SOCIAL CAPITAL IN A MULTICULTURAL SOCIETY:THE CASE OF CANADA

Richard Johnston

and

Stuart Soroka

University of British Columbia

May 1999

Paper prepared for delivery at the annual meeting of the Canadian Political ScienceAssociation, Sherbrooke, Quebec, June 1999. Data analysed in this paper are from theCanadian World Values Survey, Neil Nevitte and Ronald Inglehart, PrincipalInvestigators, data collection by Gallup Canada. Our analysis was supported by theEquality, Security, Community Project at the University of British Columbia, JonKesselman, Principal Investigator, under Social Sciences and Humanities ResearchCouncil of Canada Major Collaborative Research Initiative 412-97-0003. Commentsfrom the Project Group meeting in October 1998, especially from Keith Banting, werevery helpful. None of these institutions or individuals bears any responsibility for errorsor omissions in the final product.

1

SOCIAL CAPITAL IN A MULTICULTURAL SOCIETY:THE CASE OF CANADA

Is diversity the enemy of civil society? Some formulations of the notion of social capitalimply an affirmative answer to this question, an answer that shows the dark side of socialintegration and trust. By our reading, diversity can be argued to undermine the delicatefabric of civil society in two ways.

The first is compositional. Some groups are more ÒcivicÓ than others, more prone toparticipate in social life, more trusting of each other and, perhaps, of strangers. Thesegroup differences endure transoceanic migration and the passage of time and generations.If groups that are less ÒcivicÓ by this standard move into turf formerly dominated bymore civic groups, the overall balance in social capital account may shrink. This claim isrooted in history, of course, so we can well imagine a low-civil-society place beingenriched by the influx of new, more mobilised groups. A strong form of the argument,however, says that getting a settler society off the ground requires high levels of civiccommitment. Only when the ground has been broken can less civic groups flourish.

The second argument is contextual. This says that rendering a place more heterogeneousmakes hitherto dominant groups themselves less civic, whatever their earlier degree ofcivic commitment. If heterogeneity inhibits older groupsÕ identification with theevolving community, they may withdraw, behaviourally and psychologically, from itsinstitutions.

Both arguments are relevant to Canada, and Canada is arguably the prime site for testingthem. For several years, Canada has maintained a relatively open-door immigrationpolicy. Moreover, the country has become ethnically more diverse and has done so inwaves. Yet the country also harbours long-standing struggles among ethnically distinctÒfounding peoplesÓ, national minorities in KymlickaÕs (1995) terminology.

Both arguments assume that social capitalÕs social psychological components -- trust andother civic orientations -- generalise, that learning, positive or negative, in one domaintransfers to others. Interpersonal trust can be cashed in on support for the political systemor the nationality. Conversely, policies that keep groups apart locally -- confessionalschools anciently, multiculturalism currently -- may undermine the cohesion of thenationality or the political community. If this is true, Canadians may exhibit a limitedand conditional sense of community, and certain groups may do so more than others. Butclaims for generalisation across civic domains are mostly supposition. This paperexamines the issue directly.

The Arguments Expanded

Social Capital - General

Social capital means two different things and we consider both. The first readingemphasises connections, or networks. These can be primary ties of kin or neighbours,but the political science literature seems to place special emphasis on formalmemberships in secondary groups. Connections can obviously be important resources forindividual or family welfare, and this seems to be the social-capital emphasis amongsociologists (Coleman, 1990). For political scientists and economists, connections

2

betoken the frequency of interaction, which affects in turn the likelihood that collectiveaction problems will be solved. Secondly, social capital is a psychological property.Here it appears as trust, first in other private citizens, then in institutions, and finally incommunities, possibly including the nationality. This paper, then, considers membershipand civic orientations.

The Compositional Argument

The compositional argument sees both kinds of social capital as cultural attributes, notnecessarily difficult to shed but certainly hard to acquire. This is the view in PutnamÕscelebrated Making Democracy Work (1993). According to Putnam, the performance ofItalian regional governments reflects more the density of social capital than the level ofeconomic development in the region. And present-day inter-regional differences in socialcapital echo old patterns: ÒOne could have predicted the success or failure of regionalgovernment in Italy in the 1980s with extraordinary accuracy from patterns of civicengagement nearly a century earlierÓ (150).

PutnamÕs argument is reminiscent of BanfieldÕs (1958) work on the relationship betweeneconomic/political performance and culture. Banfield characterised the southern Italianculture as Òamoral familism,Ó and went on to speculate that Òthere is some reason todoubt that the non-Western cultures of the world will prove capable of creating andmaintaining the high degree of organization without which a modern economy anddemocratic political order are impossibleÓ (8). Drawing such a conclusion at the start ofthe 1960s, the Òdevelopment decade,Ó was unwelcome, but for Italy at least Putnamseems to bear Banfield out.

And the idea is spreading. This congenital view of social capital exists explicitly orimplicitly in much recent work linking social capital with either political or economicperformance. InglehartÕs (1988, 1990) work on the relationship between civic cultureand countriesÕ years of democracy stands as one well-known example.1 Similarly,HelliwellÕs (1996b) proposal that different growth rates in Asian economies may belinked to variations in social capital is implicitly premised on this model. Given the lackof data, Helliwell is prudent in drawing conclusions. No such prudence restrainsFukuyama (1995). According to Fukuyama, the centrality of the family in Chineseculture explains why that country has, on one hand, a mercantile history but, on the other,has been unable to develop large, economically successful corporations. The Japanese,conversely, have typically developed close ties beyond the family, and this helps buildthe social capital required for industrial capitalism. FukuyamaÕs cultural explanation isonly loosely tied to evidence, and his disregard for the impact of the countriesÕ laws on

1 For recent work on the relationship between social capital and democracy, see Boix andPosner (1998); Booth and Richard (1998); Muller and Seligson (1994); Newton (1997);Nichols (1996); Rice and Sumberg (1997)

3

economic development has drawn criticism (Fellmeth, 1996). Nevertheless, his argumentexemplifies a compositional model.2

The compositional argument takes a special form for societies built by large-scaleimmigration. Although Verba, Schlozman, and Brady (1995) tend not to employ social-capital terminology, they nonetheless find abiding ethnoreligious differences in civicengagement. In comparison with Anglo-Whites, African Americans are slightly lessactive and Latinos are significantly less active in both political and non-politicalorganisations (with the exception of religious groups and church attendance). That thesepatterns reflect abiding differences in Òold-countryÓ patterns is strongly suggested byRice and Feldman (1997), who find that US respondentsÕ scores across a number of civicculture variables are strongly correlated with the scores of respondents from their countryof origin. Living in the US has only a mild homogenising effect on respondentsÕ civicculture, as significant origin-related differences remain, even controlling otherdemographic variables. Rice and FeldmanÕs findings correspond to HelliwellÕs (1996a)identification of Canadian and American inter-regional differences in social capital.Helliwell suggests that the increase in social capital from the south to north-central US,and from east to west in Canada, may be a product of the inhabitantsÕ countries of origin.The high civic scores in the north-central US, for instance, may be due in large part to thesizeable Scandinavian population.

