which best describes you”1 to calculate the proportion of pupils not identifying as white, and used this proportion as a class-level measure of ethnoracial diversity, our main variable of interest (henceforth simply called ‘diversity’). Properly speaking this measure refers to density, which does not necessarily correspond to diversity. Density, for instance, is high and diversity low in situations where a single ethnic minority group makes up the majority of the class population. However, having carefully examined the data, we found that classes with high ethnic densities were also relatively diverse in terms of the number of children of different ethnic groups. For this reason we considered this measure of density to be a good proxy of diversity. Moreover, density measures have the distinct advantage of not being colour blind like the more traditional measure of diversity, the Herfindahl Index of Fractionalization (Tolsma et al 2008). Unlike the latter, density measures are able to distinguish a situation of an 80% native majority and a 20% ethnic minority from its mirror image (80 % ethnic minority and 20 % native majority). Being able to distinguish between the two situations obviously matters for West-European countries because differences between the ethnic majority and ethnic minorities are accentuated by gaps in socio-economic status. In subsequent analyses we will explore the effect of diversity controlling for a number of individual and class-level conditions. The individual level conditions are: (1) 1 The answer categories of this question are: (1) White; (2) Black Caribbean; (3) Black African,; (4) Black other; (5) Indian; (6) Pakistani; (7) Chinese; (8) Bangladeshi, (9) Other 17 gender [0 – girl; 1 – boy]; (2) Social status [scale with six values] (based on the item ‘number of books at home’; this item is strongly correlated with ‘education mother’ and ‘education father’; unlike Hooghe et al (2007) we chose not to create a composite index combining all three items because of high numbers of missing values on the education items); (3) civic competence (a ready-made composite measure based on the results of a civic knowledge and skills test); (4) ethnoracial identity [0 – white; 1 – non-white] (based on the “which best describes you” item). The importance of each of these conditions in shaping different components of social capital has been amply demonstrated in the literature: on the effect of gender, see Verba, Brady and Schlozman (1995) and Hooghe and Stolle (2004); on that of social status and educational attainment in particular, see Dalton (2004), Nie, Junn and Stehlik-Barry (2004) and Hagendoorn (1999); on that of civic knowledge and skills, see Galston (2001) and Delli, Carpini and Keeter (1996); on that of ethnic background, particularly of a non-western kind, see Rice and Feldman (1997). The inclusion of ethnoracial identity enables us to assess whether diversity represents a true contextual effect exerting an independent influence on the dependent variables or whether it is merely a compositional effect reflecting no more than the sum of individual differences. A class-level condition is a true contextual effect if the social capital levels of all students in that class are affected by this condition. The condition is only a compositional effect if it no longer exerts an independent effect on social capital once it is controlled for its individual level equivalent. For example, if differences in social capital are solely a function of individual ethnoracial identity, then the diversity of 18 a class is nothing more than a compositional effect. If diversity affects all students in that class, in addition to the effect of ethnoracial identity, then it is a contextual effect. We used two class-level conditions as control variables. The first of these – classroom climate – is the class average of a ready-made scale in the database labeled as ‘an open climate for classroom discussion’. Previous research by Torney-Purta (2002, 2004) on the same dataset has shown that this variable is strongly correlated to various components of social capital. The second is class status, which is the class average of the aforementioned individual-level social status. Many studies have pointed to the importance of this contextual condition for a range of social outcomes. As noted before, Letki (2008) found low neighbourhood status to be particularly harmful for both formal and informal forms of social capital. In similar vein, Oliver and Mandelberg (2000) note that residents in low status neighbourhoods are often exposed to crime, decay and disorder, leading them