#2789-ASR 71:2 filename:71202-ruiter National Context, Religiosity, and Volunteering: Results from 53 Countries Stijn Ruiter Nan Dirk De Graaf Radboud University Nijmegen Radboud University Nijmegen To what extent does the national religious context affect volunteering? Does a religious environment affect the relation between religiosity and volunteering? To answer these questions, this study specifies individual level, contextual level, and cross-level interaction hypotheses. The authors test the hypotheses by simultaneously studying the impact of religiosity of individuals, the national religious context, and their interplay on volunteering while controlling for possible confounding factors both at individual and contextual levels. Based on multilevel analyses on data from 53 countries, frequent churchgoers are more active in volunteer work and a devout national context has an additional positive effect. However, the difference between secular and religious people is substantially smaller in devout countries than in secular countries. Church attendance is hardly relevant for volunteering in devout countries. Furthermore, religious volunteering has a strong spillover effect, implying that religious citizens also volunteer more for secular organizations. This spillover effect is stronger for Catholics than for Protestants, non-Christians and nonreligious individuals. Delivered by Ingenta to : among the most common voluntary organizahurch members generally are more involved of Groningen University 29 Feb 2008 10:55:03 in voluntary organizations thanFri, nonmemtions (Gaskin and Davis Smith 1995). bers (Becker and Dhingra 2001; Curtis, Baer, Although church members volunteer more and Grabb 2001; Hodgkinson 2003; Lam 2002; than nonmembers, research suggests that volWilson and Janoski 1995).1 Not only are memunteering is not driven by church membership, bership rates among them higher, they are also but instead by levels of church attendance (De more likely to volunteer. Therefore, it comes as Hart 1999; Hodgkinson, Weitzman, and Kirsch no surprise that voluntary organizations often 1990; Watt 1991; Wilson and Musick 1997). have religious backgrounds. This is not only Nominal members who never visit church and true for the United States, where over one third nonmembers volunteer equally often, whereas of all volunteers is active in religious organizathe highest volunteer rates are found among tions (Boraas 2003). Even in more secular frequent churchgoers. Regular churchgoers are European societies, religious organizations are better integrated within religious networks than those who never attend church. Being part of such networks enhances the chance to volunteer (Bekkers 2003). Direct correspondence to Stijn Ruiter, Radboud Nonmembers volunteer less. But, what if University Nijmegen, ICS/Sociology, PO Box 9104, 6500 HE Nijmegen, Netherlands (s.ruiter@ru.nl). they are nested within a highly religious context? An earlier version of this article was presented at the Most probably, in such a scenario, they have a ESRC Research Seminar on Politics and Religion in considerable number of active church members Oxford, UK, September 13, 2004. The authors thank amongst their family, friends, and acquainTom Snijders, Ruud van der Meulen, Paul de Graaf, tances. Will this make them more likely to voland the anonymous ASR reviewers for their helpful unteer? We know that avid church members comments. volunteer more because they are better inte1 Throughout this article, we use “church” also for grated within religious networks (Bekkers 2003; non-Christian holy places (e.g., mosques and synaWilson and Musick 1997). They are more likegogues). Consequently, a non-Christian is also called ly to know about voluntary organizations, more a church member. C AMERICAN SOCIOLOGICAL REVIEW, 2006, VOL. 71 (April:191–210) #2789-ASR 71:2 filename:71202-ruiter 192—–AMERICAN SOCIOLOGICAL REVIEW likely to be asked to participate (Bryant et al. increasingly less likely for nonmembers to have 2003; Musick, Wilson, and Bynum 2000), and active church members within their social enviit will be harder for them to refuse such requests ronment. This could, indirectly, cause a nation’s (Snow, Zurcher, Ekland-Olson 1980). However, volunteer rate to drop as well. it is unclear to what extent this network arguTo test for effects of the national religious ment applies also to people who themselves context, we have to study the impact of indinever visit church. Are they affected by the relividual religiosity and national religious congiosity of their social environment too? This text simultaneously while controlling for would make the network explanation more genconfounding factors at both individual and coneral: being more strongly integrated within nettextual levels. This requires international comworks of religious people makes one more likely parative research. Smith (1994) and Wilson to volunteer. This should then hold for all peo(2000) conclude in their overview articles that ple, not just for regular churchgoers. research on the impact of the context on indiThe general network hypothesis implies difvidual volunteering is underdeveloped. ferences between countries: citizens in more Similarly, Curtis et al. argue that scholars paid devout societies are more likely to associate little attention to developing “theories of crosswith active church members (cf. Kelley and De national variation in association involvement” Graaf 1997), which should increase their chance (2001:784). Hodgkinson also concludes that to volunteer. If the religious environment indeed “future research would greatly benefit from a is an important predictor, we expect not just stronger theoretical base to explain the differhigher volunteer rates in more devout societies ences in rates of volunteering across nations” (i.e. a composition effect), but also that all cit(2003: 52). Up until now, international comizens of those societies—irrespective of their parative research on voluntary association own religiosity—are more likely to volunteer. involvement (studying either memberships or Hence, we predict that devoutness ofDelivered society isby Ingenta volunteering, to : and sometimes both) has focused positively related to participation level (i.e. a of Groningen mainly on political and economic factors (e.g., University context effect). Secular people in devout counFri, 29 Feb 2008Salamon 10:55:03and Sokolowski 2003; Hodgkinson tries should also volunteer more, since knowl2003; Schofer and Fourcade-Gourinchas 2001). edge about volunteering reaches them more However, because we hypothesize that religion easily through their social networks and they are is a key factor, we propose that international more likely to be recruited and motivated by the comparative research on volunteering should large number of religious fellow citizens. The take the national religious context into account national religious context effect could imply a as well. Although religion is included in some dampened effect of individual religiosity in recent studies (Dekker and Van den Broek 1998; more devout societies, since people who do not Halman 2003), these studies do not discriminate attend church would be more likely to volunteer compositional from contextual effects. Curtis et and frequent churchgoers would not have to al. (2001) do distinguish these effects and find invest much time to sustain high levels of volthat working memberships (religious organizaunteering. tions and unions excluded) vary with national Whether people’s religious environment influreligious composition.2 Parboteeah, Cullen, and ences their volunteering becomes an important Lim (2004) also find a strong positive effect of question if we consider that many industrialized the national religious context, but unfortunatecountries secularized in the past decades (with ly they do not control for church attendance at the United States as a possible exception: cf. the individual level. We try to add to this Norris and Inglehart 2004; De Graaf and Need 2000). If religiosity indeed is an important factor, secularization might result in declining vol2 Exactly the same World Values Survey items on unteer rates for two reasons. First, the number unpaid work for voluntary organizations are someof avid church members—the people who are times used to measure working memberships (e.g., most likely to volunteer—declines. Curtis et al. 2001), while in other cases they are said Consequently, levels of volunteering could drop to measure (formal) volunteering (e.g., Halman 2003; sharply. Second, while more and more people Hodgkinson 2003; Parboteeah, Cullen, and Lim turn their back on the church, it becomes 2004). We address this issue in the data section. #2789-ASR 71:2 filename:71202-ruiter RELIGIOSITY AND VOLUNTEERING—–193 research by providing new hypotheses on the influence of religious context and by testing them on a large-scale, international comparative dataset containing information on volunteering in 53 countries between 1981 and 2001. THEORY AND HYPOTHESES Volunteering is a form of institutionalized, unpaid helping behavior that benefits other people, groups, or organizations (Hodgkinson 2003; Wilson 2000). Although such behavior could be beneficial to volunteers themselves, they certainly do not gain financially (Dekker and Halman 2003). Thus, to initiate volunteerism, people have to be motivated in another way. To understand why people volunteer, we build upon previous research distinguishing three types of explanations: (1) altruistic norms, (2) social networks, and (3) knowledge and skills.3 First, we elaborate on the relation between individual religiosity and volunteering. Subsequently, we specify hypotheses on effects of the national religious context. knit communities, they are more likely to know about existing voluntary organizations, it increases the chance that they are asked to participate (Bryant et al. 2003; Musick, Wilson, and Bynum 2000), and furthermore, the close-knit community makes it difficult to refuse such requests. Wilson and Musick refer to this network explanation when they argue