Nones, Dones, and the Social Identity of Religion

As research on nonreligious individuals has grown (Smith and Cragun Reference Smith and Cragun2019) a significant body of work examining the differences between formerly religious members of the nonreligious community and lifelong nonbelievers has been published. This work has largely taken place in psychology but is not absent in sociology and anthropology. Researchers have coined the term religious “dones” for those who have disaffiliated and “nones” for those who have never had a religious affiliation (Van Tongeren et al. Reference Van Tongeren2021). The literature has identified residual effects of religious identification on variables such as implicit and explicit feelings towards deities (Van Tongeren et al. Reference Van Tongeren2021), scores on Haidt’s moral foundations (Schwadel et al. Reference Schwadel2021), and consumer decisions (Van Tongeren, DeWall, and Van Cappellen Reference Van Tongeren, Nathan DeWall and Van Cappellen2023).

Tongeren et al. (Reference Van Tongeren, Nathan DeWall and Van Cappellen2023) examine currently religious, former, and never religious. They find that former religious identifiers differed from the never religious and currently religious in cognitive, emotional, and behavioral processes such as prosocial behavior (charitable donations and a self-rated measure). They find religious individuals are the most pro social, former members retain some of the prosociality associated with religion and never religious respondents have the lowest prosocial ratings (Van Tongeren, DeWall, and Van Cappellen Reference Van Tongeren, Nathan DeWall and Van Cappellen2023).

Most important to this paper is the work examining identity threat among the nonreligious. Newer work in this area examines the degree to which concealment is affected by the distinction between nones and dones (Mackey et al. Reference Mackey2021; Mackey, Van Tongeren, and Rios Reference Mackey, Van Tongeren and Rios2023). This research points to a potential causal explanation for divergent ideology and political expressions between the nones and the dones: Identity Threat. Former Christian identifiers experience social alienation, a loss of belonging, and often feel the need to conceal their nonreligious identity at far greater rates than religious nones (Mackey et al. Reference Mackey2021; Mackey, Van Tongeren, and Rios Reference Mackey, Van Tongeren and Rios2023). We propose that this social pain alongside Christianity’s association with right wing politics drives former identifiers to left wing identification and beliefs. A Potential Mechanism of Disaffiliation and Political Ideology

Key to our theory are the forces driving Americans away from religious beliefs and institutions. Scholars have proposed many reasons for the rapid and sustained losses religious institutions have suffered. First and oldest is secularization theory, which posits that modernization should necessarily make religion irrelevant and thus disappear (Latré, and Vanheeswijck Reference Latré, Vanheeswijck and James2015). A second explanation posited is that the systematic enabling of and concerted efforts to cover up decades of appalling sexual abuses by religious leaders and institutions has driven Americans away from religious identity (Ballard Reference Ballard2023).

However, a third theory of disaffiliation is both the most referenced and the most relevant to our findings. Scholars such as Hout and Fisher (Hout and Fischer Reference Hout and Fischer2002; Reference Hout and Fischer2014), Evans (Evans Reference Evans2016), and (Djupe, Neiheisel, and Conger Reference Djupe, Neiheisel and Conger2018) have proposed the Christian Right’s prominent involvement in politics as an explanation for American’s disaffiliation from religion generally. The association of religious belief and institutions with right wing political ideology, beliefs, policy, and movements as a driving force behind disaffiliation has widespread support in the literature (Baker and Smith Reference Baker and Smith2015; Braunstein Reference Braunstein2022; Campbell, Layman, and Green Reference Campbell, Layman and Green2020; Djupe, Neiheisel, and Conger Reference Djupe, Neiheisel and Conger2018; Hout and Fischer Reference Hout and Fischer2002; Margolis Reference Margolis2018; Vargas Reference Vargas2012). This work has found that disaffiliation increases in states where religious institutions involve themselves in politics (Djupe, Neiheisel, and Conger Reference Djupe, Neiheisel and Conger2018). In particular, more visible Christian Right activism is associated with increased identification as nonreligious. Furthermore, the effects appear to be linked to issue specific disagreements such as same sex marriage wherein the involvement of religion in politics produces not just organized opposition to conservative positions, but disaffiliation from religion (Djupe, Neiheisel, and Conger Reference Djupe, Neiheisel and Conger2018). Moreover, this effect is likely to function across intersectional lines such as race. In light of this work the meteoric increase in nonreligious identification can be seen as a reaction to the right-wing politics of many Christian denominations in the United States. In other words, disaffiliation may be a backlash that also pushes dones further left ideologically than nones.

