Network Diversity and Political Participation: A Complication or an Asset? Allison Harell allison.harell@mcgill.ca McGill University Montreal, Canada Dietlind Stolle dietlind.stolle@mcgill.ca McGill University Montreal, Canada Ellen Quintelier ellen.quintelier@soc.kuleuven.be Catholic University of Leuven Leuven, Belgium DRAFT: Please do not cite without author s permission. Paper prepared for presentation at the European Consortium for Political Research Annual Meet in Pisa, Italy from August 5-9, 2007. The authors would like to thank the Social Science and Humanities Research Council, the Fonds Québecois de la Recherche, the Fonds Wetenschappelijk Onderzoek Vlaanderen and the Bijzonder Onderzoeksfonds K.U.Leuven for their generous support in collecting the data for this project. 1
Abstract Social capital researchers have suggested that bridging ties, i.e. networks of people with different opinions and backgrounds, are important for democratic politics. Yet, while some theoretical explanations and research shows a positive effect of network diversity on political mobilization (Rubenson 2005; Lake and Huckfeldt 1998; Huckfeldt et al. 2004b); other research tends to suggest a dampening, rather than a mobilizing, effect (Alesina and Ferrara 2000; Mutz 2002a, 2006; Costa and Kahn 2003; Hill and Leighley 1999). Given these limited and contrary findings, we propose to take a detailed look at how different types of network diversity influence political participation among youth, because young people are most susceptible to the effects of the composition of their surroundings. The study draws on the new Comparative Youth Study (CYS) with almost 10,000 young people between the ages of 15 and 17 in Canada and Belgium. We examine two questions with this data. First, how does network diversity generally related to political participation? Second, we attempt to disaggregate network effects: does network diversity function similarly across different dimensions of diversity (racial, religious, linguistic, socio-economic and political) and different types (strong and weak) of networks? Our youth sample is ideally designed for this test. We are able to isolate diversity across a number of key dimensions for individuals networks of close friends as well as for acquaintances in the wider school context. What sorts of diversity in which contexts matter most is an important question that previous research has failed to address. The findings of this study are then considered in light of increasing ethno-cultural diversity in contemporary democracies. 2
Introduction In political science, social capital researchers have suggested that bridging ties, i.e. networks of people with different opinions and backgrounds, are important for democratic politics (Putnam 2000; Marschall and Stolle 2004; Cigler and Joslyn 2002). Weak bridging ties in informal networks of friends and acquaintances as well as in voluntary organizations have been highlighted as particularly beneficial for the dissemination of information and political mobilization (Granovetter 1973; Mutz and Mondak 2006). Yet, the limited research into the impact of social diversity on political participation tends to suggest a dampening, rather than a mobilizing, effect (Alesina and Ferrara 2000; Costa and Kahn 2003; Mutz 2002a, 2006). Several problems occur in the research on the effects of diversity on political outcomes. Often diversity is examined at the aggregate level using census tract data which does not fully capture the actual social networks of individuals. Alternatively, other approaches focus exclusively on the concept of strong ties, which is often measured with the name generator technique. Finally, current research is limited in the analysis of single dimensions of diversity, either focusing on racial/religious or political diversity. Our paper attempts to overcome these weaknesses by taking a comparative look at how different dimensions of diversity and types of ties influence the political participation of young people in Belgium and Canada. In particular, we focus on political, ethno-cultural, linguistic and socio-economic diversity among networks of strong and weak ties. The focus on young citizens is particularly useful, as one can expect more powerful socialization effects from experiences in one s social networks at a young age. 3
Two main questions drive the research in this paper. First, does network diversity indeed decrease political participation? Current research suggests this is the case in the US but we will extend this finding to the comparative context and focus on young people using an extensive participation battery that captures the different ways in which young people may engage politically. The second question attempts to disaggregate network effects: does network diversity function similarly across different dimensions of diversity and different types of networks? This paper will show that despite the literatures focus on the potential demobilizing effect of diversity on political participation, young people who have more diverse social networks are in fact more likely to participate in a variety of social and political activities. The impact of political diversity in one s network is strongest, but exposure to ethno-cultural and socio-economic diversity also demonstrates positive effects. Furthermore, the mechanism by which political action is fostered seems to be increased interest, discussion and knowledge about politics that are fostered in more diverse settings. Diversity A Complication or an Asset for Political Participation? Most accounts of social capital theory emphasize the importance of social interactions in formal and informal networks in providing both personal benefits as well as the development civic values and societal engagement. The overarching insight of social capital theory is that networks have value (Putnam 2000). So, for example, sociologists stress the importance of parental social networks for the performance of school children (Coleman 1988); they write about the importance of diverse personal networks for success in the job market and job promotions (Granovetter 1973; Burt 4
