Substitution or Symbiosis? Assessing the Relationship between Religious and Secular Giving

The authors would like to thank Atlantic Philanthropies for funding collection of the 2001, 2003 and 2005 Center on Philanthropy Panel Study and the Bill and Melinda Gates Foundation for funding data dissemination. Additional thanks go to Christian Smith and the members of the RASR seminar at Notre Dame for helpful comments on an earlier draft. We also wish to thank the anonymous reviewers for their contributions. Direct correspondence to Jonathan P. Hill, Department of Sociology & Social Work, Calvin College, 3201 Burton S.E., Grand Rapids, MI 49546.E-mail: [email protected].© 2011 The Author. Published by Oxford University Press on behalf of the University of North Carolina at Chapel Hill. All rights reserved. For permissions, please e-mail: [email protected].

Social Forces 90(1) 157–180, September 2011

Substitution or Symbiosis? Assessing the Relationship between Religious and Secular Giving

Jonathan P. Hill, Calvin CollegeBrandon Vaidyanathan, University of Notre Dame

Research on philanthropy has not sufficiently examined whether charitable giving to religious causes impinges on giving to secular causes. Examining three waves of national panel data, we find that the relationship between religious and secular giving is generally not of a zero-sum nature; families that increase their religious giving also increase their secular giving. We argue that this finding is best accounted for by a practice theory of social action which emphasizes how religious congregations foster skills and practices related to charitable giving. We also argue that denominational variation in the influence of religious giving is best accounted for by the financial structuring of the denomination. We conclude with the implications for studies of religious causal influence more generally.

While recent research has established a positive relationship between religiosity and financial giving (Bekkers and Weipking 2007; Brooks 2006; Wilhelm, Rooney and Tempel 2007), it is not clear whether money donated to religious causes impinges on giving to secular causes. In this article we focus specifically on the role of orga-nized religion in enabling or suppressing household-level charitable giving to secular causes.1 Several key sociological insights about giving behavior lead to conflicting expectations about the role of religious giving and participation in decisions to increase or decrease giving to secular causes. Using panel data spanning a five-year period, we develop a critical test of these competing hypotheses. Our findings re-veal that the relationship between religious and secular giving is generally not of a zero-sum nature: households that increase their religious giving tend to increase their secular giving as well, except in particular religious traditions. We argue that this finding is best accounted for by a practice theory of social action emphasizing how religious congregations foster skills and practices related to charitable giving. Nevertheless, we also find differences between denominations which suggest that the financial structuring of religious groups may shape givers’ perceptions of the degree of substitutability between religious and secular causes. Our study challenges some prevailing assumptions about how philanthropy works and proposes important future avenues of exploration.

at Calvin C

ollege on April 24, 2013

http://sf.oxfordjournals.org/D

ownloaded from

http://sf.oxfordjournals.org/

158 • Social Forces 90(1)

Religion and Giving to Secular Charities

The religious sector in the United States plays a large role in the charitable donations of individuals. Religious giving in the form of congregational donations and donations to religious charities is estimated to make up nearly half of all charitable giving by house-holds in the United States (Independent Sector 2001; Giving USA 2009). It remains a point of debate, however, whether and to what extent religious giving contributes to the welfare of others in society, particularly to those who are economically and socially at the margins of society.

On the one hand, religious congregations, who receive the bulk of religious chari-table giving, are institutions with the primary purpose of facilitating religious ritual and education for their members, not providing social services (Chaves 2004). Indeed, sociological accounts of the institutional strength of congregations, such as the “strict” church hypothesis (Iannaccone 1994; Kelley 1977), emphasize that congregations which deviate from “costly” religious commitment (e.g., doctrinal purity, demanding lifestyle restrictions, moral conformity and so on), and instead emphasize social services such as helping the needy, begin to quickly lose members. Thus, it should not be sur-prising that even among religious congregations that have some involvement in social service activity, less than 3 percent of an average congregation’s budget is actually spent on these programs (Chaves 2004). On the other hand, studies have found that religious individuals are more likely than non-religious individuals to donate to secular causes (Bekkers and Weipking 2007; Brooks 2006). For example, people who attend religious services weekly are more likely to give to nonreligious charities than people who rarely attend, and the value of the average religious household’s donations to nonreligious charities is 14 percent more than that of the average secular household (Brooks 2006).

These studies, however, do not tell us enough about the relationship between reli-gious and nonreligious charitable giving. First, the studies that have cited these statis-tics do not typically contain information on the type of nonreligious cause to which religious individuals donate. As James and Sharpe (2007) note, studies of charitable giving typically do not distinguish between different types of charitable giving. As a result, we do not know whether religion matters differently for donations to different kinds of causes. One contribution of this article, therefore, will be to assess the relation-ship between religion and donations to non-religious educational organizations, health organizations and organizations that serve the needy. Such an investigation will help us more adequately understand the relationship between religious and secular giving.

Second, it is not clear that religious giving is the causal mechanism in these studies. Because the majority of these studies are cross-sectional, focusing on a snapshot of the population at a given point in time, it is difficult to ascertain whether religious giving is what leads to secular charitable giving. An alternative explanation is that the relation-ship is spurious, and that a third factor is causally prior to both religious giving as well as secular charitable giving. For example, religious people might often be “joiners,” who simply tend to become involved in various associations and activities (Regnerus and Smith 2005; Wann and Hamlet 1994), and this mentality could explain both their

at Calvin C



ownloaded from


Religious and Secular Giving • 159

generosity to religious organizations and their secular charitable giving.2 Such studies also cannot rule out the possibility that giving to secular organizations is actually caus-ally prior to joining religious organizations.

To address these concerns with past studies, we use panel data to examine change in families’ giving behavior over time, as opposed to comparing individuals at a single point in time. Additionally, we examine both an overall measure of nonreligious giv-ing as well as measures of giving to specific domains such as education, healthcare, the needy and to umbrella organizations that cover multiple charitable domains. We consider two possible ways in which religious participation can relate to non-religious charitable giving: (1. religion as a suppressor of secular giving, and (2. religion as a facilitator of secular giving.

Substitution Hypothesis: Religious Giving as a Suppressor of Secular Giving

The heart of the substitution hypothesis rests upon a conception of the charitable giver as homo economicus – the rational, utility-maximizing decision maker. We find this con-ception particularly appropriate because the allocation of money has been traditionally assumed to be driven by rational calculation. In the classical version of decision theory individuals primarily choose between different courses of action by subjectively weigh-ing costs and benefits (and in conditions of risk, probabilities of success) in order to maximize their subjective expected utility (Becker 1993).

One of the most important assumptions driving neoclassical rational choice theory is the importance of stable and transitive preferences (Altman 2006; Hechter and Kanazawa 1997). This particular assumption provides analytical leverage when con-sidering why we might expect increases in religious giving to be accompanied by de-creases in secular giving among individuals over time. Beginning with this assumption, behavioral change (such as shifts in charitable giving) does not originate from new and different preferences – which are assumed to be constant – but rather is accounted for by varying circumstances external to the individual (Stark and Finke 2000). Thus, indi-viduals always choose optimally (according to the tenets of SEU) but in the context of shifting constraints on knowledge, resources, opportunities, skills, etc. (Becker 1993). This leads us to the general expectation that constraints which result in increased giving to religious causes can be expected to accompany decreased giving to secular causes, given that money is, by definition, a scarce resource and preferences related to charitable giving are stable. Although there may be other theoretical resources we could draw upon that support the substitution hypothesis, we find that the neoclassical formulation of rational choice theory – the so-called “thin” model of rational choice (Hechter and Kanazawa 1997) – provides an intuitive and parsimonious account.