This theory fits well with findings in psychology that religious dones experience social alienation from their previous faith and higher levels of social identity threat from religion (Mackey et al. Reference Mackey2021; Mackey, Van Tongeren, and Rios Reference Mackey, Van Tongeren and Rios2023). If religious dones’ primary reason for disaffiliation is related to politics and they perceive religion as threatening, we should expect them to have distinct political views from religious nones. Theoretically, the perception of conservative religion as threatening would be driving a backlash that leads to disaffiliation and left-wing beliefs. Consequently, disaffiliation driven by religious institutions’ association with right wing politics generates our expectation that dones will take more progressive attitudes across a set of social justice policies and positions as well as have increased opposition to reactionary policies. Stated formally here:

Dataset and Measures

To examine the political differences between religious nones and dones, we draw upon the 2020 wave of the Collaborative Multiracial Post-Election Survey (CMPS). The CMPS 2020 includes 17,556 respondents who were surveyed between April 2 and October 2, 2021. For our analysis, we focus on the 3,594 respondents who answered the question, “When it comes to religion, do you consider yourself to be…” with “atheist,” “agnostic,” or “none.” The CMPS offers distinct advantages for our study. It includes items on both past and current religious affiliation, perceived threats posed by conservative Christians, and views on key ideological and policy issues, while also allowing for analysis across racial and ethnic groups. Crucially, no other existing dataset includes the combination of variables necessary to address our two central research questions:Footnote ⁴ (1) Does the religious past of nonreligious individuals shape their ideological positions? and (2) Do perceptions of threat from conservative Christians mediate that relationship?

Explanatory and Moderating Variables

Table 1 presents the previous religious affiliations of nonreligious individuals. Our main explanatory variable is respondents’ past religious affiliation. To identify nonreligious individuals with religious backgrounds, we first generated a religion variable based on sixteen subcategories in the dataset. Respondents were asked, “When it comes to religion, do you consider yourself to be?” Those who selected “atheist,” “agnostic,” or “none” were grouped together and categorized as “nonreligious.”

Table 1. Past religious affiliation among nonreligious respondents (CMPS Sample)

We then created a variable capturing past religious affiliation based on responses to the question, “What primary religion, if any, did your family practice or follow when you were raised?” Using this information, we divided nonreligious individuals into two groups: those with a religious upbringing and those without. The first group, referred to as “dones,” includes nonreligious individuals who were raised in a Christian tradition. We only focus on Christian past because nearly 41% of nonreligious respondents reported that they have been raised in a Christian household, as shown in Table 1. The second group, referred to as “nones,” includes respondents who are nonreligious and were not raised in any religious tradition (45%).

To explore a possible explanation for differences in policy attitudes between nones and dones, we examine perceptions of threat as a moderating variable. We use a measure from the CMPS that asks respondents to evaluate whether various groups are supportive of or threatening to their vision of American society. Responses range from 1 (“strongly supports”) to 7 (“strongly threatens”), with higher scores indicating a greater perceived threat. We focus on respondents’ evaluations of conservative Christians, based on the premise that perceived threat linked to this group may shape the political attitudes of nonreligious individuals, particularly in distinguishing dones from nones. Finally, we accounted for several control variables, including gender, education, age, income, political affiliation, and race (White, Black, Latino, Asian, Other).

