Religion and Giving for International Aid: Evidence from a Survey of U.S. Church Members

Sociology of Religion Advance Access published June 23, 2014 Sociology of Religion 2014, 0:0 1-23 doi:10.1093/socrel/sru037 Religion and Giving for International Aid: Evidence from a Survey of U.S. Church Members Allison Schnable* Princeton University This article investigates the relationship between twowell-documented patterns in charitable giving: the positive relationship between religion and generosity and the growth in the numbers of, and donations to, international aid organizations. I discuss three modes by which religion shapes Americans preferences on international aid: values, social norms, and exposure to need. Using a 2005 national survey of church members, I find that (1) altruistic values, congregational social ties, and exposure to international needs through one s congregation all are associated with giving to international causes; (2) individuals with more frequent attendance, those with more social ties in the congregation, and evangelicals and black Protestants are significantly more likely to prefer church over government aid; and (3) aid organizations affiliated with a religious tradition enjoy an in-group advantage in support. Key words: charitable giving; international aid; congregations; denominations; globalization. Americans gave nearly $16 billion to international causes in 2010, and the sprawling relief effort following the earthquake in Haiti that year demonstrated the breadth of groups that now comprise the international relief and development field (Giving USA 2011). By 2011, the field had expanded beyond stalwarts like the Red Cross and Catholic Relief Services to include more than 10,000 U.S.-based nonprofits working on international aid (Roeger et al. 2012). Faced with both humanitarian emergencies and broader challenges of global poverty and confronted with an array of organizations competing for the charitable dollars to help how do prospective givers decide which efforts deserve their support? I argue that religion provides meaningful influence, shaping not just *Direct correspondence to Allison Schnable, Princeton University, Princeton, NJ, USA. E-mail: schnable@princeton.edu. # The Author 2014. Published by Oxford University Press on behalf of the Association for the Sociology of Religion. All rights reserved. For permissions, please e-mail: journals. permissions@oup.com 1

2 SOCIOLOGY OF RELIGION Americans likeliness to give for international aid but also their attitudes about which kind of aid organizations will be most effective. Religiously affiliated aid groups enjoy continued support from U.S. donors. World War I- and II-era religious aid groups like Lutheran World Relief, the Jewish Joint Distribution Committee, and the American Friends Service Committee remain important providers of aid (Smith 1990). They are joined by a wave of organizations that emerged from the evangelical movement: World Vision (1950), Samaritan s Purse (1970), and Rick Warren s P.E.A.C.E. Plan (2004). Denominational aid groups figured prominently in the response to the Haiti earthquake. The Evangelical Lutheran Church in America raised $12.5 million, the United Methodist Committee on Relief $43 million, and Catholic Relief Services $194 million, with $50 million of that coming from special offerings collected within dioceses (Catholic Relief Services 2012; ELCA 2011; United Methodist Communications 2011). Religious groups have also provided platforms for advocacy on a number of development and human rights issues. Activists across the theological spectrum campaigned jointly for debt relief for poor countries in the late 1990s (Kurtz and Goran 2002). Since 2000, scholars and pundits have noted increased attention by evangelical leaders to international religious freedom, the environment, and human trafficking (Mead 2006; Zalanga 2004), though survey evidence shows that rank-and-file evangelicals have not surpassed other religious groups in support for what Wuthnow and Lewis (2008) call altruistic foreign policy goals. Existing literature on religion and international aid typically takes one of two approaches. The first is to situate aid as a foreign policy issue (i.e., the purview of government) and to ask how religion shapes people s opinions on it (Hertzke 2004; Kurtz and Goran 2002; Mead 2006; Wuthnow and Lewis 2008). The second is to take either churches or religious aid organizations as the units of analysis and ask how religion makes them different from secular aid providers (Bornstein 2005; Egan 1988; Hearn 2002; King 2012; Miller and Yamamori 2007; Whaites 1999). What has not been done is to consider international aid as something everyday Americans shape not just through their political voices but by their donations to different kinds of organizations working and competing for resources in the broad field of international relief and development. And while the scholarship on charitable giving has offered thoughtful attention to the role of religion, research there has not yet taken account of the increasing transnationalism of American religion and the distinctive ways that international giving may be cultivated by religious congregations. This paper aims to fill those gaps. Existing data show that religion has a strong influence on charitable giving generally and suggest that religious ties will also shape Americans patterns of giving on international issues. The annual Giving USA report documents the continued preeminence of religious institutions as beneficiaries of charitable donations: in 2011, religious causes received $96 billion, or 32 percent of all