1997); and they examine how informal social resources are utilized to achieve occupational mobility (Lin et al. 2001). Participation in informal social networks and voluntary associations has also been linked to political mobilization and participation (Galston 2004; Seligsen 1999; Teorell 2003). Associations, in particular, are believed to foster civic skills (Verba et al. 1995) and to encourage civic spirit and volunteerism (Putnam 2000, 121ff), as well as a sense of political efficacy (Berry et al. 1993, Ch. 8), and generalized trust and other civic attitudes (Putnam 2000, 1993; but see Stolle 2001; Uslaner 2002). Participation in associations also expands people s informal social networks. As a result, they are more likely to be exposed to information about politics and to be the target of appeals for political action (Teorell 2003). Similarly, the more people discuss politics with his or her acquaintances and friends in informal networks, the more likely they are to participate in politics (Verba et al. 1995; Knoke 1990; McLeod et al. 1999). The key point is that information flows through social networks (Putnam 2000, 338). But which types of networks matter most? Two important dimensions of the structure of social interactions have emerged in the literature on social capital. One relates to the strength or depth of interaction, and is captured by the distinction made by network analysts between strong ties (close friends and family) and weak ties (acquaintances) (Granovetter 1983, 1973). The other one focuses on the sorts of people with whom one interacts, which is captured by the distinction between bridging and bonding interactions (Putnam 2000). Granovetter (1983; 1973), in particular, has emphasized the strength of weak ties. He argues that casual acquaintances (weak ties) can provide people with 5
information and resources beyond those that are available from within their immediate circle of close friends and relatives (strong ties). This is because casual acquaintances often serve as bridges to social circles beyond our own, bringing us into contact with ideas and information that we might otherwise not encounter. Huckfeldt and his colleagues (1995, 1028) have drawn on this argument to explain how information about politics spreads: Information that is communicated through weak social ties typically travels farther because it is less likely to feed back to the point of origin the casual acquaintances of my casual acquaintances are not so likely to be my associates, and thus information conveyed through such patterns of interaction is likely to travel farther. This is because the more tightly knit a network is, the less likely that any one person in the network will provide new and different information. Without such information, people are argued to be less likely to participate (Teorell 2003; Delli Carpini and Keeter 1996; Eveland and Scheufele 2000). On the other hand, casual acquaintances are argued to provide information by linking a person to wider networks that they do not have direct access to. Putnam and others have emphasized that it is not only the intensity with which people interact but also the composition of social networks that has effects on political outcomes. He claims that bridging interactions with people from different social backgrounds are more conducive to the acquisition of political information than bonding interactions with people from similar backgrounds (2000). There is some evidence to support the argument about the importance of bridging ties for political outcomes. For example, Rubenson (2005) demonstrated that people who had more diverse networks were more likely to vote, sign petitions, attend political meetings, take 6
part in demonstrations or boycotts and to be active in a political group. Similarly, Oliver (1999) argues that the level of civic participation is highest in diverse, middle-income cities. How do bridging ties exert their effects on political involvement? Political discussions with people who hold different viewpoints forces people to constantly rethink and refine their issue stances as a result of potentially being challenged in their opinions by non-likeminded others, which makes them more likely to participate (Scheufele et al. 2004, 316; see also Huckfeldt et al. 2004a; Kenny 1992; Klofstad 2005; McLeod et al. 1999; McClurg 2003; Leighley 1990). Given that many political values are based in social background variables, examining the diversity of people from different social backgrounds might also capture important differences in political views. If so, than different socio-demographic characteristics in one s social network might have similar effects on participation when compared with the effects having diverse political networks. For example, Langston (1969) showed that when working-class students had higher-class friends, they differed with working-class students who had only friends of their own class. The first group was more positively inclined towards voting, more supportive of political rights for minorities and less supportive of the regime than the second group (ibid). However, not everyone agrees on the importance of bridging ties for political outcomes. There is something to be said about the value of bonding or homogeneous networks as well. The sharing of similar views and opinions in homogenous networks between like-minded people can augment the viewpoints, promote recognition of common problems, and spur one on to collective action (Hooghe 2003; Mutz 2002a; 7
Putnam 2000; Campbell 1980) 1. Particularly Mutz (2002a; 2006) argues that the discussion in strong networks with people who hold different opinions might create cross-pressures and ambivalence, which makes politics more complicated (Mutz 2002a; Scheufele et al. 2004). Since many people prefer social harmony and would therefore avoid arguments with close friends, they consequently prefer to stay out of politics and political discussions altogether when their networks contain a diversity of opinions (Mutz 2002a, 2006; Eliasoph 1998; Lane 1959). Although cross-cutting political talk makes people more tolerant (Mutz 2002b; Oliver and Wong 2003), Mutz finds that it depresses political participation (Mutz 2002a; see also McClurg 2006; Ulbig and Funk 1999; Berelson et al. 1963). What are we to make of these contradictory assumptions and findings? So far, the literature seems to suggest that whereas diversity in strong ties might lead to political inhibition; bridging weak ties might be more conducive to political mobilization. Three clarifying aspects are missing though in the current research agenda. First, as the expectations of their benefits vary, we need a more systematic account of the differential effects that diversity can have among strong and weak ties. Secondly, distinctions need to be made between the different dimensions of diversity. Whereas some studies exclusively focus on political diversity; others look only at socio-demographic variety among network members. However, it is not clear that we can use these various dimensions of diversity interchangeably. Thirdly, the political effects of social ties have been studied mostly at the aggregate level with census tract information or else in very close networks (with the name generator method). Better measurements of network diversity that are 1 Eventually, this process might lead to an overall polarization of society and the entrenching of people's initial viewpoints (Druckman and Nelson 2003). 8