Denominational Context

While we recognize that strict versions of SEU often remove decision-making from its social context, experimental research has demonstrated that social context should not be ignored (see Anand 1993; also see Mourao 2007 for similar evidence in national survey

at Calvin C



ownloaded from



data). Because of this we move away from rational choice models in order to hypothesize how the culture and structure of religious denominations may affect the degree to which religious adherents perceive religious and secular donations to be substitutable.3

The sociological evidence is clear that adherents of different religious denomina-tions vary in their degree of charitable giving (Chaves 1999; Smith, Emerson and Snell 2009). Because denominations are an important source of variation when it comes to individual allocation of financial resources, we suspect that variations in the degree of “substitutability” will be most likely evident at this level of social organization. Our analysis addresses this possibility by examining if the relationship between religious and secular giving depends on denominational affiliation.

The first source of variation across denominations might be the “strictness” of a religious group, which can be understood as the degree to which congregations within a given denomination raise the cost of nongroup activities. Iannaccone (1994:1187) contends that such exclusive churches explicitly or implicitly “penalize or prohibit alternative activities that compete for members’ resources.” The sense of exclusivity and social closure that characterizes these “strict churches” leads us to expect to find the substitution hypothesis most active here (Kelley 1977). Some strict churches, par-ticularly those that prioritize proselytizing, might believe that nonreligious charitable giving compromises the church’s mission because it ignores the religious dimension. Previous sociological studies have relied on scales ranking Christian denominations on their degree of sectarianism or exclusiveness, with “liberal” mainline denominations such as Episcopalians, Methodists and Presbyterians on one end, running through Lutherans, Catholics, to various evangelical and conservative denominations (such as Baptists), and finally, fundamentalists, Pentecostals and “sects” (into which they categorize groups such as Mormons) (see Iannaccone 1994 for a summary). Iannaccone finds that these rankings tend to be quite stable over time with a high degree of agree-ment between religion scholars on the relative ranking of strictness by denomination. The substitution hypothesis would expect more conservative and sect-like groups to encourage members to give primarily to the congregation and to religious charities approved by the congregation, and not to secular charities.

The second source of variation is the degree to which denominations mandate the support of denominational charities through donations of individuals in congrega-tions. Members of denominations which centrally control the allocation of donations to various causes might be more likely to see a zero-sum trade-off between religious and secular charitable giving. Although we know of no previous research that has ranked denominations or religious traditions according to this standard, we find evidence for three broad categories that can be ordinally arranged from strong to weak central fi-nancial control. The strongest control is evident in the Church of Jesus Christ of Latter Day Saints. All funds are collected from wards (i.e., congregations), are reallocated by LDS headquarters (Dahl and Ransom 1999), and are used to fund denominationally-administered charities such as LDS Humanitarian Services (Petty 2008). On the other end of the continuum lie Baptist and Pentecostal denominations, which value local

at Calvin C



ownloaded from



congregation financial autonomy.4 Even though charitable organizations exist (e.g., Baptist Charities of America), individual congregations are not institutionally wedded to these channels. This unavailability of denominationally-mandated charities may mean that Baptists and Pentecostals might be the least likely to view religious giving as a substitute for secular giving.

Other Christian denominations and traditions, such as Roman Catholics and most mainline Protestant denominations, fall somewhere in between Mormons and Baptists/Pentecostals. Our own contact with the financial offices of several mainline Protestant denominations confirmed that most have a number of denominational charities sup-ported by charitable contributions of members.5 Although there is typically a recom-mendation for local congregations to contribute funds earmarked for these charities to regional and national denominational structures, local congregations have considerable flexibility in the degree to which they adhere to these recommendations.

Data from the National Congregations Study (Chaves and Anderson 2008) provide some support for classifying denominations into these three categories. It is based on congregational leaders’ reports of the median percentage of total congregational income received during a fiscal year that was then given back to the denomination or convention. Pentecostal and Baptist congregations reported donating less than 5 percent of total congregational income to the denomination or convention; other Christian denominations (Presbyterians, Catholics, Lutherans, Episcopalians and Methodists) report donating 5 to 12 percent; and Mormons reported donating 98 percent. Jewish congregations reported donating 1.2 percent, but because they do not solicit donations during worship services we have excluded them from our categoriza-tion here. This measure can perhaps serve as a proxy for the degree to which members of a denomination may view religious and secular giving as substitutes.

In general, even though religious households may give more to secular charities than their nonreligious counterparts (Bekkers and Weipking 2007; Brooks 2006), the substitution hypothesis predicts they would be contributing even more if they would decrease giving to their local congregations or other religious causes. From this perspec-tive, secular charities would be financially healthier, and the poor in society would have their needs met more efficiently.

Symbiosis Hypothesis: Religious Giving as Facilitator of Secular Giving

In opposition to the substitution hypothesis, the symbiosis hypothesis holds that giving is not a zero-sum activity. Rather, charitable giving in the religious sector is likely to beget giving in the secular sector. This hypothesis questions the explanatory power of the rational, utility-maximizing conception of the individual when it is applied to giving behavior. Iannaccone (1995:81-82), a leading proponent of the rational choice approach to studying religion, has claimed that the primary alternative to the maximizing social action is “unreflective action based on habit, norms, emotion, neurosis, socialization, cultural constraints or the like – action that is largely unresponsive to changes in perceived costs, benefits or probabilities of success.” Moreover, the dominant economic theories

at Calvin C



ownloaded from



of charitable giving all draw upon a rational actor approach (Mourao 2008). While we do not believe it is necessary to consider charitable giving as somehow irrational, we do think the symbiosis hypothesis is best accounted for by reference to practice theories of social action as opposed to rational, utility-maximizing accounts.

As an alternative to the standard rational choice account of social behavior, many so-ciologists have conceived of social action as a set of habitually formed practices (Bourdieu 1990; Gross 2009; Swidler 2001). These practice-based accounts emphasize how culture and social interaction generate patterned dispositions that the actor is able to creatively use to achieve desired ends. Such an approach can explain people’s frequent lack of ar-ticulate ideological explanations for their behavior (Vaisey 2008). The primary content of socialization, in this view, is not propositional beliefs about which people have discursive consciousness, but rather, skills, habits, schemas and codes that are learned through experience and stored in procedural memory (Lizardo and Strand 2010).

