Nations of Joiners: Explaining Voluntary Association Membership in Democratic Societies

by James E. Curtis, Edward G. Grabb, Douglas E. Baer
Nations of Joiners: Explaining Voluntary Association Membership in Democratic Societies
James E. Curtis, Edward G. Grabb, Douglas E. Baer
American Sociological Review
Start Page: 
End Page: 
Select license: 
Select License


University of Waterloo University of Victoria University of Western Ontario

Levels of voluntary association membership for 33 democratic countries are com- pared using data from surveys of nationally representative samples of adults from the 1990s. Four explanations of national differences in association involvement are identified and tested: economic development, religious composition, type of polity, and years of continuous democracy. The analyses consider total and working asso- ciation memberships, both including and excluding unions and religious associa- tions. Americans volunteer at rates above the average for all nations on each mea- sure, but they are often matched and surpassed by those of several other countries, notably the Netherlands, Canada, and a number of Nordic nations, including Ice- land, Sweden, and Norway. Hierarchical linear models show that voluntarism tends to be particularly high in nations that have: (I) multidenominational Christian or predominantly Protestant religious compositions, (2)prolonged and continuous experience with democratic institutions, (3) social democratic or liberal democratic political systems, and (4)high levels of economic development. With some excep- tions for working memberships, these factors, both separately and in combination, are clearly important predictors of cross-national variation in voluntary association


SOCIAL SCIENTISTSargue that volun- tary associations are essential to the political and social vitality of democratic so- cieties. The free association of individual citizens in such organizations reinforces par- ticipatory norms, encourages cooperative in-

Direct correspondence to James E. Curtis, De- partment of Sociology, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1 (curtis@ We thank Ronald Ingle- hart and colleagues and the Inter-university Con- sortium for Political and Social Research, Uni- versity of Michigan, for making the data avail- able. Neither the original investigators nor the disseminating archive bears any responsibility for the analyses conducted here. We thank Ronald Inglehart, Shane Dixon, Doug Enright, Helga Hallgrimsdottir, Jenna Hennebry, Deborah Matthews, Kevin McQuillan, and Terry Stewart for their assistance. The Social Sciences and Hu- manities Research Council of Canada provided funding for this study.

teraction, and promotes interpersonal trust, all of which are believed to be crucial for achieving effective solutions to important problems facing the wider community (Al- mond and Verba 1963, chaps. 10-11; Pug- liese 1986, chaps. 3-4; Putnam 1993: 171- 76, 2000:20; Verba, Schlozman, and Brady 1995, chap.11; Wuthnow 1991:300-302).'

Much of the literature on voluntary asso- ciations argues that Americans have exhib- ited more involvement in these organizations than any other people. Several writers over

I Putnam (1995; 2000, chap. 3) argues that these important outcomes of voluntarism may be increasingly in jeopardy in the United States be- cause association involvement is in decline. Based on reanalyses of Putnam's data, which came from the General Social Surveys for the 1970s to early 1990s, others (Paxton 1999; Rotolo 1999) dispute the claim that association membership has declined (also see Baer, Curtis, and Grabb 2001).


200 1, VOL. 66 DECEMBER:^^^-^^^) 783

the years have posed this hypothesis, typi- cally without conducting systematic national comparisons. For example, Weber's (1911: 53) view was that the United States was "the association-land par excellence," and Tocqueville ([I8351 1961: 141) observed that Americans were a people who "constantly form associations" (also see Ryan 1999581; Schlesinger 1944; Wood 1992:329; Wright and Hyman 1958:286). Some scholars have extended this assessment to recent times. Lipset (1985) observed in the 1980s and the 1990s that, compared with people in other nations, "Americans are more likely to take part in voluntary efforts to achieve particu- lar goals" (p. 141) and are less disposed to rely on the state for such purposes (Lipset 1990: 148, 1996:277). Drucker (1994) is among recent observers to draw similar con- clusions, asserting that "outside the English- speaking countries there is not much of a voluntary tradition. . . . [Tlhe modern state in Europe and Japan has been openly hostile to anything that smacks of voluntarism" (p. 76). Increasingly, however, as comparative data have become available, more tentative and qualified assessments have been offered for how Americans rank on voluntarism compared with other peoples (Curtis 1971; Curtis, Grabb, and Baer 1992; Hallenstvedt 1974; Putnam 1995:74, 2000, chap. 3; Wuthnow 199 1:289-90).

A close examination of the few available studies based on cross-national survey data raises doubts about which countries have the greatest involvement in associations. Three relevant cross-societal studies have been conducted using national samples, albeit covering only a few countries. One study found that Canadians were the most likely of six populations to belong to at least one association, although Americans ranked about the same as Canadians if union mem- berships were excluded (Curtis 1971 ). These data were from the United States, Britain, Germany, Italy, and Mexico for 1959 to 1961 and from Canada for 1965 and 1968. Hallenstvedt (1974) used the 1959-1961 data for five of these nations (the Canadian sample was excluded) and added national survey data for different years from Finland (19721, Norway (19691, and Sweden (1971). He found that the average rates of voluntary association membership in all three Nordic nations were somewhat higher than the rates for the other five countries. The most sys- tematic cross-national analysis to date is a study of 15 countries using national survey data for 1981 to 1983 (Curtis et al. 1992). This study found that Americans ranked the highest among all samples when respondents were asked if they belonged to at least one voluntary association. However, the Ameri- cans ranked first mainly because of their high level of church or religious involve- ment. When religious memberships were ex- cluded, Canadians and people from several other countries (e.g., Sweden, the Nether- lands, and Norway) equalled or surpassed the Americans in voluntarism. In analyses that included only working memberships (excluding nominal or inactive members), respondents from several additional nations had higher levels of association involvement than did the Americans.

Apart from the paucity of comparative studies, another shortcoming of previous work on this topic is that little attention has been given to developing theories of cross- national variation in association involve- ment. Several reviews of the literature note this problem (e.g., Curtis et al. 1992: 149; Knoke 1986:17-18; Pugliese 1986, chap. 2; Smith 1975:264). The lack of theoretical de- velopment may be attributable to the wide- spread acceptance of the American exceptionalism thesis, which offered no clear rationale for going beyond explanations or theories emphasizing American uniqueness. Thus, Tocqueville, Weber, and Lipset, among others, argued that the high associa- tion activity in the United States was a re- sult of a distinctively American value sys- tem that placed a high priority on individual participation in community affairs. This value system is alleged to have originated in America's revolutionary past and in the separation of church and state, which al- lowed associations, both through the church and outside the church, to grow unimpeded (Lipset 1996:60-61; Tocqueville [I8351 1961:36; Weber 191 1 ).

We extend the existing comparative work on voluntary associations in two key ways. First, we update the cross-national analyses to the decade of the 1990s, using data from many more countries than have been consid- ered in previous studies. Second, we conduct explicit tests for a set of contextual or coun- try-level explanations of differences in vol- untary association involvement across na- tions.

Although theoretical work on national dif- ferences in voluntary association involve- ment is still underdeveloped, at least four explanations can be gleaned from the litera- ture. One explanation focuses mainly on economic causes, while a second points to religious influences; the remaining two em- phasize political factors.

First, the interpretation that emphasizes the effects of economic organization asserts that the greater, and the earlier, the industri- alization of a society, the greater the volun- tary association activity. This argument con- tends that industrialization leads to occupa- tional specialization, as well as to rising edu- cation and income levels, and that these changes increase social status distinctions in society, around which a diverse range of vol- untary interest groups will form, and in which members of the population may then participate (e.g., Almond and Verba 1963, chaps. 1, 13; Curtis et al. 1992:150; Lipset 1994:2; Smith 1972). For example, Lipset (1994) emphasizes that economic develop- ment is conducive to the establishment of a sizeable middle class "that can stand up against the state and provide the resources for independent groups" (p. 2). The re- sources available to the average citizen in industrialized societies include greater ma- terial affluence, and more time and training, all of which facilitate high levels of commu- nity group activity. A variant of the eco- nomic interpretation is found in Putnam's (1993, 2000) temporal analyses for Italy. Based on evidence from the mid- 1800s through the 1980s, he concludes that eco- nomic prosperity both shapes cultural devel- opment and is shaped by it. He suggests that economic development is associated with greater voluntarism because of the high lev- els of social capital (in the form of greater interpersonal trust and cooperativeness) that are relatively more common in economically prosperous societies or regions. This social capital leads in turn to increased voluntary group creation and participation. Using data from more than 100 societies, Smith (1972) provides supporting evidence for the eco- nomic hypothesis. He shows that gross na- tional product per capita and five other mea- sures of economic development are all posi- tively correlated with indicators of two spe- cific forms of organizational activity (level of union membership and the number of nongovernmental organizations in each so- ciety) (Smith 1972; also see Almond and Verba 1963, chaps. 1, 13; Smith and Baldwin 1983). In recent research relevant to the economic hypothesis, Inglehart and Baker (2000:20), using data on social values from 65 societies, show that GDP per capita is positively correlated with shifts toward rational, tolerant, trusting, and participatory norms.