In the Canadian case, Black (1987) found in a four-group immigrant-native matched-sample design that group differences in civic engagement overwhelmed immigrant-nativeones, a classic compositional pattern. And one of NevitteÕs (1996) models of culturaltransformation was explicitly compositional: he asked if new orientations (such as thedecline of deference) were, first, more common among new Canadians and, second,propagated by the increasing proportion of new Canadians on the demographiclandscape. The problem is that he had not read his Black: he appears to have definedÒnew CanadiansÓ as immigrants, seemingly the wrong way to proceed.

Immigrant status may stand as a structural factor, however, as an index of recency ofarrival. The less time someone has spent in Canada, the harder it may be to negotiate thechannels of participation and, correspondingly, the less opportunity there may have beento size people up. The scholarly record casts doubt on this proposition. Black (1982)found that, other things equal, newly naturalised citizens Òoverparticipated,Ó thatimmigrant adaptation was almost instantaneous.

The Contextual Argument

The contextual model of social capital is more shadowy, and requires identification bytriangulation. An elegant statement can be found in Miller (1995). Miller is concerned tosalvage the idea of nationality for liberal theory, which is normally hostile toparticularistic and history-based claims. He argues that liberals, especially on the left,should accept that for many people identification with fellow citizens facilitates solutions

2 For recent work on the relationship between economic performance and social capital,see Helliwell and Putnam (1995); Kenworthy (1997); Knack and Keefer (1997); Templeand Johnston (1998).

4

to the aforementioned collective-action problems. Love of country helps underpin thewelfare state, for example. But countries that are ethnically relatively homogeneous are,he argues, easier to love, and so he is sceptical of the prospects for a place like Canada.In this part of what is generally a philosophical analysis, Miller stakes an empirical claim.

Unfortunately, he has no evidence. And evidence directly on the question is thin to non-existent. But two literatures bracket the claim, so to speak. One looks at contact andethnic prejudice. The other looks at the relationship between diversity and the size of thestate, particularly the welfare state.

On contact and ethnic prejudice, ForbesÕ (1997) seems to be the definitive review of therelevant evidence. Although in any setting individuals who regularly come in contactwith members of other ethnic groups are less prejudiced than those who keep apart, therelationship is the opposite in the aggregate. The more likely such contact is, the greaterthe average level of prejudice. As contact threatens groupsÕ distinctiveness, greaterinvestments are made in maintaining that distinctiveness.3 And the likelihood of inter-group contact is greater the more ethnically diverse is the setting. Forbes review appearsto substantiate LehningÕs (1998) contention that, Òthe greater the number and diversity ofpersons in a group, the more that universalistic norms require altruism, and yet -- at thesame time -- the weaker the force of altruismÓ (238).

At the other end of the chain of evidence is research on the relationship between ethnicdiversity and public spending. For instance, McCarty (1993) finds that countries withhigher levels of ethnic diversity tend to spend a lower proportion of their GDP on socialsecurity and welfare programmes. Alesina, et al. (1997) also find an inverse relationshipbetween ethnic fragmentation and public spending on both welfare and public goodsamong US states.4

Is the density of social capital the key intervening variable, as Miller argues? Putnam(1995b), for one, resists such suggestions. He rejects the idea that the US civil rightsrevolution precipitated a Òwhite flightÓ from group life. Group memberships havedeclined evenly for blacks, pro-segregationist whites, and anti-segregationist whites,suggesting to Putnam that racial explanations cannot account for the decline in Americansocial capital (672-3). It is not clear that this pattern goes to the issue raised by logic of acontextual model. And Helliwell (1996a) could be marshalled for the other side. He findsthat respondents who identify themselves as belonging to a particular ethnic group -- asopposed to simply ÒAmericanÓ or ÒCanadianÓ -- score lower on social capital measures,suggesting that, Òtrust levels are lower where perceived cultural cleavages are strongerÓ(11). Neither reference is clearly on the question, however, so we shall have to look forourselves.

3 This is not to deny that protracted contact does break down differences. It is preciselyforeknowledge of this fact that motivates attempts to keep groups separate.4 On a related but different tack, Easterly and Levine (1997) find a significantrelationship between ethnic diversity and negative economic growth in sub-SaharanAfrica. Ethnic diversity, they suggest, Ò(a) encourages the adoption of growth-retardingpolicies that foster rent-seeking behaviour and (b) makes it more difficult to form aconsensus for growth-promoting public goodsÓ (1207).

5

Linkages Among Civic Orientations

Most of the time, social capital is treated as essentially homogeneous. Each specificelement-- civic engagement, interpersonal trust, political trust, love of country -- bearsthe same functional relationship to causal factors as each other one. Orientationsgeneralise, presumably from the most particular, trust in other individuals, on up. Theinterpersonal trust that greases the wheels of economic transactions is highly similar tothe trust in political institutions that might be a precondition for supporting the welfarestate.5

But should we assume that civic orientations generalise so easily? Causal relationshipsamong orientations may be asymmetric. Indeed, interpersonal trust may be contingent onattributes of the larger society. Brehm and Rahn (1997), for instance, find with US datathat confidence in government is powerfully implicated in interpersonal trust, not theopposite. Similarly, Muller and Seligson (1994) find that a countryÕs years of experiencewith democracy is a powerful predictor of its average score on the interpersonal trustmeasure.6 Both findings echo Levi (1996) and Rothstein (1998), who argue that justicein overarching institutions helps unlock cooperative possibilities in local settings.

The possibility of most immediate relevance to Canada is that each object of orientationattracts its own functional relationships. Where the political community is divided,groups or subcommunities which are internally cohesive may nonetheless see the state asan instrument of other groups and so reject its claims. Then again, the claims of the stateand the claims of political nationality may be quite distinct. Johnston (1986) found, forinstance, that Quebeckers expressed more confidence than others in the nationalgovernment even as they identified less with Canada as a national community.