RELIGION AND GIVING FOR INTERNATIONAL AID 3 charitable giving. Religious Americans give more to charity on average than their nonreligious counterparts (Flanagan 1991; Independent Sector 2002), and the strength of religious involvement, typically measured by service attendance, is positively associated with higher levels of giving (Chaves and Miller 1999; Iannaccone 1997; Regnerus et al. 1998; Smith and Emerson 2008; Vaidyanathan et al. 2011; Wuthnow 1991). The 2002 report from Independent Sector and the National Council of Churches offers further evidence that membership in a religious congregation shapes giving to international causes. Six in 10 American households give financial support to religious congregations; of that group, 10 percent also give to international causes, while only 4 percent of households who give only to secular causes support international organizations. They find that the donations to international organizations are dominantly 86 percent from households that also support religious congregations. Given that a majority of charitable donations come from religious people, it is worth examining how religion influences (1) whether individuals will donate money to any sort of international assistance project, and (2) which types of organizations secular? faithbased? affiliated with one s own religious traditions? that religious individuals are likely to support. In a 2011b review on the predictors of philanthropy, Bekkers and Wiepking offered a general set of mechanisms by which religion affects charitable giving. First, religion shapes individuals values both the general altruistic values that are at the heart of the major religious traditions, but also among some traditions conservative beliefs that discourage giving to secular organizations that conflict with those beliefs. 1 Second, religious congregations are sites where individuals are exposed to needs (see also Bekkers and Wiepking 2011a) and are likely to be solicited to give. Finally, since individuals are more likely to give when they are being observed particularly by others with whom they have close social ties (Bekkers 2004) religious congregations can be a site where enforceable social norms contribute to greater giving. I use these mechanisms as a starting point to analyze how religion shapes Americans giving for international assistance. In the next section, I elaborate on Bekkers and Wiepking s mechanisms to take account of (1) the transnational nature of international giving, or the fact that the need is distant and must be framed for givers, and (2) the ambiguity about the causes of and potential solutions to global poverty, with the result that givers must choose not only if but how to give. From there, I use a national survey of church members to examine religious influences on overall likeliness of giving for international assistance as well as support for particular sources of aid: churches, government, the World Bank/IMF, nondenominational faith-based agencies, and agencies associated with specific religious traditions. 1 See, for example, some religious congregations moves to end their sponsorship of Boy Scout troops after the Boy Scouts decision to allow openly gay members (Jenkins 2013).

4 SOCIOLOGY OF RELIGION RELIGIOUS MODES OF INFLUENCE ON INTERNATIONAL GIVING Values The major religious traditions all hold sacred the values of compassion, altruism, and charity. Jewish tradition teaches that humankind is made in god s image; loving one s neighbor is a central teaching of Christianity, while compassion for all beings is central to Buddhist traditions. Almsgiving is a common practice in most faiths, and Muslim and Christian scriptures institutionalize giving to the needy in the form of the zakat or the tithe (Eckel and Grossman 2004; Queen 1996). In at least two moments in the Gospels, Jesus enjoins his followers to broaden their notions of compassion: in the parable of the Good Samaritan and in Matthew 25, where he tells his listeners that whoever has fed the hungry or welcomed the stranger has done it unto me. Wuthnow and Lewis (2008:196) suggest that a diffuse cultural ambience exists in religious communities such that regular involvement exposes people to ideas about caring for the needy. Commitment to any of these religious traditions is apt to cultivate values of altruism and generosity that are associated with giving for international aid. In addition to the general altruism learned in religious practice, religious traditions are associated with particular ideas about causes of and solutions to poverty (Emerson and Smith 2000; Hunt 2002). Hunt found that Catholics and white Protestants were more likely to offer individualist explanations for poverty, while Jews, black Protestants, and adherents of other minority religions were more likely to give structuralist explanations. Emerson and Smith found similar patterns in churchgoers explanations for inequality. Smith (1998) suggests that evangelical Protestants are distinctive among Christian denominations in employing a personal influence strategy to social issues. Wuthnow (2004) also found denominational differences in support for government funding of religious service agencies, with evangelicals being most likely to support these initiatives. If differences among religious traditions about the role of government and private, faith-based charity extend to international assistance, we would expect that denominational affiliations will be associated with support for different types of aid organizations. Social Norms Individuals are more likely to make charitable donations and to volunteer when their social networks place norms on those behaviors (Bekkers and Schuyt 2008; Olson and Cadell 1994). Most religious congregations make giving normative in part through the weekly financial solicitations that are woven into worship services. Whether congregation members give or not is observable in several ways: through the passing of the plate, or better still, having members approach the altar with offerings; by requiring tickets for Jewish High Holy Day services, which are accessible to dues-paying members; or by engaging in stewardship campaigns that ask members to fill out financial pledge cards (on the effects of some these tactics, see Hoge et al. 1996). Social pressure within

RELIGION AND GIVING FOR INTERNATIONAL AID 5 congregations may incline members to support the international aid efforts of their own denominations. One measure of norms, then, refers to support for the aid efforts of one s own tradition. I hypothesize that denominational aid groups will enjoy significantly higher support from their own members. There is evidence that interpersonal ties within congregations are particularly significant for giving. Putnam and Campbell (2010) found that, net of controls for attendance and religious beliefs, having more close friends at church and belonging to a small group within the congregation were both associated with higher overall charitable giving. A second measure of social norms thus concerns how strong social ties in a congregation reinforce the group s desired behaviors. I expect that stronger social networks within a congregation will be associated with greater likelihood of overall giving. Bekkers and Schuyt (2008) found that social pressure was associated with a greater likelihood of giving to religious causes but not to secular ones; on the question of types of aid, I expect that stronger social ties within the congregation will be associated with support for religious aid efforts but not for secular ones. Exposure to Need In Bekkers and Wiepking s model, exposing potential givers to the need for assistance is a critical precondition for giving. While religion may cultivate the general values that predispose an individual to generosity, churchgoers are more likely to give when they are presented with specific information about the people, places, and conditions where assistance is needed. This framing of needs is particularly important for international causes because churchgoers have limited personal knowledge of development issues. Immigrants in congregations are one potential source of information about specific needs abroad. Portes s (2000) and Levitt s (2001) studies of transnationalism show that migrants actively forge connections between their community of origin and their country of reception, increasing over time the international flow of goods, money, and information. Levitt s ethnographic study observed that these links were embedded in religious communities, with money flowing from the United States to the Dominican Republic for traditional religious festivals and churches in the sending town. Wuthnow (2009) also notes that immigrant congregations sometimes send direct aid to the home country, or that churches take up aid projects in particular countries because they have an immigrant member with roots there. Even when immigrant members of a congregation are not transnational entrepreneurs in the manner of Portes or Levitt, they may be visible symbolic links, putting a face on a poor country that had before been part of a generalized, impoverished world out there. I expect that, net of other factors, respondents who report larger numbers of immigrants in their congregation will be more likely to give money to aid projects. Religious congregations can also take deliberate action to educate their members about international poverty. Wuthnow and Lewis (2008) argue that congregations cannot be imagined as simply ambient spaces or subcultures, but