that “most formal volunteers are persuaded to volunteer by family members, coworkers, or fellow worshippers” (1997:700). Thus, based on both the altruism and the network explanation, our hypothesis is as follows: Hypothesis 1: Church members volunteer more than nonmembers. The explanation that church members volunteer for altruistic reasons is under attack. Cnaan, Kasternakis, and Wineburg (1993) show that volunteers and nonvolunteers do not differ in their intrinsic religious motivation. Bekkers (2003) finds that volunteering directly varies with altruism, but altruism does not explain why church members volunteer more than nonmembers. By contrast, the social network explaDelivered by Ingenta to : nation has received strong support (Becker and University HYPOTHESES ON INDIVIDUAL RELIGIOSITY AND of Groningen Dhingra 2001; Bekkers 2003; Jackson et al. Fri, 29 Feb 2008 10:55:03 VOLUNTEERING 1995; Lam 2002; Park and Smith 2000; Yeung Two general explanations for why church mem2004). Empirical evidence suggests that church bers volunteer more than nonmembers have members have more active volunteers within been proposed. First, avid church members their social networks. Consequently, church would have internalized the social norm of altrumembers are more likely to meet other volunism (the intrinsic motivation to help others) teers and be recruited by them. Becker and Dhingra underline the power of this social netmore than nonmembers. Solidarity, love of one’s work explanation stating “social networks, neighbors, and self-sacrifice are important rather than beliefs, dominate as the mechanism virtues promoted by the world’s major religions leading to volunteering . . . the importance of (Wuthnow 1991:122). Religious people would religious beliefs plays little role in church attenacquire these altruistic norms in church and ders’ decisions to volunteer” (2001:329–30). this would lead to an increased chance of volThe fact that most volunteers were asked to unteering. Second, according to the social netjoin (Boraas 2003; Gaskin and Davis Smith work explanation, church members would 1995) is also in-line with the social network volunteer more because their social network explanation. provides stronger recruitment and motivation The network explanation suggests the imporfunctions. Given that members are part of closetance of church attendance. Regular churchgoers are more strongly connected to their religious group, making them more likely to be asked to 3 Wilson and Musick (1997) and Oesterle, Johnson, volunteer. Most probably, this recruitment mechand Mortimer (2004) group these explanations under anism is considerably weaker for those church the following headings: (1) cultural, (2) social, and members who only occasionally visit church. (3) human capital. We do not use these highly abstract This is exactly what De Hart (1999) and Bekkers terms, because their wide use have made their mean(2003) find for the Netherlands: nonmembers ing unclear, which makes accumulation of knowledge and members who attend church infrequently difficult. #2789-ASR 71:2 filename:71202-ruiter 194—–AMERICAN SOCIOLOGICAL REVIEW are almost equally less likely to volunteer, whereas church members who often visit church are much more likely to volunteer. Wuthnow also indicates this when he argues that “religious inclinations make very little difference unless one becomes involved in some kind of organized religious community” (1991:156). Because we expect church attendance to be highly influential, we predict the following: Hypothesis 2: Differences in volunteering between church members and nonmembers decline strongly after taking church attendance into account. spillover effect. On top of the network explanation, they add that people active in religious volunteering obtain specific skills that are valuable for secular voluntary organizations as well. This would make them more likely to be recruited by secular organizations, because organizations are actively “prospecting for participants” who have these skills (Brady, Schlozman, and Verba 1999). Thus, the spillover hypothesis reads as follows: Hypothesis 4a: People who do religious volunteer work are more likely to volunteer for secular organizations as well. According to Wilson and Janoski (1995), Church members and nonmembers may difsome conservative Protestant denominations fer considerably with regard to their volunteerdiscourage secular volunteering. Volunteering ing, this does not make all members equally for organizations that are directly linked to the involved. Research shows large denominationchurch would be strongly supported though. al differences among church members. In genTherefore, strong integration within these eral, Protestants seem to volunteer more than denominations should lower the chance to volCatholics (Curtis et al. 2001; Dekker and De unteer for secular organizations and raise the Hart 2001; Lam 2002; Wuthnow 1991:322). chance of religious volunteering. Consequently, Dekker and De Hart (2001) provide an instituthere should be no spillover effect for these tional explanation: the Protestant church is less denominations. Park and Smith (2000) indeed hierarchically structured than theDelivered Catholicby Ingenta to : find that high church attendance reduces the church. Furthermore, the Protestant church sub- of Groningen University divides into smaller parishes. Both the less Fri, 29hierFeb 2008probability 10:55:03 of volunteering through a nonchurch organization among churchgoing Protestants. archical structure and the smaller subdivisions However, since we cannot distinguish between would generate more involvement. Lam different Protestant denominations within this (2002:408) argues that “Protestant principles study, we expect the following: discourage the pursuit of self-interests and induce a sense of social responsibility among Hypothesis 4b: Conservative Protestants lower their adherents.” Based on these explanations, the overall spillover effect for Protestants our next hypothesis is as follows: compared to the spillover effect for other denominations. Hypothesis 3: Protestants volunteer more than Catholics. All previous hypotheses are about volunteering in general. However, it is reasonable to expect that religious involvement boosts volunteering for religious voluntary organizations to a larger extent than volunteering for secular organizations (Wuthnow 1999). Therefore, we have to study the impact of religious involvement on general volunteering apart from its effect on secular volunteering. However, the network explanation applies again. If those already involved in religious volunteering are more likely to get acquainted with people who volunteer for secular organizations, their chance to volunteer for secular organizations should be high as well. Jackson et al. (1995) and Dekker and De Hart (2002) find evidence for such a HYPOTHESES ON NATIONAL RELIGIOUS CONTEXT AND VOLUNTEERING So far, we have formulated hypotheses on the impact of individual religiosity on volunteering. Next, we elaborate on the relation between the national religious context and volunteer work. This relation is somewhat neglected in the literature. However, Kelley and De Graaf (1997) provide us with arguments to predict a positive impact of devoutness of a society. They find that people who were raised by secular parents in relatively devout countries are more religious than people who grew up with similar parents in more secular countries. According to Kelley and De Graaf, this comes about (a) through #2789-ASR 71:2 filename:71202-ruiter RELIGIOSITY AND VOLUNTEERING—–195 people’s exposure to religious culture and (b) because the pools of potential friends, teachers, colleagues, and marriage partners are predominantly devout. We expect that these two phenomena influence volunteering as well. As we have already argued, especially the social network of people influences their chance to volunteer. Under the assumption that people in devout countries have an increased number of religious people in their social networks, they likely acquaint themselves with an increased number of already active religious people who possibly recruit them for volunteer activities. Moreover, in such networks, the norm to volunteer could be stronger, as well as the social pressure to behave accordingly. Based on these arguments, we predict the following: Hypothesis 5: The chance to volunteer increases with the devoutness of the society. strongly influenced by church attendance in secular societies. Conversely, in devout societies, if nonreligious people indeed have a higher chance to be involved in volunteering (see Hypothesis 5), this relieves avid churchgoers of the task to invest much time keeping levels of volunteering high. Besides, if levels of volunteering are already high, the added value of an additional volunteer is lower. Based on these arguments we expect the differences between frequent and infrequent churchgoers to be smaller, implying that church attendance does not strongly affect volunteering in devout societies. In sum, we expect the following: Hypothesis 6: Individual church attendance influences volunteering less in more devout societies. Furthermore, both theoretical arguments DATA AND METHODS and empirical findings lead us to expect that the impact of individual church attendance on For the test of our hypotheses, we use a convolunteering varies with the national religious catenated dataset from the 1981–1984, context. Hypothesis 2 suggests thatDelivered all peopleby Ingenta to : 1990–1993, and 1999–2001 waves of the are influenced equally by church attendance. University of Groningen European Values Surveys/World Values However, we argue that the impactFri, is weaker 29 Feb 2008 10:55:03 Surveys (European Values Study Group and in more devout societies and stronger in more World Values Survey Association 2005; World secular societies. Kelley and De Graaf (1997) Values Study Group 1999). These waves confind similar cross-level interaction effects for tain similar questions on volunteering for seven religiosity: religious upbringing influences types