Generating Social Capital Civil Society and Institutions in Comparative Perspective Edited by Marc Hooßhe and Dietlind Stolle macmillan 88 Dag WolleBííek and Per Selle | Putnam, R. (1995). "Bowling AJone: America's Declining Social Capital." f Journal of Democracy (6)1, 65-78. I Putnam, R (2000). Bowling Alone, Tlie Collapse and Revival of American | Community. New York: Simon and Schuster. Quigley, K. (1996). "Human Bonds and Social Capital." Orbis40(2), 333-343. Rothstein, B. (1998). "Vanŕrän kommer det sociala kapitalet?" Socialvetensbapltg tidskrifi 3(2), 164-171. Rueschemeyer, D. (1998). "The Self-Organization of Society and Democratic Rule: Specifying the Relationship." In D. Rueschemeyer, M. Rueschemeyer and B. Wittrock (eds.), Participation and Democracy. Armonk: M. E. Sharpe. Rueschemeyer, D., M. Rueschemeyer and B. VVittrock (eds., 1998). Participation and Democracy. Armonk: M. E. Sharpe. Selle, P. and K. Stromsnes (1998). "Organised Environmentalists: Democracy as a Key Value?" Voluntas 9(4), 319-343. Selle, P. and K. Stromsncs (2001). "Membership and Democracy: Should We Take Passive Support Seriously." In P. Dekker and E. Uslaner (eds.), Social Capital and Politics in Everyday Life. London: Routledge. Seile, P. (1998). "Orgaiusasjonssamfunnet—ein statsreiskap?" In T. Gronlie and P. Selle (eds.), Ein stat? Pristillingas Fire Ansikt. Oslo: Samlaget. Stolle, D. (2001). "Clubs and Congregations: The Benefits of Joining an Association." pp. 202-244. In K. Cook (ed.), Trust in Society. New York: Russell Sage Foundation. Stolle, D. and T. Rochon (1999). "The Myth of American Exceptionalism. A Three-Nation Comparison of Associational Membership and Social Capital." pp. 192-209. In J. van Deth et al. (eds.), Social Capital and European Democracy. London: Routledge. Tocqueville, A. de (1954 [1835/40]). Democracy in America. New York: Random House. Torcal, M. and y. Montero (1999). "Facets of Social Capital in New Democracies: The Formation and Consequences of Social Capital in I Spain." pp. 167-191. In J. van Deth et al. (eds.), Social Capital and European Democracy. London; Routledge. Torpe, L. (1998). Lokal organisasjonsdeltakelse og demokrati. Meilern Individ ogfiellesskab samt borgcr og styre. Bergen: LOS-senteret. van Deth, J. (1997). "Introduction: Social Involvement and Democratic Politics." In J. van Deth (ed.), Private Groups and Public Life. London: Routledge. Whiteley, P. (1999). "The Origins of Social Capital." In J. van Deth et al. ------- (eds.), Social Capital and European Democracy. London: Routledge. Woliebajk, D. (2000). Participation in Voluntary Associations and the Formation of Social Capital. University of Bergen: Ph. D. Dissertation. Wolleba:k, D., P. Seile and H. Lorentzen (2000). Frivillig Innsats. Oslo: Fagbokforlaget. Uapter 5 Voluntary Associations and Democratic Attitudes: Value Congruence as a Causal Mechanism Marc Hoojjhe Introduction In most of the recent research on social capital, a positive correlation is observed between membership of voluntary associations and the adherence to democratic value patterns. However, we have access to few research results, which could explain the causal mechanism that is responsible for this positive correlation. Two models seem to predominate the current literature on social capital (see Stolle this volume). Some authors clearly rely on a socialization logic: Because of the interaction with others, members of voluntary associations are socialized into more democratic and more social value patterns. Subsequently, these positive attitudes are transferred to society as a whole: Members do not only learn to trust their fellow members, but they also develop a generalized trust in other citizens. In this view, voluntary associations or other societal contexts function as learning schools for democracy, as de Tocqueville called it. Putnam (1995, 666) echoes this view when he states: "the causation flows mainly from joining to trusting." Tliis line of reasoning has been received with much skepticism. Some empirical research expresses doubts about the existence of a significant relation between membership and civic attitudes (Berman 1997; Mondak and Mutz 1997), calling our attention to other societal 90 Marc Hooghe contexts in which civic attitudes are shaped, whereas other research demonstrates that the relationship is at best weak (Brehm and Rahn 1997; Claibourn and Martin 2000; Mayer this volume). Levi (1996) has drawn our attention to the fact that not all associations will have positive effects: Some might just as well have negative consequences, and the existence of this "dark side of social capital" has been acknowledged in some of the more recent studies on social capita! (Putnam 2000,350-363). Other authors are critical about the claims of the socialization school, and they stress the importance of self-selection (Newton 1997, 1999; Stolle 1998, 2002; Whiteley 1999; Uslaner this volume). Persons with antisocial attitudes will refrain from joining voluntary associations, and these associations will recruit members who are already relatively high on civic attitudes. This pattern of self-selection is taken to explain the positive correlation between membership and democratic value patterns. Uslaner (1998) follows the same logic when he states that some people have a more optimistic and more trusting outlook toward life (due to youth experiences) and therefore will be more inclined to join all kinds of associations. At the present state of research, the debate between socialization and self-selection approaches remains unsolved (Dekker 1999). The problem is that neither the authors of the socialization school nor most of their critics have explained the causal mechanism that is responsible for this admittedly weak correlation pattern (Hooghe 2000). Within the literature on civic participation, no one really denies that a process of self-selection takes place: It seems self-evident that not everyone will have the same inclination to join voluntary associations. The basic research question, however, is to know whether voluntary associations have an additional socialization effect (Stolle 2000): Even after discounting the recruitment effect, is there still some influence from participation left, and, if