Examining Gender Differences in Nonreligion

SCUXXX10.1177/2329496516686619Social CurrentsEdgell et al. research-article6866192017

Article

Social Currents 19–1 From Existential to Social © The Southern Sociological Society 2017 Reprints and permissions: sagepub.com/journalsPermissions.nav Understandings of Risk: DOI:https://doi.org/10.1177/2329496516686619 10.1177/2329496516686619 Examining Gender journals.sagepub.com/home/scu Differences in Nonreligion

Penny Edgell1, Jacqui Frost1, and Evan Stewart1

Abstract 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.

Keywords nonreligion, atheism, risk, gender

Introduction underlying assumption that women are, in general, more risk averse than men. In contemporary Western societies, women We argue that the concept of risk is useful in are generally found to be more religious than explaining women’s generally greater religios- men; explanations for this pattern include ity, but that the risk is social, not existential. gender differences in socialization, social Building on recent research regarding how roles and expectations, and biology. One of social location influences religious and nonre- the most influential explanations, risk prefer- ligious choices (Baker and Smith 2015; ence theory, argues that women are more Schnabel 2016; Ellison and Sherkat 1995) and likely to avoid the existential risk of rejecting how women are differentially sanctioned for belief in a God who may turn out to be real (e.g., Miller and Hoffmann 1995). As we 1University of Minnesota, Minneapolis, USA describe below, risk preference theory has generated an extensive body of scholarship, Corresponding Author: Penny Edgell, University of Minnesota, 1039 Social much of which modifies or extends the origi- Sciences Building, 267 19th Avenue South, Minneapolis nal formulation by Miller and Hoffmann MN 55455, USA. (1995), and some of which contests the Email: [email protected] Social Currents 2 nonreligious identification (Baker and Smith (Baker and Smith 2015; Chaves 2010; Edgell 2015; Schutz and Roth 2015), we argue that 2012; Edgell and Tranby 2007; Frost and women are more likely than men to incur Edgell 2017). Likewise, when gender is con- social costs for embracing nonreligious beliefs, ceptualized as a contextually embedded set of identities, and practices, and that this influ- practices, a more complex understanding of ences not only the choice to be nonreligious what gender means and how it intersects with but also which nonreligious identities and religiosity becomes possible (Cornwall 2009; practices women adopt. Rather than being Neitz 2014; Sullins 2006). more risk averse, women face a greater risk We approach religiosity and gender as con- than men for similarly socially stigmatized textually embedded practices and argue that, in choices, and we develop the concept of social the contemporary United States, being nonreli- risk to account for this. gious is a more socially risky practice for We draw on recent, nationally representa- women, who are more likely to be actively tive survey data to analyze how gender influ- sanctioned for rejecting religion. Although ences the propensity to be nonreligious, using social risk and stigma have been taken into measures of nonreligion that include nonin- account in research on the religiosity of other volvement in religious institutions, nonbelief, marginalized populations (e.g., Ellison and and a variety of nonreligious identifications. Sherkat 1995), they have been largely absent We analyze how gender shapes the forms of in the literature on women’s religiosity. We nonreligious identification that individuals argue that the concept of social risk enables a embrace, comparing those who identify as practice-oriented approach to nonreligious atheist with those who claim less overtly reli- choices by highlighting nonreligious identifi- gion-rejecting identities (e.g., “nothing in par- cation as a process of claiming a social identity ticular” [NIP] and “spiritual but not religious” and creating symbolic boundaries around that [SBNR]). This allows us to explore how gen- identity (e.g., Lamont 1992), allowing us to der shapes the particular way in which nonreli- examine how power and privilege shape religion is practiced and expressed, with a focus gious and nonreligious identification. on comparing more and less socially risky choices. Our research contributes to the cri- tique and reformulation of understandings of Nonreligion as Existentially gender, risk, and religiosity by reconceptualiz- Risky: Differential Risk ing the nature of the risks involved in eschew- Preference and the Gender ing religious involvement (Avishai, Jafar, and Gap in Religiosity Rinaldo 2015; Baker and Smith 2015; Cornwall 2009). Gender differences in religiosity are pervasive We also answer recent calls for research across Western societies in the modern era that goes beyond a belief-centered approach to (Carroll 2004; Voas, McAndrew, and Storm religiosity. Both the risk preference explana- 2013), and numerous explanations for this gap tion and many of its critics rely on the same have been offered, including women’s roles in asocial and belief-centered understanding of Western societies as caretakers and homemak- the risk of being nonreligious as existential; ers (Becker and Hofmeister 2001; Hastings unbelief is understood to trigger the same and Lindsay 2013); the socialization of women impersonal cognitive mechanism across con- into being emotional, caring, and thus religious texts. But we agree with those who argue that (Collett and Lizardo 2009; Devine 2013); dif- both religion and nonreligion are best under- ferences in sex-type personalities (Sherkat stood as a set of practices that are contextually 2002; Thompson 1991); and biological innate- embedded in institutions and relations of ness (Bradshaw and Ellison 2009; Stark 2002). power, intersecting with other aspects of iden- One of the most prominent explanations for tity in contingent ways that shape the relevance the gender gap in religiosity is the argument and impact of religious beliefs for individuals for differential risk preferences (Miller 2000; Edgell et al. 3