to develop feelings of anxiety, alienation, and suspicion towards strangers. Table 1 provides the descriptives of all variables. The distribution of scores on the trust variable (‘never’ – 5.1%; ‘only some of the time’ - 29.5%; ‘most of the time’ – 36.5%; ‘always’ – 12.6%) appears to approximate a normal distribution, which means that it can be analyzed as a continuous variable in a linear regression model. The ethnic tolerance scale can likewise be treated as a continuous variable. The scores on the other two dependent variables (the two participation measures) very much lean towards nonparticipation (85.1% and 85.3% not participating in religious/ethnic and environmental/human rights organizations, respectively). We treat these measures as binary variables and will consequently analyze them in logistic regression models. 19 Among the independent variables we find that ethnoracial diversity is highly skewed towards the homogenous end, the average share of non-white students being only 12.7%. Unsurprisingly, the variable on which diversity is based, ethnoracial identity, is tilted in a similar manner with a mere 12.6% not identifying as white. Table 1 about here We further note that the strength of the correlations among the contextual variables is not such that there is a risk of multicollinearity in subsequent analyses (these correlations are: -.03 between diversity and class status; .14** between diversity and classroom climate; .36** between class status and classroom climate). Remarkably, ethnoracial diversity and class status are only weakly correlated, which runs counter to the common perception that diversity is closely linked to deprivation. Finally, we examine the internal consistency of social capital and its bonding and bridging varieties. Table 2 presents the bivariate correlations among and between the three measures of bridging and the one measure of bonding social capital. It turns out that our aforementioned skepticism towards the concept is more than warranted: we see that the three indicators of bridging social capital are strikingly uncorrelated. In other words, generalized trust, participation in inclusive organizations and ethnic tolerance do not necessarily go together. By contrast, the two participation measures, although tapping different varieties of social capital, are strongly correlated, blurring the difference between bonding and bridging. The only relation supporting the notion that there is 20 tension between the two forms of social capital is the negative correlation between generalized trust and participation in ethnic/religious organizations. Table 2 about here What the pattern of correlations above all shows is the autonomous nature of the two attitudinal measures vis-à-vis each other and vis-à-vis the two participation measures. This does not only undermine the distinction between bonding and bridging forms of social capital but also the validity of the concept of social capital itself. Our findings thus strongly contrast with those of Brehm and Rahn, who, as noted before, did find a cohesive social capital syndrome of attitudes and behaviors at the individual level. Because our indicators neither cluster neatly in a bridging dimension of social capital nor in an overall syndrome of social capital, we will explore them individually in the ensuing analyses. We would like to emphasize that our continued use of the terms bonding and bridging does not suggest that we hold these theoretical concepts to refer to welldemarcated empirical phenomena. Method of analysis Since our independent variables are pitched at two levels (class and individual) and our dependent variables are at the individual level, the appropriate method to explore the relationships between diversity and social capital is a multi-level analysis. This is all the more required given the nested structure of the data. A structure of this kind, with 21 students being nested in classes, classes in schools, and schools in countries, precludes the use of more conventional multiple regression techniques since these require that observations are independent. Using such techniques to analyze nested data would result in an underestimation of the standard errors of the contextual variables (and therefore an overestimation of the effects of these variables). Aggregating the dependent variables to the level of the independent contextual variables and performing a conventional regression analysis at that level is not a solution either as this makes it impossible among the independent variables to distinguish contextual effects from effects resulting from the aggregation of