Outcome Variables

We constructed our dependent variables using a set of questions related to support for or opposition to socially and politically contested issues. First, we measured support for voting rightsFootnote ⁵ using the question: “Do you think the Voting Rights Act is necessary today to make sure that Blacks are allowed to vote, or do you think the Voting Rights Act is no longer necessary?” Respondents selected either “necessary today” or “no longer necessary.”

Next, we assessed attitudes toward immigration policies enacted under the Trump administration, particularly regarding the deportation and detention of asylum seekers. Respondents indicated whether they supported or opposed these policies.Footnote ⁶ Both the voting rights and immigration items were coded as binary variables.

We also evaluated support for abortion rights and criminal justice reform using reverse-coded 5-point Likert scale items, where higher values reflect stronger support for more liberal positions. Support for abortion rights was measured using respondents’ level of agreement with the statement, “Abortion should remain a legal option in the United States.” Support for criminal justice reform was assessed using the statement, “Law enforcement and the criminal justice system need to be completely reformed.”

Additionally, to measure attitudes toward same-sex marriage, we included a question asking respondents to indicate whether this issue should be a high or low priority for their communities, measured on a 4-point scale. To facilitate interpretability and comparability across outcomes, all variables were standardized to a 0–1 scale. Finally, we used the standard ideology scale included in the CMPS and recoded it as a conservatism scale, where higher values indicate more conservative ideological self-placement.

Intersectional Identities: Religious Past and Racial and Ethnic Identities

While the previous results indicate that both religious past and perceived threat linked to conservative Christians significantly predict how nonreligious individuals approach controversial policy issues, we also aim to assess whether these patterns hold across subgroups within this population. Racial and ethnic identity is another salient identity in the United States, frequently shaping political attitudes and behavior. One of the key strengths of the CMPS is its ability to provide high-quality data on diverse racial and ethnic groups, allowing us to explore how the intersection of religious past and racial/ethnic identity influences policy preferences among nonreligious individuals. Given the available sample sizes, our analysis focuses on four major racial and ethnic groups in the United States.: Whites, Blacks, Latinos, and Asians.

Before turning to the evaluation of outcome variables, we first examine whether the effect of religious past on perceived threat posed by conservative Christians varies across racial and ethnic groups. Table 6 shows that identifying as a religious “done” is positively associated with higher levels of threat perception among all groups except Black respondents. Among Whites, both higher education and Democratic Party affiliation are positively associated with threat perception, while age is negatively associated, suggesting that older White respondents are less likely to perceive conservative Christians as a threat.

Table 6. Predicting perceived conservative Christian threat by past religion, by racial and ethnic subgroups

^∗ p < .1, ^∗∗ p < .05, ^∗∗∗ p < .01. Standard errors in parentheses.

For Black respondents, higher levels of education, income, and age are all positively associated with perceived threat, indicating a different pattern from Whites. Among Latino respondents, both education and Democratic affiliation predict greater threat perception. In contrast, among Asian nonreligious individuals, the only significant predictor of threat perception, aside from religious past, is Democratic Party affiliation.

Table 7 demonstrates that both being a religious “done” and perceiving conservative Christians as a threat are important predictors of support for voting rights across all racial and ethnic groups. Although the effect of religious past for Asian nonreligious individuals does not reach statistical significance (p = .13), the estimated effect is substantively meaningful and aligns with the liberal positions that Asian religious dones show on other policy issues. We also assess whether the previously observed effects of religious past and threat perception on opposition to conservative immigration policies under the Trump administration hold across racial and ethnic subgroups. Due to limited sample size, Asian respondents were omitted from this analysis.Footnote ⁸ Robust logistic regression results indicate that nonreligious Latinos raised in Christian households are more likely to oppose Trump-era immigration policies. The results are statistically significant only for religious dones with Latino backgrounds. Although white religious dones show a similar pattern of support, the association does not reach statistical significance, possibly due to limited sample size. The number of Black respondents is also very small because of subsampling within a subset of the sample, leading to low statistical power and likely causing null results.