The second broad explanation for cross- national differences in association involve- ment focuses on the effects of religious tra- dition. Much of the literature suggests that, for a number of reasons, Protestant societies should have more highly developed systems of voluntary organizations than other na- tions. Writers from Tocqueville onward have discussed the pivotal role of religion, espe- cially Protestant sectarianism, in promoting voluntary action in the United States. Prot- estantism is frequently contrasted with the more corporatist, hierarchical, and often state-financed religions, most notably Ca- tholicism, but also Orthodox Christianity, Is- lam, and Confucianism (Inglehart and Baker 2000: 19,49; Lipset 1994:s). Protestantism is seen as promoting an ethic in which, rather than relying on the state or the church estab- lishment to provide for the needs of the com- munity, people are encouraged to join together voluntarily as free individuals to ful- fill various societal functions, including phi- lanthropy and the preservation of public mo- rality (Ladd 1999: 136-37; Lipset 1990: 143- 44, 1996:67-69; also see Drucker 1994; Smith 1975). Also, Verba et al. (1995:245- 46, 320) argue that, compared with Catholic churches, Protestant churches provide better training grounds for people to experience various forms of voluntarism, including po- litical and community activity, because Prot- estant churches are more egalitarian and par- ticipatory, while Catholic churches are more hierarchical or elitist (e.g., see McMullen 1994). Verba et al. (1995) compare Protes- tants and Catholics to test this theory, but only at the individual level and with national sample data for the United States alone. In our analyses, we are able to extend tests of the religious traditions hypothesis in two ways. First, we consider whether the alleged tendency in the United States for greater Protestant voluntarism at the individual level can be generalized to other national samples. Second, we assess whether nations with pre- dominantly Protestant populations, as a con- textual or country-level effect, exhibit higher levels of voluntarism than nations with other religious compositions.

Another basis for expecting religious tra- dition to have an effect on voluntarism is found in research on the sociology of reli- gion. Bibby (1987:208), for example, sees participation in churches and religious asso- ciations as largely a function of competition among churches for members. He poses this as an explanation for the comparatively high level of church participation in the United States, suggesting that the more aggressive "marketing" of religion to the American populace by large numbers of sectarian com- petitors, compared with the situation in Canada and Great Britain, makes for greater American rates of involvement. Similarly, Stark and colleagues (Finke and Stark 1998; Stark and Iannaccone 1994; Stark and McCann 1993) argue that, in societies and subsections of societies in which religious pluralism is high (and assuming that state regulation of religious groups is low), levels of religious attendance also will be high (also see Chaves and Cann 1992; Duke, Johnson, and Duke 1993; Olson 1998). In these settings, there is more competition among churches for members, which in turn generates high participation. These lines of interpretation suggest that high levels of re- ligious association membership should oc- cur, not where a single religion predominates and where there is considerable religious ho- mogeneity, but where several different reli- gious groups, Protestant or otherwise, coex- ist and must compete for followers.

It is not clear from the literature what im- pact competitive church recruitment may have on memberships in nonreligious asso- ciations. One reasonable hypothesis, how- ever, is that active recruitment of association members in one sector of the community (i.e., religion) may prompt similar recruiting in other areas. Ladd (1999:52-53) suggests that this is true for the United States, argu- ing that there is an expanding and increas- ingly competitive voluntary "group market- place" that has affected both religious and nonreligious association memberships in re- cent years.

The idea that one type of voluntary activ- ity prompts competing activities is also sug- gested in the literature on social movement organizations. This work emphasizes that activity begets oppositional activity, pro- vided there is no strong state suppression of voluntarism and the interests of opposing groups are not well addressed by state poli- cies and programs (see the review by Meyer and Staggenborg 1996.) On the other hand, it may be that the tendency for voluntary ac- tivity to stimulate further activity stems from cooperative processes as well. Several schol- ars argue that voluntary church activity should lead to more voluntary activity in the community, through example and through the creation of complementary interests (Lipset 1996:61-62; Putnam 2000:65-66; Wuthnow 199 1 :296-97).

The third theoretical interpretation we consider deals with the effects of different types of political organization across democ- racies. Janoski (1998) provides one of the most detailed statements of this explanation. Drawing on regime theory (Esping-Andersen 19901, Janoski (1998: 18-23) iden- tifies three kinds of democracies-liberal, social democratic, and traditional corporat- ist-while also allowing for mixtures of these three main types. He contends that, be- cause of their more elitist and segregated in- stitutional arrangements, traditional corpo- ratist democracies, such as Austria, France, and Italy, tend to lag behind both liberal and social democracies in forming voluntary as- sociations. He cites results from the 15-na- tion study by Curtis et a1.(1992) as evidence for his explanation. Janoski concludes that both liberal and social democracies are high in voluntary association activity, but for dif- ferent reasons. He believes that, in liberal democracies, high levels of association in- volvement occur "as a substitute for a strong welfare state," with church-related organiza- tions being a key to this process (Janoski 1998: 129, 132; also see Drucker 1994; Lipset 1990:80-8 1, 1996:60-62). Janoski emphasizes, however, that social democra- cies, in spite of their strong state presence, do not discourage or retard the growth of voluntarism. On the contrary, Janoski argues that voluntary organizations are "quite strong" in social democratic regimes, be- cause social democracies are more likely than other nations to accept or encourage the expansion of both union memberships and political organization memberships in the p~pulation.~ also

Janoski (1998:132-33) contends that, because voluntary associa- tions are frequently state-funded in social democratic regimes, members of voluntary associations in these nations can devote less working time to fund-raising, which means that they are often more independent from and critical of state activities than is the case under liberal regimes (also see Huber and Stephens 1998; Wuthnow 1991). It is likely, as well, that the provision of state funding to voluntary associations is itself a key fac- tor in the promotion of such organizations in these nations.

Finally, some scholars, while not denying the importance of distinguishing among types of democracies, have emphasized a fourth explanation for cross-national differ- ences in voluntary activity. This interpreta- tion is also political, but rather than focus- ing on the type of democracy per se, it sug- gests that the degree of stability or continu- ity of democracy is an important factor in- fluencing the level of community participa- tion and association involvement of differ- ent societies (Crenshaw 1995; Inglehart 1997: 188; Inglehart and Baker 2000; Lipset 1994; Wuthnow 1991:288). This hypothesis follows logically from the proposition that it takes time for a rich fabric of voluntary com- munity organizations to develop. Lipset (1994:3), for example, notes that such ideas as freedom of speech and the right of assem- bly "do not evolve overnight." In a similar

There also is evidence that the type of party in power has consequences for union affiliation. Western (1995) shows that, among other predic- tors, declining union density among 18 OECD nations in the 1990s was related to the electoral failure of social democratic parties in the 1980s.

vein, Inglehart's (1997: 164-65, 180-81) research in 41 societies shows that interper- sonal trust is positively correlated with the stability of democracy in those societies. This finding suggests that the widespread trust required for the mass of the population to interact voluntarily with others in their communities depends in part on how long- standing and enduring are the democratic in- stitutions of those nations. Presumably, then, the citizens of established and stable democ- racies, because they generally have had more experience with the principles and practices of free association, will tend to be more active in forming and joining voluntary organizations of different types. This leads to the prediction that the more prolonged and sustained the period of continuous democ- racy in a society, the greater the likelihood of high levels of inv~lvement.~

We test the four lines of explanation re- viewed above by exploring patterns of vol- untary association membership across the United States and 32 other nations in the 1990s. While our principal focus is on differ- ences in associational involvement among types of societies, as defined by their eco- nomic, religious, and political organization, we also compare association memberships across the individual societies. The latter analysis is necessary to check for exceptional cases or anomalies. Thus, we can replicate and extend to more nations our 15-nation comparison for the 1980s (Curtis et al. 1992).

Our 15-nation study also considered whether cross-national differences in asso- ciation involvement could be largely attrib- utable to differences in the sociodemo- graphic composition of the populations. Analyses controlled for variation in the edu- cation, gender, employment, age, size of community, and marital status profiles of the different samples. These factors proved im- portant in explaining voluntary association affiliation and activity within and across na-

Janoski and Wilson (1995) suggest that fam- ily socialization and continuity over generations may be another mechanism for increasing na- tional voluntary activity levels across time peri- ods. Their analysis of "community-oriented" associations indicates that the involvement of a person's parents in such organizations is a good predictor of that person's own involvement.

tions, although there was considerable vari- ance across societies left unexplained after introducing these controls. In the present analysis, we again control for the effects of individuals' sociodemographic backgrounds. We also control for the religious affiliation of individual respondents, given the argu- ments in the literature about Protestant exceptionalism and the effects of competi- tion among religious groups.

Based on previous research, we expect that education level is positively related to voluntary involvement; age has a curvilinear relationship, with the middle-aged the most likely to join associations, followed by older and then younger people; men are more likely to belong to associations than women; married people are more likely to be in- volved than the unmarried; and Protestants are more likely than those of other religious affiliations to join associations (Almond and Verba 1963, chap. 13; Janoski and Wilson 1995; Knoke 1986:3; Verba et al. 1995, chaps. 7-8, 11; Wilson and Musick 1997).

Our data come from the World Values Sur- veys of 1991 to 1993, a set of international surveys conducted by Inglehart and associ- ates that involve more than 40 countries (Inglehart 1997). In each country, a nation- ally representative sample was interviewed using essentially the same questionnaire (Inglehart 1997, app. 1). Only respondents 18 years old or older are included in our analyses. Our analyses are also limited to the 33 national samples for which data are avail- able on both voluntary association member- ship and the control variables. Except for Finland (N = 588) and Northern Ireland (N = 304), sample sizes ranged from 702 to 2,790. For two of the nations-Britain and Latvia-complete information was available for total memberships, but not for working memberships.

All the countries in our working sample were democratic nations as of 1991.4 Al-

Russia is one nation that may not have achieved official democratic status by the exact time of data collection. The Russian data were collected in 1991. During that year, because of various democratic initiatives, a formally demo-

though we would have preferred a broader sample of countries (one more representative of the world's diverse political systems, so- cial structures, and cultural traditions), no such data are available. However, this par- ticular set of national samples serves us well for testing the hypotheses concerning the ef- fects of both type of democracy and years of continuous dem~cracy.~

The initial dependent measure is a count of all association memberships. The count is based on a question asking respondents to look at a list of 16 types of "voluntary orga- nizations and activities" and to say "which, if any, do you belong to?" These organiza- tions were: (1) social welfare services for elderly, handicapped, or deprived people; (2) churches or religious organizations; (3) edu- cation, arts, music, or cultural activities; (4) trade unions; (5) political parties or groups,

(6) local community action (concerning pov- erty, employment, housing, and racial equal- ity); (7) human rights or Third World devel- opment; (8) conservation, environment, or ecology; (9) professional associations; (10) youth work (scouts, guides, youth clubs, etc.); (11) sports or recreation; (12) women's groups; (13) the peace movement; (14) ani- mal rights; (1 5) voluntary organizations con- cerned with health; and (16) "other group^."^

cratic government was established under Presi- dent Yeltsin.