We find this practice-based account particularly compelling as support for the symbiosis hypothesis because religious giving is most commonly situated in the insti-tutional framework of the local congregation. As Chaves (1999:182) argues in his dis-cussion of congregational giving, it is essential to understand “the ways in which they become institutionalized,” that is, “the ways in which they become relatively automatic, habitual, and unself-consciously part of what it means to be involved in a congrega-tion.” There are two primary reasons, then, why we suspect that a religious institutional context might be important in the development of giving as a routinized practice. First, from a social learning perspective (Bandura 1977) the entire process of giving is modeled and emulated within the institutional context of a religious congregation. Religious congregations provide a context in which people are made aware of the needs of others in the community, are recruited into social networks, and become involved in planning and carrying out charitable activities (Hodgkinson 1990; Wuthnow 1991). Secondly, individual giving in a congregational context is repetitive in nature, with giving opportunities occurring on at least a semi-regular basis. For most congregations, soliciting funds to support the institution is carried out weekly and is part of the ritual of worship itself. This is distinct from the one-time charitable gifts that are typical of many forms of non-religious giving. After an individual adopts practices of giving (for example, writing a check to one’s congregation on a weekly basis), these become part of the actor’s cultural repertoire and may be transferable to different settings. As Swidler (2001) argues, it is most often not utility-maximizing deductions from value-preferences that drive action, but such habits and skills that constitute actors’ cultural repertoires. If people learn charitable behavior in churches, and “if charitable behavior is habitual,” then it “could bleed over into [other] areas.”(Brooks 2003:46)

This approach is also supported by recent research on the intersection of cognitive neuroscience and sociology, particularly on the importance of automatic (rather than deliberate) cognition, and on how the mirror neuron system explains how people pick up practical competences from other actors in their surrounding environment and transfer them to other contexts (DiMaggio 1997; Vaisey 2008). In addition, consistent

at Calvin C



ownloaded from



with the symbiosis hypothesis, research on the psychological benefits of giving suggest that the very act of giving may stimulate reward centers in the brain (Harbaugh et al. 2007) providing positive emotional feedback. Andreoni (1990) has also argued that individuals, as a consequence of this giving, experience a “warm glow” effect, which can serve as an emotional motivator for future giving. If an individual begins to give more to religious causes, this may alter the psychological benefit of giving more generally and cause her to give more liberally to secular causes. However, because this “warm glow” effect is not domain specific, we would expect giving to secular causes to have the same effect on religious giving.

Unlike the substitution hypothesis, the symbiosis hypothesis developed above pre-dicts that secular organizations would benefit from individuals increasing their chari-table donations to religious congregations and organizations, despite such increases leaving less money for secular organizations.

Religious Giving as Causal Factor

Establishing the causal efficacy of religion has been a significant area of recent schol-arly inquiry in the sociology of religion (Chaves 2010). While a growing number of research articles identify religious beliefs and practices as theoretically central social mechanisms (Smilde and May 2010), the possibility of spurious self-selection processes and the plausibility of alternative explanations based on “reverse causation” continue to be contentious issues within the sociology of religion (Regnerus and Smith 2005). Self-selection arises precisely because religious practices, such as worship service at-tendance and congregational giving, are voluntary in nature. Individuals, in most circumstances, choose to practice religion. Several researchers posit that stable character traits such as risk aversion (Miller and Hoffman 1995), conformism (Ellison 1991), or the desire to belong to various social clubs and associations (Wann and Hamlet 1994) independently influence both the likelihood of choosing certain religious practices as well as many of the pro-social outcomes that researchers have identified as positively related to religious practices. This is a simple case of spuriousness resulting from factors that often go unmeasured in standard survey analysis.

The second causal concern, reverse causality, is often suspected when certain social behaviors are prescribed or proscribed by religious norms. If, for example, individuals begin to volunteer greater amounts of time and donate larger sums of money to secular charities, this might reasonably alter their friendship network and self-identity in such a way as to make church attendance and religious giving more or less frequent. In a simple cross-sectional analysis, the correlation between religious giving and secular giving would be significant, but it would be because secular giving alters religious giving and not the reverse.

Data and Methods

We use the 2001, 2003 and 2005 Center on Philanthropy Panel Study, which are modules within the biennial Panel Study of Income Dynamics (Wilhelm et al. 2005).

at Calvin C



ownloaded from



The PSID is a nationally representative, longitudinal study of economic and other demographic changes in U.S. adults and their family units, and is sponsored by the National Science Foundation and numerous other government agencies and private foundations. The sample consists of the original 1,968 core sample (which contains an equal probability selection of families from the 48 contiguous states as well as a separate sample of low-income families) as well as a refresher sample of immigrant families added in 1997 to improve sample representativeness. The survey is administered bien-nially by computer-assisted telephone interviewing. Response rates in any given survey year are typically greater than 95 percent.6 The COPPS modules contain detailed information on charitable giving, volunteer activity and religious participation since 2001. For this analysis, individuals were identified and matched by head of household to create a panel of 9,502 families.7

Dependent Variables

We use five primary dependent variables in our analyses. The first variable is derived from summing together the specific dollar amounts families reported giving to the fol-lowing types of organizations in the past year: educational organizations, health care or medical research organizations, organizations that provide food, shelter and other basic necessities, and organizations that serve a combination of (non-religious) purposes. These organizations may be run by religious groups, but not primarily concerned with proselytizing, religious education or spiritual development. The remaining outcome variables are the individual components of the additive measure. We transform all outcome variables using natural logs (adding a small constant in order to avoid taking the natural log of zero). In addition to correcting for the high positive skew of these variables, the natural log allows us to examine the percent change in money donated to secular causes associated with single unit changes of our independent variables.

Independent Variables

Our primary independent variables are measures of religious giving and religious par-ticipation. To measure religious giving, we take the log of the amount that the family reported donating to organizations for religious purposes or spiritual development.8 To measure religious participation we take the natural log of the annual number of religious worship services the head of household reported attending in the previous year. Religious service attendance was only included in the 2003 and 2005 COPPS, therefore any models that use these variables have been restricted to these years.

Similar to recent studies analyzing charitable giving (Andreoni, Brown and Rischall 2003; Guo and Peck 2009), we also include control measures for survey year,9 family composition, employment status of the head of household, and educational attainment of the head of household. Because of our fixed-effects modeling strategy, important time-invariant controls do not need to be directly included in the models. Measures of religious tradition and denomination are also included to capture differences in the effect of religious giving on secular giving by different theological and cultural commitments.

at Calvin C



ownloaded from



Plan of Analysis

In order to provide a critical test of our competing hypotheses, we estimate fixed-effects tobit models relating variance in giving to religious organizations to variance in giving to secular causes. This method largely deals with problems of spuriousness by focusing on variance within families measured at multiple points in time. Moreover, we estimate these effects for both the composite secular giving outcome as well as the individual giving components that make up the composite measure. This allows us to establish a general claim about the relationship between religious and secular giving as well as to determine whether the relationship is driven by domain-specific secular causes.

All of our models are constructed in the following way: we first estimate the influences of variation in religious giving on variation in secular giving while controlling for changes in family composition, employment status and educational attainment. Following this, we estimate the same model but substitute religious participation for religious giving. In the third model of both tables, we include both measures of attendance and measures of giving to determine whether some or all of the effect of religious service attendance is mediated through religious giving. Lastly, in our final model we examine whether the influence of religious giving is dependent on religious culture and institutional differences by including interaction terms by religious tradition/denomination.

We also include an additional set of models in a final table that addresses the issue of reverse causality. Although we cannot directly examine if our models suffer from endogeneity bias without employing instrumental variables, we can examine (1. if there is any correlation between the components of our secular measure and (2. whether these correlations are accounted for by religious giving. If there are no correlations between the different components of secular giving after religious giving is included in a multivariate model, then it is most likely that religious giving occurs causally prior to secular giving. While it is still technically possible that each domain of secular giving influences religious giving without influencing other domains of secular giving, such a scenario seems highly unlikely.