Miller and Hoffmann 1995; Miller and Stark original study, Freese (2004) found no support 2002). Building on Pascal’s Wager, Miller and for the claim that risk preferences drive gen- Hoffmann (1995) argued that being nonreli- der differences in religion (see also Freese and gious is an existentially risky choice; if one Montgomery 2007). denies belief in a God that does in fact exist, Critiques have also focused on refuting one risks eternal damnation. They found that assumptions of the universality of gender dif- women are more averse to risk taking, adven- ferences in religiosity. Sherkat (2002) found ture seeking, and danger, and they concluded that gay men are more religious than lesbian that women’s propensity to be risk averse is the women, arguing that it is individuals with main explanatory factor for their higher religi- “feminine” personality traits that are more risk osity. Miller (2000) subsequently argued that averse and more religious. Carroll (2004) the risk of nonreligion varies across cultures, detailed an historical account of the “feminiza- arguing that women in less religious cultures tion of religion” explaining that before the see less of a risk to nonreligion. Finally, Miller nineteenth century Protestantism and and Stark (2002) attempted to explain why Catholicism in the United States were prac- women are more risk averse than men. Drawing ticed more by men than by women. Similarly, on Stark’s (2002) work, Miller and Stark Sullins (2006) found that among Jews and argued that risk aversion is a biologically Muslims, men are in fact more religious than “feminine” trait and, as nonreligion is existen- are women. Baker and Smith (2015) found that tially risky, the source of women’s greater reli- the gender gap disappears under certain condi- giosity is biological. tions of political ideology and educational attainment such that highly educated, liberal Extensions and Critiques women match educated liberal men in their of the Risk Preference propensity to be nonreligious. Explanation However, the majority of extensions and critiques of risk preference theory retain the Important extensions and critiques of the dif- core idea of nonreligion as potentially risky for ferential risk preference explanation have existential reasons. That is, they focus on reli- called into question the validity and universal- gious belief as the essential, defining element ity of its claims (Carroll 2004; Collett and of religiosity and they set out to explain gender Lizardo 2009; Cornwall 2009; Hoffmann differences in theism. This belief-centered 2009; Roth and Kroll 2007; Schnabel 2015; understanding of religiosity assumes that reli- Schutz and Roth 2015; Sherkat 2002, 2014; gious beliefs are the main driver of religiosity Sullins 2006). For example, Collett and and that they have a uniform effect on social Lizardo (2009) drew on power-control theory, action (Chaves 2010). In this way, risk prefer- finding that women raised in traditional, patri- ence explanations elide the intersectionality of archal households are more risk averse than religiosity with other aspects of social identity women raised in egalitarian households; they and the ways in which religiosity is constructed argued that socialization, not biology, is the through practice (McGuire 2008; Riesebrodt root of gender-differential risk preferences. 2009), contextually embedded in institutions Roth and Kroll (2007) argued that many non- (Edgell 2012), and influenced by structural religious individuals do not believe in eternal location (Edgell and Tranby 2007; Frost and damnation, and thus perceive no risk to reject- Edgell 2017). A practice-oriented approach ing religious beliefs. Their analysis reveals takes into account social embeddedness and that a belief in life after death affects men’s understands both gender and religion as things religiosity more so than it does women’s and individuals “do,” not things that individuals that women who do not believe in heaven or “are” (Cornwall 2009; Neitz 2014; Sullins hell participate in religious institutions more 2006). As Cornwall (2009) explained, religion than men who do believe. Furthermore, in a and risk preference are gendered constructs, replication of Miller and Hoffmann’s (1995) and research should focus on “gendering Social Currents 4 processes” that account for differences in both religious and nonreligious identities, but their risk preference and religiosity. rejection of religion means they are rightfully grouped with the nonreligious. Lee (2014) Nonreligious Options and argued that individuals might choose this label because they are indifferent to religion and Choices nonreligion, or they may want to signal their In the contemporary United States, the increase disaffiliation from any engagement and “locate in nonreligious persons (Hout and Fischer 2014) themselves outside of religious culture in has led to a variety of groups, movements, and general” (p. 474). modes of identification among the nonreligious (Baker and Smith 2015; Beaman and Tomlins Rethinking Risk: Nonreligion 2015; Cimino and Smith 2014). Individuals as Socially Risky may practice nonreligion in a variety of ways or even mix elements of religious and nonreligious In the United States, religiosity is intertwined practice (e.g., Keysar 2014). Thus, the category with both national identity and civic virtue; of “nonreligion” is increasingly understood as religious belonging and practices such as an indicator of highly heterogeneous beliefs, prayer and church attendance enable religious identifications, and behaviors (Cotter 2015; Lee individuals to define themselves as moral peo- 2015). Atheists have been found to be the most ple, good neighbors, and good citizens (see vocal and active in regard to criticizing religion Edgell et al. 2016). The fact that religion is in the public sphere and promoting nonreligious voluntary and chosen makes it particularly identity politics (Baker and Smith 2009; useful for boundary maintenance, and it is LeDrew 2015); they are more often male, white, often understood as an indicator of moral highly educated, and politically liberal (Sherkat worth and the acceptance of culturally valued 2008). Agnostics are a less studied group; some practices; rejection of such practices makes use the agnostic label to indicate an openness to one a “cultural outsider,” and cultural outsider phenomenon other than what they can see or status may persist despite increasing visibility detect, while others use the label to indicate a of or contact with the outsider group (Kalkan, certain apathy toward both religion and nonreli- Layman, and Uslaner 2009). As a result, athe- gion (Cotter 2015). ists in America are disliked and distrusted Another prominent subgroup within the non- (Cook, Cohen, and Solomon 2015; Gervais, religious is the “spiritual but not religious" Shariff, and Norenzayan 2011), seen as (SBNR) Those who take on this label often immoral and/or elitist (Edgell et al. 2016; maintain religious beliefs and behaviors, includ- Wright and Nichols 2014), and are discrimi- ing attending church and belief in the supernatu- nated against in the workplace, the military, ral; the choice to distinguish their identity from and even by their friends and family (Cragun religion is typically a moral and political dis- et al. 2012; Hammer, Cragun, and Hwang tancing from organized religion (Ammerman 2013; Hammer et al. 2012; Wallace, Wright, 2013). SBNR individuals are more often and Hyde 2014; B. Wright et al. 2013). In women, older, and nonwhite (Aune 2015; short, nonreligion is stigmatized in American Brown, Taylor, and Chatters 2015; Wilcox society, especially atheism, and in this sense, 2009). While at first categorizing SBNR as rejecting belief in a God in America is not sim- “nonreligious” seems counterintuitive, and in ply existentially risky, it also entails strong fact many surveys categorize SBNR as an social risks. “other” religious group, it is an intentional dis- tancing from organized religion and should be The Causes and Consequences of categorized as a nonreligious choice. Stigma Perhaps the most understudied subgroup among the nonreligious is the “nothing in par- Stigmatization is the process whereby an ticulars” (NIP). This group eschews both attribute or trait is believed to discredit or Edgell et al. 5 taint an individual or group, resulting in the Stigmatized individuals use various strate- devaluing of their status in the eyes of others gies to cope with, lessen, and avoid stigma and (Goffman 1963; Major and O’Brien 2005). its consequences. Some may confront the Stigma is the co-occurrence of “labeling, ste- stigma head on by affirming more strongly reotyping, separation, status loss, and dis- their association with the stigmatized trait or crimination” in which a trait or characteristic group and wearing that trait proudly is first labeled and set apart as different and (Branscombe et al. 1999; Ellemers, Spears, inferior and then applied to specific individu- and Doosje 2002). Alternatively, fear of stigma als or groups (Link and Phelan 2001:363). may cause individuals to “pass” by hiding their Members of a society come to develop a stigmatized trait or to disengage with that iden- shared stereotypical understanding of a cer- tity or trait altogether when it is an option tain trait, and those with that trait soon come (Ellemers et al. 2002; Goffman 1963). Whether to understand their devalued status based on an individual confronts or avoids stigma is prior experiences and exposure to dominant largely dependent on their social status as well culture (Major and O’Brien 2005). Individuals as their commitment to that group or trait. who are perceived to have a stigmatized trait High-status or high-commitment individuals experience a variety of consequences, includ- may more strongly affirm a stigmatized identity, whereas low-status and low-commitment ing status loss, discrimination, and stress individuals rework their identity or conceal it (Link and Phelan 2001; Major and O’Brien (Ellemers et al. 2002). 2005; C. Miller and Kaiser 2001). As a result, These various responses to stigma have also stigma has been linked to “poor mental health, been found among the nonreligious. Some physical illness, academic underachievement, nonreligious individuals, particularly atheists, infant mortality, low social status, poverty, redefine and reassert their nonreligious identi- and reduced access to housing, education, and ties by claiming their beliefs are “more moral” jobs” (Major and O’Brien 2005:394). than those of the religious; they use humor and Importantly, the stigmatizing process is derision of the religious to affirm their nonreli- largely dependent on power differentials gious identities as superior (Baker and Smith among different social groups in a society. 2015; Guenther 2014). More often, however, Link and Phelan (2001) argued, the nonreligious “pass” and hide their nonreligious beliefs and identities in the presence of Stigmatization is entirely contingent on access to others (Cragun et al. 2012; Manning 2015; social, economic, and political power that allows the identification of differentness, the Schutz and Roth 2015; Zuckerman 2011). construction of stereotypes, the separation of labeled persons into distinct categories, and the The Gendered Nature of full execution of disapproval, rejection, Nonreligious Social Risk exclusion, and discrimination. (P. 367) Feminist scholars have begun to outline the This means that not only do groups with more ways in which women are more likely than power and status have more of a say in how a men to bear the social costs of stigmatized stigmatizing label gets applied, but also that nonreligious choices (Hutchinson 2011; Miller those with power have more resources for cop- 2013). By rejecting religion, women are seen ing with stigma when it is applied to them as rejecting their femininity and their rightful (Link and Phelan 2001; Major and O’Brien place in society, including their moral obliga- 2005). Women and minorities, for example, tions to be the “keepers of morality” by super- are less likely to have the power and resources vising the moral upbringing of children and to avoid or combat stigma (Link and Phelan fulfilling the emotional needs of the family (cf. 2001). Thus, stigma is not experienced by all Sherkat 2000). Atheism, then, is socially risky individuals of a stigmatized group in the same for women because it violates gendered expec- way. tations, while for men such choices are less Social Currents 6 risky because dominant understandings of alternatives” when they are available (see C. masculinity are not intertwined with nurturing Miller and Kaiser 2001 for a review). In an and religiosity. As Ashley Miller (2013) exploratory qualitative analysis detailing how explained about atheism, “[M]en do not face this process works for nonreligious women, the risk of seeming less masculine by embrac- Schutz and Roth (2015) found that nonreli- ing what is portrayed as an emotionless, scien- gious women experience higher rates of dis- tific, masculine belief” (p. 217). To be a crimination than do nonreligious men, and that nonreligious woman in America is to “unsex” the sanctions they face are more serious. For oneself (Hutchinson 2011), a risk that is par- example, they found that more women than ticularly acute for black women, given the his- men report public harassment, discrimination toric centrality of religion in the black in the workplace, and ostracism from previ- community (cf. Baker and Smith 2015; Ellison ously close friends and family members as a and Sherkat 1995; Hutchinson 2011; Lincoln result of revealing their nonreligion. and Mamiya 1990). Baker and Smith (2015) Nonreligious women are also more likely to argued that this explains why atheists are more engage in various stigma management prac- likely to be privileged (white, male, and with a tices by “pretending” to be religious, hiding college degree); privileged individuals have their nonreligion by “passing,” or taking on the resources to avoid or combat the stigma “softer” and more ambiguous nonreligious associated with nonreligion (Link and Phelan labels. 2001). Taken together, this research points to the Nonreligious women violate gendered utility of investigating not only gender differ- expectations, and women are more likely to be ences in religion but also gender differences in publicly criticized and sanctioned for noncon- embracing an array of nonreligious identities forming behaviors of all kinds. For example, a and practices as a way to negotiate the social recent study finds that students evaluate female risks that nonreligious women face. It is of professors more harshly and critically than course the case that not believing in a God may male professors; the looks and emotional entail existential risks, especially in certain expressions of female professors receive nega- social contexts. But, along with others (Baker tive comments while students focus on and and Smith 2015; Hutchinson 2011; Schutz and evaluate more positively the intellectual traits Roth 2015), we argue that in the contemporary of male professors (MacNell, Driscoll, and United States, nonreligion is socially risky Hunt 2015). Women are sanctioned more heav- because the rejection of religion is socially ily than are men for eschewing altruistic stigmatized. This stigma may be especially behaviors (Heilman and Chen 2005), for strong for atheism, the nonreligious identity choosing childlessness over parenthood that most overtly and explicitly rejects reli- (Gillespie 2000; Letherby 2002), and for tak- gion. An intersectional approach helps us ing a dominant leadership role in the work- understand why this risk may be more acute place (Brescoll and Uhlmann 2008; Rudman for women, for whom religion is congruent et al. 2012). with gendered expectations of investment in Thus, women may be judged more harshly mothering, nurturing, and moral socialization. than men for rejecting religion and may be As a result, women may be judged more more likely to face sanctions that flow from harshly than men when they reject religion, these negative judgments, just as they are in and in the American context, those doing the general more likely to be stigmatized and have judging may be more willing to directly sanc- fewer resources to combat stigmatization tion or discriminate against nonreligious (Link and Phelan 2001). Social psychologists women than they would nonreligious men. have found that women who expect that they Building on prior work, we propose that the will be stereotyped and stigmatized are more concept of social risk offers a constructive likely than men to avoid stigma-tainted situa- reframing of the gender and risk literature. If tions altogether or to seek out “prejudice-free existential risk involves the anticipated loss of Edgell et al. 7 a valued existential good (salvation), social representative online survey contracted risk involves the anticipated loss of valued through the survey company Gesellschaft für social goods; that is, social risk is driven by the Konsumforschung (GfK) with funding from knowledge of social costs associated with stig- the National Science Foundation (Croll, matized choices. Based on prior research, non- Tranby, Edgell and Hartmann 2014). The sur- religious individuals may reasonably anticipate vey was drafted during the summer of 2013 the loss of three specific social goods due to and fielded during a two-week period in early the stigmatization of nonreligion in the U.S. 2014. Participants were selected from GfK’s context: reputation or social status, valued nationally representative Knowledge Networks relationships with others, and access to infor- panel sampling frame, which uses probability- mation or opportunities (discrimination). In all based random address sampling from U.S. cases, the mechanism leading to loss is also Postal Service records to recruit respondents in social; reputational loss, relationship damage, English- and Spanish-speaking households and discrimination occur as individuals apply through direct mail, telephone follow-up, and sanctions for nonconforming behavior during online registration. GfK provides laptop com- social interactions. puters for respondent households lacking We argue that nonreligious women face a Internet access. Using probability proportional greater social risk than nonreligious men in the to size weighted sampling, GfK drew a sample contemporary United States, and the social consisting of 2,400 adults, including two risk of taking on a nonreligious identity may oversamples of 400 black and 400 Hispanic affect nonreligious choices via two pathways. respondents, from their panel. Sample weights First, the anticipation of social risk may shape were applied to the data to adjust for these eth- an individual’s initial choice in how to identify nic and language-based oversamples, in addi- as nonreligious, with less powerful individuals tion to poststratification sample weights for choosing less risky nonreligious identities and noncoverage and nonresponse biases similar to more powerful ones embracing more risky those used in comparable national surveys identities (in this case, atheism). Second, the such as the General Social Survey (GSS) and result of nonreligious choices may trigger the Current Population Survey (CPS) (see social sanctions which either cause a person to Chang and Krosnick 2002; Huggins and switch to a less risky religious identity or may Eyerman 2001 for more information on lead to a more committed affirmation of a par- Internet survey reliability) ticular nonreligious identity; again, power and While respondents were paid for their par- resources will shape this choice. Either path- ticipation, GfK caps participation in multiple way would lead to women claiming less risky surveys at two to four per month, does not nonreligious identities than men. allow for self-selection into either the frame or In our analysis, below, we show that women survey sample, and corrects for frequent sam- and men sort into less and more risky nonreli- pling of respondents on particular demographic gious choices, respectively, and that women characteristics in their sampling procedure. report more discrimination for nonreligious Before fielding, the BAM survey underwent choices than do men. Our findings highlight two rounds of pretesting and revisions with a the importance of considering the multifaceted convenience sample of undergraduates at an nature of nonreligion and the ways it is shaped East Coast university and a random sample of by social location and cultural context. 50 respondents from GfK’s panel.