Although more recent data were available from the 1995-1998 World Values Surveys, we chose to analyze the 1991-1993 data for two rea- sons. First, the dependent measure is different in these two waves of the surveys, with only the 1991-1993 surveys asking about any unpaid work that respondents have done for voluntary associations. Second, the 1991-1993 data permit us to consider a more extensive set of nations than do the 1995-1998 surveys.

This was the only indicator available to esti- mate the number of memberships of each respon- dent. A limitation is that some respondents may have had multiple memberships of the same type, but these would be counted only once. Although the extensive set of categories reduces the likeli- hood that any two organizations will fall in the same category, this limitation should be kept in mind when interpreting the results.

Analyses were first conducted using all 16 association types. Given previous research showing different patterns for union mem- berships, as well as concerns in the litera- ture that union membership is not necessar- ily voluntary (e.g., Smith 1975:249), we also performed separate analyses with union memberships excluded.' In addition, be- cause earlier studies indicate an exception- ally high level of religious membership in some nations, especially the United States, a third set of analyses was carried out with both unions and religious associations ex- cluded. We then looked at a count of work- ing memberships, based on a second ques- tion that asked respondents to consider all their reported memberships and to indicate "which, if any, are you currently doing un- paid work for?'@ As with the total member- ship measure, the working measure was con- sidered with unions included, with unions excluded, and with both unions and religious associations excluded.

Five exogenous variables were used throughout the controlled analyses and were coded as follows. Because of international differences in education systems, respon- dent's education level was measured using age at completion of schooling. This vari- able was grouped into three categories: com- pleted schooling at less than age 17, at ages 17 to 20, and at age 21 or older. Age was grouped into six categories: 18 to 24, 25 to 34, 35 to 44, 45 to 54, 55 to 64, and 65 and over. Marital status involved three catego- ries: married (including common law); wid- owed, separated, or divorced; and never- married. The gender variable was a simple dichotomy. Finally, religion was grouped

'Unions are a special type of affiliation in that union membership is a requirement for employ- ment in many occupations. Also, as Janoski (1998: 132-33) and others emphasize, some soci- eties, including social democracies and state so- cialist countries, are more likely to promote union affiliation in their populations.

Because the question was worded in this way, some respondents may have reported active in- volvements only if their activities benefited the organization or other people, and not if they ben- efited themselves.

into six categories: none, Roman Catholic, Protestant, Eastern Orthodox, Buddhist, and a combined category including all other reli- gious denomination^.^


For the initial controlled analysis of total memberships across individual nations, we used Poisson regression models (Long 1997:218-30). The coefficients estimated in these models can be used to predict logged values of the count variable. Exponentiation was used to convert these values to pre- dicted means.1° In the models for working memberships, there was evidence that the Poisson model might yield inflated test sta- tistics as a result of overdispersion; thus, for these analyses, negative binomial mod- els were employed instead (Long 1997: 230-41).

To test the extent to which the four coun- try-level explanations account for between- country differences, we employed hierarchi- cal models (Bryk and Raudenbush 1992; Raudenbush et al. 2000), in which equations are estimated at two levels: an individual- level, or level-1 (within-country) equation, and a country-level, or level-2 equation. The level- 1 equation takes the form:

Missing values were most pronounced for re- ligious affiliation (8.4 percent), education (4.8 percent), and age (1.4 percent). For each of these variables, separate additional categories were created for nonrespondents, which were then in- cluded in the analyses. Across the remaining variables, missing cases (deleted listwise) consti- tuted less than one per cent.

'O Effects coding was used for the exogenous variables, so the predicted mean counts are evaluated at the unweighted averages of the val- ues for the various categories of the exogenous variables. For the HLM analysis, we collapsed the categories of religion and age to ensure that our dummy variables did not generate singular data matrices within some countries. Religion was grouped into Protestant, Catholic, and other. Age was simply grouped into ages 34-54 versus others, because this distinction captured the ma- jor age effect.

where the subscript i refers to individual cases and j refers to countries. The X vari- ables are mean-centered; that is, Xlij is actu-

-ally Xlij -Xlej. The level-1 coefficient, Po, is specified for each of the j level-2 units (in this case, each country) and is, in turn, mod- eled as a function of country-level variables in a level-2 equation that takes the form:

where Wm,refers to the value of a country- level variable Wmfor the jth country." We use a hierarchical generalized linear model (Raudenbush et al. 2000: 11 1-3 I), with a log-link function and a Poisson sampling model at level 1.

The first level-2, or country-level, variable is the nation's level of economic develop- ment, as measured by the natural logarithm of GDP per capita in 1990. Our measure is based on data from the World Bank (1992:2 18-20) and Inglehart's (1997:360) analysis. This variable has a range of 7.39 to 10.16, with a mean of 9.04. The second country-level measure classifies the nations according to religious composition, with four groupings: mainly (70 percent or more) Protestant (Britain, Denmark, Finland, Ice- land, Norway, Sweden); mainly Roman Catholic (Argentina, Austria, Belgium, Bra- zil, Chile, France, Hungary, Ireland, Italy, Lithuania, Mexico, Portugal, Spain); mixed Protestant and Catholic, which we label "mixed Christian" (Canada, East Germany, Netherlands, Northern Ireland, the United States, West Germany); and others (Bul- garia, Estonia, Japan, Latvia, Romania, Rus- sia, South Korea), which was used as the ref- erence category.12 The religious categoriza- tion is derived from Barrett (1982) and the Encyclopedia Britannica (1994; also see Davis and Robinson 1999). The third aggre-

'I Hierarchical models can also involve level- 2 equations to model each of the level-1 equation slopes and not just the intercepts as shown here. Our hypotheses focus on the intercept effects, yo,, yo2,and so on.

l2 Slovenia is excluded from the HLM analy- sis because of the absence of individual-level variability in religion. This is not an issue with pooled data (Tables 1 and 2) but is problematic in HLM models.

gate measure is political type. Here we have adapted Janoski's (1998: 18-23) classification of democracies to include liberal de- mocracies (Britain, Canada, Ireland, North- ern Ireland, the United States), social de- mocracies (Denmark, Finland, Iceland, the Netherlands, Norway, Sweden), a third cat- egory denoting former eastern bloc state so- cialist nations (Bulgaria, East Germany, Es- tonia, Hungary, Latvia, Lithuania, Romania, Russia), and a fourth group of other democ- racies, which was the reference category and comprised the remaining 13 nations.'"he fourth country-level measure gauges each country's democratic stability, based on the number of years of continuous democracy between 1920 and 1990 inclusive. This indi- cator, which is derived from Inglehart's (1997: 358) classification, has a range of 0 to 71 and a mean of 33.

Two of the country-level variables, GDP (log) and years of continuous democracy, are highly correlated (r = .85). There is also con- siderable overlap between the social democ- racy political type and the Protestant reli- gious type: Except for the Netherlands, which is mixed Christian, all the social de- mocracies are mainly Protestant. For this reason and because of the small number of countries, interpretations of the results with controls for other country-level variables

'"ur reference category corresponds closely to Janoski's (1998) "traditional corporatist" de- mocracy type, if we provisionally include six Hispanic countries in southern Europe and the southern Americas-Portugal, Spain, Brazil, Chile, Mexico, Argentina-in this category. These nations share several key characteristics with Janoski's traditional corporatist societies. Most important, all six countries have been marked by authoritarian or elitist political regimes in their recent histories, and all have strong collectivist religious traditions that have shaped their political cultures and related social institu- tions (Bergalli 1997:36, 39-40; Esping-Andersen 1990:27; Inglehart 1997:95; Leibfried 1992:253- 54; Lipset 1967:4; Reher 1998:208-10; Wiarda 1986:265, 268). Similarly, Japan and South Ko- rea are also placed in the reference category, as various writers question their characterization as liberal democracies, given their relatively recent experience with democracy and the "heavy resi- dues of authoritarianism and elitist structures" they retain (Oh 1999:4; also see Han 1974; Huber and Stephens 1998).

should be done with some caution. The high degree of intercorrelation makes it difficult to assess the unique effect of each country- level variable, although it also makes for a rigorous standard when calculating signifi- cance tests for such effects.

WITHOUTCONTROLS. Column 1 of Table 1 shows the zero-order results for the count of all memberships (including unions and reli- gious organizations) in each country. The expected overall mean (1.03) in column 1 shows that respondents in the combined set of 33 nations average about one voluntary association membership per person. As in previous research, the level for the United States is relatively high at 1.98 member- ships, or almost twice the mean for all coun- tries. However, several nations rank higher, with the Netherlands (2.69) at the top, fol- lowed by Iceland (2.43) and Sweden (2.08). Next comes Norway, which, at 1.96, is vir- tually identical to the United States. In de- scending order, the other countries that are significantly above the average for all na- tions in column 1 are Finland, Denmark, Canada, East Germany, South Korea, Bel- gium, West Germany, Estonia, Lithuania, Northern Ireland, Britain, and Austria.