The Fixed Effect Tobit Model

Fixed effects, like the more commonly utilized random-effects models, are an effective way to deal with the problem of correlated error terms within individuals or families over time in panel data. However, unlike the random-effects model, the fixed-effects model eliminates the variance between cases by differencing out the estimating equa-tions after the fixed effects are entered. The remaining parameters represent the esti-mated effect of the covariates using only the variation that exists within cases measured at several points in time. Although we do not directly measure time-invariant variables, they are accounted for in the fixed-effects model. The most obvious advantage to a fixed-effects method over the standard random-effects model lies in the ability to control for stable, unobservable characteristics of respondents that commonly result in model misspecification (Allison 2005; Halaby 2004). In our particular application, this resolves much of the concern about the “joiner/giver” orientation. To the extent that

at Calvin C



ownloaded from



these hard-to-measure psychological constructs are stable over time, they are controlled for in our models. Although we realize this assumption cannot be fully met (i.e., it is possible to conceive of an individual becoming more or less of a giver/joiner over time), we believe it is a significant improvement over standard cross-sectional analyses and random effects models that use panel data.

We adopt the tobit model which is widely used in analyses of charitable giving (Wilhelm 2010; Duncan 1999). Because giving data are censored at zero, and because many families give nothing, standard OLS estimations are biased. The fixed-effects tobit model takes the following form:

γit* = αi + Xit β + εit

where εit ~ N(0; σ2), αi represents the unobserved time-invariant fixed effects that are differenced out of the model, Xit is the set of time-variant variables, β is a matrix of coefficients, and yit* is a latent variable representing the underlying desire or propen-sity to donate that is observed for values greater than 0. The observed y is therefore defined by the following measurement equation: yit = yit* if y* > 0; yit = 0 if yit* ≤ 0. It is important to note that for all models the coefficients are interpreted as the effect of the covariate on the latent uncensored outcome and not the observed censored outcome. Unfortunately maximum likelihood estimates for a fixed-effects, censored regression model suffer from a consistency issue known as the incidental parameters problem that results in bias in the standard errors (Neyman and Scott 1948). To overcome this issue, we use the semi-parametric estimator developed by Honoré (1992) as part of the Pantob program in Stata.

Results

Table 1 presents the results from a fixed-effects regression of the composite measure of secular giving on religious giving and participation. Model 1 estimates the effect of variance in within-family religious giving on variance in within-family secular giving net of changes in family income and other time variant control variables. The coeffi-cient for religious giving in this model is highly significant and implies that an increase of 10 percent in religious giving would result in an increase of more than 3.4 percent in secular giving, a result consistent with the symbiosis hypothesis.10

Model 2 in Table 1 replicates the analysis of Model 1, but substitutes a measure of the religious participation of head of the household for religious giving.11 These coefficients estimate that families in which the head of household increases his or her religious service attendance by 10 percent will increase the family’s secular giving by approximately 2 percent. Model 3 includes the measure of religious giving along with attendance. It is immediately clear that nearly all of the increase in secular giving related to religious service attendance is due to increases in religious giving. Although this corroborates past research, our model suggests that religious participation must lead to an increase in religious giving in order to influence secular giving. This also suggests that it is not information or other resources available within the context of a

at Calvin C



ownloaded from



religious congregation that influences the likelihood secular giving, but it is the actual act of religious giving, most likely to the local congregation, that explains variation in giving to secular causes. This finding is most consistent with the symbiosis hypothesis and our explanation of giving as a socially-learned practice.

Model 4 introduces interaction terms by religious tradition/denomination.12,13 The omitted reference category is Baptist. In a fixed-effects framework, this model

Table 1: Fixed-Effects Tobit Regression Predicting Logged Giving to Nonreligious Organizations

(1) (2) (3) (4)Giving and IncomeLn(Religious giving) .296*** .292*** .396***Ln(Family income) .372*** .367*** .289** .302***Religious ParticipationLn(# Religious services attended past year) .183** .017Yeara

2003 .079 .0402005 .274*** .269*** .212** .249***Family Number of children .068 .184 .117 -.039Divorced -.823* -1.085 -1.029 -.644Separated -.227 -.150 -.089 -.262Widowed .642 -.330 -.252 1.279Employment and EducationCurrently employedb .071 .307 .257 .017Retiredb -.258 -.086 -.043 -.266Disabledb -.292 -.532 -.506 -.669Educational attainment .122* -.905 -.783 .167*Interactions by Religious Traditionc,d

Catholic * Ln(Religious giving) -.124**Lutheran * Ln(Religious giving) -.206*Methodist * Ln(Religious giving) -.143*Presbyterian * Ln(Religious giving) -.185+Episcopalian * Ln(Religious giving) -.320*LDS * Ln(Religious giving) -.876***Pentecostal * Ln(Religious giving) -.102Jewish * Ln(Religious giving) -.340***Other religion * Ln(Religious giving) -.029N 21,481 14,456e 14,456e 14,960f

Groups 9,328 8,311 8,311 6,302Source: Center on Philanthropy Panel Study, 2001-2005Notes: aReference group is 2001 in models 1 and 4, reference group is 2003 in models 2 and 3; bReference group is unemployed; cReference group is Baptist; dThe following number of families (groups) in each religious tradition are as follows: Baptist(2,619), Catholic(1,718), Lutheran(377), Methodist(688), Presbyterian(185), Episcopalian(116), Mormon(73), Pentecostal(236), Jewish(163), Other Religion(127); eModels 2 and 3 only utilize the 2003 and 2005 waves of COPPS; f Model 4 only includes families who belong to one of the 10 major religious traditions/denominations specifically measured in the data.+p < .10 *p < .05 **p < .01 ***p < .001

at Calvin C



ownloaded from



estimates if variation in religious giving within families is associated with a different magnitude of change in secular giving depending on the religious tradition of the head of household. Examining the coefficients in this model, it is clear that the estimated main effect of religious giving associated with Baptists is larger in magnitude compared to the overall effect from religious giving for the entire population in prior models. Measured as deviations from this Baptist effect, all measured religious denominations and traditions have a weaker relationship between religious and secular giving. There is no statistically significant difference between Baptists and Pentecostals and between Baptists and “other” religious groups. Mormons exhibit the strongest negative effect. Not only is the estimated effect for Mormons significantly different from Baptists, the estimated coefficient is also statistically significantly different from zero (p < .001, results not shown) with a negative estimated effect from religious giving on secular giving. These results appear to be consistent with our expectations concerning the financial structure of the denominations but mostly inconsistent with cultural expecta-tions regarding “strictness.”

In Table 2 through Table 5, we replicate the modeling strategy of Table 1, but we predict giving to non-religious organizations that make up the various components of the composite secular giving measure. In Model 1, the estimated effect of religious giving on the various measures of secular giving is highly significant (p < .001) for each outcome. The effect size of the religious giving coefficient is roughly similar to the coefficient in Table 1 for the remaining tables with the exception of giving to non-religious, combination-purpose organizations, which has a slightly weaker relationship to religious giving (Table 5: β = .206).