Data and Measures

What orientations? What groups? What contexts? We use data from the Canadian WorldValues Survey (WVS), 1991.7 The WVS has strong representation for civic orientationsand for memberships, and for this reason has already performed yeoman service in social-capital analyses.8 Its weakness is in identification of groups.

Social Capital

Memberships are measured here by WVS questions about sixteen different types oforganisations, from political parties to social welfare groups to professional associations.

5 Trust in individuals is also likely to be critical to welfare state support, of course, to theextent that it reduces subjective estimates of cheating. See Rothstein (1998).6 Brehm and Rahn analyse the 1972-94 General Social Survey Cumulative File. Mullerand Seligson use World Values Survey data, augmented by Central American surveysand non-survey data sets with political and economic variables, to a mount a pointedcritique of Inglehart (1988).7 The Canadian WVS is described in more detail in Nevitte (1996). Exact questionwording and coding for this paper is recorded in Appendix A.8 Helliwell (1996a) is a case in point.

6

Unfortunately, the number of memberships in any given category was not recorded --only whether a respondent was or was not a member in any number of organisations of agiven type. The total memberships variable used here, then, is a sum total of the numberof different types of memberships to which a respondent belongs.9 It ranges from 0 to 16.

Six items correspond to the civic orientations of obvious interest: (1) interpersonal trust;(2) trust in government; (3) confidence in parliament; (4) confidence in the Canadianpolitical system; (5) national pride; and (6) willingness to fight in a war. Do the itemshang together, or are there clearly distinguishable subtypes? The simplest way to answerthis is to enter the six indicators into a factor analysis, whose results appear in Table 1.

Three distinct dimensions are required to account for civic orientation:10

• Interpersonal trust -- ÒGenerally speaking, would you say that most people can betrusted, or that you canÕt be too careful in dealing with people?Ó -- stands by itself.This is very hard to square with the notion that people reason from personalrelationships to more distant ones.

• The second factor makes sense logically and empirically, as it includes the item onÒtrustÓ in government and the items gauging ÒconfidenceÓ in Parliament and theCanadian political system. These are combined as an indicator of political trust.

• The third factor includes the two indicators of commitment to Canada as a country,pride in being Canadian and willingness to fight in the event of war. These two itemsform an index of national pride.

Each civic orientation measure is rescaled to the 0,1 interval.

Groups

The compositional hypothesis requires grouping by ethnicity, ideally by ethnic origin ofthe earliest North American ancestors, the standard census item. Unfortunately, the WVSemploys a very simple code, apparently with cross-national comparison in mind. Westart with a base question which asks respondents to designate themselves as French orEnglish Canadian, as an ÒEthnicÓ Canadian, Canadian first and then some other group, orCanadian first and only. We then combine this question with a racial imputation byinterviewers and language of interview to create four categories:

• French are either self-designated as such, or were interviewed in that language. Thelanguage of interview was included because many Francophones did not identify

9 In preliminary analyses we tried to develop a classification of associations, to seewhether ethnic groups or regions were particularly participant in certain types ofassociation. Results were weak, partly, we suspect, because the WVS categories areinternally heterogeneous.10 While Table 1 presents factor analysis results using a Varimax rotation, exactly thesame components emerge with an oblique rotation.

7

themselves as French Canadians.11 97% of these respondents were Canadian-born,and 82% were interviewed in Quebec.

• Non-White is based strictly on interviewer coding. 5.5% of the respondents werecoded as something other than Òwhite.Ó Slightly more than 70% of Ònon-whiteÓrespondents were born outside Canada -- by far the highest proportion of the fourethnic groups -- and these foreign-born respondents tend also to be relatively recentimmigrants. The majority of these respondents are either East or South Asian, withBlacks and Latinos as other small minorities. They, along with ÒWhite EthnicÓ (seenext paragraph) respondents, tend to be slightly better educated than average.

• Respondents coded as White Ethnic were coded ÒwhiteÓ by interviewers, butdesignated themselves as either ÒEthnic CanadianÓ or ÒCanadian first and then amember of an ethnic group.Ó This category includes slightly over 15% of therespondents. Immigrants constitute 37% this group, although most immigrants arrivedmore than 15 years before the survey was conducted.

• White Non-Ethnic is the residual category. These respondents were coded ÒwhiteÓ byinterviewers, identified themselves as either ÒEnglish CanadianÓ or ÒCanadian firstand only,Ó and answered surveys in English. About 88% were born in Canada, andthe majority of the foreign-born immigrated over fifteen years before the study.

Sometimes analysis employs this four-category classification. Sometimes the first threecategories appear as dummy variables, with the fourth (Òwhite non-ethnicÓ) as thereference category.

Context

The best we can do for ethnic context in the WVS is province of residence. We employthe standard five-region setup: Atlantic, Quebec, Ontario, Prairies, and British Columbia.If contextual arguments hold any water in Canada, Ontario and BC should stand out.Whatever the ethnic-group differences on social capital, the social diversity of these twoprovinces should give them lower readings than in a straight adding-up based on theirethnic composition. When regions appear as dummy variables, Ontario is the referencecategory. The coefficient on the dummy variable for each other region indicates thedifference between Ontario and that region. If the compositional model is borne out, allregional coefficients in an estimation that includes ethnic variables should be positive,except possibly for BC.

Results

We look at the data in two ways. First is a simple two-way comparison of membershipand civic-orientation means, simultaneously by ethnicity and region. This establishesbasic patterns, to see if there is anything to either compositional or contextual claims.The second stage takes the most arresting patterns from the first stage and attempts toaccount for them, by stage-wise controls in a regression setup. Discussion of the content

11 This measurement difficulty illustrates both the pitfalls of comparative survey researcheven as it foreshadows a substantive argument below: offering francophone Qu�b�coisÒFrench CanadianÓ as the primary mode of differentiation is a bad idea.

8

of controls will be postponed for now. For the moment suffice it to say that stagewiseregressions test whether the group or regional differences identified in the first round areintrinsic -- truly embedded in the groupÕs or regionÕs culture -- or incidental -- theproduct of group or region compositional differences on some other variable. The othervariable may also be cultural (e.g., religious affiliation) or it may not (e.g., educationalattainment or job status).

Identifying the Patterns

Table 2 compares means, with ÒlowÓ and ÒhighÓ groups highlighted.12 Group andregional differences appear in each domain, but differences among domains are asstriking as differences within each. Membership and interpersonal trust appear to form asyndrome, much as discussed in the social capital literature. But for neither does thegroup-region pattern generalise to political trust, and the political trust pattern is notreflected in national pride. And for none of these domains does the usual contextualprediction find even prima facie support.