so, what causal mechanism is responsible for this effect? Only if such an additional effect can be documented can we safely conclude that voluntary associations indeed play a key role in generating social capital and should not just be considered as indicators for the presence of social capital. In this chapter we want to explore the-relationship between participation and the attitudinal components of social capital by using data from a recent Belgian face-to-face survey with 1,341 respondents (representative of the Dutch-speaking population of the Flemish autonomous region in Belgium). Voluntary Associations and Democratic Attitudes 91 The central question in this chapter is whether membership in voluntary associations leads to a reduction of feelings of ethnocentrism, and, if so, what causa! mechanism could be responsible for this effect. Another ambition of this chapter is to introduce some of the recent social psychological concepts and research results about group processes into the social capital debate. This line of research suggests that interaction within groups does lead to socialization effects, but it does not support the claim that this kind of interaction automatically leads to the development of more social norms. Any effect groups might have is dependent upon the characteristics of the group and its members. As the Belgian survey included a large number of questions about associational participation, we can ascertain whether all kinds of associations have similar socialization effects, or whether these effects are context specific. The Effects of Group Interaction In the current research on the relation between voluntary associations and the formation of citizenship attitudes, one can witness an intriguing lack of communication. On the one hand, authors inspired by the work and the insights of de Tocqueville and Putnam claim that interaction within voluntary associations has distinct effects on the value patterns of the members, and thus should be seen as a source of social capital. But on the other hand, we have access to an extensive body of empirical research about what actually goes on in primary and other groups, and what are the effects of these kinds of interaction. This line of research, however, is all but neglected in the current social capital debate. The claim that interaction within groups has positive socialization effects is based on the assumption that members develop trust in their fellow members as a result of the sustained interaction within the association, which in turn is generalized toward society as a whole (Stolle 2000). Both assumed mechanisms are highly problematic. To start with, there is no evidence whatsoever that prolonged interaction in groups would actually lead to higher trust levels within the group.1 Furthermore, given the fundamental difference between knowledge-based trust and generalized trust (Yamagishi and Yamagishi 1994), is seems unlikely that particularized trust would be converted that easily into generalized trust. We do not develop trust in total strangers, about whom we have no information, because we trust people we actually know (Uslaner 2000). But even if this was the case, it only 92 Marc Hooghe Voluntary Associations and Democratic Attitudes 93 adds to the problem. If particularized trust could be converted into generalized trust, there is no reason to assume that interaction within voluntary associations would be a privileged source of generalized trust. In that case, interaction within families, schools, neighborhoods or work environments would function just as well as a source of generalized trust (Whiteley 1999). In sum, the causal mechanism assumed does not seem entirely plausible. We can rely on an enormous body of literature and research about the effects of group interaction (Tajfel 1981; Forsyth 1983; Turner et al. 1987; Hendrick 1987; Paulus 1989; Robinson 1996). The social-psychological literature on group interaction, however, does not offer support for any of the basic tenets of social capital theory. There is no indication whatsoever that interaction with other group members would automatically lead to the development of a more socially oriented value pattern, to a rise in trust levels, or to die abandonment of prejudices (Goslin 1969; Mills and Rosenberg 1970; Duncan and Fiske 1977; Ajzen and Fishbein 1980; Turner 1984; Turner et al. 1987). There are even some laboratory experiments showing that membership of a group can lead to an enhancement of prejudices against members of outsider groups (Skinner and Stephenson 1981; Haslam et al. 1999). We do not have any indication that group interaction automatically leads to a more socially desired value pattern, and this lack of research data should question the important role social capital theory assigns to voluntary associations. This does not imply that interaction with other group members could not produce socialization effects. It does imply that these effects will be dependent upon context characteristics: The process is endogenously induced—the value changes are not exogenous. Members of a group are subjected to socialization experiences because they are influenced by the values of other group members, resulting in a process of value congruence within the group (Tajfel 1981; Turner et al. 1987; Levine and Russo 1987; Abrams and Hogg 1991). This form of "personal influence" within groups is already described in the classical study by Katz and Lazarsfeld: "We are led to expect that an individual's opinions will be substantially affected by the opinions of others whose company he keeps, or whose company he aspires to keep" (Katz and Lazarsfeld 1955, 53}. During the past decades, this argument has received support from numerous experiments, demonstrating the existence of processes of value congruence within groups (for a review, see Levine and Russo 1987). Youth research demonstrates time and time again the importance of the effect of peer group members on the attitudes of individual actors (Adler and Adler 1998; Bankston and Caldas 1996), and although we can assume that die magnitude of these effects will be less pronounced among adults, we have no reason to expect that socialization will not occur at all (Williams 1975). Following