able to capture both weak and strong ties have been missing. Our paper attempts to improve the research on the effects of diversity on political participation on these three fronts by using data collected from young people in Belgium and Canada. Accordingly, we expect that weak bridging ties will generally be the most conducive to young people s political empowerment. We have more doubts about the potential of strong bridging ties. In addition, we assume that political diversity translates more efficiently and directly into difference of opinion and viewpoints than socio-demographic diversity, and is therefore a more powerful predictor of political mobilization. Young People and the Power of Peers While the literature on the effect of networks on political attitudes and behaviors overwhelmingly focuses on adults, we expect that an analysis of youth may provide stronger, or even perhaps different, effects. A large literature on socialization stresses the importance of early life experiences for shaping citizens attitudes about and involvement in political life. For example, those who vote soon after becoming old enough are more likely to turn out later in life (Franklin 2004; Plutzer 2002). According to Green and Shachar (2000), voting is a habit. Once it is ingrained, it becomes easier to continue voting. Voting is not the only form of political participation that seems to benefit from early experiences: those involved in youth organizations and other forms of extracurricular activities are more likely to stay involved as adults (Verba et al. 1995; Glanville 1999; Hanks and Eckland 1978). Therefore, experiences in early adulthood are clearly important predictors of future participation. 9
While a habit of participation can be instilled early in life, we also expect that one s youth experiences in networks may be consequential for one s life. Why might this be? First of all, peers are key people in the political socialization of adolescents, and they help to shape young people s attitudes about politics (Brown 2004; Torney-Purta 1995; Dawson and Prewitt 1969; Parry et al. 1992). Unlike parents or the school, peers are unique in that they interact on relatively equal footing to each other and share similar social status (Jaros 1973; Dawson and Prewitt 1969; Eisenberg and Morris 2004; McLeod et al. 1999; Brown 2004). Peer groups do develop their own subcultures, with their own values and social norms (Jaros 1973; Langston 1969). It is reasonable to assume that such subcultures can either foster or discourage involvement in politics (Roker et al. 1999; Crosnoe 2000), and this may be particularly true among adolescents, for whom peer relations take on increasing importance (Jaros 1973). We are here particularly interested in the effects of the diversity of such peer groups. Thus, a focus on youth is inspired by a history of socialization studies that acknowledges (although rarely examines in detail) the unique importance of peers for adolescent political socialization. However, a further reason for focusing on youth is that network effects suffer from a serious self-selection bias. In other words, it is very difficult to ascertain the causal importance of networks on a person s political participation because the very same factors that promote participation may also make more homogenous networks more likely. Young people, while they also may be prone to such self-selection effects, are also particularly prone to influences by peers. In addition, like adults in the workplace later in life (Mutz and Mondak 2006), adolescents spend a large portion of their time in the school environment where they have less choice about 10
the types of people they must interact with on a day-to-day basis. 2 Research has found that in the U.S., students at schools with more racial diversity are more likely to have cross-race friendships (Joyner 2000; Mouw and Entwisle 2006). This suggests that while self-selection is likely part of the explanation for the diversity of an individual s networks, opportunity is also an important aspect here. Therefore, while there is clearly evidence that young people show the same tendency as adults toward homophily, we believe that network effects on political mobilization and participation may play a more important role during adolescence than at other periods in a person s life course, given the importance of peers in the socialization process and the opportunities that young people have to encounter peers from different backgrounds. Data and Methods The data for this study comes from the Comparative Youth Study (CYS), which was a self-reported questionnaire completed among high school students in Canada and Belgium during the 2005-2006 academic year. The Canadian sample (McGill Youth Study, or MYS) included 3334 10 th and 11 th grade students sampled in seven cities in Quebec and Ontario. The cities included the two largest cities in each province, two medium sized cites of approximately 150,000 residents, and three small towns of approximately 15,000 residents. Schools were intentionally selected to vary in terms of socio-economic and racial/religious diversity. The sample is similar to the overall 2 Of course, students in both Canada and Belgium are segregated somewhat in the school system as students are tracked toward university or trade schools. This clearly limits some of the potential exposure to diversity, yet we believe there is still more opportunity for diverse interaction than might be true in other arenas (such as a youth organization or at a place of worship). 11
