Introduction
Worldviews are reifications of socially constructed understandings of the external world. They are beliefs, attitudes, and values that provide a model for how to interpret life that are often internalized to such an extent that they are taken-for-granted and unquestioned (Berger, 1967; Johnson et al., 2011; Sire, 2015). Given their typically assumed nature, how then do people come to adopt new worldviews? Berger (1967) classically argued that religious worldviews are more plausible when they are shared by entire societies and encompassed under a “sacred canopy” that protects them from the worldviews of disconfirming others. Thus, the taken-for-grantedness of a worldview depends, in part, on how well accepted it is by society. In later work he relaxed the proposition that the dissolution of a sacred canopy would undermine the plausibility of religious worldviews and, instead, argued that plausibility can be maintained even in pluralistic environments (Berger, 2014).
Worldviews are supported through social interactions and contexts even without sacred canopies (Berger, 1967; Corcoran, 2013; Smith and Emerson, 1998; Stark and Finke, 2000; Wellman, 2008). “Sacred umbrellas” (Smith and Emerson, 1998) or “sacred tents” (Wellman, 2008)—smaller plausibility structures embedded in close friendship and familial ties—can also reinforce the worldviews one holds. People look to their communities to evaluate and determine the plausibility of their worldviews (Bankston III, 2002; Corcoran, 2013; Iannaccone, 1995; Sherkat, 1997; Smith and Emerson, 1998; Stark and Finke, 2000). Those with weak plausibility structures should be more likely to adopt new worldviews compared to those with strong plausibility structures. Some cross-sectional research has examined how having religious, non-religious, or congregational friends are associated with religious beliefs and (non)affiliation (Baker and Smith, 2015, 2009a; Corcoran, 2013; Stroope, 2012; Welch, 1981; Wilkins-Laflamme, 2023; Wollschleger, 2021). It is difficult to know whether this is due to selection or influence effects (Cheadle and Schwadel, 2012), only the latter of which would involve actually changing one’s worldview.
Religious or theistic (i.e., belief in God) worldviews have primarily been the focus of much of past research examining the importance of likeminded social ties. Nonreligion and nontheism have historically been viewed as the absence of religion and belief, rather than the presence of secular meaning systems, identities, beliefs, and values that construct distinctive secular worldviews worthy of study. There have been several calls for more research on secular worldviews (Coleman III et al., 2018; Smith and Cragun, 2019) and an increasing number of studies beginning to do so (Baker and Smith, 2015; Edgell et al., 2023; Lee, 2015; Scheitle and Corcoran, 2023, 2020; Smith and Halligan, 2021; Taves et al., 2018). Just as there are multiple theistic worldviews, so too there are multiple secular worldviews. Secular, non-theistic, or non-religious worldviews, including atheistic, may include belief in humanism and science and valuing autonomy, rationalism, caring for the earth, kindness, equality, and education (Brown, 2017; Hunsberger and Altemeyer, 2006; Lee, 2015; Manning, 2015; Silver et al., 2014; Smith, 2011; Smith and Halligan, 2021; Strhan et al., 2024; van Mulukom et al., 2023). Some non-theists hold anti-theist beliefs that view religious beliefs as backward and ignorant; their opposition to religious beliefs propels them to seek to educate others on the contradictions and limitations of religion (Silver et al., 2014). Others are disinterested, apathetic, or unconcerned entirely with religious, atheist, and agnostic matters (Silver et al., 2014). And still others are the opposite of apathic and strongly value activism regarding atheist/agnostic social and political issues (Silver et al., 2014).
The worldviews of religious nones are diverse and can include belief in God, since reporting “no religion is an identity question”, whereas atheism is “a matter of belief” (Baker and Smith, 2009b, p. 720). For this study, we focus specifically on atheistic worldviews, a subset of the worldviews of religious nones. While there is diversity in atheistic worldviews, the common element connecting them as a distinct worldview category is not believing in God. As Baker and Smith (2009b, p. 721) note, “atheists make a definitive claim that God does not exist.” Whether likeminded social ties facilitate the adoption of an atheistic worldview has received less attention in the literature than the adoption of a religious worldview.
Some qualitative work identifies the importance of friendships and social interactions for becoming atheist or leaving religion (Smith, 2011; Zuckerman, 2015). Cragun et al.’s (2021) computer simulation found that the more religious people there were in one’s social network, the lower their odds of exiting religion, whereas the more secular people there were in one’s social network increased the odds that they would exit religion. They call for empirical research to further confirm these findings. Quantitatively, we still do not know whether the (non)-religious composition of friendship networks affects one’s odds of adopting an atheistic worldview, that is, moving from a theistic or agnostic worldview to an atheistic worldview, net of relevant factors. We explore this using panel data from the National Study of Youth and Religion (NSYR) to test whether the religious or non-religious composition of one’s friendship ties during adolescence increases the odds of adopting an atheistic worldview as an emerging adult in the United States.