this logic implies that the socialization effects of organizations or groups do not originate from outside the group: The interaction does not introduce qualitatively new values into the group but enforces already existing values, as Katz and Lazarsfeld (1955, 96) already stated. The presence of and the interaction with other group members will influence the values and judgments of individual actors: "The sight and sound of others doing the same thing as oneself functioned as conditioned social stimuli to release and augment learned reaction tendencies previously existing in individuals... Importandy, however, social facilitation did not represent the emergence of new group properties; individual behavior did not change qualitatively in groups, it was merely 'enhanced' so to speak" (Turner et al. 1987, 11). The fellow members, in this respect, function as a "reference group" (Merton and Kitt 1950), providing the members with cues about how to construct and possibly transform their own value patterns. Self-Categorization Theory A recent attempt to build diis line of research into a coherent theoretical framework can be found in the social identity, or self-categorization, dieory as it was developed by Henri Tajfel (1981) and John Turner et al. (1987). They assume that individuals tend to avoid cognitive dissonance but that at the same time by themselves they are not capable of developing a coherent value pattern, which is congruent with the complexity of observations from the outside world. To counter that potential source of insecurity, individuals are dependent upon the interaction and die dialogue with significant others. Turner especially stresses the fact that processes of self-categorization can reduce complexity: Individuals learn to see themselves as members of a socially defined category, and therefore they also gain access to the corresponding role and value pattern. Within this process of value congruence, one can distinguish two different levels (Turner et al. 1987, 35; Turner and Oakes 1989). Widi regard to information, group members arc dependent upon the information they retrieve from other group members to construct their own worldviews and value patterns, Thislnformation has already been selected and this can lead to a certain congruence in the value patterns of the members. Secondly, one can distinguish a normative 94 Marc Hooc.he influence, causing group members to adopt, at least to some extent, the normative positions of their fellow group members. This does not mean that the consequences of group interaction remain limited to facilitating a convergence toward a pre-existing average position, which would imply that group membership does not have an additive effect. Because of tendencies toward group polarization, the convergence will occur on a more extreme position, thus strengthening already existing values: "Uniformities in intra-group behavior result from the members' opinions becoming more extreme in the socially favored direction rather than from convergence on the average of their initial position" (Turner 1982, 35). It is assumed, and to some extent also documented, that die presence of like-minded others serves as a stimulus for the individual actor to further develop socially desired traits and values (Fraser and Foster 1984). Turner's self-categorization theory implies that socialization will be most successful when the individual is integrated in a group with a relatively homogeneous value pattern. Homogeneity within the group allows the reinforcement of the influence of the various group members, or at least means that these various sources of influence will operate in die same direction. As Verba (1961, 40) already noted: "The greater the homogeneity of primary group contacts, the greater the intensity of political participation... it was found that voters with friends of various political persuasions were less strong in their voting intentions than those whose friends were all of the same persuasion." If we confront the results of this body of empirical research with the assumptions of social capital theory, a clear problem arises. Authors stressing socialization within organizations assume tiiat the interaction within groups leads to die introduction of qualitatively new values, like tolerance or generalized mist. The results of the research using social identity or self-categorization theory indicate that no new value patterns are introduced because of the interaction but that pre-existing value patterns are made more salient or are reinforced. This insight could be important for the further development of social capital theory. Levi (1996) has called attention to the fact that if voluntary associations really have such beneficial effects, it is difficult to explain how criminal organizations could produce forms of "unsocial" capital, a point that has been acknowledged in some of the more recent writings on social capital (Putnam 2000, 350-363). Levi's remark indeed presents a dilemma for social capital theory, but not for social identity theory. Criminal organizations will recruit members attracted to a criminal way of life, and it will further socialize them into this value pattern, just as religious organizations will recruit Voluntary Associations and Democratic Attitudes 95 members who are interested in spiritual matters and will further socialize them into a more religious value pattern. The Belgian survey on participation in voluntary associations offers a unique opportunity to put these assumptions to the test. Because the survey contains numerous questions on participation, it is possible to distinguish various kinds of organizations. Of course, other authors, too, have already pointed out that not all voluntary associations will have the same beneficial effects. Sorting out what kind of interaction is associated with what specific kind of effect, however, should allow us to uncover the causal mechanism that could be responsible for the association between the structural and attitudinal components of social capital. Studying Ethnocentrism In building the argument, we will rely on the results of a face-to-face survey our research unit conducted during the spring of 1998 (n = 1,341) and that proved to be