6 SOCIOLOGY OF RELIGION actually do things. Specifically, they find some evidence that congregational activities related to social service and foreign policy are associated with churchgoers greater support for altruistic foreign policy goals. Here, I use whether a congregation has invited a missionary to speak in the last year as an indicator of exposure to need, predicting that giving to international causes will be higher for individuals whose congregations have hosted missionary speakers. To summarize, I suggest that religion encourages generosity toward the poor in other countries by cultivating altrusim, by providing a site for enforceable social norms that favor giving, and by using transnational connections to expose members to the needs of the poor in other countries. I further hypothesize that social norms within congregations and distinctive values about the causes of and solutions to poverty will shape church members preferences about types of aid. Specifically, I anticipate that stronger social ties will be associated with support for religious aid agencies; that denomination-affiliated aid will be more likely to have support from in-group members; that evangelicals will be particularly likely to support religious approaches to aid; and that aid approaches that focus on structural change will receive particular support from black Protestants. METHODS Data I use a national survey of American church members to test whether religion influences the likeliness of giving to international causes and support for specific kinds of aid through the pathways of values, social norms, and exposure to need. The data here come from the 2005 Global Issues Survey (hereafter GIS), a random-digit dial phone survey of 2,231 Americans aged 18 and over. All participants in the survey identified themselves as being church members or attending religious services at least once a month. The bulk of items in the survey are concerned with the international outreach of the respondent and the religious congregation to which the respondent belongs. The survey also included questions about the respondent s political affiliation and opinions about several public policy issues as well as basic demographic items. The cooperation rate for the study was 68.4 percent and the response rate was 56.2 percent. The data set is exceptional in the breadth of variables on religious practice and beliefs; however, the subsamples for non-christian adherents are very small (only 40 out of 2,231) and the sample does not allow for direct comparisons with nonreligious Americans. Though the sample is not nationally representative, recall that the target population, Americans who attend religious services once a month or more, constitutes about 40 percent of the U.S. population (Wuthnow 2009). The GIS categorizes respondents religious affiliation with the reltrad scheme (Steensland et al. 2000), with Jews classified here with other faiths because of the small sample size. The distribution of religious affiliation approximates

RELIGION AND GIVING FOR INTERNATIONAL AID 7 population estimates from the 2008 Pew Religious Landscape Survey, with mainline Protestants constituting about one-fifth of the sample, Catholics one-fourth, and black Protestants about 7 percent. Evangelicals are 38 percent of the sample here, exceeding Pew s estimate of 26 percent in the American population but close to the 2007 Baylor Religion Survey estimate of 33 percent (Baylor Religion Survey 2007). The large share of women in the sample (64 percent) reflects the disproportionate involvement of women in American religion. Dependent Variables I model the influence of religious values, social norms, and exposure to need on both respondents giving behavior whether or not they made a contribution to hunger and poverty projects in the last year and their attitudes toward particular sources of aid. The first model presented below measures whether respondents have actually given money for international aid in the last year. The dependent variable is a yes response to the question During the past year, have you given money for international relief or hunger projects? As table 1 shows, overall giving for international aid is quite high; 76 percent of respondents reported having made a donation in the last year. Subsequent models analyze respondents support for different sources of international aid. While the survey asked respondents to rate their confidence in particular types of aid organization, it did not ask whether or how much money they had donated to those groups. Since our interest here is in the sort of strong support that is necessary (though not sufficient) for charitable giving to aid groups, responses for subsequent models are dichotomized for strong support or less. The second model measures whether respondents prefer church aid efforts over those operated by government agencies. Respondents could disagree strongly, disagree somewhat, agree somewhat, or agree strongly with the statement Churches can do a better job of helping people in other countries than the U.S. government can. Thirty-five percent of respondents strongly agreed with the statement, and the logistic regression model predicts this dichotomous outcome. In addition to the question of church over government aid, I model churchgoers support for five specific kinds of aid organizations. GIS respondents were asked to rate their confidence (none, only a little, a fair amount, a great deal) in faith-based humanitarian organizations; the World Bank or International Monetary Fund; and organizations associated with three religious traditions Catholic, mainline, and evangelical. I take the World Bank and IMF as exemplars of a structural approach to international development, which I hypothesized would enjoy higher support from black Protestants. The models predict the odds of expressing a great deal of confidence in each type of organization. Table 1 shows that reports of great confidence were highest for faith-based humanitarian groups at 27 percent and lowest for the World Bank and IMF at 4 percent, with the denomination-linked groups falling between 15 and 23 percent.