Data and Methods Methods While risk preference theory examines gender Data differences in religiosity, we conceive of nonre- All data are drawn from the Boundaries in the ligion as socially risky and assess the effects of American Mosaic (BAM) survey, a nationally gender on the propensity to practice nonreligion Social Currents 8 in a variety of ways (cf. Keysar 2014). discrimination. Our control variables include Specifically, we use a series of binomial logistic dichotomous measures of nonwhite racial regression models focusing on nonreligious identification (black or Hispanic), political belief (rejecting theism), nonreligious behavior conservatism, whether the respondent is a par- (not attending worship services), and nonreli- ent, and whether she or he completed college. gious identification (e.g., identifying as “agnos- We also include a continuous measure of age tic” or “atheist”). To test nonbelief, we used a and a categorical measure of income. A full list forced choice dichotomous question asking of variables is presented in Table 1. whether respondents “believe in a God or a Agnostics make up our smallest subsample, higher power” with participants who responded with only 75 individuals reporting identifica- “no” coded as 1.1 To test nonattendance, we tion with the agnostic label. Following Hosmer recoded a categorical measure assessing the fre- and Lemeshow (2013), who recommended quency of church attendance into a dichotomous including only one independent variable for variable with respondents answering “never” every eight to 10 cases of a dependent variable coded as 1. Finally, to test nonreligious identifi- in logistic regression, we limit our suite of cation, we recoded a categorical measure of reli- independent variables to eight measures to gious identification from the BAM survey based maintain statistical power, and we hold this on a modified version of the Steensland et al. block constant across all the models for ease of (2000) approach which asked respondents, interpretation. “What is your current religious preference, if We report nine binomial logistic regres- any?” Respondents could choose only one label sions. Seven of our nine reported models use from a list of religious and nonreligious identifi- the full sample of the BAM survey to test the cations, and our coding reflects their choice of probability of nonbelief, nonattendance, and one of four possible nonreligious identifications nonreligious identification against believing, from the list: “atheist,” “agnostic,” “SBNR,” and attending, and identifying baselines. In our “NIP.” In our sample, 33 percent of men and 28 final two models, we compare self-identified percent of women claim one of these four forms atheists with a subsample baseline of nonbe- of nonreligious identification. lievers (N = 275) to assess who, among those Our main independent variables of interest who have already taken on Pascal’s Wager and are gender and religious/nonreligious discrimi- denied belief in a God, is more or less likely to nation. To measure the social risk of express- identify as atheist. These subsample models ing nonreligious identities, we use a variable bring additional nuance to our analysis because indicating whether respondents have ever they allow us to test not only whether gender experienced discrimination due to their nonre- associates with nonbelief, but also how that ligious beliefs and identification.2 Of course, nonbelief is represented. discrimination is just one of the many social costs one can incur for taking on a stigmatized Results identity. However, in this case, using a measure of discrimination is a good test of our A social risk perspective on nonreligious iden- theoretical approach because it is in many tification suggests that (1) nonreligious respon- ways the most serious risk associated with tak- dents experience different kinds of social ing on a stigmatized identity and this measure sanctions for nonaffiliation and (2) they captures an experience or perception that is respond to those risks with different strategies. notable and memorable. This is useful because To demonstrate these claims, we first ran a the perception of discrimination motivates series of bivariate tests for gender differences action to avoid future discrimination. To test in nonreligious identification and experiences whether women experience the social risk of of discrimination based on those identifica- discrimination differently than men, we also tions. Table 2 presents a descriptive picture of include an interaction term in our models for the gender breakdown across different kinds of gender and experiences with nonreligious religious and nonreligious identification, as Edgell et al. 9