individual characteristics (i.e. composition effects) (Hooghe et al 2007; Snijders and Bosker 1999). We used Mlwin software to analyze a two level model consisting of classes (level 2) and students (level 1) with ethnoracial diversity, class status and classroom climate entered as class-level variables and gender, social status, civic competence and ethnoracial identity entered as individual-level variables. We first explored whether the effects of the individual-level variables interacted with the contextual variables. This possibility should certainly not be ruled out since there are studies demonstrating that the effect of ethnoracial identity on tolerance changes along with the ethnic composition of the school or neighbourhood. Researching racial prejudice in the American South, Glaser (1994) for instance shows that racial hostility among respondents identifying as white increases the more racially diverse the area they live in is. By contrast, and more relevant for the current paper, Billings and Holden (2007) found ideas of racial superiority among white 15-year olds in Burnley to be particularly strong in homogenously white schools. 22 However, entering interaction terms of diversity x identity and diversity x social status in the models, we found no evidence of interaction effects (results can be obtained from the author). . In other words, ethnoracial identity and social status were correlated in more or less the same way to the four social capital outcomes across different levels of diversity. We therefore decided to employ relatively straightforward random intercept models consisting only of main effects to examine the relations of interest. These models assume individual-level effects to be constant across classes. Results We report on the results in a stepwise fashion. First we present the results of the zero model, which provides an estimate of the variance in our outcome measures at the class (L2) and individual (L1) levels (see the columns marked with 0 in Table 3). We then report on models including only diversity as explanatory variable (Columns I). Subsequently we present the results of models including the three contextual variables (Columns II). Finally, we report on models containing all explanatory variables (Columns III). This procedure allows us to assess (1) whether the variance at the class level is large enough to warrant multilevel analysis, (2) whether ethnic diversity constitutes a true contextual or merely a compositional effect, and (3) how much the variance is reduced after the inclusion of more variables in the models. Table 3 about here 23 The zero models show that the class level captures 1.7% and 11.2% of the total variance in generalized trust and ethnic tolerance, respectively (as indicated by the Intraclass Correlation Coefficients (ICC), which represent the proportion of the total variance in the outcome measures accounted for by all observed and unobserved factors operating at a distinct level). According to Duncan and Raudenbusch’s (1999) rule of thumb, these percentages represent small/medium and very large effect sizes respectively. Consequently, while the application of multilevel analysis is of marginal utility for trust, it is a must for ethnic tolerance. ICCs could not be computed for the other participation outcomes since these are explored using logistic regression. Subsequent analyses will show, however, that the contextual variables have quite a substantial effect on these outcomes, thus justifying the use of multilevel analysis. Examining the effect of diversity without any of the control variables (Model I), we see that diversity is related in quite different ways to the four social capital indicators. While diversity is negatively linked with generalized trust, it shows a positive correlation with ethnic tolerance (though not significant) and the two participation measures. This pattern of links, moreover, holds when we control for class status and classroom climate, the other two contextual variables (Model II; note that the link with ethnic tolerance becomes significant). Thus, unlike Letki (2008) we do not find social status to be the most important factor at the group level overwhelming the impact of ethnic diversity. More generally, it can be seen that the effect of diversity is not consistent across the three indicators of bridging social capital, adding further weight to our observation in the previous section that there is little empirical support for distinguishing between bonding 24 and bridging forms of social capital. We further note that these results so far are consistent with other studies