of organizations. Unfortunately, the third individual’s religiosity in devout countries only marginally, whereas its effect is strong in secwave of the World Values Survey (1995–1997) ular societies. They argue that religious parents cannot be used, because the questions on volin secular societies have to invest more to untary participation were changed too much. inculcate religious beliefs in their children, We select only those countries for which valid whereas in devout societies, the religious conscores on all dependent and independent varitext already produces much of this socializaables were available.5 Furthermore, only peotion. Consequently, the effect of parental ple between 18 and 90 years old are selected. religiosity is smaller in more religious contexts. These selections result in a dataset of 117,007 As an analogy, we argue that frequent churchindividuals distributed over 53 countries and goers in secular societies face the problem of the three waves. The distribution of all responinsufficient volunteer involvement. Therefore, dents over the countries and waves as well as building up and sustaining vibrant voluntary 4 average volunteer rates per country are disorganizations rest mainly on their shoulders. played in Table 1. As a result, the chance to volunteer should be 4 We assume that networks of secular people do not negatively affect the voluntary participation of religious people in secular countries. This seems plausible, since absence of a norm to volunteer does not mean there exists a norm not to volunteer. 5 Because in the Chinese questionnaire, serious translation errors were made with respect to our dependent variable (see codebook for World Values Study Group 1999), we do not include the Chinese dataset. #2789-ASR 71:2 filename:71202-ruiter 196—–AMERICAN SOCIOLOGICAL REVIEW Table 1. Respondents per Country and Wave and Percentage Volunteers per Country Respondents Country Albania Argentina Austria Belarus Belgium Brazil Bulgaria Canada Chile Croatia Czech Republic Denmark Estonia Finland France Germany Germany (East) Germany (West) Greece Hungary Iceland India Ireland Italy Japan Latvia Lithuania Luxembourg Macedonia Mexico Moldova Netherlands Northern Ireland Norway Peru Philippines Poland Portugal Puerto Rico Romania Russia Slovakia Slovenia South Africa Spain Sweden Tanzania Turkey Uganda UK Ukraine USA Zimbabwe Total 1981–1984 1990–1993 1999–2001 .— .— 00,904 00,529 00,592 01,210 .— 01,402 01,497 .— .— 00,977 00,976 02,624 01,762 .— 01,460 .— .— 00,945 00,965 01,187 01,664 01,868 .— 01,450 01,136 .— .— 00,912 .— .— 01,809 01,181 01,008 00,999 .— 00,977 .— 00,582 00,927 01,197 00,914 01,517 .— .— 01,999 .— 01,325 .— 01,273 02,063 .— .— .— 01,050 .— 00,666 00,973 00,797 00,678 00,944 .— .— 01,337 01,180 Delivered00,995 by Ingenta to00,934 : 01,345 University 01,964 01,926 of Groningen 01,167 Fri, 29 Feb 00,960 01,257 2008 10:55:03 .— 00,432 00,937 .— .— 00,970 .— .— 01,005 .— .— 01,055 .— 01,168 01,120 .— .— 01,008 01,116 01,002 00,991 00,310 00,302 00,926 01,208 01,180 .— .— .— 01,395 .— .— 01,181 .— .— 01,084 .— 01,080 00,895 .— .— 00,551 .— 01,089 01,051 .— 01,654 02,450 .— .— 01,327 .— 00,934 00,966 .— .— 02,621 02,297 03,670 02,215 00,832 00,922 00,948 .— .— 00,844 .— .— 00,944 .— .— 00,975 01,192 01,458 00,838 .— .— 01,116 02,214 01,536 01,135 .— .— 00,859 20,001 37,719 59,287 Volunteers Total Average % 0v0,904 002,331 002,899 000,977 005,362 001,460 001,910 004,719 002,586 000,912 001,809 003,188 000,977 001,509 003,628 001,999 001,325 003,336 001,050 001,639 002,419 001,337 003,109 005,235 003,384 001,369 000,970 001,005 001,055 002,288 001,008 003,109 001,538 002,388 001,395 001,181 001,084 001,975 000,551 002,140 004,104 001,327 001,900 002,621 008,182 002,702 000,844 000,944 000,975 003,488 001,116 004,885 000,859 117,007 42.1 15.5 17.5 14.9 19.5 21.2 14.6 31.5 25.9 15.5 18.5 14.5 12.1 25.6 13.6 10.2 28.2 18.0 31.9 12.2 22.2 33.0 17.4 17.5 11.2 17.3 10.0 20.1 25.1 25.4 31.7 25.7 17.1 20.6 33.5 42.7 09.6 11.7 41.7 17.5 09.9 31.7 13.7 44.3 11.4 29.2 74.2 05.5 57.1 16.6 09.1 38.4 58.9 21.4 #2789-ASR 71:2 filename:71202-ruiter RELIGIOSITY AND VOLUNTEERING—–197 DEPENDENT VARIABLES Although there exists considerable variation over the waves, countries with high volunteer rates in one wave generally also have high rates in the other waves. The Pearson R correlation between the volunteer rates in 1981–1984 with those of 1990–1993 is .75 (N = 17), for 1990–1993 and 1999–2001 it is .65 (N = 26), and for 1981–1984 and 1999–2001 it is .67 (N = 15). Averaged for all countries and waves, 21.4 percent of the respondents are active volunteers. For the variable nonreligious volunteering, we exclude volunteer work for religious and church organizations from general volunteering.7 Since most volunteer work is done in religious organizations (over 46 percent of all volunteering for the seven different organizations is done for religious organizations), general volunteering and nonreligious volunteering are far from identical. Although the complete list of voluntary organizations differs over the three waves, all three waves contain similar questions on volunteering for seven different types of organizations. Respondents were asked whether they were doing unpaid work for one or more of the following organizations: (1) social welfare services for elderly, handicapped, or deprived people; (2) religious or church organizations; (3) education, arts, music, or cultural activities; (4) labor unions; (5) political parties or groups; (6) conservation, environment, and animal rights groups; and (7) professional associations. Because respondents were first asked whether they were members of these organizations, some scholars refer to the unpaid work with “working membership” (e.g., Curtis et al. 2001) rather than volunteering. However, others use exactly the same survey items and call it volunteering INDEPENDENT VARIABLES (e.g., Halman 2003; Hodgkinson 2003; Inglehart 2003; Parboteeah et al. 2004). Because Basic descriptive statistics of the independent we believe the items fit the aforementioned defvariables are displayed in Table 2. Church meminition of volunteering, we refer to them asby Ingenta Delivered bershiptois: measured by the question whether University such. However, to determine whether results of Groningen people belong to a religious denomination, and Fri, Feb 2008 10:55:03 differ for memberships, we estimate our29final if so, which one. Question formulation differed models for volunteering as well as membersomewhat between wave 1981–1984 and waves ships. 1990–1993 and 1999–2001. Also, in some counWe construct the variable general volunteertries, different answer categories were used. ing by assigning a score of 1 to respondents who However, we are able to assign respondents to did unpaid work for at least one of the seven the following four categories: Catholics, organizations.6 We assign a score 0 to responProtestants, non-Christians, and nonreligious.8 dents who did not do any volunteer work. We Originally the answer categories for the quesuse this dichotomous dependent variable (cf. tion on church attendance (apart from weddings, funerals, and christenings) ranged from Curtis, Grabb, and Bear 1992; Wuthnow 1999) “(practically) never” to “more than once a instead of a count variable (cf. Curtis et al. week.” We recode that ordinal variable to the 2001; Parboteeah et al. 2004) for two reasons. First, our hypotheses pertain to the chance of volunteering and not the number of organizations. Second, a count variable does not neces7 For nonreligious voluntary memberships, we sarily correspond to level of involvement. People exclude memberships of religious organizations from who are involved in two or more voluntary voluntary memberships. 8 Church membership should not be confused with organizations do not automatically invest more membership of a religious organization. We thank one time than someone who volunteers for a single of the anonymous reviewers for providing the folorganization. 6 Voluntary membership is constructed similarly based on the seven items pertaining to the question on memberships which preceded the question on unpaid work. lowing clear example. Being a member of the Catholic church is really different from belonging to the Knights of Columbus. In fact, of all church members in our dataset (Catholics, Protestants, and nonChristians) only 23 percent is a member of a religious organization and only 12 percent volunteers for such an organization. #2789-ASR 71:2 filename:71202-ruiter 198—–AMERICAN SOCIOLOGICAL REVIEW Table 2. Descriptive Statistics for Independent Variables Individual level variables:a —Sex (female = 1) —Education —Age —Age-squared —Married —Cohabiting —Divorced —Widow —Single —Catholic —Protestant —Non-Christian —Nonreligious —Church attendance Contextual level variables:b —Average church attendance —GDP/capita (US$1,000) —Level of democracy —Welfare state expenditure (% of GNP) Range Mean Standard Deviation 0–1 0–9 18–90 324–8,100 0–1 0–1 0–1 0–1 0–1 0–1 0–1 0–1 0–1 0–104 .53 5.39 42.72 2,108.87 .59 .04 .06 .07 .23 .49 .22 .07 .22 19.19 .50 3.01 16.84 1,585.98 .49 .19 .24 .26 .42 .50 .41 .25 .42 30.88 2.13–69.74 .48–44.01 4–14 .00– 42.80 19.14 14.68 11.93 19.46 14.79 7.91 2.76 9.73 a Averages and standard deviations are calculated over all 117,007 individuals. and standard deviations are calculated over 96 country-wave combinations for all but welfare state expenditure for which they are calculated over 83 country-wave combinations b Averages Delivered by Ingenta to : University approximate number of times someone visits of Groningen have a strong positive impact on volunteering Fri, 29 Feb 2008 10:55:03 church per year. This variable ranges from 0 to (Wilson 2000). We use the question at which age 104, which corresponds to no visits at all and respondents finished their fulltime education. two visits a week, respectively. The answers are recoded to a variable ranging For every unique country-wave combination from 0 (at the age of 12) to 9 (at 21 or older). we average church attendance to obtain a measAge is