Table 1. Variable Descriptions.

Mean, Standard deviation, Variables Observations proportion standard error Minimum Maximum Dependent variables Nonbelieving 2,465 0.13 0.34 0 1 Nonattending 2,479 0.28 0.45 0 1 Nonreligious identification 2,471 Atheist 80 0.03 0.01 SBNR 191 0.08 0.01 Agnostic 75 0.03 0.01 NIP 406 0.16 0.01 Independent variables Female 2,521 0.52 0.50 0 1 Age 2,521 46.97 17.02 18 94 Nonwhite racial identity 2,521 0.33 0.47 0 1 Completed college 2,521 0.29 0.45 0 1 Income 2,521 11.87 4.50 1 19 Respondent is a parent 2,485 0.64 0.48 0 1 Political conservatism 2,466 4.14 1.52 1 7 Experienced religious 2,454 0.18 0.38 0 1 discrimination

Source. Boundaries in the American Mosaic Survey (2014). Note. Descriptive statistics do not include poststratification weights. SBNR = spiritual but not religious; NIP = nothing in particular.

Table 2. Bivariate Analyses of Nonreligious Identification and Experiences of Discrimination by Gender.

Gender Experienced discrimination

Religious and Men Women Nonreligious Identification Men Women Significant No Yes No Yes Significant Any religious affiliation 66.91% 72.19% ** 698 120 736 156 85.33% 14.67% 82.51% 17.49% SBNR 5.54% 9.89% *** 60 7 93 29 * 89.55% 10.45% 76.23% 23.77% Atheist 3.42% 2.65% *** 37 20 18 5 64.91% 35.09% 78.26% 21.74% Agnostic 4.65% 1.85% 33 9 24 9 78.57% 21.43% 72.73% 27.27% NIP 19.48% 13.42% *** 209 26 142 23 88.94% 11.06% 86.06% 13.94%

Source. Boundaries in the American Mosaic Survey (2014). Note. Chi-square tests of significant differences across gender. Descriptive statistics do not include poststratification weights. SBNR = spiritual but not religious; NIP = nothing in particular. *p < .05. **p < .01. ***p < .001 (one-tailed tests). well as the percentages of each group who respondents identify as NIP, atheist, or gener- report ever experiencing discrimination by ally nonreligious, a higher percentage of gender. While a higher percentage of male women identify as SBNR. A chi-square test Social Currents 10

Table 3. Odds of Nonreligious Beliefs and Nonreligious Behavior—Logistic Regressions: Full Sample.

Does not attend religious Does not believe in a god services

Variables Odds ratio Standard error Odds ratio Standard error Female 0.52*** 0.10 0.76* 0.09 Age 0.97*** 0.01 1.00 0.00 Nonwhite 0.55** 0.12 0.49*** 0.07 College 1.64** 0.29 0.83 0.11 Income 1.04 0.02 1.00 0.01 Parent 0.56** 0.10 0.67** 0.09 Conservative 0.66*** 0.04 0.73*** 0.03 Religious discrimination 1.44 0.35 0.84 0.14 Female × Religious Discrimination 0.99 .20 1.10 0.18 Constant 2.46* 1.07 3.07*** 0.98 N 2,381 2,399 McFadden R2 .149 .058 Wald χ2(13) 113.05*** 87.22*** BIC 1,654.24 2,740.10

Source. Boundaries in the American Mosaic Survey (2014). Note. Models include poststratification weights. BIC = Bayesian information criterion. *p < .05. **p < .01. ***p < .001 (one-tailed tests).

indicates that this difference is statistically sig- do not believe in a god or universal spirit and nificant (p < .001). Among women who claim the odds of respondents reporting that they no religious identity, 19 percent reported hav- never attend religious services. As indicated by ing experienced discrimination, whereas 15 the odds ratios, women are about half as likely percent of nonreligious men experienced dis- to say they do not believe in a god as are men, crimination, though these differences were not and they are about 24 percent less likely to statistically significant. However, among report never attending religious services (odds SBNR respondents, women report signifi- ratio of 0.76). Nonwhite respondents (those cantly higher rates of discrimination than men who self-identify as black or Hispanic) are (p < .05). While agnostic and NIP females are about half as likely as white respondents to more likely to report discrimination than their report each of these outcomes as well. Parents male counterparts, these differences were not and politically conservative respondents are statistically significant. Atheist men reported also less likely to report these outcomes, while experiencing discrimination at higher rates respondents who completed college are 1.64 than atheist women (35 vs. 22 percent), but the times more likely to claim nonbelief in this limited number of atheists in the sample does model. not render this difference statistically signifi- Table 4 tests the odds of respondents report- cant; neither chi-square nor Fisher’s exact test ing a range of nonreligious identities, with was significant for this group. respondents who do report a religious identity Building on this bivariate analysis, we serving as the reference category. The first col- report logistic regression models assessing umn treats all nonreligious respondents as the how these experiences with discrimination outcome variable, while each subsequent col- relate to various kinds of nonreligious expres- umn treats only a specific nonreligious identifi- sion when controlling for key demographic cation as the outcome. The two most important and ideological factors. Table 3 presents results findings in this table stem from the varied for the odds of respondents reporting that they effects of gender in each of these models. First,

0.12 0.01 0.12 0.62 0.04 0.29 0.09 0.90 0.35 1.48 error Standard Atheist .21 ratio 0.30** 0.96*** 0.25** 1.97* 1.03 0.79 0.60** 2.48* 0.75 1.63 2,405 Odds 60.7*** 694.6 0.33 0.01 0.14 0.60 0.04 0.22 0.06 0.38 0.42 0.16 error Standard Agnostic .10 1.05 1.00 0.36** 1.90* 1.02 0.71 0.63*** 0.90 1.37 0.22* 68.39*** 2,405 737.5 Odds ratio 0.10 0.00 0.16 0.13 0.02 0.13 0.04 0.17 0.22 0.85 error Standard particular Nothing in .05 ratio 2,405 Odds 0.69** 0.98*** 1.01 0.69* 0.95** 0.80 0.90* 0.66 1.04 2.44* 2,224.1 81.12*** 0.44 0.01 0.27 0.26 0.02 0.17 0.05 0.26 0.31 0.06 error Standard religious .06 ratio Spiritual but not 2,405 Odds 2.06*** 1.01* 1.27 1.26 0.97 0.78 0.75*** 0.77 1.51* 0.11*** 1,237.6 65.01*** 0.10 0.00 0.10 0.15 0.01 0.09 0.03 0.18 0.17 2.84 error Standard .08 0.81 0.98*** 0.76* 1.09 0.96** 0.71** 0.72*** 0.92 1.13 9.13*** All nonreligious 2,405 2,871.69 123.97*** Odds ratio 2 Odds of Nonreligious Identification—Logistic Regressions: Full Sample. (13) 2 Wald χ Female Age Nonwhite College Income Parent Conservative Religious discrimination Female × Religious Discrimination Constant N McFadden R BIC Variables Source. Boundaries in the American Mosaic Survey (2014). Note. Models include poststratification weights. BIC = Bayesian information criterion. * p < .05. ** .01. *** .001 (one-tailed tests). Table 4.