representative of the Dutch-speaking population of the Flemish autonomous region in Belgium (see appendix for technical details). In our analysis we will use a five-item balanced measurement scale on ethnocentrism as the main dependent variable. All of the items in this scale refer to negative prejudices toward migrant groups widiin Belgian society, with a typical item being "Generally speaking, migrants cannot be trusted." There are four reasons why we have chosen to study ethnocentric prejudice. First of all, one has to keep in mind that racism has become a highly salient political issue in Belgium, as in most of the European Union countries. In the Flemish part of the country, an extreme right party with a very hostile view toward die presence of ethnic and cultural minorities in the country has -taken 15 percent of die vote (Lubbers, Scheepers and Billiet 2000). Reducing ethnocentric prejudice therefore has become one of the main policy priorities, not only of the Belgian government but also for a lot of voluntary associations, and it seems worthwhile to investigate whether these associations actually succeed in reducing ethnocentrism among their members. Second, in the past decades Belgium, like most Western European countries, has evolved from an ethnically rather homogeneous nation into a multicultural society as a result of the influx of relatively large groups of migrants from the Mediterranean, Eastern Europe and Africa (Hooghe 2001). This heterogeneity implies ..that....Belgian society will have to develop new citizenship concepts in order to allow full citizen participation of these ethnic minorities (Soysal 1994). One can argue, 96 Marc Hooghe therefore, that reducing ethnocentric prejudice is of crucial importance for the future development and stability of a democratic political culture in Belgium and other Western countries. A third reason to study ethnocentrism is that this attitude is strongly correlated with other attitudinai scales, ail of Lheni related to die concept of an ''authoritarian personality" that contrasts sharply with contemporary conceptions of what democratic civic attitudes should be. In our survey, we observed a marked affinity between ethnocentrism and a utilitarian conception of individualism, feelings of authoritarianism, feelings of insecurity, and also with a vote for the extreme right party in Belgium. No other single attitude is as strongly related to all these other indicators of a lack of support for democratic values. Therefore, it seems safe to consider ethnocentrism as the core of a complex of culturally con servative values, as it has indeed been used in the literature on cultural conservatism (Lipset 1959). A fourth reason to study ethnocentrism is technical: The concept can be measured by a solid scale that was developed by Billiet, Carton and Eisinga (1995). Because it has been used extensively in previous surveys, we know it to be a reliable instrument to register negative feelings toward outsider groups. Using solid attitudinai scales is more reliable than using just a single item as a dependent variable, which is often pursued in social capital research. We know from earlier research that the feeling of ethnocentrism is not spread evenly across the Belgian population: Respondents with lower educational credentials support a more ethnocentric discourse than respondents with high educational credentials. In a standardized scale (that ranges from 0, which indicates no ethnocentrism, to 100, which indicates high ethnocentrism) the mean scores vary from 48.3 among the respondents who have only completed elementary education to 32.9 for those with higher education, The relation between education levels is not only very strong but also perfectly linear for all the various groups in the population. The fact that ethnocentrism is unevenly distributed throughout society and is more strongly present in the lower-educated strata allows us to contrast various groups in further analysis.2 To fully test the proposition on self-selection, we would need data on the kind of values people adhere to when diey enter assoeiarional lite, and these data are not available for our research setting. This forces us to approach the problem of self-selection indirectly. Given the close relation between education and ethnocentrism, we can assume that if we know the educational level of a sufficiently large group of respondents, we at least have some indication about their initial feelings of ethnocentrism. The association Voluntary Associations and Democratic Attitudes 97 between education and ethnocentrism is so strong that, if we have any group of a reasonable size of highly educated people it is very likely that their score on the ethnocentrism scale will be lower than that of a similar group with only elementary education, and vice versa, even if these groups are recruited through a process of self-selection. Self-Selection The first step we have to take in die analysis, therefore, is to demonstrate that processes of self-selection are indeed occurring. To a large extent, this step is superfluous: We have access to a large body of participation research showing that the active population of any given society is not representative for the population as a whole (Verba and Nie 1972; Verba, Schlozman and Brady 1995; Dekker 1999). Education levels in particular explain an important part of the variation with regard to participation behavior. Nie, Junn and Stehlik-Barry (1996, 2) summarize the research quite neatly: "Formal education is almost without exception the strongest factor in explaining what citizens do in politics and how they think about politics." In our survey, too, the traditional inequalities emerge: Active membership in voluntary associations is not spread evenly throughout society. In our survey, respondents were given a list of twenty-two different kinds of organizations, with five answering possibilities for each category: (1) never been a member; (2) has been a member; (3) passive member; (4) active member; (5) member of the board. If we perform a logistic regression analysis with the dichotomous variable "active member, or member of die board of at least one association/not an active member" as a dependent variable, die same picture emerges as in previous participation research. We see that gender, education and religion are strongly related to participation (Table 5,1). Men participate more intensely than women; churchgoers participate more than non-Christians or nonchurchgoing Christians. But again, as Nie et al. (1996) already wrote, education proves to be a very strong variable, with the higher-educated group participating almost twice as much as the lower-educated group. We learn from this regression that the higher-educated strata will be over-represented in voluntary associations, and given the strong relation between education and ethnocentrism, we can already expect that on average members of voluntary associations will be less ethnocentric than the rest of society. 