measured in years. We also include a ure for the religiosity of the country. Note that quadratic term for age, because the relation this variable actually varies over the waves. between age and civic participation could be Countries with a high average church attencurvilinear (Boraas 2003; Curtis et al. 1992; dance are relatively religious, whereas counKnoke and Thomson 1977). Middle-aged peotries that score low on average church attendance ple would be most active because they are more are relatively secular. strongly integrated in work and family than the Next to the variables that are relevant for young and old (Wilson 2000). Similar differtesting our hypotheses, we include sex, age, ences are found between the married and unmareducational level, and marital status at the indiried. Married people in the United States vidual level in our analyses. Women appear to volunteer more than unmarried people (Sundeen volunteer more than men in the United States 1990; Wilson 2000). Again, for Europe these (Boraas 2003; Hayghe 1991), whereas in Europe differences are less clear-cut (Gaskin and Davis the picture is less clear. In some European counSmith 1995). We control for marital status tries women are more involved while in others because it might differ significantly over relimen contribute more (Gaskin and Davis Smith gious groups and countries. We distinguish mar1995). Men and women might differ because ried, cohabiting, divorced, widowed, and single they are differently integrated in family, church, people. and work. Women are more involved in caring At the contextual level, we also include tasks, attend churches more often, but they less national economic development measured by often have a job. Consequently, their social netreal gross domestic product (GDP) per capita works differ from men. Education is reported to (Laspeyres index, US$ in 1996 constant prices) #2789-ASR 71:2 filename:71202-ruiter RELIGIOSITY AND VOLUNTEERING—–199 which we obtain from Heston, Summers, and and “civil liberties” and reverse the scale. This results in a scale ranging from 2 to 14. Aten (2002).9 In the literature (cf. Curtis et al. Another possibly relevant context is implied 2001; Halman 2003), two arguments for a posby the crowding out hypothesis: collective welitive effect of economic development on level fare state arrangements would crowd out volof volunteering can be found. First, it is argued unteering because they provide substitutes for that economic development leads to occupaindividual efforts to provide collective goods tional specialization resulting in more diverse (Menchik and Weisbrod 1987; Arts, Halman, interest groups in which people would particiand Van Oorschot 2003; Salamon and pate voluntarily. Second, affluent countries Sokolowski 2003). However, the opposite has would provide people with more resources (e.g., also been argued. According to the interdetime and training) necessary to participate. pendence theory, more state involvement in However, results are mixed. Parboteeah et al. social welfare activities would lead to more vol(2004) do find a positive effect of gross nationunteering because voluntary organizations al product (GNP) per capita, whereas Curtis et would in fact be supported by the state (Salamon al. (2001) do not find a significant effect of the and Sokolowski 2003). Although a lot of natural logarithm of GDP per capita on workresearch on the influence of the welfare state ing memberships. Although Halman (2003) uses the typology of Esping-Andersen (1990), does find a positive bivariate relation between we use a real measure for welfare state investGDP per capita and level of volunteering, it ment (cf. Salamon and Sokolowski 2003). The appears to be spurious after taking other conInternational Labour Organisation provides us textual characteristics into account. Country with such a measure, the percentage of the characteristics that turn out to be conducive to Gross National Product spent on social securivolunteering in Halman’s analysis are years of ty (International Labour Office 1968–1996).11 continuous democracy and level of democracy. This measure Delivered by Ingenta to : has four advantages over the use It is assumed that democracies provideUniversity the infra- of Groningen of Esping-Andersen’s typology. First, the typolstructure (e.g., freedom of speech and ogy is based on decommodification measures, Fri,the 29right Feb 2008 10:55:03 to assemble) necessary for voluntary organizawhich suffer from methodological weaknesses tions to flourish. We do not use years of contin(Van Voorhis 2002). Second, the data vary not only by country but also over time, whereas the uous democracy, because for many countries it typology remains fixed. Therefore, variations in is linearly related to year of survey. However, the volunteering between waves might be better variable for level of democracy is included. It is explained with our measure. Third, this measmeasured with the Gastil Index (Freedom House ure suits the theory better, since the degree in 2005).10 We sum the scores on “political rights” which the welfare state is developed is central to the hypothesis and because more welfare state arrangements require more social security expenditure. Fourth, Esping-Andersen’s 9 GDP per capita scores are unavailable for exacttypology can not easily be applied to countries ly the right years for the following countries. outside his original sample, whereas data on Therefore, we use the scores from the years in parensocial security expenditure are available for theses instead: Bulgaria (1991), Malta (1998), Puerto many more countries. However, because inforRico (1998), Russia (1991), Singapore (1996), and Vietnam (1997). Although Penn World Tables 6.1 mation on welfare state expenditures is not does not distinguish East and West Germany anyavailable for all 53 countries in our dataset, we more, these regions are distinguished in WVS, and they have had radically different policies on religion in the past. Therefore, we decide to maintain the country distinction and estimate GDP per capita figures based of Penn World Tables 5.6 data. 10 No scores are available for exactly the right years for the following countries. We therefore use the scores for the years in parentheses: East Germany (1989), Latvia (1991), Russia (1991), Slovenia (1991), and West Germany (1989). 11 Because these data are unavailable for some countries for the years 1981, 1990, we calculate them on the basis of linear inter- and extrapolation of long term trends. All data for 2000 are based on linear extrapolation of the trends. Curve estimation per country show that linear trends are good approximations of the true developments. #2789-ASR 71:2 filename:71202-ruiter 200—–AMERICAN SOCIOLOGICAL REVIEW report the results based on 42 countries in a church members and nonmembers, we set the nonreligious as reference category. By doing so, separate section after the analyses on our comthe beta parameters for the religious groups plete sample.12 should be interpreted in terms of how much the Correlations between the contextual varilogit deviates from the one for nonreligious ables show that more prosperous countries are people. Next to church membership all control generally more democratic (R = .60; N = 96), variables are added. less religious (R = –.29; N = 96), and spend All religious affiliates appear to be much more on social security (R = .58; N = 83). More more likely to volunteer than the nonreligious, democratic societies are less religious (R = which supports Hypothesis 1. Moreover, there –.24; N = 96) and have more extensive welfare are large differences among the denominations. states (R = .43; N = 83). Extensive welfare Clearly, Protestants have a higher expected probstates appear to be less religious (R = –.40; N ability to volunteer (.31) than Catholics (.25), = 83). which corroborates Hypothesis 3. 13 NonSince we hypothesize individual and conChristians (.32) are equally engaged in voluntextual level effects as well as cross-level interteer work as Protestants (differences between action effects, we use multilevel analyses them are not significant), whereas the nonrelitechniques (Bryk and Raudenbush 1992; gious (.20) have a considerably lower probability Snijders and Bosker 1999). Because our data to volunteer. consist of repeated cross-sectional surveys, we Besides the clear effect of church memberdistinguish three levels (cf. Duncan, Jones and ship, the control variables are influential as Moon 1996; Subramanian, Jones, and Duncan well. The chance of volunteering increases 2003): level 1 is the lowest level and consists of strongly with educational level. As expected, the the individual respondents; level 2 comprises the relation between age and volunteering is curvisurvey waves within countries; and level 3 is linear. However, the effect of age cannot simcomposed of the countries. Delivered by Ingenta to : ply be interpreted as life course effects, since For the estimation of our models we use the of Groningen University explanations could be involved. Putnam Fri, 29 Feb 2008cohort 10:55:03 statistical program HLM, version 6.02a (2000) argues that declining levels of volun(Raudenbush et al. 2004). Because we want to teering might be due to cohort effects, i.e. older explain the probability of volunteering and our birth cohorts are more active than younger dependent variables are dichotomous, we esticohorts. This might partly explain the positive mate hierarchical logistic regression models. effect of age. The effect of marital status is also The procedure we use is penalized quasi-likeas expected. Married people are the most active lihood estimation and all but the dummy varivolunteers compared with cohabiting, divorced, ables are mean-centered in the analyses. or widowed individuals, but singles are similarly active. RESULTS In model 2, we include church attendance and allow its effect to vary for countries (level GENERAL VOLUNTEERING 3) and waves (level 2). In general, attending We start our analyses with the estimation of the religious ceremonies influences the chance of null model with random intercepts only (not volunteering considerably. People who attend shown in Table 3). From the null model it is clear church twice a week are more than 5 that the probability of volunteering varies over (exp(104*.016)) times more likely to volunteer countries and survey waves. than people who never visit church.14 However, In model 1, we include church membership and test whether the probability to volunteer 13 Expected probabilities of volunteering are caldiffers among religious denominations. Because culated for married men with average age (42.7 years) Hypothesis 1 focuses on the difference between 12 Data are unavailable for Albania, Croatia, Estonia, Latvia, Lithuania, Macedonia, Moldova, Puerto Rico, Slovenia, South Africa, and Zimbabwe. and average educational level (5.4). 