Spiritual Struggles and Psychological Distress: Is There a Dark Side of Religion?

Readers may be surprised that our models do not include measure(s) of social integration or support. As noted in our description of the GSS data, since 1987 NORC General Social Survey has utilized a “split-ballot” interview design, in which (a) a relatively small core set of items are asked of all GSS respondents, and (b) the other items are asked of a randomly selected subset of respondents, usually roughly two-third of the total sample. In practice, this means that combining variables from several different “ballots” can result in quite small sample sizes, and indeed, this was the case when we included indicator(s) of social integration into the models presented in Table 3. However, in ancillary analyses (not shown, but available upon request), we did explore correlations between each of the spiritual struggle variables and a three-item index tapping the frequency with which respondents reported socializing with (a) neighbors, (b) friends, and (c) relatives. This indicator of secular social participation was virtually uncorrelated (i.e., r < .05) with each of the spiritual struggle variables, which suggests that it is unlikely to confound the associations between spiritual struggles and feelings of distress.

The issue of missing data deserves comment. Briefly most missing values occur on items tapping spiritual struggles or other aspects of religious involvement, or on the measure of psychological distress. These missing cases are handled via listwise deletion, which accounts for nearly all of the 20% case loss (298 of the initial 1,445). A smaller number (roughly 11%) of respondents failed to provide useable information on the family income item. To retain those cases in the analysis sample, we followed the longstanding practice advocated by Cohen and his associates (2002), substituting a fixed value (i.e., the valid sample mean on the variable) for the missing data, and then adding a dummy variable flag to identify those cases that were initially missing. Like the missing data flag for non-attendance at religious services, this dummy variable was never a significant predictor of psychological distress, and thus it was dropped from the final regression models. Extensive analyses were conducted to assess any potential biases associated with this approach and none were found.

Although our study does not focus on denominational differences in distress, spiritual struggles, or the relationships between these constructs, readers may be interested in the religious composition of the GSS (sub)sample. Based on the classificatory scheme proposed by Steensland and his associates (2000), our 1,147 respondents consisted of approximately 29% conservative (i.e., fundamentalist, evangelical, and charismatic) Protestants, 27% Catholics, 18% mainline (i.e., moderate and liberal) Protestants, 12% persons with no religious preference at all, 5% members of various other Christian groups (e.g., Mormon or LDS, Jehovah’s Witness, Mennonite or Amish), 4% adherents of various non-Christian traditions, and 5% persons who reported hard-to-classify or indeterminate religious or spiritual groups.