8 SOCIOLOGY OF RELIGION TABLE 1 Descriptive Statistics Variable Mean (SD) Measurement Dependent variables Gave to international aid project 0.76 (0.42) 1 ¼ Yes Church over government aid 0.35 (0.47) 1 ¼ Strongly agree Faith-based humanitarian orgs 0.27 (0.44) 1 ¼ Great confidence World Bank/IMF 0.04 (0.20) 1 ¼ Great confidence Aid efforts of Catholic Church 0.23 (0.42) 1 ¼ Great confidence Aid efforts of mainline denominations 0.15 (0.35) 1 ¼ Great confidence Aid efforts of evangelical organizations 0.22 (0.41) 1 ¼ Great confidence Independent and control variables Moral responsibility abroad 5 categories Disagree strongly (excluded) 0.01 (0.11) Don t know/refused 0.01 (0.10) Disagree somewhat 0.05 (0.23) Agree somewhat 0.38 (0.48) Agree strongly 0.55 (0.50) Close friends in the congregation 9.14 (7.48) Missionary speaker 0.74 (0.44) 1 ¼ Yes Immigrant members 5 categories Don t know/refused 0.09 (0.29) None 0.23 (0.48) Only a few 0.35 (0.42) Some 0.19 (0.40) A lot 0.13 (0.33) Catholic 0.26 (0.44) 1 ¼ Yes Mainline 0.19 (0.40) 1 ¼ Yes Evangelical 0.39 (0.49) 1 ¼ Yes Black Protestant 0.07 (0.26) 1 ¼ Yes Other faith 0.08 (0.27) 1 ¼ Yes Service attendance 4 categories Twice monthly or less 0.23 (0.42) Nearly every week 0.35 (0.48) Every week 0.19 (0.40) More than once a week 0.25 (0.43) White 0.81 (0.39) 1 ¼ Yes Black 0.10 (0.30) 1 ¼ Yes Hispanic 0.09 (0.29) 1 ¼ Yes Asian 0.01 (0.11) 1 ¼ Yes Female 0.64 (0.48) 1 ¼ Yes Foreign-born 0.08 (0.28) 1 ¼ Yes College graduate 0.40 (0.49) 1 ¼ Yes Republican 0.44 (0.50) 1 ¼ Yes Age 50.18 (16.94) Continued

TABLE 1 Continued Variable Mean (SD) Measurement Income,$10,000 0.05 (0.22) $10,000 20,000 0.08 (0.27) $20,000 30,000 0.12 (0.32) $30,000 40,000 0.11 (0.31) $40,000 50,000 0.10 (0.30) $50,000 75,000 0.16 (0.36) $75,000 100,000 0.11 (0.32).$100,000 0.14 (0.34) Missing 0.13 (0.34) RELIGION AND GIVING FOR INTERNATIONAL AID 9 9 categories Independent and Control Variables I operationalize the values influence as respondents stated belief that people of faith have a responsibility to help the poor. To measure this, I use a series of dichotomous variables measuring respondents agreement with the statement People of faith have a moral responsibility to learn about problems in other parts of the world and do what they can to help. Religious affiliation is coded with dichotomous variables using the reltrad scheme (again, with Jews here classified with other faiths because of the sample size). Congregational social ties are measured by the respondent s report of how many close friends he or she has in the congregation. Since the survey item had respondents choose among categories of uneven intervals, the responses were midpoint-coded and treated as a continuous variable. The ways in which congregations could expose members to international need concerned visiting missionary speakers and immigrant members. The missionary dichotomous variable was coded as 1 if the respondent said his or her congregation had hosted a missionary speaker in the last 12 months. The number of immigrants in the congregation is measured as a series of dichotomous variables for the following responses: none (the excluded category), only a few, some, a lot, or don t know/refused. The don t know/refused responses are modeled as a category to avoid losing cases for missing data. Frequency of attendance at religious services is included as a control variable. Response categories are twice monthly or less (the excluded category), nearly every week, every week, or more than once a week. (Recall that the survey screen-in question asked for church members or those attending church at least once a month.) Other variables typically correlated with charitable giving and religious behavior are included as controls in all models: sex, race, age, level of education (a dummy variable for attaining a Bachelor s degree or higher), income, being born outside the United States, and identification with the Republican party. White is the excluded racial category for all models. Household income is measured in eight

10 SOCIOLOGY OF RELIGION categories ranging from less than $10,000 to greater than $100,000 per year. An additional category to model missing income responses is also included to avoid loss of 13 percent of cases who refused to respond to the income question. Where a respondent refused to give a specific age but responded to an age category (e.g., 25 34), a single year value was imputed using the mean age for respondents in the given age range. Table 1 shows descriptive statistics for all variables used in the analyses. Analytic Strategy I model all dependent variables with logistic regression and display the odds ratios and 95 percent confidence intervals for predictor variables in the tables. Because of the importance of religious tradition and congregation variables to the analysis, I dropped from the sample the 60 individuals who said they were unaffiliated with any particular religious tradition or refused the question. Except for the income, values, and immigrants-in-congregation variables, where don t knows/refusals are modeled as categories, missing data are light. The N for all but one model is 2,173. For the model predicting confidence in the World Bank/IMF (table 3), 21 cases are dropped because one of the values categories perfectly predicts a 0 outcome, leaving an N of 2,152. RESULTS Giving to International Causes Table 2 shows logistic regression models that predict the likelihood of a respondent having given to an international hunger or poverty project in the last year. The altruistic values and social norms hypotheses find support. Respondents who agree somewhat that people of faith have responsibilities abroad have twice the odds of making a donation compared with those who disagree strongly, and those who agree strongly have triple the odds vis-à-vis the reference group. The number of close friends in the congregation is also a significant predictor of giving. Holding other factors equal, each additional close friend in the congregation is associated with 1.6 percent greater odds of making a donation to international causes. There are no statistically significant differences between mainline Protestants and the reference group of Catholics in the odds of giving, but evangelicals and respondents in the other traditions category (which includes Mormons and Orthodox Christians as well as non-christian adherents) had roughly 50 percent lower odds than Catholics of having given money for international aid. The model also supports the hypothesis that exposure to need through congregations is associated with a stronger tendency to give. All else equal, a respondent who attends a congregation with some or a lot of immigrants has nearly double odds of giving compared with a respondent who attends a congregation with no immigrants. And respondents whose congregations had hosted a missionary in the last year had, ceteris parabus, 83 percent greater odds of donating.