For both membership (Table 2A) and interpersonal trust (Table 2B) the sharpest contrastis not between ÒoldÓ and ÒnewÓ Canadians but between the newest Canadians and the notquite so new. ÒNon-white ethnicÓ respondents (most of whom are immigrants) exhibitthe fewest memberships on average, while Òwhite ethnicÓ respondents claim the most.Roughly the same is true for interpersonal trust, although here Ònon-white ethnicÓrespondents are not the lowest but the second-lowest group. Just to confuse matters, onboth indicators the closest companions to Ònon-white ethnicÓ respondents areFrancophones.13

If the structural correspondence between memberships and interpersonal trust shouldreassure social capital theorists,14 for neither indicator is diversity of context obviouslyproblematic. The low-end provinces are Quebec and the Atlantic region. QuebecÕsposition simply reflects the provinceÕs majority-Francophone composition. The lowscore for the Atlantic provinces is clearly not the result of the regionÕs ethnic makeup, as

12 Highlighting does not necessarily imply a statistically significant difference. Wherethe significance of a contrast, in the analysis-or-variance sense, is at issue, we discuss itin a footnote.13 On memberships, Ònon-white ethnicÓ respondents are significantly lower than all othergroups, Òwhite ethnicsÓ are significantly than higher than all others, and French and ÒAllOthersÓ are statistically indistinguishable. On interpersonal trust, francophone and non-white ethnic respondents are indistinguishable from each other. ÒAll OthersÓ aresignificantly different from both groups, but are also significantly from white ethnicrespondents, the high-end outlier. Contrasts are extracted from a general factorialanalysis of variance; confidence intervals are not adjusted for multiple comparisons.Such adjustment (e.g. a Bonferroni or Scheff� correction) tends to mute differences.14 PutnamÕs (1993, 1995a, 1995b) discussions of memberships as a component of socialcapital, for instance, are premised on the notion that there are strong links betweenmemberships and trust.

9

attested by the patterns in the ÒAll OthersÓ column. If there is something intrinsic to theAtlantic-provinces context, it cannot be backlash against diversity.

The pattern for political trust is much simpler. Only one group and one province standout. ÒNon-white ethnicÓ respondents trust the political system the most; other groups areessentially indistinguishable. Among provinces, Quebec has the highest score.15 Bear inmind what the political trust items refer to: parliament, the Canadian political system,and politicians in Ottawa . Respondents are not invited to assess some abstractdecontextualised entity, but the government of Canada. That Quebeckers should be themost supportive is striking but hardly novel, as the 1991 WVS pattern essentiallyreproduces the finding in Johnston (1986, 29-33).

National pride evokes yet another structure. ÒNon-white ethnicÓ respondents again seemthe most proud to be Canadian, but only marginally so. What really stands out is thatfrancophone Qu�b�cois are the least committed to the Canadian nationality. This is notsurprising in itself, given sovereignist sentiment in Qu�bec. It is jarring, though, whenjuxtaposed to the pattern for political trust. Again, however, the WVS pattern onlyconfirms JohnstonÕs (1986, 44-6) finding.

For neither political trust nor national pride does the contextual model hold water. On thefirst orientation, the two most diverse provinces are the most trusting, apart from Quebec.On the second orientation, no interpretable pattern appears, again Quebec aside.

Explaining the Patterns

Given the differences that do stand out, can we explain them, or explain them away?Table 3 condenses output from stepwise regressions aimed at investigating thepersistence of noteworthy initial differences in the face of controls. Each independentvariable is a 0,1 or ÒdummyÓ variable, indicating membership in the named category.16

The leftmost column in the body of the table presents coefficients on groups or regions ofinterest. For example, in the memberships panel, only the two Òethnic CanadianÓcoefficients appear, as only these ethnic groups stood out in Table 2. In the estimationsthat underpin Table 3 a complete set of ethnic and region dummies was entered;coefficients of no interest are suppressed for readability.

The coefficients in the first column essentially reproduce the Table 2 pattern. The rest ofcolumns address the question of why that pattern appeared in the first place. The

15 Pairwise comparisons, on the model described in note 13, confirm that these are theonly significantly distinct categories.16 A dummy variable is scored 1 for possession a particular trait, and 0 otherwise. AÒPrairiesÓ dummy variable, for instance, is scored 1 for respondents from PrairiesProvinces and 0 for all other respondents. Dummies are created for all but one value on agiven dimension -- regional dummies, for example, were created for all five regionsexcept Ontario. Ontario then serves as the point of reference -- a coefficient of 0.10 forthe ÒPrairiesÓ dummy indicates that residents of that region score 0.10 higher on thedependent variable than do Ontario residents. For ethnicity, the reference category is ÒAllOthers.Ó One exception to the dummy-variable rule is the membership indicator, which isdiscussed further below.

10

coefficients in columns further to the right are for the same variables as in the firstcolumn. The issue is whether and how much a coefficient shrinks as plausible controlsare added. For example if what really distinguishes Ònon-white ethnicÓ respondents isthat so many are immigrants, the coefficient on the ethnic variable should shrink whenplace of birth is controlled. If no such shrinkage occurs, then immigrant status fails as anexplanatory factor. And so on. Variables are added cumulatively, so each consecutivecolumn includes the new variables along with all the previous ones. The adjusted R2 alsoappears at each stage as an interpretative tool.

Now, what are the controls?

• Membership plays a dual role. The number of memberships is an aspect of socialcapital in its own right and so is treated as a dependent variable. But membershipalso commonly appears as a causal factor in civic orientations, especiallyinterpersonal trust. So the number of memberships is the first control in each ofPanels B, C, and D. And once membership enters, its own coefficient in the equationis recorded, so that the evolution of membershipÕs ceteris paribus effect on theorientation as other controls are added can also be monitored.

• Immigrant status allows us to gauge how much of what appears to be an ethnicdifference is really the product of recency of arrival. Similarly, it tests how much aregionÕs distinctiveness is simply the by-product of its immigrant-native makeup.

• Religious variables test a rival cultural hypothesis, that differences as expressed aremainly the result of ethnic differences in religious composition. An obvious examplewould be French-Other or Quebec-Other contrasts, where a Catholic/Non-Catholicdifference is obviously implicated.