composition of schools in each city and reflect on average the socio-economic situation of the larger population from which the students were sampled. The Belgian sample (Belgian Youth Study, or BYS) included a representative sample of 6,300 10 th grade students in Belgium. The main dependent variable for this analysis results from a series of fifteen questions asking the respondent how often they have participated in a given activity in the previous 12 months. The activities were selected to vary in terms of the ways in which young people in contemporary democracies might be involved in politics, from volunteering for a cause, being involved in a political party, or boycotting a product. The items are included in an additive scale (Political Activity Scale) that varies from 0 to 13 and represents the number of activities the respondent has participated in during the previous 12 months. The main independent variables are questions about the composition of the respondents networks. Five different types of diversity (official language, religious, racial, socio-economic and political) were queried in two different contexts that measure strong and weak ties (close friends and other people the respondent talks to at school other than their close friends). Each respondent indicated how many of their close friends (strong ties) and how many of the other people at school they speak with other than their close friends (weak ties): 1) Are from a different race 2) Are from a different religion (or was religious if the respondent was not) 3) Are from a family with more or less money than the respondent s 4) Often disagree about politics or public affairs 5) Speak the other official language than the respondent 12
Answers varied on a seven point scale from 1=none to 7=all. Factor analysis of the ten items shows four distinct scales (not shown). The race and religion questions loaded together for close friends and acquaintances. These four items make up our racial/religious diversity scale. The remaining questions about close friends and acquaintances loaded together on each dimension. Four additive scales were created that vary from 1 to 7 and can be interpreted as the average response on the original scale that varies from none to all. 3 Higher scores always indicate more diversity. In addition, separate scores for weak and strong ties were created, by combining the five dimensions for each type of tie together to make an aggregate diversity measure among close friends and an aggregate measure among acquaintances. Earlier we discussed how the literature has focused on a variety of causal mechanisms through which exposure to diversity may influence political participation. Being exposed to a diversity of opinions may make political discussion less likely as people attempt to avoid controversy and conflict. At the moment, we are unable to test the argument about conflict avoidance. Alternatively, network diversity may foster political discussion and knowledge as well as interest in political issues by making political discussion more likely and providing new information through exposure to dissimilar others; this in turn might spur political action. We have therefore included three measures of political engagement: political knowledge, political interest and the frequency of political discussion in order to be able to test these as potential intermediary variables in the relations between diversity and political participation. In addition, 3 The four factors were racial/religious (alpha =.832), socio-economic (alpha=.703), official language (alpha=.695) and political (alpha=.661) diversity; for all the close friends scale and school acquaintance scale load together. Identical results emerge when this is analyzed separately by country of survey. 13
several control variables are included in our multivariate analyses, such as gender, urban/rural status, visible minority status, parental education, citizenship and the number of close friends the respondent has; they all were shown to correlate with political participation as well. Data and Findings To begin, it is important to determine what types of diversity are present among young people s networks. Our assumption is that young people are more frequently exposed to diversity in their social networks, because they do not always have a choice about where they go to school and they might generally have fewer prejudices coloring their interaction with other people. As discussed above, we have created two different types of diversity measures: one that measures various dimensions of diversity and another that distinguishes diversity in strong and weak ties (see details in previous section). In Table 1, we present these measures. As expected, ties among close friends are less diverse than those created with acquaintances at school; in fact, for all separate dimensions of diversity we find significant differences at the <.01 level, with acquaintances being more diverse on average (Detailed results not shown). Generally, though, the diversity of youth networks seems lower than expected, hovering around 2 indicating an average of almost none of one s peers being different, with the exception of Canadian acquaintances at school. 4 Regarding the individual diversity dimensions, racial and religious diversity is highest in Canada, as one might expect given the over-sampling of urban youth and the higher levels of racial and ethnic diversity in the country more generally; economic 4 Complete details of the exact question wording are available in the appendix. The seven categories from 1 to 7 are: none, almost none, a few, about half, many, most, all. 14