finding a negative relation between diversity and trust at the aggregate level (e.g Pennant 2005; Soroka et al 2004; Delhey and Newton 2005). Moreover, Putnam (2007) also found ethnic diversity to be negatively related to trust and positively related to alternative forms of political participation across neighborhoods in the US. Will these relations hold, however, once we start controlling for individual level variables? (Model III). We see a marked difference arising between the attitudinal and behavioral outcomes. While ethnic diversity no longer shows a significant relation with trust and tolerance, it retains its positive link with each of the two participation measures. The effect of ethnic diversity is thus spurious with regard to the two attitudinal outcomes, reflecting merely the sum of the individual-level effects. In other words, it represents a compositional and not a true contextual effect for these outcomes. Among the individual-level controls, ethno-racial identity emerges as a particularly strong condition affecting trust and tolerance in contrasting ways: non-whites are significantly less trusting and significantly more tolerant than whites. As the ethnic tolerance measure reflects attitudes on immigrants it is not surprising to find non-whites showing higher levels of tolerance. These higher levels are likely to reflect an awareness among minority students that they are themselves (descendents of) migrants and are seen as such by the dominant group. The non-white students may in other words have identified with immigrants. Having restrictive opinions on immigrants as a non-white would thus entail agreeing to be placed in a subordinate position with respect to the ethnic majority, which, understandably, few minority students would find appealing. In 25 this sense, our ethnic tolerance measure is likely to have only tapped tolerance levels among whites. The negative link between minority identity and trust is more puzzling, all the more so since it has been controlled for gender, social status and civic competence. Possibly, it can be explained by the specific wording of the item on trust. Students may have understood the words “the people who live in this country” as referring to the native majority. If so, the trust item has actually functioned as a measure of in-group trust for white students and of out-group trust for non-white students. However, if it has been interpreted this way, then this interpretation must have been unique to Britain because Janmaat (2008), using the same data set, has not found a similar negative relation between minority identity and generalized trust in several other West-European states with large immigrant populations. Why are the attitudinal outcomes so unrelated to ethnic diversity once we control for individual-level conditions? We can only speculate about the reasons. Possibly, youngsters have become so much used to the presence of ethnic minorities in the media or to direct contact with cultural others in their neighbourhood that the ethnic composition of the class or school matters less than it used to. Indeed, in Ford’s (2008) view, the daily exposure to black and Asian Britons on the television and in other environments is one of the principle factors explaining the steady decline in racial prejudice among the native majority in Britain. Alternatively, schools’ attempts to encourage independent thinking or the process of individualization more generally may have caused the opinions of individual students to be less guided by those of their 26 classmates than before, thus undermining the impact of contextual conditions such as ethnic diversity. The positive link of ethnic diversity with the two participation outcomes is no less difficult to interpret. After all, with participation in ethnic and religious organizations reflecting bonding social capital and participation in environmental and human rights organizations representing bridging social capital, these findings are neither in line with the contact nor with the conflict perspective. Both perspectives assume the two forms of social capital to be mutually exclusive, but the former expects ethnic diversity to be positively linked to bridging and negatively linked to bonding social capital while the latter expects the very opposite pattern of relationships. Moreover, the findings contrast totally with Putnam’s constrict perspective which anticipates a negative effect of diversity on both forms of social capital. What could explain the positive relation of diversity with participation in environmental and human rights organizations? Again we cannot do more