11 Social Currents 12 there is no statistically significant difference in twice as likely to identify as atheist or agnos- the odds of female respondents identifying as tic. Income and age are significant in some of nonreligious in general. Instead, female respon- the models presented, but their effects are not dents are 31 percent less likely to identify as substantively large.3 “NIP” than male respondents, and they are 70 Table 5 presents the results from logistic percent less likely to identify as atheists. regressions for a subsample of only respon- However, female respondents are twice as dents who say they do not believe in a god or a likely to identify as SBNR as are men. universal spirit (N = 275). With this small sub- The second important gender finding relates sample of respondents, we do not mean to to experiences with discrimination. A one stan- draw highly generalizable conclusions; rather, dard deviation increase in reported frequency we add context to extend our understanding of of nonreligious discrimination significantly gendered expressions of nonreligiosity. We ran associates with about 2.5 times higher odds of this model twice, once with the full model and identifying as an atheist, while the same again without the gender and religious dis- increase in our interaction term for gender and crimination interaction term for a slight experiences of discrimination associates with improvement in model fit. In the full model, 1.5 times higher odds of identifying as SBNR. only race remained significant and substantive, These are the only two times in the full sample suggesting that nonwhite respondents who do models where experiences with discrimination not believe in a god are about 66 percent less are statistically significant. Together, with the likely to identify as an atheist than white findings above, we can conclude not only that respondents who said they do not believe in a women are more likely than men to identify as god. SBNR, but also that women who report fre- In the reduced model, gender becomes sig- quent religious discrimination are more likely nificant at a similar magnitude, suggesting that to do so. At the same time, men are more likely women who do not believe in a god or univer- to identify as atheists, and men who report sal spirit are still less likely to identify as athe- more experiences with religious discrimina- ists. As our data are ordinal, we used the tion are more likely to do so. This supports the Karlson, Holm, and Breen method for media- idea that for women and men, the choice to tion analysis (KHB mediation) to assess embrace a specific religious identity (SBNR whether gendered experiences with discrimi- for women, atheist for men) may be part of a nation mediate the effect of gender on atheist gendered strategy for dealing with experiences identification (Breen, Karlson, and Holm of discrimination. 2013). The test identified a significant direct Political conservatism has the most consis- path from gender to atheist identity in the tent relationship across all of the models in reduced model (p = .020), and a significant, Table 4, associating with lower odds of any but small, mediation effect in a full model kind of nonreligious identification. This find- including gendered discrimination (p = .034), ing is in line with existing literature that sug- but the difference between the two models was gests a primary motive for leaving religious not significant. This means that experiences groups is dissatisfaction with the link between with discrimination might mediate the rela- religion and conservative political views (Hout tionship between a nonbeliever’s gender and and Fischer 2002, 2014). Parents are less likely their odds of identifying as an atheist, but our to identify as nonreligious in general but show small subsample of respondents does not show no specific patterns in identification prefer- definitively that such mediation effects are sta- ences. Nonwhite respondents are less likely to tistically significant. identify as nonreligious overall, and much less likely to identify as atheist or agnostic. While Discussion and Conclusion college completion does not significantly relate to the odds of identifying as nonreligious Our results point toward an understanding of overall, those who complete college are almost gender differences in nonreligion grounded in Edgell et al. 13

Table 5. Odds of Atheist Identification—Logistic Regression: Nonbelieving Subsample.

Variables Odds ratio Standard error Odds ratio Standard error Female 0.34* 0.15 0.40 0.20 Age 0.97* 0.01 0.97* 0.01 Nonwhite 0.34* 0.17 0.34* 0.17 College 1.65 0.66 1.65 0.67 Income 0.98 0.04 0.98 0.04 Parent 1.16 0.51 1.11 0.49 Conservative 0.86 0.15 0.86 0.15 Religious discrimination 1.64 0.59 1.94 0.88 Female × Religious Discrimination — — 0.74 0.37 Constant 3.12 3.34 2.91 3.15 N 275 275 McFadden R2 .11 .11 Wald χ2 17.72* 18.25* BIC 394.2 399.4