diversity also is elevated in both countries in comparison to other dimensions. Linguistic diversity is clearly the lowest in both settings. It should be noted, though, that there are significant correlations between diversity scales, so that those who are exposed to some types of diversity are also likely exposed others (correlations between.139 and.293). [table 1 about here] While diversity levels tend to be low, there is also reason to believe that these levels vary among young people. Do certain young people do have more diverse networks than others? A long-standing finding in sociological research is that one s social position structures the composition of one s networks (Blau 1977). In Table 2, we examine how visible minority status, urban/rural, gender and parental education structure network diversity on the four dimensions. The relationships, in general, are in the expected direction across the four dimensions. Visible minorities, urban dwellers, and those from more educated families have significantly higher reported diversity scores on every dimension. This is consistent with past research that shows minorities, those living in the city, and those from higher socio-economic backgrounds tend to have more access to heterogeneous others (Ibarra 1995; Marsden 1987; McPherson et al. 2001; Hallinan and Williams 1989). These results also hold when we examine the breakdown by strong and weak ties (results not shown). The one exception to our expectations is gender. While research tends to suggest that young men and women have similar patterns of network composition, and that in adulthood men tend to gain an advantage in terms of network diversity (McPherson et al. 2001, 422-4; Maccoby 1998), our samples report higher levels of network diversity among young women on two dimensions: ethnocultural diversity and official language diversity. 15
Why might we find these divergent results? One reason may be that most research on gender differences tend to focus on the gender diversity of networks, instead of other dimensions of diversity. However, one study conducted in the U.S. found that boys tend to have a greater tendency toward racial heterogeneity in their networks than girls (Maccoby 1998). An alternative explanation may be that young women tend to have larger networks that provide more opportunity for diverse ties. In our sample, this is just not the case when we look at the size of close friend networks. When asked the number of close friends they had, young men reported small, but significantly higher levels. 5 Further research may suggest that inter-country or intergenerational explanations may be more satisfying in explaining these differences. The differences are noteworthy, nonetheless, and amongst others suggest the importance of controlling for these factors when examining the impact of network diversity on political participation. Recall that the main dependent variable of interest is an additive scale of the ways in which young people have been involved politically. The full scale varies from 0 to 13 activities. 6 Unsurprisingly, the mean number of activities is skewed toward the low end of this scale with young people reporting on average engaging in 2.6 of the activities listed on the scale. 7 For the analysis, the full scale has been truncated at 10 to partially address this skewness and to equalize the scale categories. Therefore, all those who reported 10 or more activities have been given a score of 10. While the scale is still skewed toward 0, this provides a more normal distribution for analysis. 8 5 On a 1 to 7 scale, boys reported on average 4.27 vs. young women who reported 3.82 (p=.00). 6 The scale is constructed with 15 dichotomous items, although 13 is the maximum number reported. 7 The mean reported by the Canadian youth was 3.7 types of activities and the Belgian youth reported on average 2.0 activities. See Appendix for complete details. 8 The mean of the truncated scale is 3.2. Significant differences remain between the Canadian sample (mean=3.95) and the Belgian sample (mean=2.63) at the <.01 level. 16
[Figure 1 about here] Figure 1 provides a first look at the relationship between network diversity and political participation. The top graph divides the network diversity measures by dimensions, and the bottom graph divides network diversity by type (strong or weak). Clearly, whether we examine dimensions or types of diversity, those who report higher levels of diversity are also on average involved in more political activities. 9 If we consider different dimensions of diversity, we can see that the strongest relationship emerges for political diversity. Whereas those with no political diversity in their networks engage on average in 1.9 activities; in contrast, young people who have half or more of their friends with different political views tend to participate in over 3.3 activities. This provides preliminary support for our contention that network diversity can serve as a mobilizing force among young people. As we expected, this seems to be the case in particular with political diversity. If we consider the breakdown by types of networks (strong vs. weak ties), we see that a similar strong relationship exists. Having more diversity among one s close friends or among one s acquaintances appear to be positively and linearly related to increased political involvement. 10 The level of involvement is similar across types, and there is no evidence that diversity among close friends has a different impact than diversity found among weaker ties; contrary to our previous expectations. This may partly reflect the fact that those who report having diversity among their close friends are more likely to have diverse acquaintances as well, which could be related to certain background characteristics which make diverse ties more likely. 9 A comparison of means shows that the increases are significant at the p<.01 level in all cases. 10 The slight dip in the final network diversity category among close friends is a function of sample size. 17
Therefore, in order to examine the effects of these network variables more fully, Tables 3 and 4 present a multivariate analysis in which we control for other factors that relate to our dependent and independent variables. The analysis is conducted in two steps. In the first step, we include network composition variables with a host of controls. In the second step, we add in three variables that we expect to mediate the relationship between network diversity and political participation, namely, political knowledge, interest and discussion. The results of the bivariate analysis lead us to expect positive and significant coefficients for the network diversity variables in the first step, especially for political and economic diversity. In the second step, we expect that political knowledge, interest and discussion should decrease the impact of network diversity on participation. Table 3 provides the results of these two-step regressions in both the Canadian and Belgium samples for the four dimensions of network diversity. Consistent with the bivariate results, we see that increased political and economic diversity among one s networks are positively and significantly related to higher levels of political participation. This holds across both samples and the results are highly significant (p=<.01). Official language diversity seems to have little impact on participation once other variables are controlled for. The one inter-country difference is the impact of racial and religious diversity. In the Belgian sample, the coefficient is positive and significant. There is no evidence of a similar effect among the Canadian youth. In the second step of the analysis, all three intervening variables are positive and significant at the <.01 level. What is interesting for our purposes is that the inclusion of these variables results in a drop in the impact of the political diversity variable by about half in the Belgian sample, and by over two-thirds in the Canadian sample. This lends 18