• For our purposes, education is a structural rather than a cultural factor. Participationin voluntary associations is assisted by possession of skills that the education systemconfers or identifies. Alternatively, education may encourage identification with thesystem. Or it may index life chances, which in turn affect willingness to volunteerand optimism about neighbourhood and polity. If any or all of these forces are atwork, what appear initially as group or regional differences may only be proxies fordifferences in educational attainment.

• Job status has similar implications to education.

• Age has two obvious roles to play. Life experience confers skills, and so olderrespondents may be more participant and more civic generally than younger ones, upto a point. By treating age as a set of dummy variables we allow for a change of signin later years. But age can also signal a generational shift. Inglehart and Nevitte,among others, alert us to this possibility. If a generational change is underway, groupswhose members are younger or older, on average, than others may appear distinct. Ifeither process operates, group distinctiveness may not inhere in the group, but mightbe only a by-product of demography.

For a complete description of the variables, see the Appendix.

11

For memberships, stagewise analysis produces sharply contrasting Òwhite ethnicÓ andÒnon-white ethnicÓ stories.17 For the former, no controls make the original difference goaway. The coefficient is just as big in the rightmost column as in the leftmost. The onlyshift of any note is in the ÒwrongÓ direction: the coefficient actually grows when religionis controlled, only to be dragged back by education controls. This group participatesheavily even though its members affiliate disproportionately with Catholicism, adenomination that inhibits group membership.18

Religion is similarly part of the story for Ònon-white ethnicÓ respondents. Theserespondents report the lowest rate of denominational affiliation and the affiliations thatare reported are disproportionately outside the Christian tradition. Both kinds ofaffiliation depress group membership even more than Catholicism does.

A big part of the Ònon-whiteÓ story is purely demographic, however, and casts doubt onthe interpretability of the original finding. Once age is controlled, the Ònon-whiteÓcoefficient effectively disappears; it is still negative but is now only two-thirds the size ofits standard error. ÒNon-whitesÓ are disproportionately young, and youth is the enemy ofcivic engagement. This may be a generational pattern, but it may also track sociallearning over the life course. Either way, it indicates that a big part of the originaldifference is not cultural at all.19

The story for interpersonal trust has many of the same features. Age controls reduce theÒnon-white ethnicÓ coefficient, although the power of controls to explain the originaldifference is more modest here than with memberships. The other ethnic difference,between Francophones and others, is sharply reduced. The key control is for religiousidentification, as the Catholicism of Francophones is clearly a factor in their lower levelof interpersonal trust.

The strong form of the compositional argument says that newcomers subtract from socialcapital just by their presence. This is clearly not true. ÒWhite ethnics,Ó adisproportionately immigrant group overparticipate and exhibit high levels of

17 Coefficients in this panel are bigger than in the other three purely as an artifact: wherein the other three panels the range on the dependent variable is 0,1 -- here it is 0,16.18 This group is about twice as likely as any other non-Francophone group to be Catholic.The low participation rate of Catholics is one of the clearest patterns in Verba, et al.(1995).19 The 1996 Census shows that just under 22% of Ònon-whitesÓ are between the ages of15 and 24; the corresponding share in our sample Ð which excludes respondents under 18years of age, and so understates the bias Ð is almost 38%. This discrepancy explains whyage accounts for a large part of the Ònon-whiteÓ difference.Keep in mind, however, that religion also lowered the coefficient. And the order inwhich controls enter affects their apparent power. If we control age first, we knock over0.10 off the initial coefficient. If we then control religion we knock off another 0.14.The fact that religion plays such a role indicates that part of the story is truly in thecultural domain. Note, however, that the biggest jump in overall explanatory powercomes when structural factors, starting with education, enter.

12

interpersonal trust. ÒNonwhite ethnicsÓ exhibit high levels of commitment to the nationalcommunity and to national institutions, and their apparent underparticipation in CanadaÕsassociational life is in large part a demographic accident. For neither group is immigrantstatus itself a factor. The general picture here confirms a pattern identified by Black(1982), that immigrant adaptation to Canadian political life is very rapid. What makesthis so, we suspect, are the very multicultural programmes commonly blamed for keepingnew Canadians apart.

The relatively high trust levels of the two Western regions simply remain unexplained. Ifthis were a product of ethnic composition, the ceteris paribus coefficients would not beso clearly positive in the very first estimation, and no control cuts the coefficient foreither region. For BC in particular, this is an embarrassment for a contextual argument.

Finally consider the pattern for memberships. Adding the membership variable (secondstage) reduces not one of the ethnic or regional coefficients in the lines above, eventhough memberships themselves vary by region and ethnicity. The impact ofmemberships, although clearly non-zero in the early stages, is modest. The coefficient issmall20 and adding the membership terms improves the overall explanatory power of theequation only slightly, much less impressively than does the later addition of structuralvariables. And once structural variables enter, the membership coefficient disappears.This pattern, together with evidence from the membership estimations in Panel A,suggests that associational membership and interpersonal trust do in a sense partake of asyndrome. But membership is not persuasive as an intervening variable on the road fromstructural factors to trust. Rather each of memberships and trust is an outcome in its ownright, separately manifesting a property we might call social capital, but neither sits in aconvincing explanatory relationship directly with the other.21

For political trust and national pride there is little to say. Membership hardly affectseither civic orientation, and, obviously, cannot help explain the original ethnic or regionalpatterns away. But then, neither does any other factor, as gaps are just as wide at the endof the exercise as at the beginning.

Discussion

The answer to the opening question, at least for Canada, must be No: diversity is notobviously the enemy of social capital. Diversity of the provincial context does not seemto induce backlash among older groups. Ethnic differences in social capital do exist andusually cannot be banished by statistical control, but recency of arrival and cultural

20 We need to be careful in interpreting a coefficientÕs absolute size, of course. In themembership estimations ( Panel A), dummy variables on the right-hand were affecting a0,16 variable on the left-hand side; this inflates coefficients. In Panels B through C, hememberships line involves a 0,16 variable moving outcomes in the 0,1 range; thisdeflates coefficients. But practically speaking, most movement in membership numberswas over a small range of association types.21 This is in contrast to the Brehm-Rahn finding for a quite powerful effect ofmemberships on trust, notwithstanding controls. Of course, we also do not allow foreffects in the other direction.

13

distance from the traditional ÒcoreÓ Canadian ethnicities do not track linearly into weakassociational involvement or into uncivic attitudes. We discuss these in reverse order.

The contextual argument -- the more diverse a place, the smaller its stock of social capital-- is simply not borne out. The most diverse provinces are never the most uncivic places.Indeed the reverse is more nearly true, and associational participation and interpersonaltrust tend to go up as one goes west.22 It is conceivable that an analysis that capturesindividualsÕ ethnicity more precisely than is possible with the WVS will also allow amore precise estimation of contextual effects, and change the picture. But the refutationin these data is so resounding that we doubt it.