Source. Boundaries in the American Mosaic Survey (2014). Note. Models include poststratification weights. BIC = Bayesian information criterion. *p < .05. **p < .01. ***p < .001 (one-tailed tests). social risk. One of our contributions is to build Our survey data are cross-sectional, and on previous explanations for this gender differ- thus we cannot definitively settle the question ence by highlighting the ways that men and of time ordering regarding experiences with women have different experiences with nonreli- discrimination and nonreligious affiliation. gious stigma and discrimination. We find that Prior research suggests that atheists are an women more frequently report experiences with “embattled minority,” and that experiences of nonreligious discrimination than do men, and discrimination push people into embracing and that men and women who experience religious sharpening their atheist identity (Doane and discrimination choose different nonreligious Elliott 2015; cf. Smith 1998). However, it is identities: men are more likely to identify as possible that people first declare an atheist atheists, and women are more likely to identify identity and then experience discrimination, as spiritual but not religious (SBNR). Notably, leading them to either “double down” and these significant gender differences show up embrace that identity more strongly or to when we look at specific nonreligious identities, switch to a less risky nonreligious identity. Our not the odds of a respondent claiming a nonreli- findings do show, however, that men’s and gious identity in general. This supports our con- women’s experiences with the social risk of ception of both religion and nonreligion as a nonreligion associate with different cultural varied set of everyday practices that intersect strategies for nonreligious identification. with other aspects of identity and experience. Research has found that women often have a And these gendered effects show up even more pragmatic, practice-oriented approach to among women who express nontheistic beliefs. religion that privileges spirituality over formal Women who do not believe in a god or universal identification, allowing them to choose strate- spirit are still less likely to identify as atheist gies of religious expression that avoid aspects than are men with the same beliefs, and their of religious institutions they perceive as prob- experiences with discrimination might mediate lematic for women, while avoiding the stigma the relationship between gender and identifica- of avowing an explicit nonreligious identity tion. This suggests that a perception of social (cf. Aune 2015; Schutz and Roth 2015; Wilcox risk persists for women regardless of their level 2009). This makes sense when we consider our of existential risk aversion. bivariate analyses in which we find that, Social Currents 14 despite tending to avoid the most stigmatized that women are not any less likely to identify as nonreligious identity (atheist), women in our nonreligious than men, but that they do choose sample still report experiencing more discrimi- less stigmatized identities and practices—iden- nation for their nonreligious identification than tifying as “SBNR,” continuing to attend church, do men. At the same time, we find that women and claiming some form of theistic belief. The prefer alternative labels for their nonreligiosity concept of social risk helps us to understand such as “SBNR” rather than “NIP”. these patterns. In the United States, we know that nonreli- The concept of social risk will also be use- gion is stigmatized, and qualitative research ful in explorations of racial differences in non- (Manning 2015; Schutz and Roth 2015) shows religiosity. If a focus on differences between that experiences of discrimination can result men and women has, in the past, led to viewing from this stigmatization. But discrimination is risk preferences as innate and biological, a not the only social cost associated with stigma- broader focus on marginality and power shifts tization (and some instances of discrimination the analysis to the social, including dynamics may not be due to stigma). Further research of discrimination and stigma management. that combines qualitative and quantitative This approach will be particularly useful for methods and includes better measures of understanding the nonreligious choices of stigma and a wider range of social costs will be members of historically marginalized racial needed to fully understand experiences with groups. Future research should examine in nonreligious stigmatization and the strategies more detail how racial minorities experience individuals use to combat it. and practice nonreligion, which identities and We also show that social context plays a practices they embrace, how gender and race role in how individuals express nonreligious intersect to shape nonreligious practice, and beliefs. Although less likely to identify as non- the specific social costs incurred by people of religious overall, nonwhite respondents who color who embrace nonreligious identities (cf. are nonreligious are more likely to identify as Brown et al. 2015). SBNR and less likely to identify as atheist or It will also be useful in future research to agnostic. Political conservatism significantly explore the differences in religious choices and associates with lower odds of nonbelief, nonat- the ways in which they intersect with social tendance, and nonreligious identification of all location and religious discrimination. Prior kinds. However, nonreligious respondents who research, including a large body of work on reli- are politically conservative are more likely to gious switching, largely focuses on the positive call themselves “NIP.” These findings support and expressive factors that shape religious a theoretical perspective that understands non- choices in adulthood (e.g., social mobility, a religion as embedded in other social contexts desire for religious homogamy with one’s and aspects of identity which shape nonreli- spouse, or finding a community more compati- gious identification, in addition to nonbelief ble with one’s lifestyle or other beliefs; Sherkat (cf. Baker and Smith 2015). 2014). When “push” factors are identified, they We show that nonreligion is socially risky, often have to do with differences in the strength that atheism is more socially risky than other or duration of religious socialization among dif- forms of nonreligion, and that it is the most ferent religious traditions. The implication is privileged who embrace the most socially risky that people choose freely in switching religious forms of nonreligion. Atheists are distinguished identities, and in a largely voluntaristic religious from the much broader population of “Nones” landscape that is not an unreasonable interpreta- by distinct markers of privilege—they are tion. But it is the fact that religion is understood young, white, male, and college educated (cf. as “chosen” in the American context that makes Baker and Smith 2015); even among the smaller individuals interpret religious choices as a group of nonbelievers in our sample, those who reflection of character, morality, and trustwor- also claim an atheist identity are the most privi- thiness, and therefore a useful marker of impor- leged—young, white males. However, we find tant symbolic boundaries (Kalkan et al. 2009; Edgell et al. 15 cf. Lamont 1992). It would be fruitful to analyze embrace nonreligious identities or are instead when and how social risk may shape religious the result of the choice to “come out” as nonre- choices and how this may work differently for ligious. Such research must be mindful of the men and women, whites and nonwhites, and the changing context of nonreligious practice and more and less educated. explore more systematically how experiences We have sought to build on and extend of discrimination vary with age, gender, race, recent critiques of risk preference explanations and other aspects of social location. This is for gender differences in religiosity, which necessary because religion is more than a mat- have argued that these differences are not uni- ter of believing in a god or an afterlife; it is a versal and that experiences of existential risk social practice that involves an array of con- may vary. We agree. We also argue that it textually embedded choices that intersect in makes sense to develop a more complex and different ways with other aspects of social socially embedded understanding of the risks identity. Particularly in contexts in which reli- of nonreligion, one that focuses on the differ- giosity is a choice, and one perceived by many ential social costs of embracing stigmatized as socially relevant and an index of moral char- nonreligious identities, beliefs, and behaviors acter, further investigation of the social risks of for different individuals (e.g., Baker and Smith nonreligion is essential. 2015; Schutz and Roth 2015). In a society in which religiosity is more normative for women Declaration of Conflicting Interests than it is for men, in which women are more The author(s) declared no potential conflicts of likely to be sanctioned for behavior that vio- interest with respect to the research, authorship, lates gender norms, and in which atheism is the and/or publication of this article. most stigmatized nonreligious choice, it makes sense for women to manage the social stigma Funding of nonreligion to reduce social risk. It is impor- The author(s) disclosed receipt of the following tant to understand that the risk being managed financial support for the research, authorship, and/ is derived from social power and stigma, rather or publication of this article: The authors appreciate than philosophical or existential concerns the generous support for data collection and research alone. Our subsample analysis of nonbelievers assistance given by the National Science Foundation shows that even women who have already (Grants 1258926 and 1258933) and the Edelstein taken Pascal’s Wager and chosen not to believe Family Foundation. are still reluctant to openly identify as atheists. The fact that findings for black and Hispanic Notes Americans mirror those for women supports 1. We acknowledge that a force-choice question the idea that sociologists should consider the for religious belief is not ideal and loses the different social costs of nonreligion that make nuance gained in questions offered by surveys nonreligious choices risky, depending on how like the General Social Survey (GSS) that religiosity intersects with other aspects of allow for more ambiguous responses (Bishop one’s social identity. 1999; Hout and Fischer 2002:173). Although Our findings point to the need for further the BAM survey used a dichotomous question investigation of the socially risky nature of assessing belief in God(s) in the style of Gallup nonreligion, including more qualitative poll data, the nuance that these authors argue research on the nature of sanctions faced by we lose is actually beneficial for our analysis nonreligious men and women in different because we are not interested in differentiat- ing the certainty of belief. Instead, our measure social and professional contexts and the vari- captures only those who are willing to explic- ous “risk management” strategies they employ itly declare their nonbelief in a survey; for our as a result. Furthermore, more and better sur- purposes, this is a conservative and valid mea- veys are needed that allow for an assessment of sure of nonbelief. whether experiences of discrimination either 2. Respondents received a question condition- serve as a turning point prompting people to ally worded on their response to their current Social Currents 16

