Beyond Clergy: Congregations’ Sponsorship of Social Services for People with Mental Disorders

Defining religious traditions is a difficult task due to the need to take into account denominations’ history, religious beliefs, and culture. For this paper, the state-of-the-art classification scheme devised by Steensland et al. (2000) is used to construct our hypotheses and the religious tradition dummy variables used in our regression analysis. Our analysis requires us to separate Protestant congregations into three categories: white, mainline Protestants; white, evangelical Protestants, and black Protestants. White, mainline Protestants are defined as having, “…accommodating stance toward modernity, a proactive view on issues of social and economic justice…” (Steensland et al. 2000, pp. 293–294). In contrast, white, evangelical Protestants are denominations that, “…have typically sought more separation from the broader culture, emphasized missionary activity and individual conversion, and taught strict adherence to particular religious doctrine (Steensland et al. 2000, p. 294).” Although black Protestants as a whole share many similarities with white, evangelical Protestants, black Protestant churches have emphasized different aspects of Christian doctrine including freedom and justice due to their marginalization in U.S. society (Steensland et al. 2000, p. 294). Despite using terms such as “white” and “black,” categorization within these schemes does not necessary indicate the racial composition of the congregations.

Drawing a nationally representative probability sample of congregations is not otherwise possible, because there is no complete list of U.S. congregations (Chaves 2004).

For the purposes of our study, social services sponsorship is defined as engaging in any type of activity intended to help people who are not members of their congregations and who have mental disorders.

None of these groups targeted people with mental disorders.

The “other” category is primarily composed of non-Christian religious groups (e.g., Jewish congregations). There are not enough of these congregations in the sample to constitute their own unique dichotomous variables in the regression analysis.

Details about how specific denominations are coded can be found in the “Appendix” of Steensland et al. (2000, pp. 314–316).

For the independent variables, a random multiple imputation technique is used. It retains cases otherwise lost and provides efficient statistical estimates. Multiple imputation involves imputing values for each missing value in a data set m times. In each of the m created data sets, the observed (non-missing) values are not transformed while the missing data are drawn randomly from the conditional joint distribution of the other variables in the data set. Each of the m data sets is analyzed separately. The results from these analyses are combined to account for within sample variation in the data and between sample variation created by multiple imputation. The random imputation process retains the original variability found in the data and thus reflects the fundamental uncertainty of our estimates, thus making it preferable to other techniques such as imputing the mean or using linear regression models. The number of imputed datasets is set at thirty to generate more efficient coefficient estimates (Bodner 2008).

Frenk and Trinitapoli’s (2012) finding is the only appropriate comparison, because no other study has examined congregations’ sponsorship of social services that target people with a specific medical condition.

The DSM-IV and all the key informants used the term “mental retardation.” However, the term “intellectual disability” is used in this paper, because it covers the same population and is now the accepted name for this disorder (Schalock et al. 2007, p. 116).

Due to the concern that the small number of congregations that have social services for people with mental disorders would produce biased estimates, the models reported in this paper were rerun using penalized likelihood (FIRTHLOGIT) to reduce small-sample bias in maximum likelihood estimation (King and Zeng 2001). The results are similar to those reported in this paper and available upon request.

The models were rerun with a modified dependent variable that did not include congregations that only sponsored social services for “other mental health issues.” The findings mirror those reported in this paper and are available upon request.

Using data from the National Comorbidity Survey-Replication, Kessler et al. (2005b, p. 620) found that the 12 month prevalence of any substance disorder was 3.8 percent. This is lower than the percentage of respondents with any type of anxiety disorder, mood disorder, or impulse control disorder (as well as several individual disorders such as major depressive disorder).

Research suggests that is due to the unwillingness of health insurers to cover treatment for substance disorders and the fragmented nature of systems of care for people living with those disorders (McCarty and Rieckmann 2010, p. 285).

The fact that congregations mention Alcoholic Anonymous is not surprising given prior research on congregational sponsorship of social services. As noted by Chaves (2004, p. 68), “…congregational social service activity is mainly done in collaboration with other organizations. Eighty-four percent of congregations that perform social services have at least one collaborator on at least one program. Seventy-two percent of all programs are carried out in collaboration with others.”

The NCS did not ask about clergy’s secular education.