Models 2 and 3 in the remaining tables examine the influence of religious service attendance on giving to secular causes. We see a similar pattern for Table 2 (giving to nonreligious organizations that help the needy), Table 4 (giving to nonreligious health organizations), and Table 5 (giving to nonreligious combination purpose organiza-tions). In each of these cases the coefficients for religious service attendance are positive in Model 2 and drop to near zero in Model 3. This is similar to the results in Table 1 using the composite giving outcome. However, the coefficients for religious service attendance are not statistically significant in Table 2, Table 4 or Table 5. In Table 3 (giv-ing to nonreligious educational organizations) there is a negative effect from religious service attendance that is strengthened after religious giving is included in Model 3. This suggests that net of the positive increase from religious giving, increases in reli-gious service attendance by the head of household are associated with significantly less giving to educational organizations.

Lastly, Model 4 tests for interaction terms by religious tradition/denomination. In all remaining tables, Baptists continue to exhibit a stronger positive influence from religious giving when compared to most other denominations and religious tradi-tions. Pentecostals are the only religious group that is never significantly different from Baptists in Table 2 through Table 5. The strong difference between Mormons and Baptists, evident in the composite measure (Table 1), appears to be driven primarily

at Calvin C



ownloaded from



by differences in the effect of religious giving on nonreligious giving to organiza-tions that provide for the needy (Table 2). Mainline Protestant denominations and Roman Catholics have negative interaction coefficients in tables 2 through 5, with Episcopalians being the most different from Baptists in Table 4 (giving to health orga-nizations) and Table 5 (giving to combination purpose organizations).

Table 2: Fixed-Effects Tobit Regression Predicting Logged Giving to Nonreligious Organizations Helping the Needy

(1) (2) (3) (4)Giving and IncomeLn(Religious giving) .269*** .271*** .414***Ln(Family income) .382*** .342* .281* .316**Religious ParticipationLn(# Religious services attended past year) .127 -.017Yeara

2003 .264* .1702005 .402** .197 .133 .328*Family Number of children .075 -.006 -.077 -.066Divorced -.002 -.388 -.521 -.110Separated .251 -.569 -.338 .488Widowed 1.007 1.235 1.485 1.881Employment and EducationCurrently employedb -.242 .149 .116 -.280Retiredb -.770+ -.621 -.660 -.665Disabledb -.685 -.126 -.106 -.510Educational attainment .320*** -.265 1.359 .393***Interactions by Religious Traditionc,d

Catholic * Ln(Religious giving) -.168*Lutheran * Ln(Religious giving) -.181Methodist * Ln(Religious giving) -.154Presbyterian * Ln(Religious giving) -.467*Episcopalian * Ln(Religious giving) -.174LDS * Ln(Religious giving) -1.211***Pentecostal * Ln(Religious giving) -.060Jewish * Ln(Religious giving) -.222Other religion * Ln(Religious giving) -.127N 21,481 14,456e 14,456e 14,960f

Groups 9,328 8,311 8,311 6,302Source: Center on Philanthropy Panel Study, 2001-2005Notes: aReference group is 2001 in models 1 and 4, reference group is 2003 in models 2 and 3; bReference group is unemployed; cReference group is Baptist; dThe following number of families (groups) in each religious tradition are as follows: Baptist(2,619), Catholic(1,718), Lutheran(377), Methodist(688), Presbyterian(185), Episcopalian(116), Mormon(73), Pentecostal(236), Jewish(163), Other Religion(127); eModels 2 and 3 only utilize the 2003 and 2005 waves of COPPS; fModel 4 only includes families who belong to one of the 10 major religious traditions/denominations specifically measured in the data.+p < .10 *p < .05 **p < .01 ***p < .001

at Calvin C



ownloaded from



Table 6 addresses issues of causal order in our model. To compare the influence of religious giving with each measure of secular giving, we use fixed-effects tobit models to regress combinations of three measures of secular giving on the fourth measure of secular giving and religious giving (including the standard control variables from Model 1 of previous tables). We also report percent change in effect size by calculat-ing the proportional change in the size of the coefficient from a model with only the

Table 3: Fixed-Effects Tobit Regression Predicting Logged Giving to Nonreligious Educational Organizations

(1) (2) (3) (4)Giving and IncomeLn(Religious giving) .282*** .313*** .414***Ln(Family income) .529*** .557** .419** .622***Religious ParticipationLn(# Religious services attended past year) -.135 -.283*Yeara

2003 .268+ .276+2005 .549*** .413** .336* .525**Family Number of children .175 .376 .289 .183Divorced -2.120** -1.501 -1.444 -2.726**Separated -.976 .265 .004 -.877Widowed 1.669 .573 .895 3.508**Employment and EducationCurrently employedb .324 .658 .488 .529Retiredb .256 .372 .349 .472Disabledb 1.490* .895 .989 1.304Educational attainment .163 .144 .129 .157Interactions by Religious Traditionc,d

Catholic * Ln(Religious giving) -.117Lutheran * Ln(Religious giving) -.520+Methodist * Ln(Religious giving) -.291*Presbyterian * Ln(Religious giving) -.125Episcopalian * Ln(Religious giving) -.549**LDS * Ln(Religious giving) -.590Pentecostal * Ln(Religious giving) .043Jewish * Ln(Religious giving) -.274*Other religion * Ln(Religious giving) -.448*N 21,481 14,456e 14,456e 14,960f


at Calvin C



ownloaded from



variable shown (plus the standard control variables) to the model shown with two covariates measuring charitable giving. Across all four models religious giving stands out as the most robust influence on measures of secular giving. However, in all four models measures of secular giving are also independently related to the other measures of secular giving, even after controlling for religious giving. Additionally, although some of the influence of secular giving can be explained by religious giving (ranging

Table 4: Fixed-Effects Tobit Regression Predicting Logged Giving to Nonreligious Health Organizations

(1) (2) (3) (4)Giving and IncomeLn(Religious giving) .282*** .284*** .393***Ln(Family income) .471*** .699*** .593*** .286+Religious ParticipationLn(# Religious services attended past year) .112 -.022Yeara

2003 .021 -.0632005 .543*** .607*** .534*** .443**Family Number of children -.120 .093 -.012 -.323+Divorced -.244 1.861 2.038 .127Separated -.344 .692 .686 -.885Widowed 1.237 2.706 2.501 1.234Employment and EducationCurrently employedb .313 .158 .039 .299Retiredb .385 .360 .315 .465Disabledb .429 -.663 -.745 .654Educational attainment -.065 -1.642*** .082 -.006Interactions by Religious Traditionc,d

Catholic * Ln(Religious giving) -.074Lutheran * Ln(Religious giving) -.134Methodist * Ln(Religious giving) -.101Presbyterian * Ln(Religious giving) -.047Episcopalian * Ln(Religious giving) -.517*LDS * Ln(Religious giving) -.313Pentecostal * Ln(Religious giving) -.054Jewish * Ln(Religious giving) -.443**Other religion * Ln(Religious giving) -.323N 21,481 14,456e 14,456e 14,960f


at Calvin C



ownloaded from



from 10.62% to 28.45% reduction in coefficient size with the inclusion of religious giving), an independent influence from secular giving still exists. Of the measures of secular giving, giving to educational causes (Model 2) seems to be the most strongly related to the other measures of secular giving. The findings in this table lend partial support to religious giving as likely being causally prior to most secular giving.