than suggest possible causal mechanisms. Perhaps the experience of being in a diverse class and learning about other cultures triggers an individual pupil’s interest in and sense of involvement with broader issues, which then results in membership of the named organizations. This mechanism, obviously, would be in line with contact theory. Perhaps the initiative comes from the school and it is the relatively diverse schools which invest more time and effort in establishing links with the named organizations than the homogenous schools. Finally, the positive relation could be the result of selection. Students participating in these organizations may have parents who value diversity (or do not object to it) and send their children to ethnically mixed schools (or keep them in these 27 schools), accordingly. We could likewise propose mechanisms linking ethnic diversity to participation in ethnic and religious organizations However, because of the crosssectional nature of the Cived data, we cannot explore all these possibilities. Whatever the precise mechanisms linking diversity to the two participation outcomes, our main finding is that diversity relates differently to the attitudinal and behavioral components of social capital. This is another sign that not only the bonding and bridging varieties but also the concept of social capital itself need to be reevaluated. After all, the core assumption of social capital theory, that participation automatically generates a specific set of attitudes, trust and tolerance including, is not supported by our findings, neither those pertaining to the interrelations among the social capital indicators nor those relating to the drivers of social capital outcomes. Thus, while our findings may be welcome news for those convinced that ethnic mixing is conducive for participation and engagement, they simultaneously question the assumption that participation has wider benefits for social cohesion. Concluding remarks Does ethno-cultural diversity in the classroom help in creating bridging social capital? This paper revealed that the answer to this question very much depends on the component of bridging social capital under investigation. Controlling for various conditions at the individual and class level, ethno-racial diversity showed no relation to attitudes like generalized trust and ethnic tolerance while it is positively related to participation in 28 environmental and human rights organizations. Thus, at least for the behavioral component of bridging social capital this study supports the contact perspective: ethnic diversity makes people open up rather than hunker down as Putnam claims. Trust and tolerance, as the attitudinal component, were mainly shaped by individual-level conditions, with ethno-racial identity being the most prominent among them. In other words, the effect of ethno-racial diversity on these attitudes was found to be spurious, reflecting a mere compositional (i.e. sum of individual-level properties) and not a true contextual effect. This study has thus demonstrated that it is vital to include individuallevel variables in the analysis when assessing the impact of contextual properties on some social capital related outcome of interest. However, a more important finding in our view is the lack of coherence among the social capital indicators. Participation in environmental and human rights organizations was found to be uncorrelated with trust and tolerance, the other two indicators of bridging social capital, while it did show a strong positive link with participation in ethnic and religious organizations, our indicator of bonding social capital. Trust and tolerance moreover proved to be unrelated. These results raise serious doubts about the empirical utility of the concept of social capital and its bonding and bridging forms, and the study thus adds to a growing body of research critical of these concepts. The effort to develop inclusive concepts reducing the complexity of social reality is understandable from the point of view of parsimony but this should not result in concepts that are as multidimensional and complex as the reality they seek to simplify. If reality resists parsimonious descriptions it is better to acknowledge its complexity and act accordingly. In the case of social capital this would mean focusing on its individual 29 components: trust, tolerance and participation. The continued use of multidimensional terms like social capital is even harmful, we would argue, because it might give policy makers the false impression that measures can be devised that foster a range of social capital qualities