14 Note that originally the church attendance variable was ordinal. So, be careful when interpreting the effect of a single day increase. We checked whether the positive effect of church attendance is less strong at the higher end of the scale. Indeed, we found evi- #2789-ASR 71:2 filename:71202-ruiter RELIGIOSITY AND VOLUNTEERING—–201 Table 3. Hierarchical Logistic Regression Models for General Volunteering and Voluntary Memberships Voluntary Membership General Volunteering Model 1 Intercept Level-1 Variables: —Sex (female = 1) —Education —Age —Age-sq. (⫻100) —Married (ref.) —Cohabiting —Divorced —Widow —Single —Nonreligious (ref.) Model 2 Model 3 Model 4 Model 5A –1.570*** –1.271*** –1.278*** –1.326*** –1.317*** (.094) (.085) (.077) (.072) (.073) –.003 (.030) .113*** (.008) .059*** (.005) –.054*** (.005) .— –.103*** (.030) .116*** (.007) .062*** (.004) –.061*** (.005) .— Model 5B –.241* (.096) –.104*** (.030) .116*** (.007) .062*** (.004) –.061*** (.005) .— –.105*** (.031) .118*** (.007) .063*** (.004) –.062*** (.005) .— –.104*** (.032) .118*** (.007) .063*** (.005) –.062*** (.005) .— –.214*** (.030) .113*** (.006) .061*** (.004) –.063*** (.005) .— –.273*** –.184** –.184** (.078) (.065) (.065) –.144*** –.080** –.081** (.032) (.029) (.029) –.105** –.144*** –.144*** (.034) (.031) (.031) –.013 –.016 –.016 (.038) (.030) Delivered(.030) by Ingenta to : .—University .—of Groningen .— –.191** (.068) –.082** (.031) –.145*** (.033) –.018 (.031) .— –.191** (.069) –.083** (.031) –.148*** (.033) –.018 (.032) .— –.039 (.057) –.093** (.033) –.072* (.029) –.068* (.034) .— .018 (.035) .307*** (.050) .260** (.097) .016*** (.001) .020 (.037) .313*** (.052) .266** (.104) .017*** (.001) .018 (.038) .310*** (.053) .263** (.104) .017*** (.001) .030 (.039) .477*** (.054) .339*** (.096) .013*** (.001) .014 (.009) –.054* (.023) .018*** (.006) .015 (.009) –.057* (.023) .017** (.006) .030* (.012) –.059 (.031) .003 (.006) –.027*** (.005) –.014* (.006) Fri, 29 Feb 2008 10:55:03 —Catholic —Protestant —Non-Christian .335*** (.041) .631*** (.080) .646*** (.116) —Church attendance .018 (.035) .305*** (.050) .259** (.097) .016*** (.001) Level-2 Variables: —GDP/capita .— .— —Level of democracy .— .— —Average church attendance .— .— .011 (.009) –.051* (.023) .— .— .— .— .— .137*** .385*** .— .138*** .294*** .004*** .124*** .307*** .004*** .123*** .234*** .004*** .127*** .231*** .004*** .301*** .396*** .036*** .— .052*** .052*** .053*** .030*** .043*** Cross-Level Interactions: —Average church attendance ⫻ ——Church attendance (⫻100) Variance components: —Level-2 variance —Level-3 variance —Random effect Church ——attendance level-2 (⫻1000) —Random effect Church ——attendance level-3 (⫻1000) Note: Numbers in parentheses are robust standard errors. N1 = 117,007; N2 = 96; N3 = 53. *p < .05; **p < .01; ***p < .001 (two-tailed tests). #2789-ASR 71:2 filename:71202-ruiter 202—–AMERICAN SOCIOLOGICAL REVIEW the effect of church attendance varies over men, this difference was not visible in model 1. countries and waves. It can be calculated that When we add GDP per capita and level of the most frequent churchgoers are—depending democracy in model 3, all other effects are on the specific country-wave combination— virtually unchanged. The effect of level of between 1.2 (exp(104*.002)) to 24.7 democracy is contrary to the theoretical expec(exp(104*.031)) times more likely to volunteer tation. People appear to volunteer more in less than people who never go to church.15 This democratic societies. This contradicts findmeans that, at one extreme, church attendance ings by Halman (2003) and Parboteeah et al. does not influence volunteering much; in those (2004). Halman finds a positive effect for years countries, frequently attending church does of continuous democracy as well as political not really increase the probability to volunteer. rights (subscale of the Gastil Index used here). At the other extreme, church attendance has a However, his analysis is restricted to European strong impact on volunteering; in those councountries (a subset of the dataset we use) and tries, frequent churchgoers are much more he does not distinguish individual level from likely to volunteer than people who do not go contextual level effects. Consequently, the to church at all. smaller variation in level of democracy in his After controlling for church attendance, the dataset as well as compositional differences effect of church membership drops considerbetween the countries could distort his results. ably. The expected probability to volunteer is Parboteeah et al. find a significant positive the same for Catholics and nonreligious peoeffect of degree of liberal democracy as well, ple who attend church equally often.16 The but their analysis is also restricted to a smalleffects for Protestants and non-Christians are er sample of countries.17 We do not find a sigstill substantial but half as strong as in model nificant effect of economic development.18 1. Apparently, the differences in volunteering This is in line with Curtis et al. (2001) and to : between religious and nonreligious Delivered people areby Ingenta Halman (2003), but contradicts Parboteeah et University of Groningen for a large part due to differencesFri, in church (2004). 29 Feb 2008al.10:55:03 attendance. This supports Hypothesis 2. Model 4 shows that the national religious However, not all differences between the relicontext has an additional positive effect. People gious denominations disappear. Even after living in the most devout country are, accordcontrolling for church attendance, Protestants ing to the model, almost four times are still more likely to volunteer than Catholics, (exp(69.74*.018)/exp(2.13*.018)) more likewhich supports Hypothesis 3. The parameter ly to volunteer than people living in the most estimates of all control variables, except the secular country. This supports Hypothesis 5. Consequently, people who never visit church one for sex, do not change much. Women are slightly less active in volunteer work than men. Because they visit church more often than dence for such a curvilinear effect. However, because it resulted only in a slight adjustment of the expected probabilities to volunteer, we decided not to present it in our models. The additional analyses are available upon request. 15 This calculation is based on the fact that the estimated beta parameter is assumed to be normally distributed. Therefore, 95 percent of all beta parameters are between ± 1.96 times the standard deviation. So, .016 ± 1.96 ⫻ (.000004+.000052).5 gives a range of [.002 ; .031]. 16 In our large dataset we have a considerable number of people who say that they do not belong to a church but still visit church often (table available upon request). 17 To test whether our larger sample with more variation in level of democracy causes us to find a negative effect of level of democracy contrary to findings by Halman (2003) and Parboteeah et al. (2004), we reestimated model 5A on two subsets of countries. Our sample has 31 and 16 countries in common with the analyses of Halman and Parboteeah et al. respectively. Indeed, results from these subsets differed considerably from model 5A. The sign of the effect of level of democracy was positive but it did not reach significance in both subsets. Other modeling differences were probably the reason why the effects in these subsets did not reach significance. 18 We also checked whether an increase of national wealth matters more for poor countries than for richer countries by including a quadratic term. Again, no significant results were found. #2789-ASR 71:2 filename:71202-ruiter RELIGIOSITY AND VOLUNTEERING—–203 Figure 1. Effect of Church Attendance on General Volunteering Delivered by Ingenta to : University of Groningen Fri, 29probFeb 2008added 10:55:03 to the model is in line with a dynamic but live in a devout country have a higher interpretation. We agree that this is not strong ability to volunteer than similar people in secevidence, but we lack the data to provide a ular societies. The voluntary participation of more rigorous test. However, if the dynamic these nonreligious people is, in effect, elevatinterpretation holds, declining volunteer rates ed in more devout countries. A dynamic intershould not come as a surprise when societies pretation of this result would imply that change continue to secularize. in the national religious context causes change In model 5, we test whether the effect of in volunteer rates.19 Unfortunately, the World church attendance varies with the national reliValues Survey data are not really suitable for gious context. As can be seen, the effect of a trend analysis to test this claim because the church attendance is smaller in more devout number of items on voluntary associations difcountries. This implies that the differences fers over the survey waves. This has unknown between secular and devout people are subconsequences for the estimation of volunteer stantially smaller in religious countries than in rates based on items that remained the same secular countries. This corroborates Hypothesis over all waves. Besides, for half the countries 6. The effects of the religious context and the only one survey wave is available. However, the cross-level interaction are depicted in Figure fact that not only variance at