support to our contention that one of the ways in which network diversity fosters political engagement is through the political discussion and information that is transmitted through such networks. 11 There is no similar drop in the impact of socio-economic diversity in either sample, or for racial and religious diversity in the Belgian sample. These coefficients remain positive and significant at about the same level, even after controlling for the potential mediating impact of political knowledge, interest and discussion. This suggests that the causal mechanism between socio-economic and racial/religious diversity in one s networks and political participation is different from political diversity. This finding should be, of course, put in the context of the relative strength of the impacts of these different dimensions. Political diversity has the largest impact, and that impact remains equal to or greater than the other dimensions of diversity, even after controlling for the intervening variables. [table 3 about here] The lack of impact for racial and religious diversity in the Canadian sample also requires some discussion. When the samples are compared, the Canadian sample has a far higher percentage of minorities than the Belgian sample. Furthermore, an examination of attitudinal variables suggests that the Belgian youth exhibit far higher levels of ethnocentrism and in-group favoritism than the Canadian sample. 12. Given the salience of foreigners in Belgium, one explanation may be that those who do have racially and religiously diverse networks are being exposed to different messages than the 11 See Figure 2 in the appendix for bivariate results of the relationship between network diversity and political knowledge, interest and discussion. As one might expect, knowledge, interest and discussion are all correlated with each other (.251 to.447). In the regression models, however, the inter-correlation does not seem to pose a significant threat to model validity, as VIF scores for these three items are below 1.5. 12 It is difficult to ascertain whether this results from sampling differences in the two countries, or from actual differences in the two contexts that result from their different histories with immigration. 19
young people in the Canadian sample, who exhibit less evidence that racial and religious diversity is salient to them. What is clear from these findings is that in these samples of young people there is no evidence of a negative impact on political participation, as much of the literature on this dimension of diversity has suggested (Hill and Leighley 1999; Costa and Kahn 2003; Alesina and Ferrara 2000). [table 4 about here] In table 4, we take our analysis one step farther and examine the types of networks in which young people are exposed to diversity. As we suggested earlier, research tends to suggest that network diversity among close friends may have a dampening, rather than a mobilizing effect on political participation. Weaker ties, in contrast, may be more likely to foster the types of political discussion and information that contribute to participation. Using aggregate measures of overall diversity among close friends and among acquaintances at school, we find that network diversity in both settings are positively and significantly related to political participation, even after controlling for a host of other variables. Several observations can be made from these results. First of all, despite our expectations, the size of the coefficient for strong ties is larger than the coefficient for weak ties. This suggests that close friends are equally or more important for political participation. As with the previous table, we see that the inclusion of political knowledge, interest and discussion reduces the impact of both types of network diversity, although the impact of network diversity among close friends remains significant at the <.01 level. Network diversity among acquaintances, on the other hand, becomes less important after controlling for these other variables, although it remains significant at the <.05 level in the Belgian sample. In general, both tables 3 and 20
4 provide convincing evidence that the impact of network diversity among the youth sampled here appears to be positive. This seems particularly the case when we consider political diversity, and to a lesser extent socio-economic and racial/religious diversity. Contrary to what one might expect, this finding holds among both strong and weak tie networks. Conclusion In this paper, we went beyond the current literature and examined the relationship between various dimensions of diversity in strong and weak tie networks and political participation of young people in Belgium and Canada. Based on the current status of the research on diversity and political outcomes, we expected that diverse networks among strong or close relationships might have demobilizing effects on young people s political activities, mostly because they have been linked people s tendency to avoid conflicts with their friends (Mutz 2002a, 2006). On the other hand, we hypothesized positive relationships between diversity and political outcomes in weak-tie networks, because they are able to introduce a degree of otherness and provide new political views and opinions in a more distanced setting. The nature of contact in weak tie networks, we argued, might be converted into heightened political interest and knowledge without the uncomfortableness that such discussion may create among close friends. Such information and knowledge, we argued, would consequently support more political involvement. Our research with the CYS in Belgium and Canada has confirmed the expectations of the second hypothesis: diverse ties do have positive and mobilizing 21