98 Marc Hooghe Table 5.1 Logistic regression for active membership in voluntary associations exp.(3) Age ns. 18-35 yr 36-55 yr 1.19 56-75 yr 0.92 Sex *** Male Female .56 Education * ** Low Middle 1.29 High 1.92 Income ns. —750 euro/month 750-1.500 euro/momh .64 1.500-2.500 euro/month .95 +2.500 euro/month 1.07 Religion *** Not religious Religious, not churchgoing 1.11 Churchgoing 2.16 Time on television ns. <14 h/wk 14-20 h/wk 1.17 +20 h/wk 0.82 Family ns. Alone or with parents With partner 0.74 Divorced/widow(er) 0.59 Note: Entries arc logistic regression coefficients (exp. (J), representing difference from reference category (= 1). * = ^<0.05; **=j><0.01; *** = p<0.001. Source: Flemish Citizen Survey 1998, n - 1,341. Effects of Membership Experience The next step examines the relation between membership and feelings of ethnocentrism (Table 5.2). Our first regression model is a baseline model, simply integrating background variables as independent variables. In this model we also included the time spent on watching television, and the religious affiliation of the respondent, because we Voluntary Associations and Democratic Attitudes 99 Table 5.2 Explaining ethnocentrism Model 1 Model 2 widiout associations associations included E (SE B) & B (SE B) Sex -.82 (1.04) -.02 -1.21 (1.03) -.03 Age .11 (0.04)*** .09 .14 (0.04)*** .12 Income -.12 (0.19) -.02 -.03 (0.19) .00 Education -1.17 (0.20)*** -.21 -.84(0.20)*** -.15 Membership — — -1.36 (0.22)*** -.19 Television time .24 (0.04)**-* - ... ig. .21 (0.04)*** .16 Church involvement 1.68 (1.16) .04 2.03 (1.14) .05 Constant 45.50 (3.87)*** 45.39 (3.80)*** adj. ? .14 .17 Note: Ordinary least squares regression; entries are nonstandardized (with standard deviation) and standardized regression coefficients. *=p<0.05; ** = ?<0.01; ***=£<0.001. Source: Flemish Citizen Survey 1998, n — 1,341. know from previous research that both these variables are significantly related with participation levels and with feelings of ethnocentrism (Hooghe 2002). We see that even this model can explain 14 percent of the variance, and again education proves to be the most influential factor. While including the effect of associational membership in the regression (model 2), we have operationalized the variable as the sum of both current and previous memberships. This has been done because a previous analysis has shown that the effect of membership on civic attitudes is not dependent just on the current membership level but also on previous-participation- experiences of the respondent (Hooghe 2003). This finding can be easily explained: One of the tilings we expect from socialization experiences is that their effects remain discernible long after the experience itself has ended. Having been a member of a scouts group should have effects, even when the person involved has matured and is no longer a member (Stolle and Hooghe 2002). Furthermore, including previous memberships seems a more valid method for measuring participation habits, avoiding the risk of measuring momentary alternations of this behavior pattern, due, for example, to a recent relocation or to child-rearing responsibilities. Building on this analysis, we introduced "ever membership" as an independent variable in model 2, whereby "ever" is simply the sum of current and previous, that is, concluded, memberships. 100 Marc Hooghe The results of model 2 indicate that membership of voluntary associations indeed is negatively related to ethnocentrism. Explained variance rises from 14 to 17 percent, and membership becomes the most important variable, even more important than education. The Company We Keep We now come to the final piece of our argument, by assessing whether the negative relation with ethnocentrism is uniform for all kinds of associations or is dependent upon member and group characteristics (Stolle and Rochon 1998). Therefore we now leave the aggregate measurement level of membership to look at the figures for each kind of organization separately. Because a few categories were poorly represented in the survey, we had to regroup the original twenty-two categories into eighteen somewhat larger groups. For each kind of association, we can list the average number of years of schooling its members have completed (Table 5.3), and here one can observe distinct differences. Organizations involved with peace, human rights, environment, and school and neighborhood councils seem to attract highly educated members, with on average more than thirteen years of completed education. Cultural, political, social and sport associations seem to recruit mainly from a middle group, with on average something like twelve years of education. We also have some associations that seem to recruit mainly from lower-educated population groups, like trade unions or the Red Cross. Only organizations for women, pensioners and clubs affiliated with local cafes have members with lower educational credentials than those of the population as a whole. One can observe that the average education level is extremely low for the organizations of pensioners, and this is due to the fact that the educational system in Belgium has expanded enormously from the 1950s onwards. Given the fact that the average age of the members of this kind of organization is sixty-six years, they have received fewer educational opportunities during their youth phase than subsequent cohorts. In the previous section we have argued that including "ever" membership (that is, current and