level 3 but also 1. This graph shows that church attendance variance at level 2 declines (albeit only marbarely affects general volunteering in the most ginally) when average church attendance is devout country, whereas the impact is strongly positive in the most secular society. The interaction effect suggests that religious peo19 We stress that this is a ceteris paribus arguple who live in secularizing countries might mentation. Not only new predictors might become relbecome more active in volunteering. So, if the evant in the future, but declining volunteer rates dynamic interpretation of the national relicaused by secularization might also to some extent gious context effect and the cross-level interbe compensated by for example rising levels of eduaction effect holds, increased civic cation. #2789-ASR 71:2 filename:71202-ruiter 204—–AMERICAN SOCIOLOGICAL REVIEW participation of religious people might compensate for the secularization effect.20 In order to test whether these effects exist only for specific organizations, we estimated separate models for all seven types of organizations. Although we do not present these seven separate models here, it should be stressed that for all but one type of organization, we found both religious context and cross-level interaction effects (results available from author upon request). Only for trade unions, neither individual church attendance nor the national religious context affected volunteering. find a negative though nonsignificant effect for democracy. They argue that their effects might reach significance in a larger sample; a claim we are unable to support. SPILLOVER For the test of the spillover hypothesis we estimate different models, which are displayed in Table 4. Although model 6 looks quite similar to model 3 in Table 3, we change two aspects. First, the dependent variable under consideration is now nonreligious volunteering. Second, we change the reference category for church membership to Protestants in order to test MEMBERSHIPS VERSUS VOLUNTEERING whether the spillover effect is smaller for In order to examine whether determinants of Protestants than for other denominations. voluntary memberships differ from the ones From model 6 it follows that people from we find for volunteering, we reestimate model different denominations (the nonreligious 5A for voluntary memberships. Results are disincluded) are equally involved in volunteering played in the last column of Table 3 (model for nonreligious organizations when control5B). Although some effects differ, similarities ling for church attendance. Consequently, all at level 1 are striking. When we turn to level 2 residual denominational differences in model 3 however, some substantial differences stand out. (Table 3) are caused by differences in religious Levels of voluntary memberships are not at all volunteering. to : affected by the national religiousDelivered context.by Ingenta In model 7, we include religious volunteer Apparently, these looser ties to civicUniversity organiza- of Groningen as a predictor for nonreligious volunteerFri,context, 29 Feb 2008work 10:55:03 tions are not influenced by the religious ing. It has a strong positive effect. People who whereas the stronger ties of volunteers are. This are involved in religious volunteering are almost fits the network theory, because getting people 3.6 (exp(1.281)) times more likely to do nonto volunteer requires more recruitment effort religious volunteer work as well. This is in line than just signing someone up. For memberwith the spillover Hypothesis 4a. Although we ships, economic development appears to be cannot be sure that participation in religious important though. In more prosperous societies volunteering actually causes people to start volpeople are more likely to join voluntary assounteering for other organizations as well, we ciations. The fact that we do not find this effect believe that if spillover really happens, the direcfor volunteering is in accordance with Putnam’s tion of causation seems most plausible. Because (2000) claim that ‘checkbook memberships’ most volunteer work is done for religious organmight have risen in Western societies. In affluizations, it is unlikely that the direction is the ent societies more people are just nominal memother way around. bers of voluntary organizations without being When Catholics volunteer for religious organactively involved in volunteering. Again, we izations, they are, compared to Protestants, more find a negative effect of level of democracy, likely to do nonreligious volunteer work as well. which is contrary to findings by Curtis et al. Model 8A shows that Catholics who are active (2001). However, it does not reach significance. in religious volunteering are over 4 Schofer and Fourcade-Gourinchas (2001) also (exp(1.059+.351)) times more likely to volunteer for nonreligious organizations than Catholics who are not volunteering for reli20 Although, we have three surveys over the perigious organizations. For Protestants, this od 1981–2001 for some countries, the parameter spillover effect is considerably smaller, which estimate for the interaction effect is based predomiis in accordance with Hypothesis 4b. Protestants nantly on differences between countries. This makes who are active in religious volunteering, are the empirical support for the dynamic interpretation almost 3 (exp(1.059)) times more likely to do of the cross-level interaction effect less strong. #2789-ASR 71:2 filename:71202-ruiter RELIGIOSITY AND VOLUNTEERING—–205 Table 4. Hierarchical Logistic Regression Models for NRV and NRV Memberships NRV Model 6 Intercept Model 8B –1.558*** –1.855*** –1.780*** (.074) (.069) (.067) –.594*** (.094) —Education —Age —Age-sq. (⫻100) —Married (ref.) —Cohabiting —Divorced —Widow —Single —Protestant (ref.) —Catholics —Non-Christian NRV Memberships Model 8A Level-1 Variables: —Sex (female = 1) Model 7 –.161*** (.036) .143*** (.007) .064*** (.005) –.063*** (.006) .— –.174*** (.038) .143*** (.007) .062*** (.006) –.060*** (.006) .— –.176*** (.038) .143*** (.007) .062*** (.006) –.060*** (.006) .— –.314*** (.031) .138*** (.008) .072*** (.005) –.075*** (.006) .— –.134 (.066) –.038 (.032) –.160*** (.043) .005 (.030) .— –.111 (.068) –.015 (.032) –.150** (.047) .015 (.029) .— –.111 (.068) –.016 (.032) –.150** (.047) .011 (.029) .— .031 (.054) –.057 (.033) –.128*** (.028) –.049 (.032) .— –.038 (.042) –.008 (.116) .052 (.046) .001 (.001) 1.059*** (.065) .— –.126*** (.034) –.085 (.128) .019 (.037) –.002*** (.000) .— .069 Delivered –.032 by Ingenta to : (.037) University(.038) of Groningen .018 10:55:03 .066 Fri, 29 Feb 2008 —Nonreligious —Church attendance —Religious volunteer work (yes = 1) (.085) –.007 (.044) .006*** (.001) .— (.085) .133** (.048) .001 (.001) 1.281*** (.074) .— —Membership religious organization (yes = 1) .— Level-1 Interactions: —Catholic ⫻ Religious volunteer work .— .— —Catholic ⫻ Membership religious organization .— .— —Non-Christian ⫻ Religious volunteer work .— .— —Non-Christian ⫻ Membership religious organization .— .— —Nonreligious ⫻ Religious volunteer work .— .— —Nonreligious ⫻ Membership religious organization .— .— Variance components: —Level-2 variance —Level-3 variance —Random effect Church attendance level 2 (⫻1000) —Random effect Church attendance level-3 (⫻1000) .131*** .108*** .011*** .011*** .141*** .180*** .007*** .017*** .351*** (.087) .— .201 (.201) .— .270 (.155) .— .132*** .102*** .011*** .011*** .665*** (.044) .— .255*** (.058) .— .223 (.230) .— .020 (.080) .220*** .353*** .001*** .015*** Note: Numbers in parentheses are robust standard errors. N1 = 117,007; N2 = 96; N3 = 53. NRV = nonreligious volunteering. *p < .05; **p < .01; ***p < .001 (two-tailed tests). #2789-ASR 71:2 filename:71202-ruiter 206—–AMERICAN SOCIOLOGICAL REVIEW nonreligious volunteer work than those who are significant and in sign even opposed to what is not active. Although the parameters for the nonexpected from the crowding-out hypothesis. Christians and the nonreligious are positive, However, this does not support the alternative they do not reach significance. So, the spillover interdependence theory either. Although effect is equally strong for Protestants, nonSalamon and Sokolowski (2003) do find supChristians, and the nonreligious. This refutes port for the interdependence theory, they conHypothesis 4b. We conclude that the spillover trol neither for compositional differences effect is stronger for Catholics than for the other between the countries nor for possible condenominations. founding factors at the country level. In our Whether members of religious voluntary analysis, most other parameters were similar to organizations are also more likely to join nonthe ones from model 5A. However, the countries religious voluntary organizations is tested in for which we lacked welfare state expenditure model 8B. We reestimate model 8A but replace data were on average less democratic. Leaving religious volunteer work with membership of a them out of the analysis reduced the variance in religious voluntary organization and change the level of democracy considerably, which caused dependent variable to nonreligious voluntary the effect of level of democracy to be nonmemberships. Clearly, the spillover effect is not significant. limited to volunteering. Joining a nonreligious voluntary organization is more likely among CONCLUSIONS AND DISCUSSION members of religious voluntary organizations than among nonmembers of religious volunIn this article, we study the impact of individtary organizations. Interestingly, it shows that ual religiosity, the national religious context, Catholics who are not a member of a religious and their interplay on volunteering. Curtis et al. voluntary organization are less likely to join (2001) show in their general study on contexnonreligious voluntary organizations than tual determinants of associational involvement Delivered by Ingenta to : Protestants. However, if they have joined a relithat working memberships (religious organizaUniversity of Groningen gious voluntary organization, their chance to be and unions excluded) vary