These findings suggest some initial sup- port for the four theoretical explanations out- lined earlier. Most of the above-average countries are relatively high in economic de- velopment (as judged by GDP per capita), have Protestant or mixed Christian religious traditions, are either liberal or social democ- racies, and have relatively long-standing ex- perience with democratic political institu- tions. The main exceptions to this assertion are South Korea and three former eastern bloc countries-East Germany, Estonia, and Lithuania. Another former eastern bloc na- tion, Russia, is at the expected mean for all nations, as is Ireland. As anticipated, the na- tions with the lowest levels of involvement tend to have less developed economies (al- though Japan, France, and Italy are excep- tions), and have predominantly Catholic or Eastern Orthodox religious traditions in most cases. In addition, none of the low-ranking nations is a liberal or social democracy, based on Janoski's classification. Finally, the ma-

jority have had little continuous experience with democracy. Included here are five former eastern bloc countries-Romania, Slovenia, Bulgaria, Latvia, and Hungary- and all four Latin American nations in the analysis-Argentina, Mexico, Chile, and Brazil. Overall, the results resemble those in our previous 15-nation study for the 1980s, although one notable difference is that the United States, not the Netherlands, ranked first in total memberships in that analysis.14

Column 2 of Table 1 reconsiders the na- tional rankings in association involvement, this time excluding union memberships. The patterns are similar to those in column 1. The Netherlands still ranks first. Probably the key change that occurs when unions are taken out of the membership count is that, except for East Germany, all the former eastern bloc nations now rank significantly below the av- erage count for all nations. This indicates that union membership, which is more compul- sory than voluntary in many societies, is largely responsible for our initially finding high levels of overall association member- ship in some former eastern bloc nations.

Column 3 of Table 1 provides national comparisons for total memberships, exclud- ing unions and religious memberships. These findings suggest that the omission of religious associations from the count

l4 If we consider the 14 nations included in both the 1980s and the 1990s surveys, the top seven nations in the 1980s for total memberships were, in descending order, the United States, Sweden, Northern Ireland, the Netherlands, Nor- way, Canada, and Britain. The bottom seven, in ascending order, were Italy, France, Japan, Spain, Belgium, West Germany, and Ireland (Curtis et al. 1992). There are some significant discrepancies in the measurement of association membership between the 1980s and 1990s data sets. The key differences are that, in the 1980s data, the checklist of organization types was shorter and lacked the residual "other" category. These differences may have raised the average count of memberships in the 1990s surveys com- pared with the average count in the 1980s, mak- ing it difficult to know whether there were actual changes in a nation's membership level between the two time ~eriods (see Baer et al. 2001, for

I comparisons o> the 1980s and 1990s data).
Table 1. Expected Counts of Total Association Memberships, with and without Controls: World

Values Surveys, 33 Nations, 1991 to 1993
Without Controls         With Controls
Excluding             Excluding
Excluding     Religious1         Excluding     Religious1
Total     Union     Union     Total     Union     Union
Nations     Memberships Memberships Memberships     Memberships Memberships M    emberships
All Nations     1.03     .80     .67     .95     .76     .65
Argentina     .35**     .34**     .27**     .33**     .33**     .27**
Austria     1.11"     .92**     .76**     1.11**     .95**     .82**
Belgium     1.39"     1.25**     1.13"     1.36"     1.26"     1.17**
Brazil     .87**     .80     .58**     .99     .96**     .74*
Britain     1.12**     .97**     .8 1 **     1.20"     1.08"     1.01"
Bulgaria     .69**     .50**     .48**     .70**     .54**     .46**
Canada     1.70"     1.57"     1.32"     1.44"     1.35"     1.15**
Chile     32"     .76     .58**     .72**     .68**     .53**
Denmark     1.75"     1.26"     1.20"     1.29"     .91**     .9 1 **
East Germany     1.67"     1.11"     .91**     1.85"     1.29**     1.03**
Estonia     1.24**     .63**     .59**     .98     .5 1 **     .42**
Finland     1.79"     1.43**     1.26**     1.25"     1 .00**     .91**
France     .73**     .68**     .62*     .75**     .72     .65
Hungary     .73**     .41**     .30**     .71**     .42**     .30**
Iceland     2.43"     1.83"     1.33"     1.79**     1.33**     1.04"
Ireland     .98     .89**     .75**     .99     .93**     .84**
Italy     .61**     .55**     .47**     .72**     .69**     .70
Japan     .49**     .41**     .35**     .46**     .41**     .30**
Latvia     .90**     .46**     .42**     .74**     .38**     .33**
Lithuania     1.21"     .69**     .66     .89     .5 1 **     .44**
Mexico     .70**     .66**     .53**     .70**     .68**     .56**
Northern Ireland     1.17'     1.05**     .go**     1.16**     1.06"     39"
Netherlands     2.69"     2.50**     2.15"     2.54**     2.43"     2.03**
Norway     1.96**     1.54"     1.43"     1.41"     1.09"     1.06**
Portugal     .55**     .5 1 **     .40**     .6 1 **     .58**     .56*
Romania     .42**     .22**     .17**     .45**     .25**     .19**
Russia     1.03     .42**     .41**     1.01     .43**     .37**
Slovenia     .6 1 **     .41**     .39**     .57**     .40**     .38**
South Korea     1.47"     1.39"     1.01**     1.14"     1.09"     .75**
Spain     .36**     .34**     .28**     .37**     .35**     .32**
Sweden     2.08"     1.50"     1.39**     1.65**     1.18"     1.15"
United States     1.98"     1.89**     1.40**     1.59"     1.51**     1.15"
West Germany     1.37"     1.22**     1.06"     1.36"     1.23"     1.16"
Dispersiona     .57     .87     1.04     .49     .73     .88
Probability     c ,001     c ,001     < ,925     c ,001     c ,001     < ,001

Note: Control variables are age, religion, gender, marital status, and education. The significance test for each country is against the mean for all nations.

a The probability for the dispersion measure comes from a two-tailed z-test against the null hypothesis that the dispersion parameter equals 1 .O.

*p< .05 **p < .O1 (two-tailed tests)

changes the basic patterns very little from those in column 2. The ranking for the United States is slightly lower with religious associations excluded. The rankings of some other countries, including South Korea, Bra- zil, and Chile, also decline to some degree when religious associations are excluded, primarily because all of these countries are above the average for religious association memberships (data not shown).

WITHCONTROLS.The right side of Table 1 provides controlled comparisons for the nations. For the most part, these results are similar to those without controls. The Neth- erlands is easily the highest ranking nation on all measures. With controls, there is an even clearer tendency for those countries that are relatively prosperous and stable lib- eral and social democracies and that have Protestant or mixed Christian heritages to predominate among the above-average nations. South Korea continues to be an excep- tion to this general pattern, ranking above average on all three measures of total mem- bership, as before. Brazil is another excep- tion in this case, rising to average or above average levels on all measures. One notable difference with controls is that East Ger- many is now the only former eastern bloc nation to be significantly above average on any of the three dependent variables, and in fact is second overall when unions and reli- gious memberships are included (column 4). All of the other eastern bloc nations are be- low average, especially when union mem- berships are excluded (column 5).


CONTROLS. In columns 1, 2, and 3 of Table 2 we repeat the analysis without controls, but only for memberships in which respondents are actively engaged in unpaid work for the organizations. Although there are a few notable differences compared with Table 1, the same categories of nations tend to be the highest in association activity. Once more, most are economically devel- oped countries with Protestant or mixed Christian religious heritages and stable so- cial or liberal democratic political systems. Again, East Germany and South Korea are the main exceptions to this assertion. We also see the same decline in membership lev- els in the former eastern bloc societies when unions are excluded.

One change from Table 1 is that the Neth- erlands does not rank highest in these com- parisons-the United States is first in over- all working memberships and in working memberships with unions excluded, and Canada is first when both unions and reli- gious organizations are excluded. In general, the patterns are less pronounced compared with those for total memberships. Another change is that some highly ranked nations in Table 1 (e.g., Denmark, Northern Ireland, and Ireland) drop to average levels on work- ing memberships, while some of the lower ranked countries in Table 1, including sev- eral eastern European nations (e.g., Estonia and Lithuania) and some Latin American na- tions (e.g., Chile and Mexico), rise to aver- age or above-average levels. This may re- flect differences in the extent to which the countries have "professionalized" associations, which have paid staff and lesser needs for unpaid volunteer help (see Curtis et al. 1992: 149).15

WITHCONTROLS.The right side of Table 2 shows national differences in working memberships, this time controlling for the effects of the five exogenous variables. With controls, Canadians rank first for all three measures of association activity, with the United States a close second and essentially equal to Canadians if religious organizations are included and unions are excluded (col- umn 5). The broad patterns are similar to those obtained without controls, although there are some deviations from the results reported previously. In particular, we find that, for all three measures, Denmark is sig- nificantly below average for working mem- berships and Norway no longer ranks among the above-average countries. Also, Brazil

l5 In supplementary analyses, we found that the proportion of working memberships relative to total memberships declines as we move from the less prosperous to the more prosperous nations in our study. A possible alternative expla- nation, beyond one based on the professional- ization of associations, is that more affluent people, especially in more affluent societies, have more material resources and more opportu- nities for joining associations, but ultimately lack sufficient time or energy to be actively involved in most of them.

Table 2. Expected Counts of Working Association Memberships, with and without Controls: World Values Surveys, 31 Nations, 1991 to 1993
Without Controls         With Controls
Excluding             Excluding
Total     Excluding Religious1-    Total     Excluding-    Religious1
Working     Union ~iion     Working     Union     ~Gon
Nations     Memberships Memberships Memberships     Memberships Memberships M    emberships
All Nations                     
East Germany                     
Northern Ireland                     
South Korea                     
United States                     
West Germany                     

Note: Control variables are age, religion, gender, marital status, and education. The significance test for each country is against the mean for all nations.

aThe probability for the dispersion measure comes from a two-tailed z-test against the null hypothesis that the dispersion parameter equals 1 .O.

*p< .05 **p < .01 (two-tailed tests)

and Italy are now above average for all three measures of working memberships, while France is above average on two of these measures. Thus, although the same general patterns tend to hold, they appear less pro- nounced when the effects of the five exog- enous variables are controlled.