previous combined) produces stronger socialization effects than limiting ourselves to a...meas--urement of current memberships. Therefore, we also included in Table 5.3 the average education level of these "ever" members. There is very little difference with the current members; Only for environmental groups does the average drop, while for religious groups it goes up slightly. The next column shows the average score -a -a N "J o y o H ^ I/) N N ui 6 fl\ U) H «' K Ö H N Ifl ni fi & "i (N H K) N + N O O ■o o ON I—1 ~s thirteen years of schooling), like peace, human rights, environment, school and neighborhood organizations. The second category includes organizations with members who have an upper-middle level of education (twelve and a half to thirteen years), like family, art, religious, political and youth organizations. Thirdly, we have organizations with a lower-middle level of education (twelve to twelve and a half years), like sports, caring and social/cultural organizations. The last category includes the organizations with a relatively low level of education (< twelve years) like hobby associations, clubs associated with a local pub, trade unions, the Red Cross and women's organizations. Again, the organizations for old age pensioners have been excluded from the analysis. Membership in of one these four categories has been recoded as a dichotomous variable, and all four have been entered simultaneously. The regression reported in Table 5.4 confirms that, controlling for the education level of the respondent, only the organizations with high and upper-middle average education levels effectively-reduce feelings of ethnocentrism. Organizations with lower-middle or low average education levels do not have a significant effect. One could argue that this analysis does not offer sufficient proof for the claim that value effects of voluntary associations are dependent on member characteristics. Because ethnocentrism is such a politically Voluntary Associations and Democratic Attitudes 105 Table 5.4 Relation among membership, ethnocentrism and political powerlessness Ethnocentrism Political powerlessness B (SE B) P B (SE B) Gender -0.S6 (1.06) -.02 2.12 (1.07)* .06 Age 0.11 (0.04)** .09 0.04 (0.04) .04 Income 0.03 (0.20) .01 -0.22 (0.20) -.04 Education -0.82 (0.20)*** -.15 -0.64(0.21)** -.12 Member organization high -6.35 (1.31)*** -.15 -4.64 (1.32)*** -.11 education level (> 13.0yr) Member organization -3.83(06]**" -.10 -3.14(1.17)** -.09 upper middle education level (12.5-13.0) Member organization -1.53 (1.14) -.04 -0.S6 (1.15) -.03 lower middle education level (12.0-12.5) Member organization low 1.81 (1.11) .05 -0.37 (1.12) -.01 education level (< 12.0yr) Television time 0.22 (0.04)*** .16 0.15 (0.04)*** .12 Church involvement 1.38 (1.15) .04 -5.38 (1.17)*** -.14 Cte. 44.59 (3.92)*** 71.70 (3.99)*** adj. r2 .18 .11 Note: Entries arc nonstandard!zed ordinary least squares regression coefficients (with standard deviation) and standardized regression coefficients. * =?<0.05; ** =p<0.0l; *** = ?< 0.001. Source: Flemish Citizen Survey 1998, w- 1,341. salient issue in Belgium, the results presented thus far could be idiosyncratic. The way people talk about ethnic diversity is seen as politically sensitive in Belgium, and therefore we might expect that this kind of discourse will be subjected to more social pressure than in other countries. Therefore -we-repeated-the same analysis, but this time with a different attitudinal scale, measuring political powerlessness (Table 5.4). Although it is often assumed that all kinds of associations boost feelings of political efficacy, here we observe the same pattern: Only the membership of associations with highly educated members is strongly and significantly related to the scale of political powerlessness. Conclusion Our analysis makes clear that not all voluntary associations automatically produce democratic attitudes, as some proponents of social capital theory would predict. At the aggregate level we observe 106 Marc Hooghe Voluntary Associations and Democratic Attitudes 107 a negative effect of participation in voluntary associations on ethno-centrism, but when we look at it more closely, we see that this effect is not uniform for all kinds of organizations. Only those organizations that, given the average education level of their members, create interaction environments that are hostile to the expression of ethnocentric stereotypes effectively reduce ethnocentrism levels, even after introducing controls for the education level of the respondent. This would indicate that the socialization effects of interaction within voluntary associations are not uniform but are context dependent, as we would expect following self-categorization theory. This implies that not all voluntary associations will actually contribute to the formation of social capital, but only those associations in which a democratic culture is present. This finding also allows us to reconcile two of the main currents of thought within the debate on social capital. Too often it is assumed that processes of self-selection and socialization exclude one another. However, if social identity theory is correct, and if our findings would be confirmed in other research settings, this could imply that voluntary associations too are subject to a "selection and adaptation" dynamic. While members select themselves into an interaction setting, subsequently they adapt to the values that are being upheld in that setting. If socialization really occurs as part of a recursive relation between participation and attitudes, this does not imply that associations do not add anything to a pre-existing average of attitudes and opinions and simply allow a convergence around this average. Research on group polarization processes demonstrates that convergence will occur not on a pre-given mean level but on a more extreme level. The selection and adaptation mechanism implies that actors make a deliberate choice to join an interaction sphere, but that subsequently they are influenced by that sphere. It therefore corresponds to the way de Tocqueville originally described the function of voluntary associations as 'Taction lente et tranquille de la societe sur elle-meme-" (de Tocqueville 