The compositional argument hardly fares better. Within our crude ethnic categorisation,differences do stand out and almost all resist statistical control: as far as we can see, thedifferences deserve to be called cultural. Most patterns resist ready interpretation andthose which are interpretable also answer the opening question in the negative. It is truethat the group with the largest percentage immigrant is the least involved and the leastinterpersonally trusting. But a big fraction of the initial difference disappears withdemographic controls, a reflection of the non-white subsampleÕs relative youth. Thegroup with the next largest immigrant component is the most participant and trusting.This group, Òwhite ethnics,Ó is not dominated by archetypal Scandinavian cooperative-movement activists. In fact, the group is disproportionately Catholic, and so exhibits therelevant traits arguably in spite of its makeup.

The issue is even more controverted when we get beyond the core universals of the socialcapital model. Neither political trust nor national pride is clearly related to interpersonaltrust, nor are they much linked to each other. New Canadians are the one group whichstands out positively for both political trust and for national pride. This finding takes usback to propositions first articulated by Richmond (1967). He argued, and our findingsconfirm, that the greater the difference between objective conditions in Canada and in theorigin country and the greater the cultural distance traversed to get here, the morecommitted to Canada individuals tend to be.

The context that matters in the Canadian case is not the ethnic diversity of a province somuch as the history of the whole country. To the extent that the Canadian WVS dataexhibit any reflection of general arguments about social capital, it is in the microcosmicworld of interpersonal relations. Once outside that world, patterns make sense only whenwe bring Canadian history back in. It is very relevant that one Canadian in four belongsto a national minority with a deeply equivocal relationship to the overarching politicalnationality. It matters that Canada is almost evenly divided at the ProtestantReformation, a turning point in the evolution of civil society in the European world,including its overseas projections (Seligman, 1997). And it matters that CanadaÕsmulticultural policy has made the country easier to join, not harder.

22 This, of course, is the pattern in Helliwell (1996a).

14

Appendix A: Question Wording

All variables used in the analyses were taken directly from the 1991 World ValuesSurvey (WVS) Canada sample. In many cases, however, variables used are recodedand/or combined. The actual question wording and coding, the variable numbers, and Ðin some cases Ð the construction of measures are noted below.

EthnicityThe construction of ethnicity variables, based on the following questions, is described inthe text.v350 To which of the following groups do you belong above all? Just call out one of the

letters on this card. (1 French Canadian; 2 English Canadian; 3 Ethnic Canadian;4 Canadian first and then member of ethnic group; 5 Canadian first and only)

v369 Ethnic group [code by observation] (1 Caucasian/white; 2 Negro/Black 3 SouthAsian Indian, Pakistani, etc. 4 East Asian Chinese, Japanese, etc. 5 Arabic 6 Other)

v373 Language spoken by respondent (1 English 2 French)

Civic OrientationsFor all the civic orientation measures, the original variables are rescaled from 0 to 1,where 1 is equal to the most trust/confidence. Civic orientations measures are describedin the text. The variables used are as follows:Interpersonal Trustv94 Generally speaking, would you say that most people can be trusted or that you canÕt

be too careful in dealing with people? (1 Most people can be trusted; 2 CanÕt be toocareful 9 DonÕt know)

Political Trustv279/v285 Please look at this card and tell me, for each item listed, how much confidence

you have in them, is it a great deal, quite a lot, not very much or none at all? Ðresponses for ÔParliamentÕ, and ÔCanadian Political SystemÕ

v289 How much do you trust the government in Ottawa to do what is right? Do you trustit almost always, most of the time, only some of the time, or almost never?

National Pridev263 Of course, we all hope that there will not be another war, but if it were to come to

that, would you be willing to fight for your country? (1 Yes; 2 No; 9 DonÕt know)v322 How proud are you to be Canadian? (1 Very proud; 2 Quite proud; 3 Not very

proud; 4 Not at all proud; 9 DonÕt know)

MembershipsThe total memberships variable is simply a sum total of the number of groups to whichthere was a positive response for the following question.v19-34 Please look carefully at the following list of voluntary organizations and activities

and say (a) which, if any, do you belong to? Ð Social welfare services for elderly,handicapped or deprived people; Religious or church organizations; Education,arts, music or cultural activities; Trade unions; Political parties or groups; Localcommunity action on issues like poverty, employment, housing, racial equality;Third world development or human rights; Conservation, the environment, ecology;Professional associations; Youth work (e.g. scouts, guides, youth clubs, etc.);

15

Sports or recreation; WomenÕs groups; Peace movement; Animal rights; Voluntaryorganizations concerned with health; Other groups.

Demographic ControlsEach demographic control used in the step-wise regressions was constructed either as asingle dummy variable, or a collection of dummy variables. The dummy variables used,along with the original question asked, are listed below.Immigrant statusSingle dummy created for immigrant based on the following variable:v348 Were you born in Canada? (1 Yes; 2 Latin America; 3 North America; 4 Asia; 5

Europe; 6 Africa; 7 Other)ReligionDummies created for (1) Catholic, (2) Protestant, and (3) Other; residual category is noreligion. Based on the following variable:v143-145 (a) Do you belong to a religious denomination? (b) If yes, which one? (0 None

1 Roman Catholic; 2 Mainline Protestant; 3 Fundamentalist Protestant; 4 Jew; 5Muslim; 6 Hindu; 7 Buddhist; 8 Other; 9 N.A.)

EducationDummies created for (1) completed education at 17 or 18, (2) completed education at 19or 20, and (3) completed education at 21 or over; residual category is those completingeducation at 16 or younger. Based on the following variable:v356 At what age did you or will you complete your full time education, either at school

or at an institution of higher education? Please exclude apprenticeships. (0 N.A.; 1Completed formal education at 12 years of age or earlier; 2 Completed educationat 13 years of age; 3 Completed education at 14; 4 Completed education at 15; 5Completed education at 16; 6 Completed education at 17; 7 Completed educationat 18; 8 Completed education at 19; 9 Completed education at 20; 10 Completededucation at 21 years of age or older; 99 N.A./ D.K.)