Table 5: Fixed-Effects Tobit Regression Predicting Logged Giving to Nonreligious Combination Purpose Organizations

(1) (2) (3) (4)Giving and IncomeLn(Religious giving) .206*** .202*** .262***Ln(Family income) .576*** .435** .375* .405**Religious ParticipationLn(# Religious services attended past year) .158 .047Yeara

2003 -.338*** -.327**2005 -.254* .161 .116 -.197Family Number of children .066 .321 .257 .107Divorced -1.058* -2.215** -2.082** -.844Separated -.415 -1.021 -.961 -.338Widowed .034 -.502 .192 .703Employment and EducationCurrently employedb .285 .518 .483 .148Retiredb -.207 .007 .090 -.475Disabledb -.936 -1.990* -1.827* -1.521*Educational attainment -.087 1.122 .981 -.024Interactions by Religious Traditionc,d

Catholic * Ln(Religious giving) -.063Lutheran * Ln(Religious giving) -.087Methodist * Ln(Religious giving) -.233*Presbyterian * Ln(Religious giving) .059Episcopalian * Ln(Religious giving) -.500+LDS * Ln(Religious giving) -.330Pentecostal * Ln(Religious giving) -.201Jewish * Ln(Religious giving) -.122Other religion * Ln(Religious giving) -.063N 21,481 14,456e 14,456e 14,960f


at Calvin C



ownloaded from



Discussion

Overall, the results suggest greater support for the symbiosis hypothesis over the substitu-tion hypothesis. Organized religion does not appear to be siphoning individual donations away from secular charities. We suspect this is largely accounted for by the organizational and cultural conditions of giving in a local congregation. Although it is not possible to conclude with certainty that a social practices account of charitable giving explains the social mechanisms responsible for our empirical findings, we find this account more compelling than the account which emphasizes rational utility maximization. This find-ing has broader implications for how social scientists conceive of the overall pool of charitable donations. Rather than treating this sum as a fixed resource that can potentially be distributed among various nonprofit and charitable organizations, it is important to incorporate the actual micro processes that underlie giving transactions and the institu-tional context that supports giving behavior. There is no comparable set of institutions, outside of religious congregations and other places of worship, which routinely model the giving process coupled with a high normative giving expectation (Chaves 2004). Because of this, we suspect, and our analysis also suggests, that philanthropy to secular causes in the United States would suffer if religious giving declines in the future.

Table 6: Fixed Effects Tobit Regression Predicting Logged Giving to Secular Causes(1) (2) (3) (4)

Education,Health &

Combination

Needy,Health &

Combination

Needy,Education &Combination

Needy,Education &

HealthLn(Religious giving) .262*** .268*** .273*** .297***

[-2.71] [-4.51] [-3.32] [-1.29]Ln(Family income) .447*** .352*** .375*** .343***Secular GivingLn(Giving to needy) .085***

[-16.53]Ln(Giving to education) .203***

[-10.62]Ln(Giving to health) .118***

[-18.79]Ln(Giving to combination) .048*

[-28.45]N 21,481 21,481 21,481 21,481Groups 9,328 9,328 9,328 9,328Source: Center on Philanthropy Panel Study 2001-2005Notes: The following control variables are included in all models, but not shown: year 2003, year 2005, number of children, divorced, separated, widowed, currently employed, retired, disabled, educational attainment.Percent change [in brackets] is a calculation of the proportional change in the size of the coefficient from a model with only the variable shown (plus the standard control variables) to the model shown with two covariates measuring charitable giving.+p < .10 *p < .05 **p < .01 ***p < .001

at Calvin C



ownloaded from



However, it is important to note that we find statistically significant differences by religious tradition and denomination in the effect of religious giving on secular giving. If the practice-based social mechanism provides a convincing account of the general positive correlation between religious and secular giving, we are not certain that it provides a compelling account of denominational variation. The frequency and intensity of opportunities and normative pressures related to giving would have to substantially vary between denominations, and we do not have clear evidence that this is the case (although we do not discount the possibility of this). However, we can conclude that our statistical models are consistent with our hypothesis regarding the financial structuring of denominations.

As we outlined in our initial articulation of the substitution hypothesis, some reli-gious traditions, such as the Mormonism, have centrally-controlled denominational charities, which could be conceived as competing with similar secular charities, and are largely supported by individual contributions to their local congregation. Compared to financially unstructured denominational groups, such as Baptists and Pentecostals, adherents to these traditions are likely to perceive giving to their religious congregation as a substitute for giving to secular charities in similar domains as their denominational charities. Denominations that fall in between Mormons and Baptists/Pentecostals typically display a stronger symbiosis effect than the former, but a weaker effect than the latter. However in giving to healthcare organizations and combination purpose organizations, Episcopalians come out as the most different from Baptists. It is pos-sible that, when it comes to these categories, Episcopalians are more likely to have denominationally-sponsored substitutes than Mormons (despite the central control of finances in the LDS church).

This perception of substitutable giving appears to be a weaker effect, on aver-age, than the overall symbiotic relationship between religious and secular giving. In most instances it merely mutes the positive correlation between religious and secular giving. However, for the Mormons in our sample, this substitution effect is very strong and actually results in an overall negative relationship between reli-gious and secular giving. Religious “strictness,” at least as previously conceptualized (Iannaccone 1994), does not provide a compelling account of our findings. Although it is consistent with our findings concerning Mormons, other religious groups such as Pentecostals and Baptists, which the earlier literature categorizes as being simi-larly strict, are more likely to give to secular organizations as a result of increases in religious giving than less strict religious groups.

In sum, although we suspect the practice-based social mechanism is operating in most religious congregations, hierarchical financial structuring of religious denomina-tions likely tempers this effect. It is also noteworthy that denominational differences seemed most pronounced for giving to nonreligious educational organizations and nonreligious organizations that help the needy. We suspect this is due to the likeli-hood that some denominations are more likely to have mandated substitutes for these categories compared to healthcare-related charities or combination-purpose charities.

at Calvin C



ownloaded from



Our primary finding also speaks to the debate over the causal vs. spurious influ-ence of religion in civic life (Regnerus and Smith 2005). Although prior studies found that religious individuals give marginally more to secular causes, they are unable to establish that religious conviction or participation is somehow the cause of this differ-ence in giving. Arguably the same psychological orientations that predispose some to join and participate in religious congregations could also be the primary motivator of other civic and pro-social activity (Wann and Hamlet 1994). Our findings, however, clearly demonstrate a positive correlation between religious and secular giving within the same families measured at several points in time. Our modeling strategy allows us to have significantly greater confidence that the relationship between religious and secular giving is not spurious.

We also address the possibility of reverse causation – shifts in secular giving alter the incidence of religious giving. Because the symbiosis hypothesis emphasizes the importance of religious institutions in producing giving practices which influence rates of secular giving, evidence of strong reverse causation would certainly undermine the plausibility of the practice-based account. Our findings suggest that religious giving is likely causally prior in most cases. However, we cannot rule out that sometimes there may be moderate to small independent effects from secular giving on religious giving. In particular, giving to educational causes maintains a substantial correlation with other forms of secular giving even after religious giving is accounted for, suggesting that giving to educational causes could partly be causally prior to religious giving. We suspect that the recipients of most educational giving are institutions that family members currently attend or have attended in the past. Familial ties to schools may activate similar institutional mechanisms as those suspected to operate in religious congregations, although probably to a lesser degree. Further investigation is needed to understand why educational giving appears to have some parallel effects as religious giving. It is clear however, that variations in specific domains of giving have different effects on further giving, suggesting that general psychological processes, such as the “warm-glow” phenomenon (Andreoni 1990) that are not domain-specific, are not sufficient to explain the findings.