An analysis of the effects of the five control variables reveals that country is a far better predictor of association involvement than any of them, and that this is true for all de- pendent measures.16 Education level has the strongest impact among the controls: As ex- pected from previous studies, the most highly educated individuals exhibit the highest membership rates. Religion has the next larg- est effect, for both total and working mem- berships, with membership levels generally highest for Protestants and others, and low- est for the Orthodox respondents and those with no religion. Age has a larger effect than religious affiliation on association activity if religious organizations are excluded from the count of memberships. Memberships are generally higher for middle-aged people and lower for both younger and older individu- als; males tend to belong to more voluntary organizations than do females. Finally, con- trary to expectations, single (never-married) people have relatively high membership lev- els, equal to those of the married respondents for all measures of association membership, although the effects of marital status are weak in both cases. l7

l6 To save space, findings for the individual- level exogenous variables are not shown, but are available on request.

l7 The likely reason why the results for marital status are different from those of our earlier re- search (Curtis et al. 1992) is that married respon- dents were contrasted with all nonmarried respondents in that analysis, so that single, never- married individuals were not analyzed separately from the widowed and divorced. We should note as well that the exogenous effects of both occu- pation and community size were considered in preliminary analyses of the 1990s data. As ex- pected from previous studies, people in profes- sional occupations generally have the highest as- sociation involvement, and community size tends to be negatively related to association member-

As the comparisons within Tables 1 and 2 indicate, controls for the effects of the five individual-level exogenous variables alter somewhat the patterns of cross-national variation in association membership. Al- though the results with controls are not dra- matically different from the uncontrolled findings, they do indicate the value of keep- ing the controls in place for the rest of the analysis. Cross-national variations in socio- demographic characteristics are related to differences among the nations in association membership, but additional factors must also be considered if we are to account more completely for national differences in asso- ciation membership levels.

Tables 3a and 3b directly assess the explana- tory power of the four country-level expla- nations for association membership across the various nations, using hierarchical linear models. Again, the six different measures for total and working memberships are em- ployed. Table 3a displays the effects of each of the country-level (level-2) variables, con- trolling for the effects of the five individual- level variables. Results in Table 3b show the effects of each country-level variable on each measure of association membership, with simultaneous controls for the five indi- vidual-level variables and for the other country-level variables. To conserve space, the coefficients for the individual-level vari- ables are not shown. The individual-level variables generally have the same effects in the HLM models as they did in the analyses in Tables 1 and 2.l8

ship, although the latter pattern is weak. Data on these two variables were not collected in all the samples. Thus, we did not use them as controls, in order to maximize the number of countries available for comparison.

l8 The one exception was for the religious af- filiation variable, and then only when the depen- dent variable excludes religious associations. With country-level controls, including religious composition, religious affiliation at the indi- vidual level did not have a significant effect on voluntary association membership. A table con- taining coefficients for the effects of the indi- vidual-level predictors is available on request.

Table 3a.Coefficients from Level-2 HLM Poisson Models Evaluating Characteristics of "Country" as a Predictor of Voluntary Association Memberships: Level-2 Variables Modeled One at a Time, World Values Surveys, 33 Countries, 1991 to 1993
Total Memberships     Working Memberships
Excluding     Excluding
Country     Total     ExcludingUnion     Religious1Union     Total     Excluding Union     Religious1 Union
Charactersitic     Memberships Memberships Memberships     Memberships Memberships Memberships
GDP (log base e)     .265**     .294**     .273**     ,048     ,092     ,099
    (.048)    (.052)     (.057)     (.043)     (.060)     (.055)
Years of democracy     .008**     .009**     ,009"     .003*     .005*     .006**
    (.001)    (.001)     (.001)     (.001)     (.002)     (.002)
Religion a         
Mixed Christian     .581**     .777**     305"     .219     .436*     ,498'
    (.135)    (.142)     (.157)     (.112)     (.197)     (.183)
Protestant     .588**     .662**     .715**     .247     ,222     ,230
    (.105)    (.097)     (.122)     (.154)     (.178)     (.173)
Catholic     ,047     .323**     .428**     .072     ,161     ,195
    (.128)    (.110)     (.140)     (.109)     (.158)     (.157)
Political Type a         
Social democracy     .728**     .566**     .761**     .267*     ,213     ,217
    (.103)    (.117)     (.110)     (.109)     (.144)     (.132)
Liberal democracy     .401**     .272     .293     .54**     .645**     ,739"
    (.112)    (.272)     (.103)     (.073)     (.116)     (.142)
Eastern bloc     ,173     -.374**     -.455**     .172*     -.I10     ,020
    (.114)    (.104)     (.158)     (.082)     (.101)     (.127)
Block Chi-Square Tests         
GDP     28.16**     34.40**     22.36"     .94     2.19     2.66
Years of democracy     28.28**     40.64**     36.81"     4.33*     7.69**     12.82"
Religion     33.12**     30.55**     33.61**     3.91     9.04*     10.61'
Political type     32.63**     39.71**     66.94**     20.34**     23.74**     30.60**

Note: Numbers in parentheses are standard errors. All models involve level-1 controls for age, religion, gender, marital status, and education. a "Other" is the reference category for "religion" and "political type." Degrees of freedom for GDP equals 1; for years of democracy d.f = 1; for religion d.f. = 3; for political type d.f. = 3, for the overall model d.f. = 8. *p< .05 *p< .Ol (two-tailed tests)

In Table 3a we see that, as expected, eco- across the nations. As shown in Table 3a, nomic development, as measured by the log the block chi-square tests for religion are of GDP per capita, has a significant positive significant for five of the six dependent effect on total association memberships, measures. As expected, Protestant countries with unions included (.265), with unions ex- are relatively high in total association mem- cluded (.294), and with unions and religious berships, overall, with unions excluded, and associations excluded (.273). However, the with unions and religious associations ex- effect of economic development is weak and cluded; the significant coefficients repre- not statistically significant for any of the senting the differences between Protestants measures of working membership (columns and the reference category are .588, .622, 4, 5, and 6). and .715. The mixed Christian nations also

Religious composition is clearly associ- have comparatively high total membership ated with variations in membership levels counts, with coefficients of .581, .777, and

.805. The difference between the mixed Christian and Protestant categories, on one hand, and Catholic countries, on the other hand, is significant for all three measures of total membership. l9 For working member- ships, the patterns are not as clear. When union memberships are included (column 4), the effect of religious type is not statisti- cally significant. When union memberships are not included (columns 5 and 6), mixed Christian countries have membership counts that are significantly higher than the reference category (coefficients of ,436 and .498). However, in no instance is the differ- ence between Protestant and Catholic coun- tries statistically significant, and the differ- ence between the mixed Christian and Catholic countries is only statistically sig- nificant when religious organizations are excluded (column 6) (x2= 4.01, d.f. = 1, p < .04).

For the two hypotheses concerning politi- cal organization, Tables 1 and 2 suggested that both type of political system and length of sustained experience with democratic government may affect association membership levels. The findings in Table 3a pro- vide support for these two hypotheses, al- though the effects are more consistent for years of continuous democracy than for po- litical type. The coefficients for years of continuous democracy are significant for all six measures of association membership. To illustrate the meaning of these coefficients, consider the coefficient of .008 for years of democracy in column 1. This indicates the expected effect on association membership of a one-year difference in experience with democracy. If we multiply this coefficient by 70 years, for example, we arrive at the expected difference of approximately .56

l9 This same difference is not statistically sig- nificant for two of the working membership mea- sures (with religious organizations included, col- umns 4 and 5, p > .lo). It is statistically signifi- cant (p < .04) for working memberships with unions and religious organizations excluded, al- though the major difference in this case is be- tween mixed Christian countries and all others, because Protestant and Catholic countries are not significantly different from each other. These tests, and others involving comparisons between two groups other than the reference category, are not shown in the table.

between a country with only 1 or 2 years of prior experience with democracy and a country with 70 years of continuous experi- ence. Because the dependent variable is the logged value of the count of memberships, we thus expect a membership count that is e.56, or 1.75 times as high for a country with 70 years of democratic experience compared with a country with 1 or 2 years of democratic experience. Interestingly, this difference is almost equal in magnitude to the difference between the lowest-ranked and the highest-ranked nations on the GDP variable.

Finally, we have tests for type of political system as a possible explanation of national differences in association membership lev- els. As expected from Janoski's (1998) theory, the results in Table 3a show that both the social democratic and the liberal demo- cratic polities tend to be highest in associa- tion memberships, with former eastern bloc nations lowest and other democracies (the reference group) falling in between. This pattern varies somewhat for total as opposed to working memberships. For total member- ships (columns 1, 2, and 3), social democra- cies have the highest membership counts, with liberal democracies ranking second (al- though they are not significantly lower when religious associations are included), and other democracies ranking third. Former eastern bloc countries rank last, unless unions are included (column I), and they rank significantly lower than all the other categories. For total memberships, social de- mocracies are significantly higher than lib- eral democracies when unions are included (p < .001) and when both unions and reli- gious organizations are excluded (p < .04), but not when religious organizations are in- cluded but unions excluded (p > .08). For working memberships, though, liberal de- mocracies are the highest-ranking category and are significantly higher than social de- mocracies for all three measures (p < .04 or better).