RELIGION AND GIVING FOR INTERNATIONAL AID 11 TABLE 2 Model) Giving to International Relief or Hunger Projects (Logistic Regression Odds ratio (95% CI) Religious tradition Catholic (excluded) Mainline 0.994 (0.687 1.438) Evangelical 0.465*** (0.343 0.629) Black Protestant 0.550 þ (0.273 1.106) Other tradition 0.526** (0.340 0.814) Moral responsibility abroad Disagree strongly (excluded) Don t know/refused 0.819 (0.246 2.729) Disagree somewhat 0.874 (0.367 2.084) Agree somewhat 2.285* (1.034 5.049) Agree strongly 3.009** (1.366 6.626) Close friends in congregation 1.016* (1.001 1.031) Missionary speaker 1.838*** (1.440 2.347) Immigrants in congregation None (excluded) Don t know/refused 1.381 (0.924 2.064) Only a few 1.513** (1.147 1.996) Some 1.952*** (1.383 2.755) A lot 1.906** (1.272 2.857) Service attendance Twice monthly or less (excluded) Nearly every week 1.469 þ (0.966 2.234) Every week 1.353* (1.030 1.779) More than once per week 1.292 (0.941 1.774) Black 1.328 (0.732 2.408) Hispanic 0.790 (0.522 1.194) Asian 0.981 (0.351 2.744) Female 1.256* (1.002 1.575) Age 1.008* (1.001 1.014) BA or higher 1.643*** (1.279 2.110) Foreign-born 0.788 (0.495 1.253) Republican 1.177 (0.934 1.484) Income,$10,000 (excluded) $10,000 $20,000 0.841 (0.499 1.417) $20,000 $30,000 1.366 (0.828 2.255) $30,000 $40,000 1.336 (0.799 2.234) $40,000 $50,000 1.803* (1.053 3.090) $50,000 $75,000 1.605 þ (0.968 2.660) $75,000 $100,000 1.472 (0.852 2.544) Continued

12 SOCIOLOGY OF RELIGION TABLE 2 Continued Odds ratio (95% CI).$100,000 3.251*** (1.794 5.894) Missing 1.293 (0.781 2.139) Constant 0.202** (0.0742 0.550) Observations 2,173 Pseudo-R 2 0.112 þ p,.10; *p,.05; **p,.01; ***p,.001. Preference for Church over Government Aid Table 3 supports the hypothesis that social norms and denominationally specific values about alleviating poverty predict a preference for church over government aid. Evangelicals have 38 percent greater odds than Catholics of expressing a preference for church over government, even when accounting for service attendance and identification with the Republican party. The support for church aid is even stronger for black Protestants relative to Catholics. 2 This may be accounted for by black churches institutional centrality in their communities, and their historical experience of providing for social needs when government institutions failed to do so. Attitudes about the greater ability of churches over government to aid people of color may have perhaps in a very conscious way have carried over to views of global aid. Social norms through congregational friendships also seem to support a preference for church over government aid: each increase of one friend in the congregation is associated with a 1.5 percent increase in the odds of favoring church aid. Exposure to need has less influence on individuals preference on church versus government aid than it did on their likeliness to give. The odds ratio for the presence of immigrants in a congregation is positive for most categories but does not reach statistical significance, and the odds ratio for having a missionary speak at one s place of worship is nonsignificant. Several control variables also merit note. Holding other factors equal, Republicans odds of preferring church aid are 20 percent higher than non-republicans at the marginally significant threshold of p,.10. As income rises, respondents are significantly less likely to prefer church to government aid, and women s odds of this preference are 33 percent less than men s. 2 This result should be interpreted with some caution because of the negative, significant effect for the black variable. In the survey sample, 150 of 219 black respondents identified as black Protestants. The rest were almost evenly distributed among mainline, evangelical, Catholic, and other traditions. Removing black respondents who were not black Protestants from the analysis, one religious group at a time did not materially change the significant negative coefficient for the black variable.

RELIGION AND GIVING FOR INTERNATIONAL AID 13 TABLE 3 Model) Strong Preference for Church over Government Aid (Logistic Regression Odds ratio (95% CI) Religious tradition Catholic (excluded) Mainline 0.986 (0.727 1.338) Evangelical 1.379* (1.067 1.782) Black Protestant 2.437** (1.263 4.702) Other tradition 1.291 (0.878 1.897) Moral responsibility abroad Disagree strongly (excluded) Don t know/refused 1.347 (0.337 5.374) Disagree somewhat 1.448 (0.522 4.013) Agree somewhat 1.349 (0.528 3.448) Agree strongly 3.138* (1.237 7.964) Close friends in congregation 1.015* (1.002 1.028) Missionary speaker 1.096 (0.870 1.379) Immigrants in congregation None (excluded) Don t know/refused 0.890 (0.606 1.305) Only a few 1.053 (0.817 1.359) Some 0.739 þ (0.545 1.003) A lot 1.170 (0.824 1.659) Service attendance Twice monthly or less (excluded) Nearly every week 1.020 (0.696 1.495) Every week 1.283 þ (0.994 1.657) More than once per week 1.646*** (1.234 2.196) Black 0.561* (0.318 0.991) Hispanic 1.361 (0.936 1.979) Asian 2.049 þ (0.884 4.749) Female 0.670*** (0.550 0.815) Age 1.002 (0.996 1.008) BA or higher 1.018 (0.825 1.257) Foreign-born 0.631* (0.413 0.965) Republican 1.203 þ (0.984 1.471) Income,$10,000 (excluded) $10,000 $20,000 0.530* (0.319 0.880) $20,000 $30,000 0.704 (0.440 1.125) $30,000 $40,000 0.539* (0.333 0.871) $40,000 $50,000 0.548* (0.336 0.893) $50,000 $75,000 0.454*** (0.284 0.724) $75,000 $100,000 0.410*** (0.248 0.677) Continued