Future Research

Our research points to the importance of further understanding the institutional context of giving practices. We believe that future research should focus on specifying and testing the social mechanisms presented in this article more precisely. Although the present research is able to establish the existence of a general symbiotic relationship between religious and secular giving in the American population, we believe that a stronger case could be made for linking our proposed mechanisms to giving outcomes through focused, localized ethnographic research. More precisely, by analyzing the frequency and intensity of giving opportunities in religious congregations, along with cataloging charitable giving and accompanying accounts of charitable behavior, future researchers would be better equipped to document the social mechanisms that generate

at Calvin C



ownloaded from



the empirical findings of this article. In addition, strategic selection of congregations based on denominational identity would allow researchers to systematically compare congregations that have similar demographic features but have differing institutional constraints. We believe that this basic strategy – establishing broad, population-level, empirical findings first, followed by localized research that is better able to assess the likelihood of particular social mechanisms – is the most promising avenue for assessing how religious giving is related to patterns of charitable giving.

Moreover, specifying the impact of organizational structures and institutional cultures on the set of skills, habits and practices that are available to a social actor may be an important avenue forward in sociological accounts of religious influence more generally. Chaves (2010), for example, makes several important suggestions for overcoming the “religious congruence fallacy” or the tendency of social scientists who study religion to uncritically assign causal efficacy to religious beliefs despite decades of research that demonstrate weak correlations between attitudes and behaviors. He states the challenge in the following way: “If we want to claim religious internalization, we would have to show that people have had enough relevant, reinforcing religious expe-riences to forge an internalized connection strong enough to produce an automatic, habitual response in a particular situation.”(11) Furthermore, he points to the impor-tance of paying attention to decision-making situations by “investigating the features of the situations that activate religious schema in people.”(12) Future research, then, should be oriented to cataloging what religious schemas become internalized through the ritual of worship and congregational social interaction, and documenting how these internalized schemas become activated in non-religious settings. We suspect that such an approach, which can be pursued through ethnographic as well as experimental methodologies, would provide insight into the social mechanisms that operate in link-ing religious beliefs and practices to a host of important social outcomes.

Notes1. In this article we define and operationalize nonreligious causes as charitable activities whose

primary end is not religious in nature, even if many of these charitable organizations are sponsored by religious groups. Previous research supports this distinction between religious purposes (e.g., religious services, education, etc.) and other activities such as helping the needy, which are not considered the primary function of religious congregations (see Kelley 1977; Chaves 2004).

2. Ideally we would control for the “joiner” tendency by including measures of the number and type of voluntary organizations our sample members belong to. Unfortunately these measures are not available in our data.

3. Although we recognize that there are alternative rational choice models that explicitly model social context. For example, the new institutional economics (Coase 1937; Williamson 2000) has modified the traditional neoclassical assumptions to explicitly take into account organizations and institutional factors that might influence decision making.

4. See http://www.sbc.net/aboutus/clwhydenomination.asp for the Southern Baptist Convention’s statement on congregational autonomy. See http://ag.org/top/About/structure.cfm for a similar statement from the Assemblies of God.

at Calvin C



ownloaded from



5. Details are available upon request.6. For information on survey design and administration, consult the extensive online

documentation available at http://psidonline.isr.umich.edu.7. The label “family” is used in a very precise manner in the PSID. Family units typically

refer to individuals living in the same household who are related by blood, marriage or adoption. Individuals who live alone are therefore considered family units and included in the sample. While this complicates the translation from theory (primarily focused on individuals’ charitable giving) to empirical test (where more than one person may make decisions to give and financial resources are shared), we expect the hypothesized mechanisms relating religious and secular giving to be the same.

8. While we do not have a direct measure of congregational giving, estimates of religious giving suggest that the overwhelming majority is through religious congregations (see the denominational model in Brown, Harris and Rooney 2004). Our own estimates using the Panel Study of American Religion and Ethnicity suggest that 89% of all individual financial gifts to religious organization are through religious congregations.

9. Apart from controlling for survey year, we are uninterested in modeling any other temporal effects in our models. Time is simply the occasion for collecting multiple points of data within families, and it matters little to our models directly.

10. In order to translate the coefficient to a percentage difference, the following expression was used: 100(Exp(β) – 1) (Hardy 1993). For example, 100(Exp(.296) -1) = 34.4. We simply divide by 10 to estimate the effect of a 10 percent increase, as opposed to a 100 percent increase, on the outcome. This method is used throughout this article to derive intuitively meaningful estimates from the coefficients.

11. The religious attendance measure was only included in the 2003 and 2005 waves of COPPS. Therefore models 2 and 3 in both tables only use data from these waves. Additional analyses reveal that the coefficients for religious giving are remarkably stable between models that use the full three waves of sample data and identical models that only use data from 2003 and 2005. The largest fluctuation in the coefficient estimate for religious giving was found in Table 4. Using all three waves the estimate is .282 for Model 1 in Table 4. The coefficient changes to .300 in a model that restricts the data to the years 2003 and 2005.

12. The N in these models is reduced due to restricting the sample to only families who identify with a major religious tradition/denomination. We utilize this strategy to make the reference category meaningful.

13. The main effects of religious tradition cannot be estimated in a fixed effects model because religious tradition is time-invariant.

ReferencesAnand, Paul. 1993. Foundations of Rational Choice Under Risk. Oxford University Press.Allison, Paul D. 2005. Fixed Effects Regression Methods for Longitudinal Data Using SAS. Cary,

NC: SAS Institute.Altman, Morris. Foundations of Contemporary Behavioral Economics: Foundations and

Developments. M.E. Sharpe.Andreoni, James. 1990. “Impure Altruism and Donations to Public goods: A Theory of Warm-

Glow Giving.” Economic Journal 100(401):464-77.

at Calvin C



ownloaded from



Andreoni, James, Eleanor Brown and Isaac Rischall. 2003. “Charitable Giving by Married Couples: Who Decides and Why Does it Matter?” Journal of Human Resources 38(1):111-33.

Bandura, Alfred. 1977. Social Learning Theory. Prentice-Hall.Becker, Gary S. 1993. “The Economic Way of Looking at Life.” Journal of Political Economy

101(3):385-409.Bekkers, Rene, and Pamela Wiepking. 2007. Generosity and Philanthropy: A Literature Review.

Available at: http://ssrn.com/abstract=1015507.Bourdieu, Pierre. 1990. The Logic of Practice. Stanford University Press.Brooks, Arthur C. 2003. “Religious Faith and Charitable Giving.” Policy Review 121:39-50.______. 2006. Who Really Cares: The Surprising Truth About Compassionate Conservatism.

Basic Books.Brown, Melissa S., Joseph C. Harris and Patrick M. Rooney. 2004. “Reconciling Estimates

of Religious Giving.” Available at: http://www.philanthropy.iupui.edu/workingpapers/reconciling_religious_giving_estimates.pdf.