The sizes of the chi-square statistics in Table 3a provide some indication of the relative effects of the four country-level predictors on association involvement. The effects of all four of these predictors are similar for total memberships (column 1) and for total memberships excluding unions

Table 3b.Coefficients from Level-2 HLM Poisson Models Evaluating Characteristics of "Country" as a Predictor of Voluntary Association Memberships: Level-2 Variables with Added Controls for Other Level-2 Variables, World Values Surveys, 33 Countries, 1991 to 1993
Total Memberships     Working Memberships
Excluding     Excluding
Country     Total     ExcludingUnion     ReligiouslUnion     Total     Excluding Union     Religiousl Union
Charactersitic     Memberships Memberships Memberships     Memberships Memberships Memberships
GDP (log base e)     ,073     ,005     -. 123     -.095     -.214     -.052
    (.099)    (.099)     (.108)     (.113)     (.146)     (.139)
Years of democracy     .010*     .009**     .0 10**     .006     .004     .005
    (.004)    (.002)     (.002)     (.004)     (.006)     (.005)
Religion a         
Mixed Christian     .452**     .444**     .603**     .203     ,180     .288
    (.170)    (.125)     (.173)     (.199)     (.253)     (.239)
Protestant     ,270     ,161     ,361     -.043     -.245     -.448
    (.230)    (.165)     (.218)     (.318)     (.409)     (.370)
Catholic     ,130     ,235     .390*     ,136     ,073     ,201
    (.121)    (.098)     (.143)     (.137)     (.174)     (.171)
Political Type a         
Social democracy     -.I16     ,042     ,181     ,154     .455     .555
    (.253)    (. 150)     (.169)     (.324)     (.424)     (.379)
Liberal democracy     -.288*     -.249     -.210     ,265     ,424     ,378
    (.224)    (.173)     (.161)     (.250)     (.334)     (.295)
Eastern bloc     .496**     ,164     .067     .299     -.282     ,143
    (. 135)    (.109)     (.131)     (.149)     (. 190)     (.184)
Block Chi-Square Tests         
Overall model     72.40**     47.10**     55.01**     22.79**     25.37**     33.99**
GDP     .55     .OO     1.03     1.81     2.15     .14
Years of democracy     6.20*     3.92*     3.93*     .7 1     .5 1     .81
Religion     8.15*     7.93*     12.45**     2.61     2.57     9.86*
Political type     16.06**     4.13     4.64     5.14     4.05     2.77

Note: Numbers in parentheses are standard errors. All models involve level-1 controls for age, religion, gender, marital status, and education. a "Other" is the reference category for "religion" and "political type." Degrees of freedom for GDP equals 1; for years of democracy d.f = 1; for religion d.f. = 3; for political type d.f. = 3, for the overall model d.f. = 8. * *

*p< .05 p < .O1 (two-tailed tests)

(column 2); however, political type has a strong collective effect on the membership stronger effect than the other three country- counts, even after individual-level differ- level variables for the other four measures ences among respondents are statistically of association membership (columns 3,4, 5, controlled. Along with the results from and 6). Table 3a, these findings indicate that the

In Table 3b we add controls for the ef- country-level political, economic, and reli- fects of the country-level variables them- gious factors are of real consequence in de- selves. The block chi-square tests for the termining which countries have high levels overall model (d.f. = 8) yield highly signifi- of association activity. cant results for each of the six membership Attributing unique effects to each of the measures. This means that, taken together, four country-level variables while control- the four country-level predictors have a ling for other country-level variables is a more difficult task. In Table 3b, there are two dependent variables (working member- ships, columns 4 and 5), for which none of the country-level predictors is significant at p < .05, despite the fact that the overall model is significant. As mentioned earlier, there are high correlations among the coun- try-level variables; with only 31 or 33 countries, it is not possible to separate out effects completely. Certain constellations of country characteristics-mixed Christian nations that tend to be liberal democratic and Protestant nations that are social demo- cratic-are associated with high voluntary participation. Nevertheless, in some cases, unique effects can be discerned by control- ling for other country-level variables, as follows.

In Table 3b, the block chi-square tests for the effects of religious composition are again significant for all three total member- ship measures, including and excluding re- ligious associations and unions (columns 1, 2, and 3). Mixed Christian countries still have the highest membership levels, but the differences between Catholic and Protestant countries are not statistically significant. Differences between the mixed Christian and Catholic countries are significant when unions are included (p < .04), but not when unions are excluded (p > .09 in each case). For working memberships excluding both unions and religious associations, there is a significant difference (p < .01) between the coefficients for mixed Christian countries (.288) and Protestant countries (-.448).

With controls, political type remains a significant predictor of total membership only for the measure that includes unions. Here, former eastern bloc countries have the highest membership counts. This result is explained by the fact that, in these coun- tries, respondents report very high union membership rates; this pattern is not ob- served with union memberships excluded. For all three measures of working member- ship, political type is not statistically sig- nificant. Finally, years of democracy con- tinues to have a significant effect for total memberships, but not for working member- ships; the coefficients are approximately the same as the zero-order coefficients reported in Table 3a, but the standard errors are higher.

This analysis for 33 individual democratic countries in the 1990s confirms and extends the results of our earlier study of 15 societ- ies (Curtis et al. 1992). The United States and Canada are again comparatively high in voluntary association membership, but they are not always the highest ranking nations. Also, as in the earlier study, the Netherlands, Sweden, and Norway generally exhibit high membership levels. The larger sample of na- tions available for the present analysis makes it even more apparent than before that it is primarily the democracies in northern and western Europe, along with the two En- glish-speaking North American nations, that are highest in voluntary association involve- ment. Our results using alternative measures of membership, with or without individual- level controls for social background, reveal only a few exceptions to these generaliza- tions. Thus, the findings clearly challenge the often-cited "American exceptionalism" thesis and also contradict the less common but long-standing observation that European countries are not receptive to voluntary ac- tivity (Drucker 1994; Lipset 1990, 1996; Tocqueville [I8351 1961). The analysis also shows a relatively consistent rank-order of nations over about a 10-year period, at least for the 14 countries included in both the

1980s and the 1990s data sets.

Although the United States is not unique in its high level of total association activity, there is strong evidence of "American exceptionalism" when church membership alone is considered. In results not shown here, the United States ranks highest among the 33 countries when it comes to joining churches and other religious associations. Interpretations of this pattern include Lipset's (1990:74-89) view that individual- istic values in the United States encourage high participation and diversity in voluntary associations, including churches. Other ex- planations point to the aggressive competi- tion for members among the many Protes- tant sectarian churches (Bibby 1987; Ryan 1999).

We have also made progress in initial tests of four explanations of cross-national varia- tion in association membership. Because of a shortage of relevant cross-national sur- veys, these tests have not been possible in the literature on voluntary association activ- ity until now. We have tested hypotheses on the effects of economic development, reli- gious tradition, political type, and number of years of continuous democracy. For each hy- pothesis, we conducted two forms of multi- variate tests: one controlling for individual- level social background characteristics, and another with added controls for the other country-level variables themselves. The first set of tests supports all four explanations. The second set suggests that religious tradi- tion has the strongest and most consistent unique effects, and that years of continuous democracy has effects for measures of total membership.

Various writers have suggested that high levels of economic and industrial develop- ment are positively associated with high lev- els of voluntary organization affiliation. This proposition is supported in the HLM analy- ses that applied individual-level controls, showing positive effects of GDP per capita on total memberships. As the literature sug- gests, this pattern probably occurs for either or both of two reasons. First, both inside and outside the workplace, economically devel- oped nations usually have greater structural heterogeneity, which generates more av- enues for specialized interests around which voluntary associations can arise and de- velop. Second, more prosperous nations nor- mally provide above-average financial well- being for their populations, and these greater material resources may give people the time, energy, and opportunity to engage in volun- tary activity. We must be cautious in endors- ing these interpretations, however, because the GDP effects are not statistically signifi- cant in the most rigorously controlled HLM analysis, which includes controls for the in- dividual-level predictors and the other coun- try-level predictors. Even so, the GDP ef- fects are in the expected direction in this HLM analysis.

The results of both forms of HLM analy- sis provide support for the effects of reli- gious tradition. Earlier, we described two variants of this explanation, having to do with whether we should find higher volun- tary activity in either Protestant or mixed Christian societies. Each variant receives support in the findings. As some have pre- dicted, Protestant countries are comparatively high in association involvement. However, the societies of the mixed Chris- tian type are often highest in membership rates, with Protestant countries ranking sec- ond. We noted that this pattern might be ex- pected from previous research in the sociol- ogy of religion that shows that societies or regions that are more heterogeneous and "competitive," in terms of the diversity of church types available for people to join, tend to have the highest religious attendance. In these contexts, church organizations ag- gressively recruit participants, thereby in- creasing attendance. We hypothesized that there might be a "spillover" effect into other types of voluntary association activity in these circumstances. This appears to be the case. The spillover may occur through the civic education function of church participa- tion, and also because additional (competing or complementary) bases for association joining are likely to arise in a climate of vig- orous church activity and involvement (Meyer and Staggenborg 1996; Verba et al. 1995). The question of which of these two processes provides the more appropriate ex- planation for our findings must be addressed in future research. Some who have argued that Protestant countries are high in associa- tion activity have also emphasized that this religious tradition places more emphasis on voluntary assistance to others and on the separation of state and religious functions. There is likely some validity to this interpre- tation, as well, because Protestant nations are comparatively high in association in- volvement. However, the fact that we have at times found even higher affiliation rates in the mixed Christian countries suggests other factors are also at work.

In keeping with another theory posed by various scholars, our analyses also confirm that number of years of continuous democ- racy has a significant positive effect on voluntarism, particularly for total member- ship levels. The proposed interpretation here is that it takes time for a social environment with an abundance of available associations to develop, and that this is best accomplished in stable democratic polities where there are few constraints on free association and on the formation of voluntary organiza- tions.

In the comparisons of association affilia- tion rates for different types of political sys- tems, the results are less consistent or con- clusive than for years of democracy. Based on the four political types used here, we find, as expected, that liberal and social de- mocracies rank highest in association activ- ity relative to the other political systems. Which of these two types ranks first de- pends, though, on the membership measure that is used. The hypotheses drawn from Janoski's (1998) work receive the most sup- port in the first set of HLM analyses, which show that social democracies tend to rank highest for total memberships but are second to liberal democracies for working member- ships. This reversal may occur because of better state funding of associations in the so- cial democracies, where organizations would then have less need for volunteer workers.

With controls for the effects of other coun- try-level characteristics, the differences across political types are generally not sig- nificant. At least for the present sample of democracies, we tentatively conclude that the length of experience with democracy, of whatever variety, is more important for pre- dicting association activity than is the type of democracy.