effects in weak tie networks. However, strong tie networks seem to be equally or even more important for political participation. This is particularly the case for political and socio-economic differences, which seem to spur increased political action, in close friend networks as well as among acquaintances at school. Linguistic diversity (the rarest form of diversity) and ethno-cultural diversity do not have consistent positive effects in these two countries. How can we explain these findings vis-à-vis the existing literature? Certainly we cannot exclude that perhaps diversity might work differently among young people in Canada and Belgium as most studies finding demobilizing effects were conducted among adults in the US. We also should add that our measurements of diverse network ties are different from the ones most often employed in the literature. We are able to measure the perceived composition of networks along various dimensions. Whereas Mutz measures the political diversity of three close friends that the respondent identifies for the interview, our measures give a more general picture of the entire friendship network and the composition of acquaintances at school. It is therefore possible that conflict avoidance does not play such an important role in these wider networks as it does among one s three closest friends. Only a direct comparison with different network measures will be able to identify the extent to which the network measurements are driving the results. Certainly, it will be a future extension of our study to include measurements of conflict avoidance in order to better understand this causal mechanism. Instead of conflict avoidance, we have identified a different causal mechanism: namely, the extent to which diverse networks strengthen political exchange, political information and interest, which in turn shape political involvement. This is particularly 22
visible, as the coefficients of diverse networks diminish when we include these political variables in our equations. This is consistent with some previous research that emphasizes the importance of network composition for fostering politically-mobilizing information and discussion (McClurg 2006; Lake and Huckfeldt 1998). There is no doubt that our results need to be viewed with caution. For starters, we must be aware of the issue of reverse causality, as those who are politically active might create more diverse friendship networks through their activities. An interesting point of our findings is that not only political but also socio-economic diversity matters for political mobilization. While this might also stem from increased involvement, its less direct association with political involvement is promising for our argument that network diversity fosters mobilization, and not vice-versa. There is also some possibility of spuriousness, that is to say that politically mobilized people discuss politics more frequently with everyone, and are therefore more aware of political otherness in their surroundings, more than inactive young people who might remain silent on political issues. Interestingly though, the effect of network diversity remains fairly strong and significant even when we control for a variety of political mobilization variables. In fact, a story seems likely in which diverse networks increase and strengthen political awareness and knowledge. Our findings then have positive implications for multiculturalism and its effects on democratic politics. Our research suggests that diversity may not in fact have such disastrous effects for participatory democracy (Putnam 2007). If young people can make friends with people from different backgrounds, this not only seems to foster tolerance (Mutz 2002b; Hooghe et al. 2007; Cigler and Joslyn 2002), but also mobilizes young 23
people to get involved in their communities in ways that exposure as adults seems unable to do. While young people, like adults, tend to have fairly homogenous social networks, this research points to the potential benefits of fostering people from different backgrounds and different beliefs to interact with one another. Such interaction seems to be an important predictor of the ways in which young people engage in the larger communities in which they live. 24
TABLES AND FIGURES Table 1: Descriptives of Aggregate Diversity Measures by Sample Mean Std. Deviation N McGill Youth Study Racial/Religious Diversity of Network 3.38 (1.31) 3166 Official Language Diversity of Networks 2.25 (1.19) 3200 Political Diversity of Networks 2.70 (1.05) 3147 Economic Diversity of Network 3.21 (1.17) 3151 Network Diversity (close friends) 2.75 (0.91) 3238 Network Diversity (acquaintances) 3.22 (0.89) 3139 Belgian Youth Study Racial/Religious Diversity of Network 2.28 (1.10) 6060 Official Language Diversity of Networks 1.60 (0.88) 6221 Political Diversity of Networks 2.29 (1.02) 5981 Economic Diversity of Network 2.79 (1.01) 6090 Network Diversity (close friends) 2.12 Network Diversity (acquaintances) 2.36 Note: Using independent samples t-test, MYS sample has signif networks than the BYS, with the exception of the Political Divers (0.74) 5975 (0.75) 5868 icantly more diverse ity variable. 25
Table 2: Mean Network Diversity By Demographic Variables Racial/Religious Diversity of Network Official Language Diversity of Networks Political Diversity of Networks Economic Diversity of Network Mean (s.d.) Mean (s.d.) Mean (s.d.) Mean (s.d.) Visible Minority Status visible minority 3.75 (1.30) 2.11 (1.15) 2.59 (1.05) 3.06 (1.19) otherwise 2.39 (1.14) 1.75 (1.00) 2.39 (1.04) 2.91 (1.06) *** *** *** *** Urban/Rural Status rural 2.13 (1.02) 1.57 (0.85) 2.31 (1.02) 2.82 (1.04) urban 3.33 (1.28) 2.13 (1.17) 2.59 (1.06) 3.09 (1.12) *** *** *** *** Gender female 2.75 (1.29) 1.88 (1.08) 2.42 (0.97) 2.93 (1.07) male 2.55 (1.27) 1.75 (0.99) 2.45 (1.11) 2.94 (1.11) *** *** p=.167 Parent University Educated? Neither 2.45 (1.22) 1.69 (0.97) 2.34 (1.02) 2.88 (1.05) One or Both 2.93 (1.33) 2.01 (1.08) 2.65 (1.03) 3.11 (1.12) *** *** *** *** Note: Network Diversity Scales run from 1 to 7, where higher scores indicate more network diversity. Full details of variables are available in the Appendix. 26