Social Network Composition and Worldviews
The core of Berger’s (1967) argument was that a religious worldview is more plausible when everyone—one’s family, friends, enemies, close colleagues, and acquaintances—holds it creating a cocoon of universal support and legitimation for it. Smith and colleagues (1998) and Wellman (2008) argue that one only needs a small group of close positive ties with like-minded others to support the plausibility of a religious worldview. Disconfirming others only undermine one’s worldview when they are close ties (Facciani and Brashears, 2019; Smith and Emerson, 1998). Several studies have found a positive association between having more or a higher percentage of religious/congregational friends and stronger or more certain religious beliefs (Facciani and Brashears, 2019; Stroope, 2012; Welch, 1981), including certainty in God’s existence (Corcoran, 2013; Wollschleger, 2021). Some research has identified a positive association between having non-religious friends and identifying as atheist, being less certain of God, or being non-religious (Baker and Smith, 2015, 2009a; Scheitle et al., 2019; Wilkins-Laflamme, 2023; Wollschleger, 2021). Due to the cross-sectional nature of these studies, these relationships may be due to selection effects, in that, people may seek out friends that share their worldviews (i.e., homophily), or, due to influence effects (i.e., socialization), people’s worldviews may become more similar to the worldviews of their friends (Cheadle and Schwadel, 2012).
Using longitudinal network data, Cheadle and Schwadel (2012) found both selection and influence effects related to religious participation, identification, and devotion with socialization having a larger impact for most of the religiosity measures. This study did not examine belief. Hill (2014) found that a social network of co-religionists is important for the maintenance of creationist beliefs over time. A considerable amount of research, primarily qualitative, has identified the importance of social ties for recruitment to religious groups (Abel, 2006; Langston et al., 2019; Lofland and Stark, 1965; Smilde, 2005; Snow et al., 1980; Stark and Bainbridge, 1980). Lofland and Stark (1965, p. 871) found that “final conversion was coming to accept the opinions of one’s friends.” Although past research has primarily focused on recruitment to religion, becoming friends with atheists or those who are not religious has been found to play an important role in leaving religion or developing an atheist identity (Smith, 2011; Zuckerman, 2015). As Facciani and Brashears (2019, p. 4) identify “A strong plausibility structure consists of shared ideological agreement among the close ties of an individual. A weak plausibility structure consists of ideological disagreement among the close ties of an individual.” Those with weak plausibility structures should be more likely to be influenced by their ties to adopt new worldviews. We expect the same for adopting an atheistic worldview. Among those at risk for adopting an atheistic worldview (i.e., those with non-atheistic worldviews), we hypothesize that a stronger non-religious plausibility structure (or, alternatively, a weaker religious plausibility structure) will increase the odds of adopting an atheistic worldview.
Data and Measurement
We use two waves (waves 2 and 4) of panel data from the National Study of Youth and Religion (NSYR) to test our hypothesis (Smith, 2021; Smith and Pearce, 2019). NSYR is a nationally representative panel survey of U.S. teenagers funded by Lilly Endowment, Inc. Wave 1 of the NSYR occurred from July 2002 through April 2003 and surveyed via telephone 3,290 English- and Spanish-speaking teenagers (13–17 years old) through random-digit dialing. The three subsequent waves resurveyed the English-speaking respondents included in Wave 1. Wave 2 of the survey was fielded from June 2005 to November 2005 to the teenage respondents who were then between the ages of 16 and 21, including those who were no longer in the country. The response rate for Wave 2 was 78.6%. Wave 4 was fielded from February 2013 through December 2013 both online (85% of respondents) and through the phone (15% of respondents). Respondents were between the ages of 23 and 28. After removing from the eligible population those who had died, were incarcerated, deployed, or receiving in-patient medical treatment, or who no longer belonged in the sample due to age, there were 3,267 eligible potential respondents with 2,144 responding. Wave 2 is the first wave including information on the religious or non-religious nature of respondents’ close friends, which we use to measure (non)religious plausibility structures. We use Wave 4 as our outcome because most respondents will be out of college, if they attended, and fully in adulthood. Additionally, because adopting an atheistic worldview is a rare event, Wave 4 provides the longest exposure period within which someone could adopt an atheistic worldview, which maximizes the number of ‘events’ (i.e., adopting an atheistic worldview). The dependent variable thus represents those who adopted an atheistic worldview and maintained it through Wave 4 whether they adopted it during Wave 4 or at some time point prior but after Wave 2. Moreover, as not all respondents are present across waves, we use only two waves of data to maximize the overall sample size.