Overall, the HLM analyses suggest that economic, religious, and political factors are all helpful in explaining or understanding why countries differ in their levels of volun- tary association activity. The most fertile ground for associational involvement seems to be found in nations with considerable re- ligious diversity and sectarian autonomy, such as is found in societies that are either mixed Christian or predominantly Protes- tant, that have long-established democratic institutions and political systems, and that are relatively prosperous. The nations that epitomize this type of society in our sample of countries are mainly those in northern or western Europe, as well as the United States and Canada. At the other end of the con- tinuum, the countries with low levels of membership fit a profile involving a strong Orthodox, Catholic, and/or non-Christian re- ligious tradition, limited experience with a democratic political system, and relatively low economic development.

A few nations are considerably higher or lower in association involvement than we

would expect from their economic, religious, and political characteristics. In these in- stances, explanations based on other factors, such as unique features of the social struc- ture or culture of these societies, should be assessed in future studies. We conclude our discussion by considering what seem to be atypical or exceptional cases and by specu- lating on factors that may make them excep- tional.

Japan is perhaps the prime anomaly in our study. Despite being highly prosperous and democratic for several decades, Japan re- mains low in association activity. Possible reasons for this are its historical emphasis on the central role of the emperor, its highly centralized patterns of social organization, and the lingering influences of traditional belief systems, including Confucianism, Buddhism, and Shintoism, that shaped Japan's political structure. Some believe that these factors have probably slowed the de- velopment of voluntary associations in Japan (Azumi 1974; Hardacre 199 1 :2 19-22, 239; Lipset 1996:212-20.) If we consider, in ad- dition to these factors, that Japan lacks the diverse Protestant or multidenominational Christian religious composition that we have found is conducive to voluntarism, then Japan's low membership levels seem less anomalous.

Another apparent exception is South Ko- rea. Notwithstanding its relatively low rank- ing on GDP in the sample of nations consid- ered here, its fairly recent history of democ- racy, and its non-Christian heritage, South Korea is above average in association activ- ity. Religion, however, may be the key to explaining this finding. Although South Ko- rea has been a Buddhist nation historically, as of 1990 the combined Protestant and Catholic minorities almost equalled Bud- dhists among those professing a religious belief (Korean Overseas Information Service 1990:133).20 South Korea's history is also marked by many Christian, especially Prot- estant, influences. These include the influx of evangelical Presbyterian and Methodist

20 The 1990s World Values Survey data for South Korea are consistent with this pattern. Buddhists accounted for 36 percent of those voic- ing a religious preference, followed by Protestants at 30 percent and Catholics at 23 percent.

missionaries as early as the 1880s, the rapid growth of some 70 different Protestant de- nominations since the 1950s, and the promi- nent role played by Protestant and other Christian groups in the development of the Korean nation-state, in certain social move- ments, and in volunteer activities (Korean Overseas Information Service 1990: 14 1-43; Wells 1990). These considerations suggest that, especially in regard to religious diver- sity, South Korea has much in common with the mixed Christian nations in our analysis. When seen in this light, South Korea's high level of association involvement is not that surprising after all.

East Germany might seem to be another atypical case because, despite only recently moving to a democratic system, this nation is above average on all six measures of as- sociation involvement. However, when we note that East Germany has a mixed Chris- tian religious composition and that our analysis shows this factor to be related to high levels of association, the East German results are not so unexpected. In addition, East Germany, while not a highly prosper- ous country, is not poor either, ranking in the middle of the 33 nations on GDP per capita. The moderately high economic development in East Germany may also help explain its above-average level of voluntary activity. The East German case parallels that of South Korea, reminding us that an enduring and continuous democratic system is not a pre- requisite for association activity in all in- stances.

Another possible anomaly concerns some Latin American or Hispanic countries, most notably Brazil, Chile, and occasionally Mexico. These nations are unusual in that, at least for working memberships, they are higher than expected, given their predomi- nantly Catholic religious compositions, their lack of sustained experience with democratic institutions, and their less developed econo- mies. However, when religious associations are subtracted from the count of member- ships, the ranks of these nations fall signifi- cantly. In other words, membership in church-related organizations is largely re- sponsible for their somewhat higher work- ing membership rates. The finding of rela- tively high working involvement in church- based associations is consistent with analy- ses of "base Christian community" organi- zations in Brazil and other parts of Latin America (Azevedo 1987; Hewitt 1991; Mainwaring and Wilde 1989; Van Vugt 1991). This research shows that church-re- lated associations account for much of the voluntary activity at the local community level, suggesting that some qualification may be necessary when proposing that multidenominational or Protestant religious systems are always the best at promoting voluntarism.

Another interpretation involving the reli- gious factor, however, is suggested by Mar- tin (1990, chap. l), who reports an "explo- sion" of new Protestant religions, especially pentecostal and evangelical faiths, in Latin America in recent decades. These religions have stimulated the growth of self-help movements, mutual aid organizations, and social capital "networks between the state and the individual" (Martin 1996:37; also see Martin 1990, chaps. 4-6). This is espe- cially true in Brazil, Chile, and to some ex- tent Mexico, but not in Argentina, Martin argues. It is possible, then, that the Protes- tant influence, or the effect of increased re- ligious pluralism more generally, helps ex- plain the higher-than-expected levels of working memberships in Brazil, Chile, and Mexico. In addition, the relatively weak role of Protestantism in Argentina may ac- count for this Latin American nation's com- paratively low rank in working member- ships in our analysis.

A general alternative interpretation for those nations with low membership rates is that, regardless of religious, political, or eco- nomic factors, societies characterized by ex- tended family systems may be lower in vol- untary association activity than societies without this characteristic. In other words, the large families in these societies may be more likely to address needs served else- where by voluntary associations. Reher's (1998:203, 209) analysis of family ties across societies suggests that this interpreta- tion is worthy of consideration. In particu- lar, his research indicates a "dividing line" between southern European societies, with their history of depending on strong and ex- tended families to care for the elderly and the poor, for example, versus northern Euro- pean and North American societies, with their weaker family systems and greater re- liance on public and private organizations to provide social assistance. This finding is generally consistent with our results, for we found that these same northern European na- tions, along with the United States and Canada, rank highest in voluntary organiza- tion activity, while southern European coun- tries typically rank among the lowest. Deter- mining whether family structure is a key fac- tor behind the low level of involvement in southern Europe, and whether this explana- tion helps account for the low membership levels in Japan and the former eastern bloc nations, is an intriguing hypothesis to con- sider in future research (see Janoski 1998: 116-17).

Our comparative analyses confirm that, among democratic societies, nation of resi- dence is a relatively strong predictor of as- sociation involvement. In addition, our find- ings show the value of juxtaposing country- by-country comparisons with analyses that consider the effects of country-level charac- teristics. The former are crucial for identify- ing nations with involvement rates that are lower or higher than expected based on ex- isting macro-level theories, while the latter provide tests of theories that explain the overall patterns. The country-by-country analyses guard against overgeneralization, reveal alternative interpretations, and sug- gest more qualified explanations of cross- national differences. Closer consideration of the exceptional cases in our analysis sug- gests, in particular, the importance of taking into account the additional effects of reli- gious composition.

A key task for the future is to apply both country-by-country analyses and the various country-level explanations to even more so- cieties, especially to nations that are not de- mocracies. Besides providing a wider inter- national context, such analyses will make it possible to elaborate on and add to the four explanations for cross-national differences in association membership considered here.

James E. Curtis is Professor of Sociology at the University of Waterloo, Waterloo, Ontario, Canada. His research has dealt with various as- pects of Canadian society, including works on social inequality, politics, ideology, sports, and voluntary association involvement. His current research includes cross-national analyses of pre- dictors of social capital and social participation. Recent publications include Social Inequality in Canada: Patterns, Problems, Policies (edited with Edward Grabb and Neil Guppy, 3d ed., Prentice- Hall Canada, 1999) and The Vertical Mosaic Re- visited (with Rick Helmes-Hayes, University of Toronto Press, 1998).

Douglas E. Baer is Associate Professor and Chair at the University of Victoria, Victoria, British Columbia, Canada. His recent research produced the book, Political Sociology: Canadian Perspectives (Oxford University Press, 2002). He has also completed a study of the relationship between religion and voluntary association par- ticipation in cross-national perspective, and a work on the relationship between political ideol- ogy and left party support in Western democra- cies.

Edward G. Grabb is Professor of Sociology at the University of Western Ontario, London, Ontario, Canada. He has recently published

Theories of Social Inequality: Classical and Con- temporary Perspectives (4th ed., Harcourt Canada, 2002). His current research focuses pri- marily on comparisons of the values and behav- iors of Canadians and Americans, as well as on international variations in social participation.

Almond, Gabriel and Sidney Verba. 1963. The Civic Culture. Princeton, NJ: Princeton Uni- versity Press.

Azevedo, Marcello de C. 1987. Basic Ecclesial Communities in Brazil. Washington, DC: Georgetown University Press.

Azumi, Koya. 1974. "Voluntary Organizations in Japan." Pp. 15-26 in Voluntary Action Re- search, edited by D. H. Smith. Lexington, MA: Lexington Books.

Baer, Douglas, James Curtis, and Edward Grabb. 2001. "Has Voluntary Association Activity Declined? Cross-National Analyses for 15 Countries." Canadian Review of Sociology and Anthropology 38:250-74.

Barrett, David B., ed. 1982. World Christian En- cyclopedia. Oxford, England: Oxford Univer- sity Press.

Bergalli, Robert. 1997. "The New Order in Spain and an Hispanic Perspective on the History and Meaning of Social Control." Pp. 34-5 1 in Social Control and Political Order, edited by R. Bergalli and C. Sumner. London, England: Sage.

