Across many social contexts, women are found to be more religious than men. Risk preference theory proposes that women are less likely than men to accept the existential risks associated with nonbelief. Building on previous critiques of this theory, we argue that the idea of risk is relevant to understanding the relationship between gender and religiosity if risk is understood not as existential, but as social. The research on existential risk focuses on religious identification as solely a matter of belief; as part of the movement away from this cognitivist bias, we develop the concept of social risk to theorize the ways that social location and differential levels of power and privilege influence women’s nonreligious choices. We show that women’s nonreligious preferences in many ways mirror those of other marginalized groups, including nonwhites and the less educated. We argue that nonreligion is socially risky, that atheism is more socially risky than other forms of nonreligion, and that women and members of other marginalized groups avoid the most socially risky forms of nonreligion.
Bibliographical noteFunding Information:
All data are drawn from the Boundaries in the American Mosaic (BAM) survey, a nationally representative online survey contracted through the survey company Gesellschaft für Konsumforschung (GfK) with funding from the National Science Foundation (Croll, Tranby, Edgell and Hartmann 2014). The survey was drafted during the summer of 2013 and fielded during a two-week period in early 2014. Participants were selected from GfK’s nationally representative Knowledge Networks panel sampling frame, which uses probability-based random address sampling from U.S. Postal Service records to recruit respondents in English-and Spanish-speaking households through direct mail, telephone follow-up, and online registration. GfK provides laptop computers for respondent households lacking Internet access. Using probability proportional to size weighted sampling, GfK drew a sample consisting of 2,400 adults, including two oversamples of 400 black and 400 Hispanic respondents, from their panel. Sample weights were applied to the data to adjust for these ethnic and language-based oversamples, in addition to poststratification sample weights for noncoverage and nonresponse biases similar to those used in comparable national surveys such as the General Social Survey (GSS) and the Current Population Survey (CPS) (see Chang and Krosnick 2002; Huggins and Eyerman 2001 for more information on Internet survey reliability) While respondents were paid for their participation, GfK caps participation in multiple surveys at two to four per month, does not allow for self-selection into either the frame or survey sample, and corrects for frequent sampling of respondents on particular demographic characteristics in their sampling procedure. Before fielding, the BAM survey underwent two rounds of pretesting and revisions with a convenience sample of undergraduates at an East Coast university and a random sample of 50 respondents from GfK’s panel.
The author(s) disclosed receipt of the following financial support for the research, authorship, and/ or publication of this article: The authors appreciate the generous support for data collection and research assistance given by the National Science Foundation (Grants 1258926 and 1258933) and the Edelstein Family Foundation.