Examining Waves 1–3, Smith and Denton (Smith and Denton, 2003) found only minor differences between those who responded and those who didn’t in terms of socio-demographic variables. Comparing the NSYR to the U.S. Census and to other surveys of US adolescents (e.g., National Household Education Survey, Monitoring the Future, and the National Longitudinal Study of Adolescent Health) supports that it is nationally representative of US teenagers living in households (Smith and Denton, 2003).
Dependent Variable: Adopting an Atheistic Worldview
Our dependent variable in this study is the adoption of an atheistic worldview. In waves 2 and 4 NSYR respondents were asked, “Do you believe in God, or not, or are you unsure?”1 In order to adopt an atheistic worldview, a respondent must change their answer to this question from “yes” or “unsure/don’t know” in wave 2 to “no” in wave 4. Because of this, we exclude respondents who are not at risk of adopting an atheistic worldview in wave 4—that is, those who already had an atheistic worldview in wave 2 (i.e., respondent answered “no” to the question).
Independent Variable: (Non)Religious Plausibility Structure
Our focal independent variable concerns the plausibility structure for a non-religious worldview created by an individual’s close friends. In Wave 2, respondents were asked to…
“think of your five closest friends. These may be from your school, neighborhood, family, a religious congregation, work, a boyfriend or girlfriend, wherever, but they should not include your parents. Please tell me the names of your five closest friends or people you enjoy spending time with” or if needed, “If you do not have five close friends, then just think of the people you like and spend the most time with.”
The interviewer wrote down as many names up to five as the respondent identified. Those who reported “don’t know” were excluded. For those who identified more than one friend, they were asked how many of these friends “are not religious” as well as how many “are involved in a religious youth group.” For respondents who only identified one friend, they were asked if that friend is “not religious” as well as whether that friend is “involved in a religious youth group.” We use these responses to create measures of the (non)religious plausibility structure of respondents.
First, we create a percentage of respondents’ friends who are not religious variable and also a percentage of respondent’s friends who are involved in a religious youth group. Higher percentages of the former represent a stronger non-religious plausibility structure (or alternatively a weaker religious plausibility structure), whereas the latter represents the reverse for higher percentages. For respondents with one friend only, their values are either 0% if their response was no to the question or 100% if their response was yes.
According to Cook and colleagues (2017, p. 75) peer “influence appears to be a cumulative process where many voices are more persuasive than few.” It is possible that there is a threshold effect wherein once a majority of one’s closest friends share a particular worldview, the plausibility structure then exerts influence either supporting retaining that worldview or adopting it. Thus, we also created two binary indicator variables—one for a majority of one’s friends are not religious where 1 = 60% or more of one’s closest friends are not religious and 0 = fewer than 60% of one’s friends are not religious and one for a majority of one’s friends are involved in a religious youth group where 1 = 60% or more of one’s closest friends are involved in a religious youth group and 0 = fewer than 60% of one’s friends are involved in a religious youth group. Note, we use 60% because the distributions of the percentage of respondent’s friends who are not religious and the percentage of one’s friends who are involved in a religious youth group are such that after 50% the next possible higher value is 60%. Thus, a majority (i.e., above 50%) is necessarily 60% or more.
Additionally, Facciani and Brashears (2019, p. 5) argue that the complete worldview homogeneity of one’s friendship ties is important for strong plausibility structures: “while total homogeneity with close associates provides a powerful shield against doubt, even a single associate with a different ideology creates a hole in one’s sacred umbrella.” Accordingly, they measure strong plausibility structure based on 100% of one’s friendship network sharing a worldview. Following this approach, we also create two binary indicator variables reflecting complete non-religious homogeneity in friendship ties (1 = 100% of one’s closest friends are not religious, 0 = otherwise) and complete religious homogeneity in friendship ties (1 = 100% of one’s closest friends are involved in a religious youth group). However, Facciani and Brashears (2019) do note that they would have also liked to measure homogeneity based on percentage but the data they used were insufficient for that purpose as nearly half of their respondents had completely homogeneous networks in terms of worldview. The NSYR has sufficient variation to allow for percentage, majority/threshold, and complete homogeneity variables.2
Control Variables
In addition to our focal independent variables, we include controls for number of friends nominated, gender (0 = male; 1 = female), teen age in years, and teen race/ethnicity (responses of don’t know were set to missing). Wave 2 does not include the respondent’s race, so we use a variable in wave 4 that represents the respondent’s race as reported in Wave 1. We combine the Other and Islander response categories due to small sample sizes: 1) non-Hispanic white; 2) non-Hispanic Black; 3) Hispanic; 4) non-Hispanic Asian; 5) Islander and Other; 6) Native American; 7) Mixed. As past research has found that people in the U.S. South are less likely to be atheist and more likely to experience discrimination if they are atheist (Scheitle and Corcoran, 2018; Sherkat, 2008), we control for U.S. census region. We created this variable by collapsing the Wave 2 U.S. census division variable into the following census regions: 1) Northeast; 2) Midwest; 3) South; and 4) West.