14 SOCIOLOGY OF RELIGION TABLE 3 Continued Odds ratio (95% CI).$100,000 0.442** (0.270 0.722) Missing 0.432*** (0.269 0.696) Constant 0.290* (0.0978 0.857) Observations 2,173 Pseudo-R 2 0.0709 þ p,.10; *p,.05; **p,.01; ***p,.001. Preferences for Types of Aid Organizations I argued above that differences in tradition-specific values about the causes and solutions to poverty, combined with social pressure to support denominational aid efforts, would affect people s preferences for the types of aid organizations. The analyses in table 4 offer some support for those predictions. Model 4A predicts support for faith-based humanitarian organizations such as World Relief and Church World Service. Controlling for other variables, mainline Protestants and evangelicals have 59 and 54 percent greater odds, respectively, than Catholics of having great confidence in faith-based humanitarian organizations. Belonging to a congregation that has hosted a missionary speaker and strong belief in people of faith s responsibilities abroad are both associated with a higher likelihood of supporting faith-based humanitarian organizations. But neither the number of close friends nor the number of immigrants in the congregation is associated with greater support for faith-based humanitarian groups. Religion variables generally play very little role in predicting support for the World Bank/IMF. Altruistic values, religious tradition, and congregational exposure to need variables have no significant effects. Contrary to predictions, black Protestants are no more likely than the reference group to support the World Bank/IMF, which are the GIS best exemplars of structural solutions to poverty. Only the control variable for foreign birth is statistically significant; respondents born abroad have triple the odds of expressing great confidence in international financial institutions. For confidence in organizations associated with mainline, Catholic, and evangelical traditions, table 4 shows evidence of an in-group advantage. Net of other factors, mainline Protestants have 64 percent greater odds than the reference group of Catholics to support the aid efforts of mainline denominations. 3 Higher altruistic values are positively associated with support for mainline efforts at marginal significance, but other religious factors number of friends in the 3 Rotating the reference category reveals significant differences between mainline Protestants and all denominations except black Protestants.

TABLE 4 Strong Confidence in Types of Aid Organizations (Logistic Regression Models) Faith-based humanitarian orgs World Bank/IMF Mainline orgs Catholic Church Evangelical orgs Religious tradition Catholic 4.151*** (2.989 5.766) Mainline 1.585** (1.154 2.177) 1.370 (0.712 2.637) 1.637* (1.116 2.401) 1.623** (1.129 2.333) Evangelical 1.542** (1.172 2.028) 0.800 (0.436 1.469) 1.190 (0.842 1.683) 0.609** (0.432 0.859) 2.156*** (1.588 2.926) Black Protestant 1.835 þ (0.917 3.671) 0.797 (0.246 2.581) 0.918 (0.331 2.542) 0.680 (0.323 1.434) 2.432* (1.169 5.062) Other tradition 0.965 (0.626 1.488) 0.690 (0.267 1.784) 0.986 (0.578 1.682) 1.039 (0.648 1.668) 1.312 (0.827 2.083) Missionary speaker 1.341* (1.045 1.721) 0.872 (0.532 1.429) 1.311 (0.947 1.815) 1.063 (0.818 1.381) 1.516** (1.147 2.002) Close friends in congregation 1.000 (0.986 1.014) 0.982 (0.951 1.013) 1.012 (0.994 1.029) 1.015 þ (0.999 1.030) 1.010 (0.995 1.025) Moral responsibility abroad Disagree strongly (excluded) Don t know/refused 0.481 (0.0457 5.049) 1.367 (0.0789 23.67) 0.246 (0.0421 1.433) 2.067 (0.395 10.81) Disagree somewhat 1.810 (0.494 6.626) 0.551 (0.0909 3.340) 1.861 (0.217 15.96) 0.412 þ (0.147 1.154) 0.814 (0.206 3.220) Agree somewhat 1.833 (0.541 6.211) 0.375 (0.0787 1.783) 2.452 (0.325 18.50) 0.512 (0.207 1.264) 1.399 (0.407 4.809) Agree strongly 3.317 þ (0.985 11.17) 0.862 (0.191 3.899) 6.197 þ (0.829 46.33) 0.997 (0.407 2.439) 2.501 (0.734 8.523) Immigrants in congregation None (excluded) Don t know/refused 1.013 (0.671 1.531) 1.127 (0.448 2.836) 0.728 (0.419 1.266) 1.065 (0.683 1.659) 1.182 (0.763 1.831) Only a few 1.089 (0.828 1.431) 1.581 (0.846 2.958) 0.997 (0.711 1.398) 0.992 (0.720 1.366) 1.093 (0.814 1.467) Some 1.282 (0.937 1.752) 0.940 (0.432 2.044) 0.973 (0.655 1.445) 1.013 (0.703 1.460) 1.387 þ (0.992 1.939) A lot 1.215 (0.836 1.768) 1.042 (0.452 2.406) 1.181 (0.741 1.882) 0.875 (0.576 1.330) 1.117 (0.742 1.683) Service attendance Twice monthly or less (excluded) Nearly every week 1.354 (0.919 1.995) 1.601 (0.730 3.513) 0.786 (0.460 1.345) 1.284 (0.834 1.976) 1.194 (0.767 1.861) Every week 1.249 (0.951 1.640) 1.244 (0.686 2.257) 1.145 (0.811 1.616) 1.406* (1.043 1.896) 1.388* (1.027 1.877) More than once per week 1.329 þ (0.975 1.811) 1.653 (0.845 3.235) 1.292 (0.877 1.902) 1.256 (0.874 1.804) 1.418* (1.014 1.982) Continued RELIGION AND GIVING FOR INTERNATIONAL AID 15 Downloaded from http://socrel.oxfordjournals.org/ at Pennsylvania State University on September 17, 2016