all at once. Our study is limited in three important ways however. First, although our indicators of bonding and bridging have nicely tapped the scope of both kinds of social capital, they could only partially address the crucial issue of within (i.e. bonding) and between (i.e. bridging) group cohesion. It is clearly the in/out group dimension which is of most concern to policy makers. It is vital that future survey studies such as the International Civic and Citizenship Education Study (ICCS – the successor of the IEA Cived study) include items on in- and out-group trust, solidarity and cooperation. Second, due to the single point in time character of the data source, it was impossible to determine the direction of causation or to explore the causal mechanisms linking (or not linking) ethnic diversity to the four social capital outcomes. Finally, the Cived study’s restriction to a single age group invalidated an assessment of age, cohort and period effects, meaning that we could not explore the stability of behaviors and attitudes acquired during adolescence. Although, as noted previously, there are several studies suggesting that civic attitudes and behaviors obtained in one’s formative years are relatively enduring, there is no consensus on this issue. A longitudinal research design involving parallel panels or repeated cross-sections of multiple age groups would be able to address these limitations. 30 Literature Alesina, A., R. Baqir, and W. Easterly (1997). Public Goods and Ethnic Divisions. NBER Working Paper no. 6009. Cambridge, MA: NBER. Alesina, A. and La Ferrara, E. (2002). ‘Who Trusts Others’. Journal of Public Economics, 85 (2), 207-234. Billings, A. and A. Holden (2007). Interfaith Interventions and Cohesive Communities: The Effectiveness of Interfaith Activity in Towns Marked by Enclavisation and Parallel Lives (report prepared for the Home Office). Bobo, L.D. (1999). ‘Prejudice as Group Position: Microfoundations of a Sociological Approach to Racism and Race Relations’, Journal of Social Issues 55, 445-72. Bourdieu, P. (1986). ‘The Forms of Social Capital’, in J. Richardson (ed.), Handbook of Theory and Research for the Sociology of Education. Westport, CT: Greenwood Press, pp. 214-58. Brehm, J. and Rahn, W. (1997). ‘Individual-level evidence for the causes and consequences of social capital’, American Journal of Political Science, 41 (3), 9991023. Bruegel, I. (2006). ‘Social Capital, Diversity and Education Policy’. Paper for Families and Social Capital ESRC Research Group. http://www.lsbu.ac.uk/families/publications/SCDiversityEdu28.8.06.pdf. Accessed on 3 June 2008. 31 Cantle, T. (2001). Community Cohesion: A Report of the Independent Review Team Chaired by Ted Cantle. http://image.guardian.co.uk/sys-files/Guardian/documents/2001/12/11/communitycohesionreport.pdf. Accessed on 2 June 2008. Coleman, J. (1988). ‘Social Capital in the Creation of Human Capital’. The American Journal of Sociology, Vol 94 supplement, S95-S120. Dalton, Russell J. 2004. Democratic Challenges, Democratic Choices: The Erosion of Political Support in Advanced Industrial Democracies. Oxford: Oxford University Press. Delhey, J. and K. Newton (2005). Predicting Cross-National Levels of Social Trust. Global Pattern or Nordic Exceptionalism? European Sociological Review, 21, 4, 311-327. Delli Carpini, M.X. and Keeter, S. (1996). What Americans Know about Politics and Why it Matters. New Haven: Yale UP. Duncan, G.J. and S.W. Raudenbusch (1999). ‘Assessing the Effects of Context in Studies of Child and Youth Development’. Educational Psychologist, 34, 29-41. Ellison, C.G. and D.A. Powers (1994). ‘The Contact Hypothesis and Racial Attitudes among Black Americans’. Social Science Quarterly, 75, 2, 385-400. Flanagan, C.A. and L.R. Sherrod (1998). ‘Youth Political Development: an Introduction’. Journal of Social Issues, 54 (3), 447-56. Ford, R (2008). ‘Is Racial Prejudice Declining in Britain?’ British Journal of Sociology, Vol 59, No. 4, 609-636. 