religious preference. If respondents identified Baker, Joseph O. and Buster G. Smith. 2015. with a religious denomination, they received American Secularism: Cultural Contours of the question, “Have you ever experienced any Nonreligious Belief Systems. New York: New discrimination because of your religion?” If the York University Press. respondent identified as an atheist, an agnostic, Baker, Joseph O’Brian and Buster Smith. 2009. nothing in particular (NIP), or spiritual but not “None Too Simple: Examining Issues of religious (SBNR), she or he received the ques- Religious Nonbelief and Nonbelonging in the tion, “Have you ever experienced any discrimi- United States.” Journal for the Scientific Study nation because you are not religious?” To build of Religion 48(4):719–33. our measure of discrimination, we interacted this Beaman, Lori and Steven Tomlins, eds. 2015. dichotomous measure with an ordinal follow-up Atheist Identities: Spaces and Social Contexts. question: “Would you say this has happened to Switzerland: Springer International Publishing. you frequently, occasionally, or rarely?” Our Becker, Penny Edgell and Heather Hofmeister. 2001. final measure of religious discrimination is ordi- “Work, Family, and Religious Involvement for nal and coded such that 0 = never experiencing Men and Women.” Journal for the Scientific discrimination, 1 = rarely experiencing discrimi- Study of Religion 40(4):707–22. nation, 2 = occasionally experiencing discrimina- Bishop, George. 1999. “Trends: Americans’ tion, and 3 = often experiencing discrimination. Belief in God.” The Public Opinion Quarterly 3. We ran an analysis that replicates the analy- 63(3):421–34. sis in Table 4 using a baseline of nonreligious Bradshaw, Matt and Christopher Ellison. 2009. respondents alone as the reference category. “The Nature-nurture Debate Is over, and Both All respondents in these models identified as Sides Lost! Implications for Understanding one of the four specific nonreligious groups, Gender Differences in Religiosity.” Journal for and each regression compares those who chose the Scientific Study of Religion 48(2):241–51. a specific designator with a baseline of all Branscombe, Nyla, Michael Schmitt, Richard Harvey, the others. Results did not vary from those in and Ed Diener. 1999. “Perceiving Pervasive Table 4. We also treated the identity categories Discrimination among African Americans: as multinomial indicators against a religiously Implications for Group Identification and Well- identified baseline. For these models, respon- being.” Journal of Personality and Social dents who claimed a religious identity were Psychology 77(1):135–49. coded into a baseline 0 category; respondents Breen, Richard, Kristian Bernt Karlson, and Anders who chose “NIP,” “SBNR,” or “agnostic” Holm. 2013. “Total, Direct, and Indirect Effects were coded into a second category which we in Logit and Probit Models.” Sociological label “nonreligious”; and self-identified athe- Methods & Research 42(2):164–91. ists were coded into a third category, “atheist.” Brescoll, Victoria and Eric Uhlmann. 2008. “Can an We repeated this coding process to generate a Angry Woman Get Ahead? Status Conferral, separate dependent variable for each specific Gender, and Expression of Emotion in the nonreligious identity label (atheist, agnostic, Workplace.” Psychological Science 19(3): SBNR, and NIP), swapping the other non- 268–75. religious identities into the second category. Brown, R. Khari, Robert Taylor, and Linda Patterns of significant covariates remained Chatters. 2015. “Race/Ethnic and Social- consistent across all three sets of models. demographic Correlates of Religious Non- involvement in America: Findings from Three References National Surveys.” Journal of Black Studies Ammerman, Nancy T. 2013. “Spiritual but Not 46(4):335–62. Religious? Beyond Binary Choices in the Study Carroll, Michael. 2004. “Give Me That Ol’ Time of Religion.” Journal for the Scientific Study of Hormonal Religion.” Journal for the Scientific Religion 52(2):258–78. Study of Religion 43(2):275–78. Aune, Kristin. 2015. “Feminist Spirituality as Lived Chang, L. and J. Krosnick. 2002, September. “A Religion: How UK Feminists Forge Religio- Comparison of the Random Digit Dialing spiritual Lives.” Gender & Society 29(1):122–45. Telephone Survey Methodology with Internet Avishai, Orit, Afshan Jafar, and Rachel Rinaldo. Survey Methodology for Studying American 2015. “A Gender Lens on Religion.” Gender & Presidential Elections: Comparing Sample Society 29(1):122–45. Representativeness and Response Quality.” Paper Edgell et al. 17

presented at the Annual Meeting of the American Non-religious in the United States.” Social Forces Political Science Association. Boston: MA. 95:607–38. Retrieved December 18, 2016 (http:// Chaves, Mark. 2010. “SSSR Presidential Address, sf.oxfordjournals.org/content/early/2016/08/17/ Rain Dances in the Dry Season: Overcoming sf.sow063). the Religious Congruence Fallacy.” Journal for Edgell, Penny and Eric Tranby. 2007. “Religious the Scientific Study of Religion 49(1):1–14. Influences on Understandings of Racial Cimino, Richard and Christopher Smith. 2014. Inequality in the United States.” Social Problems Atheist Awakening: Secular Activism and 54(2):263–88. Community in America. New York: Oxford Ellemers, Naomi, Russell Spears, and Bertjan University Press. Doosje. 2002. “Self and Social Identity.” Collett, Jessica and Omar Lizardo. 2009. “A Power- Annual Review of Psychology 531(1):161–86. control Theory of Gender and Religiosity.” Ellison, Christopher G. and Darren E. Sherkat. Journal for the Scientific Study of Religion 1995. “The ‘Semi-involuntary Institution’ 48(2):213–31. Revisited: Regional Variations in Church Cook, Corey L., Florette Cohen, and Sheldon Participation among Black Americans.” Social Solomon. 2015. “What If They’re Right about Forces 73(4):1415–37. the Afterlife? Evidence of the Role of Existential Freese, Jeremy. 2004. “Risk Preferences and Gender Threat on Anti-atheist Prejudice.” Social Differences in Religiousness: Evidence from Psychological & Personality Science 6:840–46. the World Values Survey.” Review of Religious Cornwall, Marie. 2009. “Reifying Sex Difference Research 46(1):88–91. Isn’t the Answer: Gendering Processes, Risk, Freese, Jeremy and James D. Montgomery. 2007. and Religiosity.” Journal for the Scientific “The Devil Made Her Do It: Evaluating Risk Study of Religion 48(2):252–55. Preference as an Explanation of Sex Differences Cotter, Christopher. 2015. “Without God yet in Religiousness.” Advances in Group Processes Not without Nuance: A Qualitative Study of 24:187–229. Atheism and Non-religion among Scottish Frost, Jacqui and Penny Edgell. 2017. University Students.” Pp. 171–96 in Atheist “Distinctiveness Reconsidered: Religiosity, Identities: Spaces and Social Contexts, Structural Location, and Racial Inequality.” edited by Lori Beaman and Steven Tomlins. Journal for the Scientific Study of Religion. Switzerland: Springer International Publishing. Gervais, Will, Asim Shariff, and Ara Norenzayan. Cragun, Ryan, Barry Kosmin, Ariela Keysar, Joseph 2011. “Do You Believe in Atheists? Distrust Hammer, and Micheal Nielsen. 2012. “On the Is Central to Anti-atheist Prejudice.” Journal Receiving End: Discrimination toward the of Personality and Social Psychology Non-religious in the United States.” Journal of 101(6):1189–206. Contemporary Religion 27(1):105–27. Gillespie, Rosemary. 2000. “When No Means No: Croll, Paul R., Eric Tranby, Penny Edgell, and Disbelief, Disregard and Deviance as Discourses Douglas Hartmann. 2014. “Collaborative of Voluntary Childlessness.” Women’s Studies Proposal: Boundaries in the American Mosaic: International Forum 23(2):223–34. Inclusion and Exclusion in the Contemporary Goffman, Erving. 1963. Stigma: Notes on the United States.” National Science Foundation. Management of Spoiled Identity. New York: Devine, Paula. 2013. “Men, Women, and Religiosity Simon & Schuster. in Northern Ireland: Testing the Theories.” Guenther, Katja. 2014. “Bounded by Disbelief: Journal of Contemporary Religion 28(3):473–88. How Atheists in the United States Differentiate Doane, Michael and Marta Elliott. 2015. “Perceptions Themselves from Religious Believers.” Journal of Discrimination Among Atheists: Consequences of Contemporary Religion 29(1):1–16. for Atheist Identification, Psychological and Hammer, Joseph, Ryan Cragun, and Karen Hwang. Physical Well-Being.” Psychology of Religion 2013. “Measuring Spiritual Fitness: Atheist and Spirituality, 7(2):130-41. Military Personnel, Veterans, and Civilians.” Edgell, Penny. 2012. “A Cultural Sociology of Military Psychology 25(5):438–51. Religion: New Directions.” Annual Review of Hammer, Joseph, Ryan Cragun, Karen Hwang, Sociology 38(1):247–65. and Jesse Smith. 2012. “Forms, Frequency, Edgell, Penny, Douglas Hartmann, Evan Stewart, and Correlates of Perceived Anti-atheist and Joseph Gerteis. 2016. “Atheists and Other Discrimination.” Secularism & Nonreligion Cultural Outsiders: Moral Boundaries and the 1:43–67. Social Currents 18