TABLE 4 Continued Faith-based humanitarian orgs World Bank/IMF Mainline orgs Catholic Church Evangelical orgs BA or higher 0.981 (0.786 1.223) 1.153 (0.712 1.866) 1.203 (0.914 1.583) 1.430** (1.117 1.829) 0.959 (0.754 1.219) Foreign-born 1.410 (0.926 2.147) 3.164** (1.567 6.388) 0.801 (0.452 1.417) 1.447 (0.924 2.265) 1.709* (1.095 2.668) Republican 1.334** (1.079 1.648) 0.869 (0.535 1.413) 1.146 (0.883 1.488) 1.245 þ (0.983 1.578) 1.450** (1.153 1.824) Income,$10,000 (excluded) $10,000 $20,000 1.267 (0.729 2.204) 1.187 (0.447 3.150) 0.896 (0.432 1.857) 0.548* (0.303 0.990) 0.971 (0.559 1.687) $20,000 $30,000 1.338 (0.797 2.249) 0.628 (0.228 1.730) 0.957 (0.486 1.884) 0.611 þ (0.353 1.059) 0.753 (0.444 1.278) $30,000 $40,000 1.191 (0.701 2.024) 0.887 (0.338 2.327) 1.172 (0.598 2.294) 0.841 (0.487 1.452) 0.994 (0.587 1.684) $40,000 $50,000 1.309 (0.769 2.231) 0.671 (0.243 1.857) 0.882 (0.438 1.776) 0.661 (0.377 1.162) 0.751 (0.435 1.296) $50,000 $75,000 0.956 (0.570 1.601) 0.521 (0.190 1.424) 1.268 (0.661 2.431) 0.575* (0.336 0.986) 0.623 þ (0.368 1.052) $75,000 $100,000 0.872 (0.502 1.514) 0.801 (0.294 2.186) 0.896 (0.443 1.814) 0.608 þ (0.344 1.073) 0.697 (0.398 1.222).$100,000 1.128 (0.661 1.927) 0.637 (0.230 1.761) 1.308 (0.666 2.565) 0.466** (0.264 0.823) 0.818 (0.475 1.407) Missing 0.842 (0.494 1.436) 0.471 (0.164 1.354) 0.719 (0.361 1.432) 0.387*** (0.221 0.678) 0.555* (0.323 0.955) Constant 0.0574*** (0.0146 0.225) 0.185 þ (0.0272 1.258) 0.0102*** (0.00115 0.0897) 0.138*** (0.0448 0.427) 0.0398*** (0.00990 0.160) Observations 2,173 2,152 2,173 2,173 2,173 Pseudo-R 2 0.0497 0.0806 0.0718 0.134 0.0672 16 SOCIOLOGY OF RELIGION All models also include controls for Black, Asian, Hispanic, female, and age. Coefficients are excluded for space. 95 percent confidence intervals in parentheses. þ p,.10, *p,.05, **p,.01, ***p,.001. Downloaded from http://socrel.oxfordjournals.org/ at Pennsylvania State University on September 17, 2016

RELIGION AND GIVING FOR INTERNATIONAL AID 17 congregation, 4 missionary speaker, and immigrants in the congregation are nonsignificant. Model 4D shows support for Catholic aid efforts, and the difference between Catholics and the reference mainline Protestants is remarkable: Catholics have four times greater odds of supporting Catholic aid. 5 Evangelicals are even less likely than mainliners to support these efforts, with odds 39 percent smaller. The number of congregational friends is associated with higher support at a marginal level of significance. The pattern of in-group support holds for evangelicals and evangelical organizations. Again using Catholics as the reference group, evangelicals have double the odds of expressing strong confidence in evangelical aid groups. 6 Mainline and black Protestants are also more likely than Catholics to support evangelical aid groups, though the increase in odds is smaller than that for evangelicals. Higher service attendance is also associated with stronger support for evangelical aid, as is membership in a congregation that has hosted a missionary speaker. The association between a missionary speaker and support for evangelical organizations is the only case where the exposure to needs variables are significantly associated with greater support for tradition-linked aid organizations. Republicans have significantly greater odds of supporting faith-based humanitarian organizations, Catholic groups, and evangelical groups than do non-republicans. To sum up, these multivariate analyses revealed that (1) altruistic values, the number of friends in the congregation, and exposure to international needs through one s congregation are all associated with giving to international causes; (2) individuals with more frequent attendance, those with more friends in the congregation, and evangelicals and black Protestants are significantly more likely to prefer church over government aid; and (3) aid organizations affiliated with a religious tradition enjoy an in-group advantage in support. I elaborate on these findings below. DISCUSSION Ample research has shown that religious people tend to be charitable givers. Given the increases in giving to international causes and the number of aid groups in the last two decades, this paper has asked how religion influences 4 To evaluate the possibility that congregational social ties would increase the odds of giving for only for members of the in-group, I ran models predicting only the in-group members support for their denomination efforts (mainline respondents support for mainline organizations, etc.). The coefficient for friends in the congregation remained nonsignificant. 5 Rotating the reference category maintains significant differences between Catholics and all other groups. The increase in odds for Catholics versus other traditions is four-fold; the odds are six times greater for Catholics versus evangelicals or black Protestants. 6 Rotating the reference category shows that evangelicals also have significantly higher odds than other traditions of supporting evangelical aid. The odds ratios versus mainline Protestants and black Protestants were positive but just short of statistical significance.