32 Frankenberg, E., Lee, C. and Orfield, G. (2003). A Multiracial Society with Segregated Schools: Are We Losing the Dream. Cambridge, MA: The Civil Rights Project, Harvard University. Galston, W. (2001). ‘Political Knowledge, Political Engagement and Civic Education’. Annual Review of Political Science, Vol. 4, 217-234. Glaser, J. (1994). ‘Back to the Black Belt: Racial Environment and White Racial Attitudes in the South’. Journal of Politics, 56, 21-41. Granovetter, M. (1978)’ The Strength of Weak Ties’, American Journal of Sociology, 78 (6), 1360-80. Green, Andy, John Preston and Jan G. Janmaat (2006). Education, Equality and Social Cohesion: A Comparative Analysis. Basingstoke: Palgrave Macmillan. Hagendoorn, L. (1999) ‘Introduction: a model of the effects of education on prejudice and racism’ in Hagendoorn, L. and Nekuee, S. (eds) Education and Racism: A Cross National Inventory of Positive Effects of Education on Ethnic Tolerance. Aldershot: Ashgate, 1-20. Hall, P. (1999). ‘Social Capital in Britain’, British Journal of Politics, 29, 417-61. Herbert, I. (2001). ‘”Apartheid” fears over first Muslim secondary school in state sector”, The Independent, 2 April: 11. Hero, R. and C. Tolbert (1996). A Racial/Ethnic Diversity Interpretation of Politics and Policy in the States of the US. American Journal of Political Science, 40, 851-871. Holme, J., Wells, A., and Revilla, A. (2005). ‘Learning Through Experience: What Graduates Gained by Attending Desegregated High Schools’. Equity and Excellence in Education 38; No 1, 14-25. 33 Hooghe, M. (2003). ‘Value Congruence and Convergence within Voluntary Associations: Ethnocentrism in Belgian Organizations’. Political Behavior, Vol. 25, No. 2, 151175. Hooghe, Marc, and Stolle, Dietlind (2004). ‘Good Girls go to the Polling Booth, Bad Boys go Everywhere. Gender Differences in Anticipated Political Participation Among US 14 Year Olds. Women and Politics, Vol 26, Nos 3-4, 1-23. Hooghe, M., Reeskens, T., Kavadias, D., Claes, E. (2007). ‘Determinanten van Burgerschapsvorming bij Jongeren: Een Internationaal Vergelijkend Perspectief naar Kiesintenties bij 14-Jarigen’, Pedagogische Studiën. Tijdschrift voor Onderwijskunde en Opvoedkunde, Vol 84, No 4, 259-276. Janmaat, J.G. (2006). “Civic Culture in Western and Eastern Europe.” European Journal of Sociology / Archives Européennes de Sociologie 47: 363-393. Janmaat, J.G. (2008) ‘The Civic Attitudes of Ethnic Minority Youth and the Impact of Citizenship Education’, Journal of Ethnic and Migration Studies. Vol 34, No 1, pp 27-54. Letki, N. (2008). ‘Does Diversity Erode Social Cohesion? Social Capital and Race in British Neighbourhoods’. Political Studies, 21 (1), 109-123. Luttmer, E. (2001). Group Loyalty and the Taste for Redistribution. Journal of Political Economy, 109, 3, 500-528. McGlynn, C., Niens, U., Cairns, E. and Hewstone, M. (2004) 'Moving out of conflict: the contribution of integrated schools in Northern Ireland to identity, attitudes, forgiveness and reconciliation', Journal of Peace Education, 1:2, 147 — 163 34 Newton, K. and Norris, P. (2000). ‘Confidence in Public Institutions: Faith, Culture or Performance?’, in Susan J. Pharr and Robert D. Putnam (eds) Disaffected Democracies: What’s Troubling the Trilateral Countries? Princeton, NJ: Princeton UP, 52-73. Nie, Norman H., Junn, Jane, and Stehlik-Barry, Keith (1996). Education and Democratic Citizenship in America. Chicago: Chicago University Press. Oliver, J.E. and Wong J.(2003). ‘Intergroup Prejudice in Multiethnic Settings’. American Journal of Political Science, 47 (2003), 567-82. Oliver, J.E. and Mandelberg, T. (2000). ‘Reconsidering the Environmental Determinants of White Racial Attitudes’, American Journal of Political Science, 44, 574-89. Orfield, G. (1978). Must We Bus? Segregated Schools and National Policy. Washington: Brookings Institution. Pennant, R. (2005). Diversity, Trust and Community Participation in England. Home Office Findings 253, Research, Development http://www.homeoffice.gov.uk/rds/pdfs05/r253.pdf. and Statistics Directorate. Accessed on 1 June 2008. Pettigrew, T.F. (1997). ‘Generalized Intergroup Contact Effects on Prejudice’. Personality and Social Psychology Bulletin, 23 (February), 173-85. Putnam, R. (1993) Making Democracy Work: Civic Traditions in Modern Italy. Princeton, NJ: Princeton University Press. Putnam, R. (1995). ‘Bowling Alone: America’s Declining Social Capital’. Journal of Democracy 6 (January): 65-78. Putnam, R. (2000) Bowling Alone: The Collapse and Revival of American Community. New York: Simon and Schuster. 35 Putnam, R.D. (2007). E Pluribus Unum: Diversity and Community in the Twenty-first Century, The 2006 Johan Skytte Prize Lecture, Scandinavian Political Studies, 30, 2, 137-174. Quinlan, L. (1995). ‘Prejudice as a Response to Perceived Group Threat: Population Composition and Anti-Immigrant and Racial Prejudice in Europe’, American Sociological Review 60, 586-611. Rice, T. W. and Feldman, J. L. (1997) ‘Civic culture and democracy from Europe to America’, The Journal of Politics, 59 (4): 1143-72. Rumberger, R.W. and Willms, J.D. (1992). The Impact of Racial and Ethnic Segregation on the Achievement Gap in California High Schools. Educational Evaluation and Policy Analysis, 14 (4), 377-396. Salter, F. K. (2004). Introduction: The Symposium Target Paper in Broader Context. In F. K. Salter (Ed.) Welfare, Ethnicity and Altruism: New Findings and Evolutionary Theory. London: Frank Cass, 3-26. Schofield, J. W. (2001). ‘Review of Research on School Desegregation’s Impact on Elementary and Secondary School Students’, in Banks, J.A. and McGee Banks, C.A. (eds) Handbook of Research on Multicultural Education. San Francisco: Jossey-Bass, 597-616. Sigelman, L. and T. Bledsoe, S. Welch, and M.W. Combs (1996). ‘Making Contact? Black-White Social Interaction in an Urban Setting’. American Journal of Sociology, 101, 1306-32. Snijders, T. and Bosker, R.J. (1999). Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modelling. London: Sage Publications. 36 Soroka, S. N., R. Johnston, and K. Banting (2004). Ethnicity, Trust and the Welfare State. In: P. van Parijs (Ed.) Cultural Diversity versus Economic Solidarity. Brussels: De Boeck, 33-58. Stolle, D. and Hooghe, M. (2004). ‘The Roots of Social Capital: Attitudinal and Network Mechanisms in the Relation Between Youth and Adult Indicators of Social Capital’. Acta Politica, 39, 422-441. Tolsma, J., van der Meer, T. and Gesthuizen, M. (2008). ‘Unity in Diversity: The Impact of Neighbourhood and Municipality Characteristics on Dimensions of Social Cohesion in the Netherlands’. Paper presented at the Politicologenetmaal in Berg en Dal, The Netherlands, 29-30 May 2008. Torney-Purta, Lehmann, J. R., Oswald, H. and Schulz, W. (2001) Citizenship Education in Twenty-eight Countries: Civic Knowledge and Engagement at Age Fourteen. Amsterdam: The International Association for the Evaluation of Educational Achievement. Torney-Purta, J. (2002). ‘The School’s Role in Developing Civic Engagement: A Study of Adolescents in Twenty-Eight Countries’. Applied Developmental Science, Vol. 6, 203-212. Torney-Purta, J. (2004). ‘Adolescents’ Political Socialization in Changing Contexts: An International Study in the Spirit of Nevitt Sanford’. Political Psychology, Vol. 25, No. 3, 465-478. Uslaner, E.M. (2000). Producing and Consuming Trust. Political Science Quarterly, 115, 569-590. 37 Verba, S., Schlozman, K. L., and Brady, H. (1995). Voice and Equality: Civic Voluntarism in American Politics. Cambridge, MA: Harvard University Press. 38 Table 1. Descriptive statistics of dependent and independent variables. Values of categorical variables Dependent Generalized rust Ethnic Tolerance Part in eth&reli org Part in env&hr org Independent Ethnoracial diversity Class status Classroom climate Social status Civic competence Ethnoracial identity Gender Minimum Maximum 1 4.04 No Yes No Yes Standard deviation N 2.68 9.73 .80 2.24 2547 2737 2985 85.1% 14.9% 85.3% 14.7% 2985 0 3.12 7.85 1 40.62 white non-white girl boy 4 14.17 Mean 87.4% 12.6% 50% 50% 94.4 5.87 13.56 6 165.04 12.68 4.40 9.98 4.40 99.40 18.1 .55 .84 1.32 18.86 3043 3043 3014 2982 3013 2972 2957 39 Table 2. Relations between social capital components (bivariate correlations) Bonding Part in eth and rel org Bonding social capital Participation in ethnic and religious organizations Bridging social capital Participation in environmental and human rights organizations Generalized trust Ethnic tolerance - .34** -.05* .02 - .01 - -.01 .00 Bridging Part in env & hr org Generalized trust 40 41 Table 3. The determinants of social capital (coefficients of multilevel linear and logistic regressions) Class level (L2) Ethnoracial diversity Class status Climate Bonding social capital Participation in ethnic and religious organizations I II III .020*** .021*** .605*** .142 Individual level (L1) Ethnoracial identitya Social status Genderb Civic competence ICC L2 (%) ICC L1 (%) (residual) variance L2 (residual) variance L1 Explained var L2 (%) Explained var L1 (%) N L2 N L1 .008* .453** .136 Bridging social capital Generalized trust Participation in env and human rights organizations I II III .013*** .013*** .016 .025 1.23*** .195*** -.167 -.001 0 .010* -.096 .019 128 2933 128 2933 II III -.004** -.096** .001 .001 -.064 .007 .224 .165** -.181 -.003 128 2933 128 2933 0 I II III .007 .009* -.160 -.209 -.008 -.312 -.287** -.255*** -.013 .012 -.002* 1.7 98.3 .011 .631 128 2933 I -.004** Ethnic tolerance 128 2933 128 2498 .007 .631 36.4 0 128 2498 .004 .631 63.7 0 128 2498 .004 .625 63.7 1 128 2498 1.53*** .081* -.60*** .011*** 11.2 88.8 .574 4.451 128 2688 Note: * P < .05; ** P < .01; *** P < .001; the data were weighted by houseweight to make the results nationally representative a Reference category: white b Reference category: girl 42 .560 4.449 2.4 0 128 2688 .522 4.448 9.1 0 128 2688 .558 4.140 2.4 7 128 2688