Job statusDummies created for (1) upper, (2) middle, and (3) skilled manual workers; residualcategory is unskilled manual workers. Based on the following variable:v364 Interviewer [code by yourself] ÐSocio-economic status of respondent (1 Upper,

upper-middle class; 2 Middle, non-manual workers; 3 Manual workers - skilled,semi-skilled; 4 Manual workers - unskilled, unemployed)

AgeDummies created for those aged (1) 31-50, (2) 50-65, and (3) 66 and over; residualcategory is those aged less than 31, based on the following variable:v354-355 (a) Can you tell me your date of birth please? (b) This means you are É years

old.

16

Table 1: The Structure of Civic Orientations

Variables Components1 2

Interpersonal Trust 0.12 0.23

Trust in Government 0.76 0.06Confidence - Parliament 0.81 0.08Confidence- Cdn Pol System 0.79 0.14

National Pride 0.14 0.74Willing to Fight in War -0.11 0.80

Extraction Method: Principle Components AnalysisRotation Method: Varimax with Kaiser Normalization

17

Table 2: Ethnicity, Region, and Social Capital Measures Ð Comparisons of Means

Ethnicity All

Region French ÒNon-WhiteEthnicÓ

ÒWhite EthnicÓ ÒWhite Non-EthnicÓ

(A) MembershipsAtlantic b b b 1.29 (110) 1.36 (142)

Quebec 1.54 (382) b b b 1.56 (426)

Ontario b 1.54 (51) 2.15 (151) 1.65 (416) 1.75 (653)

Prairies b 1.98 (58) 1.74 (219) 1.78 (317)

BC b 1.31a (30)2.36 (32) 1.63 (146) 1.68 (192)

non-Quebecc 1.58 (82)

All 1.55 (464) 1.32 (95) 2.21 (262) 1.63 (908) 1.68 (1729)

(B) Interpersonal TrustAtlantic b b b .45 (110) .45 (142)

Quebec .35 (382) b b b .37 (426)

Ontario b .40 (51) .55 (151) .57 (415) .55 (652)

Prairies b .64 (58) .63 (218) .63 (316)

BC b .58a (30).81 (32) .65 (146) .67 (192)

non-Quebecc .45 (82)

All .36 (464) .44 (95) .62 (262) .59 (906) .52 (1727)

(C) Political TrustAtlantic b b b .37 (105) .37 (136)

Quebec .43 (379) b b b .44 (421)

Ontario b .43 (49) .43 (143) .39 (404) .40 (627)

Prairies b .38 (56) .34 (213) .35 (305)

BC b .40a (28).38 (31) .40 (143) .40 (188)

non-Quebecc .37 (76)

All .42 (454) .47 (89) .41 (251) .38 (282) .40 (1676)

(D) National PrideAtlantic b b b .84 (108) .84 (140)

Quebec .58 (376) b b b .59 (409)

Ontario b .85 (42) .82 (142) .81 (395) .82 (613)

Prairies b .80 (56) .77 (211) .78 (305)

BC b .82a (30).74 (31) .77 (143) .77 (188)

non-Quebecc .80 (80) b

All .61 (456) .83 (81) .80 (248) .80 (870) .75 (1656)a Prairies and BC respondents were combined.b Too few cases for analysis.c Calculated only for ÔFrenchÕ ethnicity.

Comparatively low scoreComparatively high score

18

Table 3: Ethnicity, Region, and Social Capital Measures Ð Step-wise Regressions

Independent ControlsVariables None

(BasicSetup)

É +Member-

ships

É +Immigrant

Status

É +ReligiousVariables

É +EducationVariables

É +Job

Status

É +Age

Variables

(A) MembershipsÒNon-WhiteEthnicÓ

-0.36(0.22)

N/a -0.43(0.23)

-0.28(0.23)

-0.32(0.23)

-0.25(0.23)

-0.14(0.23)

ÒWhiteEthnicÓ

0.53***(0.14)

N/a 0.50**(0.15)

0.59***(0.15)

0.53***(0.15)

0.53***(0.15)

0.54***(0.15)

R2-adj 0.01 N/a 0.01 0.03 0.08 0.09 0.10

(B) Interpersonal TrustFrench -0.15**

(0.05)-0.15**(0.05)

-0.15**(0.05)

-0.10(0.06)

-0.10(0.05)

-0.09(0.05)

-0.09(0.05)

ÒNon-WhiteEthnicÓ

-0.12*(0.05)

-0.11*(0.05)

-0.09(0.06)

-0.10(0.06)

-0.12*(0.06)

-0.10(0.06)

-0.07(0.06)

Prairies 0.10**(0.03)

0.09**(0.03)

0.09**(0.03)

0.09**(0.03)

0.09**(0.03)

0.09**(0.03)

0.09*(0.03)

BC 0.11**(0.04)

0.11**(0.04)

0.11**(0.04)

0.10*(0.04)

0.10*(0.04)

0.11*(0.04)

0.11*(0.04)

Member-ships

0.03***(0.01)

0.03***(0.01)

0.03***(0.01)

0.02*(0.01)

0.01*(0.01)

0.01(0.01)

R2-adj 0.05 0.06 0.06 0.06 0.09 0.10 0.11

(C) Political TrustÒNon-WhiteEthnicÓ

0.08***(0.02)

0.08***(0.02)

0.08**(0.02)

0.07**(0.02)

0.07**(0.03)

0.07**(0.03)

0.08**(0.03)

Quebec 0.05*(0.02)

0.05*(0.02)

0.05*(0.02)

0.04(0.02)

0.04(0.02)

0.04(0.02)

0.05*(0.02)

Prairies -0.05***(0.01)

-0.05***(0.01)

-0.05***(0.01)

-0.05**(0.01)

-0.05**(0.01)

-0.05**(0.01)

-0.05**(0.01)

Member-ships

0.01**(0.00)

0.01**(0.00)

0.01*(0.00)

0.01(0.00)

0.01(0.00)

0.01(0.00)

R2-adj 0.03 0.03 0.03 0.04 0.04 0.04 0.05

(D) National PrideÒNon-WhiteEthnicÓ

0.06*(0.03)

0.06*(0.03)

0.06(0.03)

0.06(0.03)

0.06(0.03)

0.06*(0.03)

0.07*(0.03)

Quebec -0.19***(0.03)

-0.19***(0.03)

-0.19***(0.03)

-0.19***(0.03)

-0.19***(0.03)

-0.19***(0.03)

-0.19***(0.03)

Prairies -0.05**(0.02)

-0.05**(0.02)

-0.05**(0.02)

-0.04*(0.02)

-0.05*(0.02)

-0.05*(0.02)

-0.05*(0.02)

BC -0.05*(0.02)

-0.05*(0.02)

-0.05*(0.02)

-0.04(0.02)

-0.04*(0.02)

-0.04*(0.02)

-0.04*(0.02)

Member-ships

0.00(0.00)

0.00(0.00)

0.00(0.00)

0.00(0.00)

0.00(0.00)

0.00(0.00)

R2-adj 0.14 0.14 0.14 0.14 0.14 0.14 0.15

p < .05; ** p < .01; *** p < .001N = (A) 1579; (B) 1535; (C) 1518; (D) 1581.Reference category is ÒWhite Non-EthnicÓ for ethnicity and Ontario for region.