18 SOCIOLOGY OF RELIGION likeliness to give and preferences in the type of international aid. Building on Bekkers and Wiepking s model of how religion affects giving, I developed three pathways by which religion shapes people s attitudes and behaviors toward international assistance. First, religion cultivates values: both the general altruistic values that would favor giving in some broad sense, and the specific values about the causes and solutions to poverty that lead people to support particular kinds of aid or aid organizations over others. Second, religious congregations can be sites for enforceable social norms where one is pressured to give (perhaps a particular amount, or to a particular organizations) or else lose face. Third, thanks to the increasing transnationalism of American religion, congregations provide sites where individuals can be exposed to specific needs and asked to give. The nature of the data do not allow me to formally test whether these three modes of influence mediate the general relationship between religiosity and giving. 7 This paper accomplished the more limited goal of showing that values, norms, and exposure to need predict giving and the forms of aid religious Americans prefer. I look forward to future work that employs novel data to formally test this mediation. A second limitation of this study is that survey evidence can only imperfectly distinguish the modes of influence discussed here. 8 Without an in-depth interview perhaps even with one it would be hard to say whether an evangelical woman who wrote a check to World Vision each month did so because of tradition-specific views about aid that were consistent with World Vision approach (values) or because a member of her Bible study asked her to donate (social norms). But there is some evidence in this study to support the distinctive contribution of each of these pathways. Net of personal characteristics, denomination, service attendance, number of friends in the congregation, or how much exposure respondents had to international need, those who more strongly agreed with the statement People of faith have a moral responsibility to learn about problems in other parts of the world and do what they can to help were more likely to give for international aid. This statement of altruistic values also had positive associations with support for church over government aid, and with higher odds of expressing confidence in two of four types of religious aid organizations. These results support a relationship between altruistic values cultivated by religion the belief that people of faith have a moral responsibility to help and overall support for international assistance. 9 7 Mediation analysis demands showing that the independent variables are causally prior to the mechanisms. Using cross-sectional data, it is impossible to determine that service attendance is causally prior to denomination and other characteristics of respondents congregations examined here. 8 But for an strong effort at distinguishing religious mechanisms using a survey data set with extensive items on psycho-social dispositions, see Bekkers and Schuyt (2008). 9 Note that though altruistic values and religious commitment measured as service attendance were positively associated with giving for international aid, Republican identification was not. This is consistent with Vaidyanathan et al. s (2011) finding, contra Brooks (2006),

RELIGION AND GIVING FOR INTERNATIONAL AID 19 Particular religious traditions can work both through distinctive values and denomination-specific social norms. The statistically significant differences between traditions about church over government aid suggest a difference in values. The wording of the question, [Do you agree that] Churches can do a better job of helping people in other countries than the U.S. government can invokes many of the distinctions about motivation, relationships, and effectiveness that pervaded the discourse about faith-based social services in the United States. Evangelicals preference for the church over government in international assistance is consistent with what we would expect from the private and relational approaches to domestic social problems that have been documented in the literature. Smith (1998) defines this perspective as the belief that the only truly effective way to change the world is one-individual-at-a-time through the influence of interpersonal relationships (187). An assistant pastor at evangelical Saddleback Church, which operates a large international aid program, illustrates the globalized version of this approach: And a lot of these post-colonial and independent countries, they kind of have this feeling, they need to have an American come in to solve the problems, or they need foreign aid to do something. The thought that Hey, we can solve our own problems, sometimes they just don t have confidence they can do something, or they re fatalistic and they wait for the government to do something for them. We re trying to say, No, you can solve your own problems. You don t need anybody else. Let s just get started.... And we saw that just spending some time with some people, show that you care, give them a little bit of vision and some encouragement, bring a little bit of training, you can get them on the right track. (Bob Bradberry, personal interview, 2007) While the literature s predictions about evangelicals and church-based approaches to aid were supported, the predicted relationship between black Protestants and structural approaches to global poverty did not find support here. While the World Bank and IMF represent a structural approach to alleviating poverty, the structural alternative they represent may not align with the values held by this religious tradition. The consistent in-group advantage enjoyed by aid organizations linked to particular religious traditions particularly the very strong Catholic advantage seem to exceed the differences one would expect if support was based merely on aid values. The particularly strong intradenominational support for Catholic aid might be a consequence of churchgoers long tradition of funding Catholic schools, hospitals, and other social institutions. Alternatively, it could reflect an especially clear sense of shared religious identity with aid recipients in developing countries. In either case, the in-group advantage findings are worth noting in light of the overall declining denominationalism of American religion (Wuthnow 1990). The Catholic case points to how denominational structures that the influence of political ideology on charitable giving is mediated by religious and civic involvement.