Figure 1: Network Diversity and Political Participation 5 n Political Activities (0 to 10) Mea 4 3 2 1 0 none almost none a few about half many, most, all Network Diversity Racial/Religious Official Language Political Economic 5 Mean Political Activities (0 to 10) 4 3 2 1 0 none almost none a few about half many, most, all Network Diversity Close Friends Acquaintances 27
Table 3: Impact of Dimensions of Diversity on Political Activities Scale McGill Youth Study Belgian Youth Study B s.e. B s.e. B s.e. B s.e. (Constant) 0.46 (0.28) -1.47 (0.27) *** -0.15 (0.18) -1.44 (0.18) *** Are you a Canadian/Belgian citizen? 0.44 (0.18) ** 0.41 (0.16) ** 0.34 (0.13) ** 0.22 (0.12) * Female 0.95 (0.09) *** 0.90 (0.08) *** 0.38 (0.05) *** 0.49 (0.05) *** Visible Minority -0.27 (0.11) ** -0.13 (0.10) 0.03 (0.11) -0.02 (0.10) Parent(s) University Educated? 0.51 (0.09) *** 0.34 (0.08) *** 0.26 (0.06) *** 0.15 (0.06) ** Urban -0.30 (0.12) ** -0.47 (0.11) *** -0.18 (0.07) ** -0.24 (0.07) *** How many close friends do you have? 0.09 (0.03) ** 0.08 (0.03) ** -0.01 (0.02) 0.01 (0.02) Racial/Religious Diversity of Network 0.00 (0.04) 0.01 (0.04) 0.14 (0.03) *** 0.13 (0.03) *** Official Language Diversity of Networks -0.01 (0.04) 0.00 (0.04) -0.03 (0.03) -0.02 (0.03) Political Diversity of Networks 0.57 (0.05) *** 0.16 (0.05) *** 0.46 (0.03) *** 0.23 (0.03) *** Economic Diversity of Network 0.18 (0.04) *** 0.16 (0.04) *** 0.14 (0.03) *** 0.12 (0.03) *** Political Knowledge Scale (standardized 0-1) 0.38 (0.14) *** 0.71 (0.09) *** How interested are you in public issues and politics? 0.67 (0.06) *** 0.41 (0.04) *** How often do you discuss public issues and politics with close friends? 0.80 (0.07) *** 0.54 (0.05) *** Adjusted R-Squared 0.132 0.281 0.099 0.191 N Note: Truncated dependent variable is used. See text for details. 2543 4530 28
Table 4: Impact of Types of Diversity on Political Activities Scale McGill Youth Study Belgian Youth Study B s.e. B s.e. B s.e. B s.e. (Constant) 1.18 (0.28) *** -1.27 (0.27) *** 0.08 (0.18) -1.39 (0.18) *** Are you a Canadian/Belgian citizen? 0.42 (0.18) ** 0.41 (0.16) ** 0.46 (0.13) *** 0.25 (0.12) ** Female 0.86 (0.09) *** 0.88 (0.08) *** 0.34 (0.06) *** 0.48 (0.05) *** Visible Minority -0.47 (0.10) *** -0.19 (0.09) ** -0.05 (0.11) -0.04 (0.10) Parent(s) University Educated? 0.57 (0.09) *** 0.35 (0.08) *** 0.29 (0.06) *** 0.15 (0.06) ** Urban -0.40 (0.12) *** -0.52 (0.11) *** -0.30 (0.07) *** -0.30 (0.06) *** How many close friends do you have? 0.05 (0.03) 0.07 (0.03) ** -0.05 (0.02) ** -0.01 (0.02) Diversity of Close Friends 0.34 (0.07) *** 0.17 (0.06) *** 0.49 (0.05) *** 0.36 (0.04) *** Diversity of Acquaintances 0.24 (0.07) *** 0.08 (0.06) 0.24 (0.05) *** 0.12 (0.04) ** Political Knowledge Scale (standardized 0-1) 0.41 (0.14) *** 0.74 (0.09) *** How interested are you in public issues and politics? 0.69 (0.06) *** 0.43 (0.04) *** How often do you discuss public issues and politics with close friends? 0.84 (0.07) *** 0.60 (0.05) *** Adjusted R-Squared 0.088 0.277 0.069 N Note: Truncated dependent variable is used. See text for details. 2543 4530 0.186 29
Figure 2: Network Diversity and Political Engagement 0.6 Mean Political Knowledge (0-1) 3 Discuss Politics (1-4) 0.4 2 0.2 0 none a few some about half many, most, all 1 none a few some about half many, most, all Political Interest (1-4) 3 2 Racial/Religious Official Language Political Socio-Economic 1 none a few some about half many, most, all 30
Bibliography Alesina, Alberto, and Eliana La Ferrara. 2000. "Participation in Heterogeneous Communities." Quarterly Journal of Economics 115 (3):847-904. Berelson, B.N., P.F. Lazarsfeld, and W.N. MacPhee. 1963. Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago: University of Chicago Press. Berry, Jeffrey, Kent Portney, and Ken Thomson. 1993. The Rebirth of Urban Democracy. Washington D.C.: Brookings Institution. Blau, Peter M. 1977. Inequality and Heterogeneity: A Primitive Theory of Social Structure. New York: Free Press. Brown, B.B. 2004. "Adolescents' Relationships with Peers." In Handbook of Adolescent Psychology, ed. R. M. Lerner and L. Steinberg. New Jersey: John Wiley and Sons. Burt, Ronald S. 1997. "The Contingent Value of Social Capital." Administrative Quarterly 42:339-65. Campbell, B.A. 1980. "A Theoretical Approach to Peer Influence in Adolescent Socialization." American Journal of Political Science 24 (2):324-44. Cigler, Allan, and Mark R. Joslyn. 2002. "The Extensiveness of Group Membership and Social Capital: The Impact on Political Tolerance Attitudes." Political Research Quarterly 55 (1):7-25. Coleman, James. 1988. "Social Capital in the Creation of Human Capital." American Journal of Sociology 94:95-120. Costa, D., and M. Kahn. 2003. "Civic Engagement and Community Heterogeneity: An Economist's Perspective." Perspectives on Politics 1 (1):103-11. Crosnoe, R. 2000. "Friendships in Childhood and Adolescents: The Life Course and New Directions." Social Psychology Quarterly 63 (4):377-91. Dawson, R.E., and K. Prewitt. 1969. Political Socialization. Boston: Little Brown. Delli Carpini, Michael, and Scott Keeter. 1996. What Americans Know About Politics and Why It Matters. New Haven: Yale University Press. Druckman, James N., and K.R. Nelson. 2003. "Framing and Deliberation: How Citizens' Conversations Limit Elite Influence." American Journal of Political Science 47 (4):729-45. Eisenberg, N., and A Morris. 2004. "Moral Cognitions and Prosocial Responding in Adolescents." In Handbook of Adolescent Psychology, ed. R. M. Lerner and L. Steinberg. New York: Wiley. Eliasoph, Nina. 1998. Avoiding Politics: How Americans Produce Apathy in Everyday Life. New York: Cambridge University Press. Eveland, W.P., and D.A. Scheufele. 2000. "Connecting News Media Use with Gaps in Knowledge and Participation." Political Communication 17 (3):215-37. Franklin, Mark N. 2004. Voter Turnout and the Dynamics of Electoral Competition in Established Democracies Since 1945. Cambridge: Cambridge University Press. Galston, William A. 2004. "Civic Education and Political Participation." PSOnline 37:263-6. Glanville, J.L. 1999. "Political Socialization or Selection? Adolescent Extracurricular Participation and Political Activity in Early Adulthood." Social Science Quarterly 10 (2):279-90. 31