Chaves, Mark. 1999. “Financing American Religion.” Pp. 169-88. Financing American Religion. Mark Chaves and Sharon L. Miller, editors. AltaMira Press.

______. 2004. Congregations in America. Harvard University Press.______. 2010. “SSSR Presidential Address. Rain Dances in the Dry Season: Overcoming

the Religious Congruence Fallacy.” Journal for the Scientific Study of Religion 49(1):1-14. Chaves, Mark, and Shawna Anderson. 2008. National Congregations Study. Cumulative data

file and codebook. Available at: http://www.thearda.com/Archive/Files/Descriptions/NCSPANEL.asp.

Coase, Ronald H. 1937. “The Nature of the Firm.” Economica 4(16):386-405.Dahl, Gordon B., and Michael R. Ransom. 1999. “Does Where You Stand Depend on Where You

Sit? Tithing Donations and Self-Serving Beliefs.” American Economic Review 89(4):703-27.DiMaggio, Paul. 1997. “Culture and Cognition.” Annual Review of Sociology 23:263-87.Duncan, Brian. 1999. “Modeling Charitable Contributions of Time and Money.” Journal of

Public Economics 72(2):213-42.Ellison, Christopher G. 1991. “Bringing it Back In: The Life Stress Process and the Study of

Depression.” Ph.D. Dissertation, Duke University.Giving USA. 2009. Annual Report. Available at: http://www.givingusa.org.Gross, Neil. 2009. “A Pragmatist Theory of Social Mechanisms.” American Sociological Review

74(3):358-79.Guo, Chao, and Laura R. Peck. 2009. “Giving and Getting: Charitable Activity and Public

Assistance.” Administration & Society 41(5):600-27.Halaby, Charles N. 2004. “Panel Models in Sociological Research: Theory Into Practice.”

Annual Review of Sociology 30:507-44.Harbaugh, William, Ulrich Mayr and Dan Burghart. 2007. “Neural Responses to Taxation

and Voluntary Giving Reveal Motives for Charitable Giving.” Science 316(5831):1622-25.Hardy, Melissa A. 1993. Regression with Dummy Variables. Sage. Hechter, Michael, and Satoshi Kanazawa. 1997. “Sociological Rational Choice Theory.”

Annual Review of Sociology 23:191-214.Hodgkinson, Virginia Ann. 1990. “The Future of Individual Giving and Volunteering: The

Inseparable Link Between Religious Community and Individual Generosity.” Pp. 284-312. Faith and Philanthropy in America: Exploring the Role of Religion in America’s Voluntary Sector. Robert Wuthnow and Virginia A. Hodgkinson, editors. Jossey Bass.

at Calvin C



ownloaded from



Honoré, Bo. 1992. “Trimmed LAD and Least Squares Estimation of Truncated and Censored Regression Models with Fixed Effects.” Econometrica 60(3):533-65.

Iannaccone, Laurence. 1994. “Why Strict Churches Are Strong.” American Journal of Sociology 99(5):1180-211.

______. 1995. “Voodoo Economics? Reviewing the Rational Choice Approach to Religion.” Journal for the Scientific Study of Religion 34(1):76-88.

Independent Sector. 2001. Giving and Volunteering in the United States. Independent Sector.James III, Russell N., and Deanna L. Sharpe. 2007. “The ‘Sect Effect’ in Charitable Giving:

Distinctive Realities of Exclusively Religious Charitable Givers.” American Journal of Economics and Sociology 66(4):697-726.

Kelley, Dean M. 1977[1972]. Why Conservative Churches Are Growing: A Study in Sociology of Religion. Harper and Row.

Lizardo, Omar, and Michael Strand. 2010. “Skills, Toolkits, Contexts and Institutions: Clarifying the Relationship Between Different Approaches to Cognition in Cultural Sociology.” Poetics 38(2):204-27.

Miller, Alan S., and John P. Hoffman. 1995. “Risk and Religion: An Explanation of Gender Differences in Religiosity.” Journal for the Scientific Study of Religion 34(1):63-75.

Mourão, Paulo Reis. 2007. “Portuguese Public Collections and the Economic Cycle: A Seminal Study.” International Journal of Social Economics 34(12):961-76.

______. 2008. “What Has the Economics of Giving Given to Economics?” Journal of Economic and Social Research 10(1):1-33.

Neyman, Jerzy, and Elizabeth Scott. 1948. “Consistent Estimates Based on Partially Consistent Observations.” Econometrica 16(1):1-32.

Petty, Miles. 2008. “Mormonism and Private Aid.” Hinckley Journal of Politics 9:51-58. Regnerus, Mark D., and Christian Smith. 2005. “Selection Effects in Studies of Religious

Influence.” Review of Religious Research 47(1):23-50.Smilde, David, and Matthew May. 2010. “The Emerging Strong Program in the Sociology

of Religion.” SSRC Working Paper. Available at http://blogs.ssrc.org/tif/wp-content/uploads/2010/02/Emerging-Strong-Program-TIF.pdf.

Smith, Christian, and Michael Emerson, with Patricia Snell. 2008. Passing the Plate: Why American Christians Don’t Give Away More Money. Oxford University Press.

Stark, Rodney, and Roger Finke. 2000. Acts of Faith: Explaining the Human Side of Religion. University of California Press.

Swidler, Ann. 2001. Talk of Love: How Culture Matters. University of Chicago Press.Vaisey, Stephen. 2008. “Socrates, Skinner, and Aristotle: Three Ways of Thinking About

Culture in Action.” Sociological Forum 23(3):603-13.Wann, Daniel L., and Michael A. Hamlet. 1994. “The Joiners Scale: Validation of a Measure

of Social Complexity.” Psychological Reports 74(3/1):1027-34.Wilhelm, Mark O. 2008. “The 2005 Center on Philanthropy Panel Study User’s Guide.” Mimeo,

IUPUI. Available at: https://oncourse.iu.edu/access/content/user/mowilhel/Web_page/Main_Data/2005%20Center%20Panel%20Giving%20and%20Volunteering%20Users%20Guide.pdf.

Wilhelm, Mark O., Eleanor Brown, Patrick M. Rooney and Richard Steinberg. 2005. The Center on Philanthropy Panel Study [machine-readable data file]. Available at: http://www.philanthropy.iupui.edu/research/copps.

at Calvin C



ownloaded from



Wilhelm, Mark O., Patrick Rooney and Eugene Tempel. 2007. “Changes in Religious Giving Reflect Changes in Involvement: Age and Cohort Effects in Religious Giving, Secular Giving, and Attendance.” Journal for the Scientific Study of Religion 46(2):217-32.

Wilhelm, Mark O. 2010. “Giving to Organizations that Help People in Need: Differences across Denominational Identities.” Journal for the Scientific Study of Religion 49(3):389-413.

Williamson, Oliver E. 2000. “The New Institutional Economics: Taking Stock, Looking Ahead” Journal of Economic Literature 38(3):595-613.

Wuthnow, Robert. 1991. Acts of Compassion: Caring for Others and Helping Ourselves. Princeton University Press.

at Calvin C



ownloaded from


Substitution or Symbiosis? Assessing the Relationship between Religious and Secular Giving

Documents