Table 7. Predicting policy attitudes by past religion and perceived conservative Christian threat, by racial and ethnic subgroups

^∗ p < .1, ^∗∗ p < .05, ^∗∗∗ p < .01. Standard errors in parentheses. Full models are presented in the appendix.

While findings for immigration policy are not strong across groups, the effects of religious past and threat perception are more consistent for other contested issues, including abortion rights, same-sex marriage, and criminal justice reform. Across all racial and ethnic subgroups, both being a religious “done” and perceiving conservative Christians as a threat are positively associated with support for abortion rights. Moreover, nonreligious individuals in all groups maintain liberal positions on same-sex marriage and justice reform. Specifically, both religious past and threat perception are negatively associated with opposition to same-sex marriage and positively associated with support for criminal justice reform across all racial and ethnic groups.

Finally, we assess how conservatism among nonreligious individuals is predicted by their religious past, conditional on perceived threat linked to conservative Christians, across racial and ethnic groups. Table 8 presents OLS estimates for each group, including models with main effects and interaction terms.

Table 8. Predicting conservatism by past religion and perceived conservative Christian threat, by racial and ethnic subgroups

Model 1 shows that among Whites, both religious past and perceived threat are negatively associated with conservatism, although the effect of religious past is not statistically significant. Model 2 includes an interaction term between religious past and threat perception. In this model, being a religious “done” has no significant effect on conservatism when perceived threat is absent (i.e., Threat Perception = 0), while threat perception significantly decreases conservatism among religious “nones” (Religious Past = 0). The interaction term is negative, indicating the expected direction of conditionality, but it does not reach statistical significance.

Among Black respondents, Model 3 shows that both religious past and threat perception are significantly and negatively associated with conservatism. Model 4 adds the interaction term and reveals a similar pattern: threat perception significantly reduces conservatism among religious nones, while being a religious done has no effect when threat perception is absent. However, the interaction term remains statistically insignificant, indicating no evidence of conditionality in this group.

Models 5 and 7 show that religious past has no direct effect on conservatism among Latino and Asian nonreligious individuals, respectively, while threat perception continues to have a significant negative effect. In Model 6, we find that among Latinos, being a religious done does not affect conservatism in the absence of perceived threat, but threat perception significantly decreases likelihood of conservatism among religious nones. As in previous models, the interaction term is negative but not statistically significant. Model 8 follows the same pattern for Asian respondents: being a religious done has no effect when threat perception is absent, while threat perception itself is associated with lower conservatism. Again, the interaction term is not significant.

Taken together, interaction models across all four racial and ethnic groups (Models 2, 4, 6, and 8) do not provide statistical evidence for conditionality between religious past and threat perception in predicting conservatism. However, Figures 3 and 4 suggest more nuanced dynamics. In particular, the visualizations indicate that differences in conservatism between dones and nones become more pronounced as perceived threat increases for some groups, offering suggestive evidence of an interaction effect that is not easily observed in the regression models.

Figure 3. The interactive effect of past religion and perceived conservative Christian threat on conservatism, by racial and ethnic subgroups.

Figure 4. The marginal effect of past religion on conservatism across levels of perceived conservative Christian threat, by racial and ethnic subgroups.

Figure 3 presents results from the analysis across racial and ethnic groups and shows no significant effect for the interaction of religious past and threat perception on conservatism. However, Figure 4 highlights a conditional relationship: religious dones become significantly less conservative than religious nones as their perceived threat from conservative Christians increases—though this pattern emerges only for some groups.

More specifically, the marginal effects illustrated in Figure 4 show that religious past has no meaningful effect on ideological orientation when individuals perceive conservative Christians as posing little or no threat. However, when perceived threat is high, White, Black, and Latino religious dones are significantly less likely to endorse conservative positions compared to those reporting lower levels of threat perception. In contrast, Asian religious dones do not exhibit variation in conservatism across different levels of perceived threat.