|
Article Excerpt
I. INTRODUCTION
Faith-based organizations provide social services for more than 70 million Americans annually (Johnson, Tompkins, and Webb 2002). Of the more than 350,000 congregations in the country, more than 85% support some type of social service activity (Solomon 2003). Biddle (1992) estimates that each year, congregations spend 24 billion dollars on philanthropic activities.
These actions have gained new relevance for policymakers as the availability of public funding for charitable church activity has dramatically increased in the past decade. Starting with the "charitable choice" provision of the 1996 welfare reform law and continuing with the creation of Centers for Faith-Based and Community Initiatives located in 11 different federal agencies, policymakers have instituted a "devolution" of federal funds intended to provide community services. In 2005, more than 2.1 billion dollars in government grants were awarded to religious organizations by federal agencies (Office of the Press Secretary 2006). (1)
Some observers fear that this policy trend will encourage racial discrimination by congregations (Leonard 2002). As Bartkowski and Regis (2003) write, "Congregational definitions of legitimate [social] problems that members care about are, in actuality, influenced by various considerations--among them, racialized perceptions of welfare recipients." Past studies by economists have postulated, but not substantiated, that churchgoers warrant special consideration as individuals whose actions are sensitive to community demographics (e.g., Alesina and La Ferrara 2000; Gruber 2005).
Alternately, some policymakers have championed faith-based organizations because they are not likely to display sensitivity to community demographics when providing services. As President Bush stated in a speech in 2002, "Faith-based charities work daily miracles because ... above all, they recognize the dignity of every citizen and the possibilities of every life" (Office of the Press Secretary 2002). John DiIulio, the former director of the White House Office of Faith-Based and Community Initiatives, argued for charitable choice by saying, "government at all levels blatantly denies the equal protection of the laws to low income, urban African American and Latino children and families who depend disproportionately on faith-based organizations for both basic life necessities and the delivery of specific social welfare benefits" (DiIulio 2002). There is also (noneconomic) research consistent with this view (e.g., Billiet 1995; Johnson 1977).
The goal of this article was to examine whether congregations are affected by the racial composition of the local community when deciding whether and how much charitable activity to provide. Exploring this relationship is valuable not only because it addresses a key policy debate but also because it contributes to existing economic work on diversity and redistribution. Much prior work on socioeconomic conditions and tastes for redistribution relies on hypothetical questions asked to respondents in the General Social Survey or in Gallop Polls (examples include Alesina and La Ferrara 2005b; Luttmer 2001; Fong 2001). This article examines how a community's racial composition affects actual charitable expenditures. Studying actual charitable activities will be valuable if individuals' stated preferences are different from their actual preferences or if stated preferences are sensitive to the wording of a question (as is the case in Alesina and La Ferrara 2005b).
To investigate the effect of race on charitable church activity, we compare how racially homogeneous congregations respond to the community's racial composition relative to other congregations. Several factors support this approach. First, members of homogeneous congregations have ostensibly taken race into account when selecting a congregation to attend; consequently, these congregations may be especially likely to display sensitivity to a community's racial composition when providing charitable services. Second, comparing one type of congregation's response to race to the response of other congregations offers some of the empirical robustness of a standard difference-in-difference estimator. Our identification will come from comparing the difference between homogeneous congregations in, for example, highly black communities to homogeneous congregations in other communities and examining if this difference is similar to the one observed for nonhomogeneous congregations. This greatly decreases the likelihood that unobserved phenomena correlated with race will drive our results, although we will spend considerable time addressing that concern. Third, homogeneous congregations are important because they make up a large share (about 25%, in the nationally representative data used here) of all congregations.
Three different data sets of charitable church activity are used. (2) The first two data sets contain information on the churches belonging to two predominantly white Protestant denominations. Both denominations are among the largest in the country. One data set is a cross-section and the other is a panel. The third data set is the National Congregations Study (NCS), a multifaith, nationally representative survey of congregations taken in 1998.
The only finding consistent across all three data sets is that congregations composed entirely of white members become less charitably active, relative to other congregations, as the share of black residents in the local community grows. This difference in response to share black is economically significant. If the share of black residents in the local community increased by 10 percentage points, the average all-white church in Denomination 1 would be expected to decrease per-member spending on charitable activity by about 6 percent, while other congregations in Denomination 1 would on average decrease spending by less than 1 percent. Further, the results indicate that if the share black in the community increases by one standard deviation, the probability of a congregation participating in a local charitable activity would fall by about one-fifth if the congregation were all white and would slightly increase otherwise. We do not find consistent evidence of a unique response by all-white congregations to any racial group, except blacks.
Despite the difference-in-difference approach, one might be concerned that unobserved phenomena correlated with race drive the results. For instance, all-white churches in diverse neighborhoods may have poorer members or fewer resources than other churches. A number of tests suggest that the results are robust to this sort of concern. We find a differential response to black residents by all-white congregations only when examining charitable activities and not when examining other types of church activity. In fact, there is some evidence that all-white churches spend more on noncharitable activities as the share of black residents grows, consistent with a story where all-white churches in black communities are not poorer than other churches but substitute out of charitable spending and into other types of church spending. Additionally, the result is robust to using both monetary and nonmonetary measures of charitable activity, controlling for congregation members' income, using community fixed effects, using either the county or the census tract to represent the local community, and instrumenting for share black using lagged values of share black. Results from the NCS data set indicate that all-white congregations favorably disposed toward receiving government funding do not respond differently to black residents than do congregations that are not all white.
