Evan Stewart uses survey data in his research to investigate the differences between non-belief and atheist identification. Evan finds that belief in evolution and education are not significant predictors of atheist identification when we control for political views.
To borrow a phrase from the feminist movement, the personal is political when it comes to identifying as an atheist in the United States. Classic work on nonreligious Americans in social science shows that political views are a key predictor of religious disaffiliation, but how do these ideas shape the way people express nonreligion after they disaffiliate?
Last year I published a review article laying out some of the challenges for studying atheism, particularly in the United States. We know that anti-atheist sentiment in the U.S. is persistent and durable, and that atheists do a lot of collective work to build a common sense of identity. As a result, my work argues that we have to carefully distinguish non-belief in god from atheist self-identification, since self-identification signals far more than non-belief alone. Choosing to call oneself an atheist in the United States means navigating stigma, community affinities, and relationships to authority, and it can therefore become a shortcut for a range of assumptions about matters of public concern. For example, women who are already non-believers are less likely to call themselves atheists than men, and my work with Penny Edgell and Jacqui Frost argues that this has more to do with how society polices women’s religiosity than the choice of whether or not to believe in god alone.
I used survey data to investigate the differences between non-belief and atheist self-identification, but the original analysis from the review article was not conclusive. Most surveys, even those that can measure (non)religion in detail, cannot get a large enough subsample of atheists to make strong claims. The study of nonreligion is developing into a robust field, drawing interest from across the social sciences and humanities. As it continues to grow, researchers who work in this area will continue to face questions about whether their findings are generalizable across different groups and whether they replicate in different social contexts. Luckily, new data have become publicly available that allow for a conceptual replication (more on that here). The 2014 Pew Religious Landscape Study has over 35,000 respondents, and so while only 3-4% of the U.S. population self-identifies as an atheist, that shakes out to over 1,000 respondents in this survey. Many more people say they don’t believe in god, about 10% of the sample.
For this analysis, I start with the group of respondents who say they do not believe in god. We are interested in the probability that a given respondent from this group will also call themselves an atheist, and, most importantly, we are interested in what other traits associate with a higher or lower probability of that atheist identification. Logistic regression can help us answer these questions, and results from two logistic regression models are presented in the table. One model looks at some basic sociodemographic factors like age, gender, race, and education. The other adds two common beliefs among American atheists: one regarding social issues (support for same-sex marriage) and one regarding scientific authority (belief in human evolution). I chose these variables to get as close as possible to the original models, to see if they produce similar results.
So, what separates the self-identified atheists from “atheistic” respondents? Not much, it turns out. Older non-believers are slightly, but significantly, less likely to call themselves atheists. Non-believing women are less likely to call themselves atheists as well, but it is important to note that this effect does not emerge until we control for substantive beliefs in the second model. As we have argued elsewhere, there is evidence that women are not generally “more religious” across the board, but rather express nonreligion differently in line with gendered social risk. Also, notice which measures are not statistically significant in the table of results. Belief in evolution and education are not significant predictors of atheist identification when we control for political views. In these models, the difference comes down to social and political views more than scientific literacy or support.
Non-believers who are more liberal and express stronger support for same-sex marriage are significantly more likely to call themselves atheists. Each of these measures used a scale to indicate strength of support. The liberalism measure ran from 1 (very conservative) to 5 (very liberal), and the same-sex marriage measure ran from 1 (strongly oppose) to 4 (strongly favor). To get a sense of how strong these relationships are, the graphs below show the changes in predicted probabilities of atheist self-identification at each step on these scales, while holding all other variables constant. Non-believing respondents who strongly oppose same-sex marriage have about a 15% chance of identifying as atheists, while those who strongly favor the policy have about a 35% of identifying. Similarly, each step from very conservative to very liberal on the political ideology scale associates with about a five-percentage-point increase in the probability of atheist self-identification.
There are some necessary limitations to these conclusions. The models are not perfect by any means, and their low fit statistics (the small Pseudo R-squared) suggest that other factors which weren’t measured on the Pew survey could explain more of the choice to identify as an atheist. They probably also indicate that the true effect of liberal political views is somewhat smaller than the predicted probabilities shown here. However, this conceptual replication does provide additional evidence that self-identified atheism is as much a political phenomenon as a personal belief system.
In my dissertation research, I tease out what findings like these mean for understanding the political impact of nonreligious Americans more broadly. The cultural work that goes into understanding atheism and other kinds of nonreligion has implications for everything from voting trends to public opinion on the social safety net, racial issues, and environmental policy.
Together, this work suggests that nonreligion in general, and atheism in particular, is an important case for understanding the changing role of religion in American public life.
|Probability of Atheist Identification Among Non-Believing Americans|
|Lives in the South||-0.08||(0.11)||-0.06||(0.12)|
|Education (Baseline-High school)|
|Supports Same Sex Marriage||0.38***||(0.09)|
|Believes in Evolution||0.38||(0.33)|
|* p<0.05, ** p<0.01, *** p<0.001
Source: 2014 Pew Religious Landscape Survey
Notes: Models incorporate the RLS survey weights for known population benchmarks
Evan Stewart is a doctoral candidate in Sociology at the University of Minnesota, where he has worked as an Edelstein Fellow with the American Mosaic Project and an Interdisciplinary Doctoral Fellow with the Center for the Study of Political Psychology. His research focuses on political culture, public opinion, and religion and secularism across a range of institutional and community contexts. His dissertation work focuses on the political impact of the growing nonreligious population in the United States, while other solo and collaborative research projects examine prejudice and tolerance, atheist identity formation, attitudes about contentious political issues, and visual sociology. Evan also serves as a graduate editor at The Society Pages.