with the nationFri, 29 Feb 2008tions 10:55:03 a member of a nonreligious organization is al religious composition. We provide new increased and even higher than that of hypotheses about the influence of religion by Protestants. Again, the spillover effect is lowest combining previous research on social networks for Protestants. Furthermore, the fact that we and volunteering (Becker and Dhingra 2001; find a small but negative effect for church attenBekkers 2003; Jackson et al. 1995; Lam 2002; dance indicates that frequent churchgoers are Musick, Wilson, and Bynum 2000; Park and slightly less likely to join nonreligious voluntary Smith 2000; Yeung 2004) with a study on the organizations than those who rarely attend impact of both parental religiosity and the church. Note however, that this effect is connational religious context on individual relitrolled for membership of a religious voluntary giosity (Kelley and De Graaf 1997). These new organization, something for which the chance hypotheses are tested on a large dataset conis in fact strongly increased by church attentaining information on volunteering for 53 coundance. tries in the period 1981-2001. This study not only demonstrates that frequent churchgoers volunteer more, it also shows VOLUNTEERING AND THE WELFARE STATE that there is an additional positive effect of To test the crowding-out hypothesis, we added devoutness of society: religious context matters. social security expenditure to model 5A from Interestingly, individual and contextual effects Table 3. Because these data were unavailable for strongly interact. The differences between sec11 countries, the analysis was limited to 104,555 ular and devout people are substantially smallrespondents distributed over 42 countries.21 The er in religious countries than in secular effect of welfare state expenditure was all but countries. Our findings imply that individual religiosity is hardly relevant for volunteering in devout nations. 21 We do not present the results of this analysis in Level of democracy affects volunteering, but contrary to the expectation, people in less dema separate table; they are available upon request. #2789-ASR 71:2 filename:71202-ruiter RELIGIOSITY AND VOLUNTEERING—–207 ocratic societies volunteer more. We are not Similarly, religious people have to invest more sure how to interpret this result. One reason in volunteering when a nation secularizes. Furthermore, religious volunteering seems might be that the less democratic societies to have a strong spillover effect. This implies that included in our dataset have less extensive welreligious people are also more involved in volfare states and therefore citizens of those coununteering for secular organizations. Consetries provide for certain public goods quently, secularization could even cause themselves. This explanation builds of course on declining participation in secular organizations. the crowding-out hypothesis, which we cannot This spillover effect is strongest for Catholics. corroborate in this study. We are unable to test Unfortunately the World Values Survey data do the explanation because welfare state expendinot allow us to distinguish specific groups of tures for most of these countries are missing. Protestants to test whether especially conserWe do not find significant effects for ecovative Protestants refrain from secular volunnomic development on volunteering. However, teering as suggested by Wilson and Janoski in our analysis on voluntary memberships GDP (1995). per capita shows a significant effect. In more Without doubt, the quality of the data differs prosperous societies, people join more organiover the 96 surveys involved in this study. For zations. Because we do not find this effect for this reason, we carefully tried to find countries volunteering, it seems that checkbook memthat might be influential outliers. However, berships are more common in wealthy nations whatever country was excluded, we could not (cf. Putnam 2000). find substantial changes in our results. An analyIf secularization is an ongoing process in sis on a subsection of the complete dataset to test rich post-industrial societies, we expect declinthe welfare state hypothesis shows that the ing levels of volunteering due to composition results are robust. Keep in mind though that and context effects. We find that differences only 7 topercent of the data population is nonover time are to some extent explained by difDelivered by Ingenta : Christian and that questions pertaining to reliferences in national religious context, which of Groningen University gion in the World Values Surveys might be supports this expectation. Of course, Fri,secular29 Feb 2008 10:55:03 better applicable to Christians than to nonization theory is contested (e.g., Iannaccone Christians. Therefore, we cannot make strong 1998; Stark 1999; Stark and Finke 2000). Yet claims with respect to non-Christian countries. there is a large body of evidence showing that However, we did a preliminary test to see church attendance, an important indicator in whether results are different for non-Christian the secularization debate, has indeed been countries. We reestimated model 5A with a declining over the past decades in most rich dummy variable for non-Christian countries post-industrial countries (e.g., Norris and included at level 3 as well as an cross-level Inglehart 2004). Furthermore, the dynamic interaction between this dummy and church interpretation of our results implies that change attendance at level 1.22 Results show that peoin the national religious context causes change ple living in non-Christian countries volunteer in volunteer rates. Although we are unable to to the same extent as people in Christian counprovide strong evidence for this interpretation, tries (results available from author upon the results suggest that diminished civic involverequest). The effect of church attendance is only ment goes hand in hand with ongoing secularslightly smaller in non-Christian societies. This ization. However, the cross-level interaction test is not ideal, but since the number of couneffect suggests that religious people who live in tries participating in the World Values Surveys secularizing countries might have an increased increases with every wave, we hope that future chance to volunteer. So, if the dynamic interpretation holds, increased volunteering of religious people could compensate for the secularization effect. Interestingly, Kelley and 22 In India, Japan, Tanzania, and Turkey, the nonDe Graaf (1997) also show that family reliChristians form the largest group. For these countries giosity is more important for the religiosity of the dummy variable was set to 1. In all other counchildren in secular nations than in devout tries, at least 75 percent of the population claims to nations. In a secular context, parents have to be either Protestant, Catholic, or nonreligious, and invest more to keep their children religious. therefore the dummy variable was set to 0. #2789-ASR 71:2 filename:71202-ruiter 208—–AMERICAN SOCIOLOGICAL REVIEW lijken.” Pp. 141–72 in Geven in Nederland 2003: research can provide stronger tests for nonGiften, legaten, sponsoring en vrijwilligerswerk, Christian countries as well. edited by T. N. M. Schuyt. Houten/Mechelen, In this study, we show that network theory Netherlands: Bohn Stafleu Van Loghum. especially provides important arguments for Boraas, Stephanie. 2003. “Volunteerism in the United predicting volunteering behavior of both reliStates.” Monthly Labor Review 126:3–11. gious and nonreligious people. Network theoBrady, Henry E., Kay L. Schlozman, and Sidney ry predicts that all people in devout countries Verba. 1999. “Prospecting for Participants: have more religious people within their social Rational Expectations and the Recruitment of networks. Consequently, nonreligious people Political Activists.” American Political Science should get to know more active religious peoReview 93:153–68. Bryant, W. Keith, Haekyung Jeon-Slaughter, Hyojin ple who could recruit them. Furthermore, in Kang, and Aaron Tax. 2003. “Participation in such networks the norm to volunteer could be Philanthropic Activities: Donating Money and stronger, as well as the social pressure to behave Time.” Journal of Consumer Policy 26:43–73. accordingly. Although we lack actual network Bryk, Anthony S. and Stephen W. Raudenbush. 1992. data, analyses of our large scale international Hierarchical Linear Models: Applications and dataset show that national religiosity strongly Data Analysis Methods. Newbury Park, CA: Sage. affects volunteering behavior of nonreligious Cnaan, Ram A., Amy Kasternakis, and Robert J. people. Also, the impact of church attendance Wineburg. 1993. “Religious People, Religious on volunteering is much smaller in devout Congregations, and Volunteerism in Human Services: Is There a Link?” Nonprofit and nations. Apparently, theoretical progress can be Voluntary Sector Quarterly 22:33–51. made for cross-national differences in volunCurtis, James E., Douglas E. Baer, and Edward G. teering. Further research to test the actual Grabb. 2001. “Nations of Joiners: Explaining recruitment mechanisms in social networks Voluntary Association Membership in Democratic would require network data collection in interSocieties.” American Sociological Review national surveys. 66:783–805. Delivered by Ingenta to : Curtis, James E., Edward G. Grabb, and Douglas E. University of Groningen Stijn Ruiter is Junior Researcher at the Department Baer. 1992. “Voluntary Association Membership Fri, 29 Feb of Sociology, Radboud University Nijmegen, the 2008 10:55:03 Netherlands. At the Interuniversity Center for Social Science Theory and Methodology (ICS), he is currently writing his Ph.D. thesis on individual and contextual influences on volunteering. Nan Dirk De Graaf is Professor at the Department of Sociology, Radboud University Nijmegen, the Netherlands, and chairman of the Interuniversity Center for Social Science Theory and Methodology (ICS). His main research interests are sociology of religion, social stratification, and the political and cultural consequences of social mobility and mixed marriages. He is one of the principal investigators of the Family Surveys of the Dutch Population 1998, 2000, and 2003, and the Religion in Dutch Society 2005 survey. REFERENCES Arts, Wil, Loek Halman, and Wim van Oorschot. 