The remainder of the article is organized as follows. Section II sets out the empirical approach; Section III describes and presents findings using the denominational data sets; Section IV describes and presents findings using the NCS data set; and Section V concludes.
II. EMPIRICAL APPROACH
A growing body of economic research documents that in many situations, individuals display "group preferences," whereby one's tastes for public goods, collective activities, and redistribution are sensitive to the local community's demographic composition; examples include Glaeser (2005); Alesina and La Ferrara (2002, 2000); Vigdor (2004, 2001); Luttmer (2001); Alesina, Baqir, and Easterly (1999); and Poterba (1997). See Alesina and La Ferrara (2005a) for a review.
This research has suggested but not verified that congregation members may warrant consideration as individuals likely to display group preferences. Alesina, Baqir, and Hoxby (2000) find evidence suggesting that "observed differences in preferences among denominations may be largely the result of ethnic differences"; and Gruber (2005) uses a community's ethnic makeup to instrument for religiosity. Alesina and La Ferrera (2000) find that church participation is very sensitive to a community's racial composition. Luttmer (2001) postulates that religion may affect individuals' preferences for income redistribution.
However, some research suggests that churchgoers and other individuals may not discriminate based on race when interacting with others. Work by Levitt (2004) and Antonovics, Arcidiacono, and Walsh (2005) finds little evidence of race-based discrimination in a high-stakes game show. Research outside economics has used survey data to argue that religious participation decreases prejudice among Catholic churchgoers (Billiet 1995) and young Mormons (Jacobson 1998).
As mentioned earlier, work on tastes for redistribution and community socioeconomic conditions has relied to some extent on responses given in Gallop Polls and the General Social Survey. While these studies are valuable, they may not accurately reflect the actions taken by individuals if individuals give mendacious responses or if variations in the wording of a question generate different answers. The focus here is on the actual charitable spending rather than on stated preferences.
Other studies, including Alesina, Baqir, and Easterly (1999) and Poterba (1997), provide a link between ethnic or racial fragmentation and community outcomes such as per-capita citywide public debt. But as Vigdor (2004) argues, interpretation of these studies "is frequently hampered by the complexity of the aggregation mechanism linking individual preferences to [the] community-level outcomes." Here, the focus is not on "community-level outcomes" but on the actions taken by members of particular congregations.
To investigate the role of race in affecting charitable church activity, the empirical model compares how homogeneous congregations respond to race relative to other congregations when providing charitable services. With this in mind, let y representa measure of a congregation's charitable activity such as the log of congregational charitable expenditures or a dummy variable equaling unity if a congregation participates in a local charitable activity. Ideally, we would estimate:
(1) y = [alpha] [summation over j] [delta] [sub.j][race.sub.j] + [gamma]H + [summation over j] [[delta].sup.H.sub.j] (H x [race.sub.j]) + [beta]X + [epsilon],
where [race.sub.j is the share of local community residents who are race j, H is a dummy for whether a congregation is homogeneous (composed entirely of members of one race), X is a vector of congregational and community characteristics, and [epsilon] is exogenous noise. We could think of this as holding for congregation i in community k at time t; the subscripts have been suppressed. One might consider interacting all the covariates with the homogeneity dummy, not just the racial covariates; this specification will be considered. The coefficients [[delta].sup.H.sub.j] capture the difference between homogeneous congregations' response to racial groups in the community and other congregations' response.
A certain racial group must be omitted; this complicates the comparison of coefficients between homogeneous congregations and other congregations. For example, if we omitted Hispanics and then estimated Equation (1), the interacted coefficient on share black would be ([[delta].sup.H.sub.black] - [[delta].sup.H.sub.Hispanic]). It is impossible to say whether this coefficient is driven by a differential response of homogeneous congregations to blacks or to the omitted group. However, estimating Equation (1) with all but one racial group omitted does yield a meaningful comparison. For example, if we estimate Equation (1) including only the black racial group, the interacted coefficient is ([[delta].sup.H.sub.black] - [[delta].sup.H.sub.nonblack]), which is readily interpretable. We consequently estimate:
(2) y = [alpha] + [delta] [sub.j][race.sub.j] + [gamma]H + [[delta].sup.H.sub.j] (H x [race.sub.j]) + [beta]X + [epsilon],
for each racial group j.
While Specification (2) has the advantage of providing a meaningful interpretation of [[delta].sup.H.sub.j], one issue with this specification is that it assumes that a congregation responds in the same fashion to all the omitted groups. The results of this article are robust to specifications that allow heterogeneous responses to multiple groups at once. This is discussed more below.
Additionally, one might be concerned that the racial composition of the community is correlated with unobserved phenomena that simultaneously affect charitable church activity. As explained before, one advantage of Equation (2) is that this approach will be robust to unobserved phenomena that affect all congregations...
|