1835, 412). The effect of interaction within voluntary associations does not appear as a deus ex machina, as a result of exogenously induced changes, but as an enhancement of previously existing value patterns. Therefore, we have no reason to assume that all voluntary associations in all circumstances will contribute to the formation of the attitudinal components of social capital. What makes voluntary associations an important source of social capital, however, is that in current Western liberal democracies there are many more associations producing social capital than there are producing "unsocial" capital, for example, intolerance, fanaticism or racism. Because of the fact that more civic-minded people are more easily organized than misanthropes, at the aggregate level, their norms will be spread more successfully within civil society than the values of misanthropes. This means that we should not expect each and every association to have a positive effect on the attitudinal components of social capital, but, within contemporary Western liberal democracies, associational life as a whole will be a vehicle to spread pro-social values and will thus function as a source of social capital. Appendix I. Description of the Survey The survey was conducted during the spring of 1998 by the Free University of Brussels. The survey comprised 1,341 face-to-face interviews with inhabitants of the Flemish autonomous region. Respondents were drawn randomly from the official population registers, in the eighteen to seventy-five age bracket. Three lists of respondents were drawn, matching the three samples on age and gender. The response rate was 61 percent, which is average for these kinds of surveys in Belgium. The field work resulted in an overrepre-sentation of younger- and higher-educated respondents (as is usual in these kinds of surveys), which was remedied by introducing weight factors for education, age and gender. II, Ethnocentrism Measurement Scale We used a balanced scale developed by Jaak Biiliet (Billtet, Carton and Eisinga 1995), consisting of five Likert items expressing negative feelings toward outsider groups. Factor analysis shows one factor, with an Eigen value of 2.65 and 53.0 percent explained variance. The scale is internally consistent: Cronbach's a is .78. Items: 1. Generally speaking, migrants cannot be trusted. 2. Migrants contribute to the economic prosperity of the country (reverse coding). 3. Migrants just try to profit from our social security system. 4. i am a racist. 5. The presence of different cultures is an enrichment (reverse coding). III. Political Powerlessness Scale We used a balanced scale of six Likert items. Factor analysis shows one factor with an Eigen value of 3.0 with 50.7 percent explained variance. The scale is internally consistent with a Cronbach's a of .80. Items: 1. Political parties are only interested in my vote, and not in my opinion. 2. If people like me communicate their opinion to politicians, they will take that into account (reverse coding). 3. Most politicians promise a lot, but they don't do a thing. 4. From the moment they are elected, most politicians feel way above people like me. 108 Marc Hooghe Voluntary Associations and Democratic Attitudes 109 5. In fact, there is not a single politician I would trust. 6. In general, we can rely on our political leaders to make such decisions that are best for the population (reverse coding). Notes 1. Maybe the opposite effect is just as likely. One cauuot help of thinking of die French political leader Jean Jaures who, in 1898, after fifteen years in die socialist party (still a voluntary association at that time), exclaimed: "I don't have any trouble with my political opponents. But my political 'friends,' my fellow party men, they could eat me. lis me devormtr (Harvey Goldberg 1970, 258). 2. A counterargument could of course be riiat what we are measuring with this scale is not ethnocentrism as such but ratiier cultural sophistication. Those with higher education "know" tiiat it is culturally desirable to refrain from expressing hostile feelings toward foreigners. Aldiough desirability certainly plays a role in answering patterns, the phenomenon cannot explain why diese patterns are so consistent, both within die edi-nocentrism scale and in the correlation with other scales. If respondents were giving a wrong representation of their true convictions, they at least do this in a very consistent manner, which makes it very improbable that this effect, by itself, could explain the strong correlation we observe with education levels. Furthermore, the desirability problem is present in all of the social capital-related questions used in this type of survey research. It is just as illegitimate to express hostile feelings toward foreigners as it is to say diat people in general cannot be trusted. The social desirability problem is not crucial to our argument: We want to demonstrate convergence processes within groups, and it does not really matter whether these processes take place at the level of values or at the discourse level. References Abrams, D. and M. Hogg (eds., 1991). Social Identity Theory. New York: Springer. Adler, P. and P. Adlcr (1998). Peer Power. Pre adolescent Culture and Identity. New Brunswick: Rutgers University Press. Ajzen, I. and M. Fishbein (1980). Understanding Attitudes and Predicting Social Behavior. Englewood Cliffs: Prentice Hall. Bankston, C. and S. Caldas (1996;. "Majority African American Schools and Social Injustice." Social Forces 1'5(2), 535-555. Bcrman, S. (1997). "Civil Society and Political Institutionalization." American Behavioral Scientist 40(5), 562-574. Billiet, J., A. Carton and R. Eisinga (1995). "Contrasting Effects of Church Involvement on the Dimensions of Ethnocentrism." Social Compass 42(1), 97-108. Brehm, J. and W. Rahn (1997). "Individual Level Evidence for the Causes and Consequences of Social Capital." American Journal of Political Science 41(3), 999-1023. Claibourn, M. and Martin, P. (2000). "Trusting and Joining? An Empirical Test of the Reciprocal Nature of Social Capital." Political Behavior 22(4), 267-291. Dekker, P. (ed., 1999). Vrijwilligerswerk vergeleken. Civil Society en Vrijwilligerswerk III. Den Haag: Sociaal en Cultured