2003. “The Welfare State: Villain or Hero of the Piece?” Pp. 275–308 in The Cultural Diversity of European Unity, edited by W. Arts, J. Hagenaars, and L. Halman. Leiden, Netherlands: Brill. Becker, Penny E. and Pawan H. Dhingra. 2001. “Religious Involvement and Volunteering: Implications for Civil Society.” Sociology of Religion 62:315–35. Bekkers, René. 2003. “De bijdragen der kercke- in Fifteen Countries: A Comparative Analysis.” American Sociological Review 57:139–52. De Graaf, Nan Dirk and Ariana Need. 2000. “Losing faith: is Britain alone?” Pp. 119–36 in British Social Attitudes: The 17th Report: Focusing on Diversity, edited by R. Jowell, J. Curtice, A. Park, K. Thomson, L. Jarvis, C. Bromley, and N. Stratford. London, UK: Sage. De Hart, Joep. 1999. “Godsdienst, Maatschappelijke Participatie en Sociaal Kapitaal.” Pp. 207–48 in Vrijwilligerswerk vergeleken: Nederland in internationaal en historisch perspectief, edited by P. Dekker. Den Haag, Netherlands: SCP. Dekker, Paul and Joep de Hart. 2001. “Levensbeschouwing en vrijwilligerswerk. Het belang van netwerken in een seculariserende samenleving.” Tijdschrift voor Humanistiek 8:9–17. ———. 2002. “Het zout der aarde: Een analyse van de samenhang tussen godsdienstigheid en sociaal kapitaal in Nederland.” Sociale Wetenschappen 45:45–61. Dekker, Paul and Loek Halman. 2003. The Values of Volunteering: Cross-Cultural Perspectives. New York: Kluwer Academic/Plenum Publishers. Dekker, Paul and Andries van den Broek. 1998. “Civil Society in Comparative Perspective: Involvement in Voluntary Associations in North America and Western Europe.” International #2789-ASR 71:2 filename:71202-ruiter RELIGIOSITY AND VOLUNTEERING—–209 Journal of Voluntary and Nonprofit Organizations Economics of Religion.” Journal of Economic 9:11–38. Literature 36:1465–95. Duncan, Craig, Kelvyn Jones, and Graham Moon. Inglehart, Ronald. 2003. “Modernization and 1996. “Health-related Behavior in Context: A Volunteering.” Pp. 55–70 in The Values of Multilevel Modeling Approach.” Social Science & Volunteering, edited by P. Dekker and L. Halman. Medicine 42:817–30. New York: Kluwer Academic/Plenum Publishers. Esping-Andersen, Gøsta. 1990. The Three Worlds of International Labour Office. 1968–1996. The Cost of Welfare Capitalism. Princeton, NJ: Princeton Social Security. Geneva, Switzerland: International University Press. Labour Office. European Values Study Group and World Values Jackson, Elton F., Mark D. Bachmeiser, James R. Survey Association. 2005. European and World Wood, and Elizabeth A. Crafst. 1995. Values Surveys Integrated Data File, 1999–2002, “Volunteering and Charitable Giving: Do Religious Release I [Computer file]. 2nd ICPSR version. and Associational Ties Promote Helping Cologne, Germany: Zentralarchiv für Empirische Behavior?” Nonprofit and Voluntary Sector Sozialforschung (ZA)/Tilburg, Netherlands: Quarterly 24:59–78. Tilburg University/Amsterdam, Netherlands: Kelley, Jonathan and Nan Dirk De Graaf. 1997. Netherlands Institute for Scientific Information “National Context, Parental Socialization, and Services (NIWI)/Madrid, Spain: Analisis Religious Belief: Results from 15 Nations.” Sociologicos Economicos y Politicos (ASEP) and American Sociological Review 62:639–59. JD Systems (JDS)/Ann Arbor, MI: Inter-universiKnoke, David and Randall Thomson. 1977. ty Consortium for Political and Social Research “Voluntary Association Membership: Trends and [producers]. Cologne, Germany: Zentralarchiv für the Family Life Cycle.” Social Forces 56:48–65. Empirische Sozialforschung (ZA)/Madrid, Spain: Lam, Pui-Yan. 2002. “As the Flocks Gather: How Analisis Sociologicos Economicos y Politicos Religion Affects Voluntary Association (ASEP) and JD Systems (JDS)/Ann Arbor, MI: Participation.” Journal for the Scientific Study of Inter-university Consortium for Political and Social Religion 41:405–22. Research [distributors]. Menchik, Paul L. and Burton A. Weisbrod. 1987. Freedom House. 2005. Freedom in the World Countryby Ingenta “Volunteer Delivered to : Labor Supply.” Journal of Public Ratings 1972 through 2003. Retrieved April 4, of Groningen Economics 32:159–83. University 2005 (http://www.freedomhouse.org/ratings/ Musick, Marc A., John Wilson, and William B. Fri, 29 Feb 2008 10:55:03 allscore04.xls). Bynum, Jr. 2000. Race and Formal Volunteering: Gaskin, Katherine and Justin Davis Smith. 1995. A The Differential Effects of Class and Religion. New Civic Europe?: A Study of the Extent and Social Forces 78:1539–70. Role of Volunteering. London, UK: Volunteer Norris, Pippa and Ronald Inglehart. 2004. Sacred and Centre UK. Secular: Religion and Politics Worldwide. Halman, Loek. 2003. “Volunteering, Democracy, Cambridge, UK: Cambridge University Press. and Democratic Attitudes.” Pp. 179–98 in The Oesterle, Sabrina, Monica K. Johnson, and Jeylan T. Values of Volunteering, edited by P. Dekker and L. Mortimer. 2004. “Volunteerism during the Halman. New York: Kluwer Academic/Plenum Transition to Adulthood: A Life Course Publishers. Perspective.” Social Forces 82:1123–49. Hayghe, Howard V. 1991. “Volunteers in the U.S.: Parboteeah, K. Praveen, John B. Cullen, and Lrong Who Donates the Time?” Monthly Labor Review Lim. 2004. “Formal Volunteering: A Cross114:17–23. National Test.” Journal of World Business Heston, Alan, Robert Summers, and Bettina Aten. 39:431–41. 2002. Penn World Table Version 6.1. Philadelphia, Park, Jerry Z. and Christian Smith. 2000. “ ‘To Whom PA: Center for International Comparisons at the Much Has Been Given...’ : Religious Capital and University of Pennsylvania (CICUP). Community Voluntarism among Churchgoing Hodgkinson, Virginia A. 2003. “Volunteering in Protestants.” Journal for the Scientific Study of Global Perspective.” Pp. 35–53 in The Values of Religion 39:272–86. Volunteering, edited by P. Dekker and L. Halman. Putnam, Robert D. 2000. Bowling Alone: The New York: Kluwer Academic/Plenum Publishers. Collapse and Revival of American Community. Hodgkinson, Virginia A., Murray S. Weitzman, and New York: Simon & Schuster. Arthur D. Kirsch. 1990. “From Commitment to Raudenbush, Stephen W., Anthony S. Bryk, Yuk Fai Action: How Religious Involvement Affects Giving Cheong, and Richard T. Congdon. 2004. HLM 6: and Volunteering.” Pp. 93–114 in Faith and Linear and Nonlinear Modeling. Lincolnwood, Philanthropy in America, edited by R. Wuthnow IL: Scientific Software International, Inc. and V. A. Hodgkinson and Associates. San Salamon, Lester M. and S. Wojciech Sokolowski. Francisco, CA: Jossey-Bass. 2003. “Institutional Roots of Volunteering: Toward Iannaccone, Laurence R. 1998. “Introduction to the a Macro-Structural Theory of Individual Voluntary #2789-ASR 71:2 filename:71202-ruiter 210—–AMERICAN SOCIOLOGICAL REVIEW of Comparative Research.” Journal of Sociology Action.” Pp. 71–90 in The Values of Volunteering. Cross-Cultural Perspectives, edited by P. Dekker and Social Welfare 29:3–18. and L. Halman. New York: Kluwer Watt, David H. 1991. “United States: Cultural Academic/Plenum Publishers. Challenges to the Voluntary Sector.” Pp. 243–87 Schofer, Evan and Marion Fourcade-Gourinchas. in Between States and Markets: The Voluntary 2001. “The Structural Contexts of Civic Sector in Comparative Perspective, edited by Engagement: Voluntary Association Membership Robert Wuthnow. Princeton, NJ: Princeton in Comparative Perspective.” American University Press. Sociological Review 66:806–28. Wilson, John. 2000. “Volunteering.” Annual Review Smith, David H. 1994. “Determinants of Voluntary of Sociology 26:215–40. Association Participation and Volunteering: A Wilson, John and Thomas Janoski. 1995. “The Literature Review.” Nonprofit and Voluntary Sector Contribution of Religion to Volunteer Work.” Quarterly 23:243–63. Sociology of Religion 56:137–52. Snijders, Tom and Roel Bosker. 1999. Multilevel Wilson, John and Marc Musick. 1997. “Who Cares? Analysis. An Introduction to Basic and Advanced Toward an Integrated Theory of Volunteer Work.” Multilevel Modeling. London, UK: Sage. American Sociological Review 62:694–713. Snow, David A., Louis A. Zurcher, and Sheldon World Values Study Group. 1999. World Values Ekland-Olson. 1980. “Social Networks and Social Survey, 1981–1984 and 1990–1993 [Computer Movements: A Microstructural Approach to file]. 2nd ICPSR version. Ann Arbor, MI: Institute Differential Recruitment.” American Sociological for Social Research [producer]. Ann Arbor, MI: Review 45:787–801. Inter-university Consortium for Political and Social Stark, Rodney. 1999. “Secularization, R.I.P.” Research [distributor]. Sociology of Religion 60:249–73. Wuthnow, Robert. 1991. Acts of Compassion : Caring Stark, Rodney and Roger Finke. 2000. Acts of Faith: for Others and Helping Ourselves. Princeton, NJ: Explaining the Human Side of Religion. Berkeley: Princeton University Press. University of California Press. ———. 1999. “Mobilizing Civic Engagement. The Subramanian, S. V., Kelvin Jones, and Craig Duncan. Changing Impact of Religious Involvement.” Pp. 2003. “Multilevel Methods for Public Health Delivered by Ingenta to : 331–63 in Civic engagement in American democResearch.” Pp. 65–111 in Neighborhoods and of Groningen University racy, edited by T. Skocpol and M. P. Fiorina. Health, edited by I. Kawachi and L. Fri, F. Berkman. 29 Feb 2008 10:55:03 Washington, DC: Brookings Institution Press and New York: Oxford University Press. Russell Sage Foundation. Sundeen, Richard A. 1990. “Family Life Course Yeung, Anne B. 2004. “An Intricate Triangle: Status and Volunteer Behavior: Implications for the Religiosity, Volunteering, and Social Capital: The Single Parent.” Sociological Perspectives European Perspective, the Case of Finland.” 33:483–500. Nonprofit and Voluntary Sector Quarterly Van Voorhis, Rebecca A. 2002. “Different Types of Welfare States? A Methodological Deconstruction 33:401–22.