Hastings, Orestes and Micheal Lindsay. 2013. Religious’ Categories and Populations.” “Rethinking Religious Gender Differences: Religion 44(3):466–82. The Case of Elite Women.” Sociology of Lee, Lois. 2015. Recognizing the Nonreligious: Religion 74(4):471–95. Reimagining the Secular. Oxford, England: Heilman, Madeline and Julie Chen. 2005. “Same Oxford University Press. Behavior, Different Consequences: Reactions Letherby, Gayle. 2002. “Childless and Bereft? to Men’s and Women’s Altruistic Citizenship Stereotypes and Realities in Relation to Behavior.” Journal of Applied Psychology ‘Voluntary’ and ‘Involuntary’ Childlessness and 90(3):431–41. Womanhood.” Sociological Inquiry 72(1):7–20. Hoffmann, John. 2009. “Gender, Risk, and Lincoln, C. Eric and Lawrence H. Mamiya. 1990. Religiousness: Can Power Control Provide the The Black Church in the African American Theory?” Journal for the Scientific Study of Experience. Durham, NC: Duke University Religion 48(2):232–40. Press. Hosmer, David and Stanley Lemeshow. 2013. Link, Bruce and Jo Phelan. 2001. “Conceptualizing Applied Logistic Regression. 3rd ed. Hoboken, Stigma.” Annual Review of Sociology NJ: John Wiley. 27(1):363–85. Hout, Michael and Claude Fischer. 2002. MacNell, Lillian, Adam Driscoll, and Andrea “Why More Americans Have No Religious Hunt. 2015. “What’s in a Name: Exposing Preference: Politics and Generations.” Gender Bias in Student Ratings of Teaching.” American Sociological Review 67(2):165–90. Innovative Higher Education 40:291–303. Hout, Michael and Claude Fischer. 2014. Major, Brenda and Laurie O’Brien. 2005. “The “Explaining Why More Americans Have Social Psychology of Stigma.” Annual Review No Religious Preference: Political Backlash of Psychology 56(1):393–421. and Generational Succession, 1987-2012.” Manning, Christel. 2015. Losing Our Religion: How Sociological Science 1(9):423–47. Unaffiliated Parents Are Raising Their Children. Huggins, V. and J. Eyerman. 2001. “Probability New York: New York University Press. Based Internet Surveys: A Synopsis of Early McGuire, Meredith. 2008. Lived Religion: Faith Methods and Survey Research Results.” Paper and Practice in Everyday Life. New York: presented at the Conference of the Federal Oxford University Press. Committee on Survey Methods, November, Miller, Alan. 2000. “Going to Hell in Asia: The Arlington, VA. Relationship between Risk and Religion in a Hutchinson, Sikivu. 2011. Moral Combat: Black Cross Cultural Setting.” Review of Religious Atheists, Gender Politics, and the Values Wars. Research 42(1):5–18. Los Angeles, CA: Infidel Books. Miller, Alan and John Hoffmann. 1995. “Risk Kalkan, Kerem Ozan, Geoffrey C. Layman, and and Religion: An Explanation of Gender Eric M. Uslaner. 2009. “‘Bands of Others’? Differences in Religiosity.” Journal for the Attitudes toward Muslims in Contemporary Scientific Study of Religion 34(1):63–75. American Society.” The Journal of Politics Miller, Alan and Rodney Stark. 2002. “Gender and 71(3):847–62. Religiousness: Can Socialization Explanations Keysar, Ariela. 2014. “Shifts along the American Be Saved?” American Journal of Sociology Religious-secular Spectrum.” Secularism & 107(6):1399–423. Nonreligion 3(1):1–16. Miller, Ashley. 2013. “The Non-religious Lamont, Michele. 1992. Money, Morals, and Patriarchy: Why Losing Religion Has Not Manners: The Culture of the French and the Meant Losing White Male Dominance.” American Upper-middle Class. Chicago, IL: CrossCurrents 63:211–26. The University of Chicago Press. Miller, Carol and Cheryl Kaiser. 2001. “A Theoretical LeDrew, Stephen. 2015. “Atheism versus Perspective on Coping with Stigma.” Journal of Humanism: Ideological Tensions and Identity Social Issues 57(1):73–92. Dynamics.” Pp. 53–68 in Atheist Identities: Neitz, Mary Jo. 2014. “Becoming Visible: Religion Spaces and Social Contexts, edited by Lori and Gender in Sociology.” Sociology of Beaman and Steven Tomlins. Switzerland: Religion 75(4):511–23. Springer International Publishing. Riesebrodt, Martin. 2009. The Promise of Lee, Lois. 2014. “Secular or Nonreligious? Salvation: A Theory of Religion. Chicago, IL: Investigating and Interpreting Generic ‘Not The University of Chicago Press. Edgell et al. 19

Roth, Louise Marie and Jeffrey Kroll. 2007. “Risky Stark, Rodney. 2002. “Physiology and Faith: Business: Assessing Risk Preference Explanations Addressing the ‘Universal’ Gender Difference for Gender Differences in Religiosity.” American in Religious Commitment.” Journal for the Sociological Review 72(2):205–20. Scientific Study of Religion 41(3):495–507. Rudman, Laurie, Corinne Moss-Racusin, Julie Steensland, Brian, Lynn Robinson, Bradford Phelan, and Sanne Nauts. 2012. “Status Wilcox, Jerry Park, Mark Regnerus, and Robert Incongruity and Backlash Effects: Defending the Woodberry. 2000. The Measure of American Gender Hierarchy Motivates Prejudice against Religion: Toward Improving the State of the Female Leaders.” Journal of Experimental Art.” Social Forces 79(1):291–318. Social Psychology 48:165–79. Sullins, Paul. 2006. “Gender and Religion: Schnabel, Landon. 2015. “How Religious Deconstructing Universality, Constructing Are American Women and Men? Gender Complexity.” American Journal of Sociology Differences and Similarities.” Journal for the 112(3):838–80. Scientific Study of Religion 54(3):616–22. Thompson, Edward. 1991. “Beneath the Status Schnabel, Landon. 2016. “The Gender Pray Gap: Characteristic: Gender Variations in Wage Labor and the Religiosity of High- Religiousness.” Journal for the Scientific Study earning Women and Men.” Gender & Society of Religion 30(4):381–94. 30:643–69. doi:10.1177/0891243216644884. Voas, David, Siobhan McAndrew, and Ingrid Storm. Schutz, Amanda and Louise Marie Roth. 2015. 2013. “Modernization and the Gender Gap “Reconceptualizing Religious Risk: Gender in Religiosity: Evidence from Cross-national Differences in the Atheist Experience.” Working European Surveys.” The British Journal of paper presented at the American Sociology Sociology 65(1):259–83. Association Annual Meeting, August 2014, San Fransisco, CA. Wallace, Michael, Bradley Wright, and Allen Sherkat, Darren. 2000. “‘That They Be Keepers of the Hyde. 2014. “Religious Affiliation and Hiring Home’: The Effect of Conservative Religion on Discrimination in the American South: A Field Early and Late Transitions into Housewifery.” Experiment.” Social Currents 1(2):189–207. Review of Religious Research 41(3):344–58. Wilcox, Melissa. 2009. Queer Women and Religious Sherkat, Darren. 2002. “Sexuality and Religious Individualism. Bloomington, IN: Indiana Commitment in the United States: An Empirical University Press. Examination.” Journal for the Scientific Study Wright, Bradley, Michael Wallace, John Bailey, and of Religion 41(2):313–23. Allen Hyde. 2013. “Religious Affiliation and Sherkat, Darren. 2008. “Beyond Belief: Atheism, Hiring Discrimination in New England: A Field Agnosticism, and Theistic Certainty in the United Experiment.” Research in Social Stratification States.” Sociological Spectrum 28(5):438–59. and Mobility 34:111–26. Sherkat, Darren. 2014. Changing Faith: The Wright, Jennifer and Ryan Nichols. 2014. “The Social Dynamics and Consequences of Americans’ Cost of Atheism: How Perceived Religiosity Shifting Religious Identities. New York: New Influences Moral Appraisal.” Journal of York University Press. Cognition and Culture 14(1/2):93–115. Smith, Christian. 1998. American Evangelicalism. Zuckerman, Phil. 2011. Faith No More: Why People Embattled and Thriving. Chicago: University Reject Religion. New York: Oxford University of Chicago Press. Press.