Planburcau. Duncan, S. and D. Fiske (1977). Face to Face Interaction. Research, Methods and Theory. Hillsdale: Erlbaum. Forsydi, D. (1983). An Introduction to Group Dynamics. Monterey: Brooks and Cole. Fräser, C. and D. Foster (1984). "Social Groups, Nonsense Groups and Group Polarization." pp. 473-497. In H. Tajfel (ed.), The Social Dimension, volume 2. Cambridge: Cambridge University Press. Goldberg, H. (1970). Jean Jaures. Paris: Fayard. Goslin, D. (ed., 1969). Handbook of Socialization Theory and Research. New York: Rand McNally. Haslam, S. A. et al. (1999). "Social Identity Salience and the Emergence of Stereotype Consensus." Personality and Social Psychology Bulletin 25{7), 809-818. Hendrick, C. (ed., 1987). Group Processes. Newbury Park: Sage. Hooghe, M. (ed., 2000). Sociaal Kapitaal en Democratic. Verenigingsleven, Sociaal Kapitaal en Politické Cultuur. Leuven: Acco. Hooghe, M. (2001). "Borders of Hospitality: The Difficult Birth of a Multicultural Society." The Low Countries. Arts and Society in Flanders and the Netherlands 9, 94-107. Hooghe, M. (2002). "Watching Television and Civic Engagement: Disentangling die Effects of Time, Programs and Stations." Harvard International Journal ofPress/Politics 7(2), 84—104. Hooghe, M. (2003). "Participation in Voluntary Associations and Value Indicators. The Effect of Current and Previous Participation Experiences." Nonprofit and Voluntary'Sector 'Quarterly 32(1), in press. Katz, E. and P. Lazarsfeld (1955). Personal Influence. New York: Free Press. Levi, M. (1996). "Social and Unsocial Capital." Politics and Society 24(1), 45-55. Levine, J. and E. Russo (1987). "Majority and Minority Influence." pp. 13-54. In Clyde Hendrick (ed.), Group Processes. Newbury Park: Sage. Lipsct, S.M. (1959). "Democracy and Working-Class Authoritarianism." American Sociological Review 24, 482-502. Lubbers, M., P. Scheepers and J. Billiet (2000). "Multilevel Modelling of Vlaams Blok Voting." Acta Politica 35(4), 363-398. Merten, R. and A. Kitt (1950). "Contributions to the Theory of Reference Group Behavior." In R. Mertori and P. Lazarsfeld (eds.), Continuities in Social Research. GJcncoe: Free Press. 110 Marc Hooühe Voluntary Associations and Democratic Attitudes 111 Mills, T. and S. Rosenberg (cds., 1970). Readings on the Sociology of Small Groups. Englewood Cliffs: Prentice Hall. Mondak, J. and D. Mutz (1997). "What's so great about League Bowling?" Taper presented at the Annual Meeting of the Midwest Political Science Association, Chicago. Newton, K. (1997). "Social Capital and Democracy" American Behavioral Scientist 40{S), 575-5S6. Newton, K. (1999). "Social Capital and Democracy in Modern Europe." pp. 3-24, In J. van Deth et ai. (eds.), Social Capital and European Democracy- London: Routiedge. Nie, N., J. Junn and K. Stehlik-Barry (1996). Education and Democratic Citizenship in America. Chicago: University of Chicago Press. Paulus, P. (ed., 1989). Psychology of Group Influence. Hillsdale: Erlbaum. Putnam, R. (1993). Making Democracy Work. Princeton: Princeton University Press. Putnam, R. (1995). ''Tuning In, Tuning Out. The Strange Disappearance of Social Capital in America." PS: Political Science and Politics28(4), 664-683. Putnam, R. (2000). Bowling Alone. Tlie Collapse and Revival of American Community. New York: Simon and Schuster. Robinson, W. P. (ed., 1996). Social Groups and Identity. Oxford: Butterwor th - Heinemann. Skinner, M. and G. Stephenson (1981). "The Effects of Intergroup Comparison on the Polarization of Opinions." Current Psychological Research 1(1), 49-61. Soysal, Y. (1994). Limits of Citizenship. Migrants and Postnational Membership in Europe. Chicago: University of Chicago Press. Stolle, D. (1998). "Bowling Alone, Bowling Together. Group Characteristics, Membership and Social Capital." Political Psychology 19(3), 497-526. Stofle, D. (2000). Communities of Trust. Public Action and Social Capital in Comparative Perspective. Ph.D. Dissertation, Princeton University. Stolle, D. (2002). "Social Capital—An Emerging Concept." In B. Hobson, J. Lewis 3nd B. Siim (eds.), Key Concepts in Gender and European Social Politics. Cheltenham: Edward Elgar. Stolle, D. and M. Hooghe (2002). "The Roots of Social Capital. The Effect of Youth Experiences on Participation and Value Patterns in Adult Life." Paper presented at the 98th Annual Meeting of the American Political Science Association, Boston, August 29-September 1, 2002. Stolle, D. and T. Rochon (1998). "Are All Associations Alike?" American Behavioral Scientist 42(2), 47-65. Tajfel, H. (1981). Human Groups and Social Categories. Cambridge: Cambridge University Press. Tocqueville, A. de (1835). De la Democratic en Ameriqm. Paris: Gallimard/ Pleiade, 1992. Turner, J. (1982). "Toward a Cognitive Redefinition of the Social Group." pp. 15-40. In H. Tajfel (ed.), Social Identity and Intergroup Relations. Cambridge: Cambridge University Press. I Turner, J. (1984). "Social Identification and Psychological. Group I Formation." pp. 518-538. In H. Tajfel (ed.), The Social Dimension, I volume 2. Cambridge: Cambridge University Press. j Turner, J. and P. Oakes (1989), "Self-Categorization Theory and Social I Influence." pp. 233-275. In P. Paulus (ed.), Psychology of Group Influence. 1 Hillsdale: Erlbaum. Turner, J. et al. (1987). Rediscovering the Social Group. Oxford: Basil Blacfcwcll. Uslaner, E. (1998). "Social Capital, Television and the 'Mean World.'" Political Psychology 19(3), 441-467. Uslaner, E. (2000). "Producing and Consuming Trust." Political Science Quarterly 115(4), 569-590. Verba, S. (1961). Small Groups and Political Behavior. Princeton: Princeton University Press. Verba, S. and N. Nie (1972). Participation in America. New York: Harper and Row. Verba, S., K. L. Schlozman and H. Brady (1995). Voice and Equality. Civic Voluntarism in American Politics. Cambridge: Harvard University Press. Whiteley, P. (1999). "The Origins of Social Capital." pp. 25-44. In J. van Deth et al. (eds.), Social Capital and European Democracy. London: Routiedge. Williams, T. (ed., 1975). Socialization and Communication in Primary Groups. The Hague: Mouton. Yamagishi, T. and M. Yamagishi (1994). "Trust and Commitment in the United States and Japan." Motivation and Emotion 18(2), 129-166.