19

References

Alesina, Alberto, Reza Baqir, and William Easterly. 1997. ÒPublic Goods and EthnicDivisions.Ó National Bureau of Economic Research Working Paper 6009.

Banfield, Edward C. 1958. The Moral Basis of a Backward Society. Chicago: The FreePress.

Black, Jerome H. 1982. ÒImmigrant Political Adaptation in Canada: Some TentativeFindings.Ó Canadian Journal of Political Science 15:3-28.

Black, Jerome H. 1987. ÒThe Practice of Politics in Two Settings: PoliticalTransferability Among Recent Immigrants to Canada.Ó Canadian Journal of PoliticalScience 22:731-53.

Boix, C. and D. N. Posner. 1998. ÒSocial Capital - Explaining Its Origins and Effects onGovernment Performance.Ó British Journal of Political Science 28:686-93.

Booth, J. A. and P. B. Richard. 1998. ÒCivil-Society, Political Capital, andDemocratization in Central-America.Ó Journal of Politics 60(3):780-800.

Brehm, John and Wendy Rahn. 1997. ÒIndividual-Level Evidence for the Causes andConsequences of Social Capital.Ó American Journal of Political Science 41:999-1023.

Coleman, James S. 1990. Foundations of Social Theory. Cambridge, Mass: HarvardUniversity Press.

Easterly, William and Ross Levine. 1997. ÒAfricaÕs Growth Tragedy: Policies and EthnicDivisions.Ó Quarterly Journal of Economics 112:1203-50.

Fellmeth, Aaron X. 1996. ÒSocial Capital in the United States and Taiwan: Trust or Ruleof Law?Ó Development Policy Review 14(2):151-71.

Forbes, H. D. 1997. Ethnic Conflict: Commerce, Culture, and the Contact Hypothesis.London: Yale University Press.

Fukuyama, Francis. 1995. ÒSocial Capital and the Global Economy.Ó Foreign-Affairs74(5):89-103.

Helliwell, John F. 1996a. ÒDo Borders Matter for Social Capital? Economic Growth andCivic Culture in U.S. States and Canadian Provinces.Ó National Bureau of EconomicResearch Working Paper 5863.

ÑÑÑ. 1996b. ÒEconomic Growth and Social Capital in Asia.Ó National Bureau ofEconomic Research Working Paper 5470.

Helliwell, John F. and Robert D. Putnam. 1995. ÒEconomic Growth and Social Capital inItaly.Ó Eastern Economic Journal 21(3):295-307.

Inglehart, Ronald. 1990. Cultural Shift in Advanced Industrial Society. Princeton:Princeton University Press.

ÑÑÑ. 1988. ÒThe Renaissance of Political Culture.Ó American Political ScienceReview 82:1203-30.

Johnston, Richard. 1986. Public Opinion and Public Policy in Canada: Questions ofConfidence. Toronto: University of Toronto Press.

Kenworthy, Lane. 1997. ÒCivic Engagement, Social Capital, and EconomicCooperation.Ó American Behavioral Scientist 40(5):645-57.

20

Knack, Stephen and Philip Keefer. 1997. ÒDoes Social Capital Have an EconomicPayoff? A Cross-Country Investigation.Ó Quarterly Journal of Economics112(4):1251-88.

Kymlicka, Will. 1995. Multicultural Citizenship: A Liberal Theory of Minority Rights.Oxford: Clarendon Press.

Lehning, Percy B. 1998. ÒTowards a Multicultural Civil Society: The Role of SocialCapital and Democratic Citizenship.Ó Government and Opposition 33(2):221.

Levi, Margaret. 1996. ÒSocial and Unsocial Capital: A Review Essay of Robert PutnamÕsMaking Democracy Work.Ó Politics and Society 24:45-55.

McCarty, Therese A. 1993. ÒDemographic Diversity and the Size of the Public Sector.ÓKyklos 46:225-40.

Miller, David. 1995. On Nationality Oxford: Oxford University Press.

Muller, Edward N. and Mitchell A. Seligson. 1994. ÒCivic Culture and Democracy: theQuestion of Causal Relationships.Ó American Political Science Review 88(3):635-53.

Nevitte, Neil. 1996. The Decline of Deference: Canadian Value Change in Cross-National Perspective. Peterborough: Broadview.

Newton, Kenneth. 1997. ÒSocial Capital and Democracy.Ó Amercian Behavioral Scientist40(5):575-87.

Nichols, Thomas M. 1996. ÒRussian Democracy and Social Capital.Ó Social ScienceInformation 35(4):629-642.

Putnam, Robert D. 1995a. ÒBowling Alone: AmericaÕs Declining Social Capital.ÓJournal of Democracy: 65-78.

ÑÑÑ. 1995b. ÒTuning In, Tuning Out: The Strange Disappearance of Social Capital inAmerica.Ó Political Science and Politics 28(4):664-83.

ÑÑÑ. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton:Princeton University Press.

Rice, Tom W. and Jan L. Feldman. 1997. ÒCivic Culture and Democracy From Europe toAmerica.Ó Journal of Politics 59:1143-72.

Rice, Tom W. and Alexander F. Sumberg. 1997. ÒCivic Culture and GovernmentPerformance in the American States.Ó Publius 27(1):99-114.

Richmond, Anthony H. 1967. Post-War Immigrants in Canada. Toronto: University ofToronto Press.

Rothstein, Bo. 1998. Just Institutions Matter: The Moral and Political Logic of theUniversal Welfare State. Cambridge: Cambridge University Press.

Seligman, Adam B. 1997. The Problem of Trust. Princeton, NJ: Princeton UniversityPress.

Temple, Jonathan and Paul A. Johnston. 1998. ÒSocial Capability and EconomicGrowth.Ó Quarterly Journal of Economics 113(3):965-91.

Verba, Sidney, Kay L. Schlozman, and Henry E. Brady. 1995. Voice and Equality: CivicVoluntarism in American Politics. Cambridge, MA: Harvard University Press.