Bibby, Reginald. 1987. Fragmented Gods. Toronto, Ontario: Irwin. Bryk, Anthony and Stephen Raudenbush. 1992. Hierarchical Linear Models. Newbury Park,

CA: Sage.

Chaves, Mark and David E. Cann. 1992. "Regu- lation, Pluralism, and Religious Market Struc- ture: Explaining Religion's Vitality." Rationality and Society 4:272-90.

Crenshaw, Edward M. 1995. "Democracy and Demographic Inheritance: The Influence of Modernity and Proto-Modernity on Political and Civil Rights, 1965 to 1980." American So- ciological Review 60:702-18.

Curtis, James. 1971. "Voluntary Associations Joining: A Cross-National Comparative Note." American Sociological Review 36:872-80.

Curtis, James, Edward Grabb, and Douglas Baer. 1992. "Voluntary Association Membership in Fifteen Countries: A Comparative Analysis." American Sociological Review 57:139-52.

Davis, Nancy J. and Robert V. Robinson. 1999. "Their Brothers' Keepers? Orthodox Religion- ists, Modernists, and Economic Justice in Eu- rope." American Journal of Sociology 104:1631-66.

Drucker, Peter. 1994. "The Age of Social Trans- formation." Atlantic Monthly 274:53-80.

Duke, James T., Barry L. Johnson, and James B. Duke. 1993. "Rate of Religious Conversion: A Macrosociological Study." Research in the So- ciology of Religion 5:89-121.

Encyclopedia Britannica. 1994. Britannica Book of the Year 1994. Chicago, IL: Encyclopedia Britannica, Inc.

Esping-Andersen, Gosta. 1990. Three Worlds of Welfare Capitalism. Cambridge, England: Pol- ity.

Finke, Roger and Rodney Stark. 1998. "Religious Choice and Competition." American Socio- logical Review 63:761-66.

Hallenstvedt, Abraham. 1974. "Formal Volun- tary Associations in Norway." Pp. 219-30 in Voluntary Action Research, edited by D. H. Smith. Lexington, MA: Lexington Books.

Han, Sungjoo. 1974. The Failure of Democracy in South Korea. Berkeley, CA: University of California Press.

Hardacre, Helen. 1991. "Japan: The Public Sphere in a Non-Western Setting." Pp. 217-42 in Between States and Markets: The Voluntary Sector in Comparative Perspective, edited by

R. Wuthnow. Princeton, NJ: Princeton Univer- sity Press.

Hewitt, W. E. 1991. Base Christian Communities and Social Change in Brazil. Lincoln, NE: University of Nebraska Press.

Huber, Evelyne and John D. Stephens. 1998. "In- ternationalization and the Social Democratic Model: Crisis and Future Prospects." Comparative Political Studies 3 1 :353-97.

Inglehart, Ronald. 1997. Modernization and Postmodernization: Cultural, Economic, and Political Change in 43 Societies. Princeton,

NJ: Princeton University Press.

Inglehart, Ronald and Wayne E. Baker. 2000. "Modernization, Cultural Change, and the Per- sistence of Traditional Values." American So- ciological Review 65: 19-5 l.

Janoski, Thomas. 1998. Citizenship and Civil So- ciety: A Framework of Rights and Obligations in Liberal, Traditional, and Social Democratic Regimes. Cambridge, England: Cambridge University Press.

Janoski, Thomas and John Wilson. 1995. "Path- ways to Voluntarism: Family Socialization and Status Transmission Models." Social Forces 74:271-92.

Knoke, David. 1986. "Associations and Interest Groups." Annual Review of Sociology 12: 1-21.

Korean Overseas Information Service. 1990. Handbook of Korea, 1990. Seoul, South Ko- rea: Korean Overseas Information Service.

Ladd, Everett Carl. 1999. The Ladd Report. New York: Free Press.

Leibfried, Stephan. 1992. "Towards a European Welfare State." Pp. 245-79 in Social Policy in a Changing Europe, edited by Z. Z. Ferge and

J. E. Kolberg. Boulder, CO: Westview.

Lipset, Seymour Martin. 1967. "Values, Educa- tion, and Entrepreneurship." Pp. 3-60 in Elites in Latin America, edited by S. M. Lipset and

A. Solari. New York: Oxford University Press.

. 1985. "Canada and the United States: The Cultural Dimension." Pp. 109-60 in Canada and the United States, edited by C. F. Doran and J. H. Sigler. Engelwood Cliffs, NJ: Prentice-Hall.

. 1990. Continental Divide: The Values and Institutions of Canada and the United States. New York: Routledge.

. 1994. "The Social Requisites of Democ- racy Revisited." American Sociological Re- view 59: 1-22.

. 1996. American Exceptionalism: A Double-Edged Sword. New York: W.W. Norton.

Long, J. Scott. 1997. Regression Models for Cat- egorical and Limited Dependent Variables. Newbury Park, CA: Sage.

Mainwaring, Scott and Alexander Wilde, eds. 1989. The Progressive Church in Latin America. South Bend, IN: Notre Dame Univer- sity Press.

Martin, David. 1990. Tongues of Fire: The Ex- plosion of Protestantism in Latin America. Cambridge, MA: Basil Blackwell. . 1996. Forbidden Revolutions. Pente- costalism in Latin America, Catholicism in Eastern Europe. London, England: Society for Promoting Christian Knowledge.

McMullen, Mike. 1994. "Religious Polities as Institutions." Social Forces 73:709-28.

Meyer, David S. and Suzanne Staggenborg.

1996. "Movements, Countermovements, and the Structure of Political Opportunity." American Journal of Sociology 101:1628-60.

Oh, John Kie-Chang. 1999. Korean Politics. Ithaca, NY: Cornell University Press.

Olson, Daniel V. A. 1998. "Religious Pluralism in Contemporary U.S. Counties." American Sociological Review 63:759-6 1.

Paxton, Pamela. 1999. "Is Social Capital Declin- ing in the United States? A Multiple Indicator Assessment." American Journal of Sociology 105:88-127.

Pugliese, Donato. 1986. "Introduction." Pp. 3-14 in Volunteer Associations: An Annotated Bib- liography. New York: Garland.

Putnam, Robert. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton, NJ: Princeton University Press. . 1995. "Bowling Alone: America's De- clining Social Capital." Journal of Democracy 6:65-78. . 2000. Bowling Alone: The Collapse and Revival of American Community. New York: Simon and Schuster.

Raudenbush, Stephen, Anthony Bryk, Yuk Fai Cheong, and Richard Congdon. 2000. HLM 5: Hierarchical Linear and Non-Linear Model- ing. Chicago, IL: Scientific Software Interna- tional.

Reher, David Sven. 1998. "Family Ties in West- ern Europe: Persistent Contrasts." Population and Development Review 24:203-34.

Rotolo, Thomas. 1999. "Trends in Voluntary As- sociation Participation." Nonprofit and Volun- tary Sector Quarterly 28: 199-212.

Ryan, Mary P. 1999. "Civil Society as Demo- cratic Practice: North American Cities During the Nineteenth Century." Journal of Interdis- ciplinary History 29:559-84.

Schlesinger, Arthur M. 1944. "Biography of a Nation of Joiners." American Historical Re- view 50: 1-25.

Smith, David Horton. 1972. "Modernization and the Emergence of Voluntary Organizations." International Journal of Comparative Sociol- ogy 13: 113-34. . 1975. "Voluntary Action and Voluntary Groups." Annual Review of Sociology 1:247


Smith, David Horton and Burt R. Baldwin. 1983. "Voluntary Organization Prevalence in the States of the U.S.A.: Presentation and a Test of the Revised Smith Model." Pp. 50-62 in International Perspectives on voluntary Action Research, edited by D. H. Smith and J. Van Til. Washington, DC: University Press of


Stark, Rodney and Lawrence Iannaccone. 1994. "A Supply-Side Reinterpretation of the 'Secu- larization' of Europe." Journal for the Scien- tific Study of Religion 33:230-52.

Stark, Rodney and James C. McCann. 1993. "Market Forces and Catholic Commitment: Exploring the New Paradigm." Journal for the Scientific Study of Religion 32: 11 1-23.

Tocqueville, Alexis de. [I8351 1961. Democracy in America. New York: Knopf.

Van Vugt, Johannes P. 1991. Democratic Orga- nization for Social Change: Latin American Christian Base Communities and Literacy Campaigns. New York: Bergin and Garvey.

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

Weber, Max. 19 11. "Deutscher Sociologentag" (German sociology today). Verhandlungen 1:39-62.

Wells, Kenneth. 1990. New God, New Nation. North Sydney, Australia: Allen Unwin.

Western, Bruce. 1995. "A Comparative Study of Working-Class Disorganization: Union De- cline in Eighteen Advanced Capitalist Coun- tries." American Sociological Review 60: 179-


Wiarda, Howard J. 1986. "Toward a Framework for the Study of Political Change in the Iberic- Latin Tradition: The Corporative Model." Pp. 249-78 in Political System and Change, edited by I. Kabashima and L. T. White 111. Princeton, NJ: Princeton University Press.

Wilson, John and Marc Musick. 1997. "Who Cares? Toward an Integrated Theory of Vol- unteer Work." American Sociological Review 62:694-7 13.

World Bank. 1992. World Development Report 1992. Development and the Environment. New York: Oxford University Press.

Wood, Gordon. 1992. The Radicalism of the American Revolution. New York: Knopf.

Wright, Charles and Herbert H. Hyman. 1958. "Association Memberships of American Adults: Evidence from National Survey Samples." American Sociological Review 23:284-94.

Wuthnow, Robert. 1991. "Tocqueville's Question Reconsidered: Volunteerism and Public Discourse in Advanced Industrial Societies." Pp. 288-308 in Between States and Markets: The Voluntary Sector in Comparative Perspec- tive, edited by R. Wuthnow. Princeton, NJ: Princeton University Press.

  • Recommend Us