In wave 4, respondents were also asked for their annual income before taxes and provided with the following response choices: 1) I did not earn any money; 2) Less than $10K; 3) $10K-$19,999; 4) $20K-$29,999; 5) $30K-$39,999; 6) $40K-$49,999; 7) $50K-$59,999; 8) $60K-$69,999; 9) $70K-$79,999; 10) $80K-$89,999; 11) $90K-$99,999; 12) $100,000-$124,999; 13) $125,000-$149,999; 14) $150,000-$174,999; 15) $200,000 or more;17) I don’t know/prefer not to say. Those who responded with don’t know/prefer not to say were treated as missing.
We also control for education during wave 2, as American atheists are more likely to be college graduates than religious Americans and American religious nones (Pew Research Center, 2024). Respondents were asked for their highest year of school completed. Having at least some college education was coded as 1 and all other values were coded as 0 (i.e., vocation/technical school, high school, elementary school, and no formal schooling).
In terms of religious controls, we include the religious tradition of the respondent in Wave 2, which is provided by the NSYR. We combined the Black Evangelical Protestant and Black Mainline Protestant categories into a “Black Protestant” category due to the small sample size for Black Mainline Protestants. This results in the following categories: 1) Evangelical Protestant; 2) Mainline Protestant; 3) Catholic; 4) Jewish; 5) No affiliation; 6) LDS/Mormon; 7) Other religion; 8) Indeterminate; 9) Black Protestant. We also include the respondent’s religious service attendance at the first congregation they identified (Wave 2): 0) Never; 1) Few times a year; 2) Many times a year; 3) Once a month; 4) 2–3 times a month; 5) Once a week; 6) More than once a week.
Analysis Plan
We use logistic regression models to predict adopting an atheistic worldview. We used a longitudinal weight in all models, which adjusts for each region-income stratum and takes into account sample attribution been Waves 2 and 4. Cases with invalid longitudinal weights were dropped.
Results
Table 1 presents descriptive statistics for all of the variables. Nine percent of the sample adopted an atheistic worldview between Waves 2 and 4. The average percentage of one’s friends who are non-religious is approximately 23%. Sixteen percent of respondents have mostly non-religious friends and 6% have all non-religious friends. The average percentage of one’s friends who were in a religious youth group is a little over 29%. For 24% of respondents, most of their friends were in a religious youth group and for 8% of respondents all of their friends are in a religious youth group. Twenty-eight percent of the sample were conservative Protestant in Wave 2 and 22% were Catholic. Most respondents are non-Hispanic White (73%), live in the South (38%), and have a household income in the range of $20,000-$29,999.
Table 1
Descriptive Statistics (N = 1,554), Weighted.
| PROPORTION OR MEAN | STD. DEV. | MIN | MAX | |
|---|---|---|---|---|
| Outcome | ||||
| Adopted Atheistic Worldview Between Wave 2 and 4 | 0.09 | 0 | 1 | |
| Focal Predictors (Plausibility Structures) | ||||
| Non-religious friends percentage (Wave 2) | 22.77 | 29.64 | 0 | 100 |
| Non-religious friends majority (Wave 2) | 0.16 | 0 | 1 | |
| Non-religious friends 100% (Wave 2) | 0.06 | 0 | 1 | |
| Religious youth group friends percentage (Wave 2) | 29.39 | 32.56 | 0 | 100 |
| Religious youth group friends majority (Wave 2) | 0.24 | 0 | 1 | |
| Religious youth group friends 100% (Wave 2) | 0.08 | 0 | 1 | |
| Controls | ||||
| Number of friends nominated (Wave 2) | 4.93 | 0.39 | 1 | 5 |
| Gender (Wave 2) | ||||
| Male | 0.45 | 0 | 1 | |
| Female | 0.55 | 0 | 1 | |
| Age (Wave 2) | 17.76 | 1.35 | 16 | 20 |
| Household Income (Wave 4) | 4.34 | 2.36 | 1 | 16 |
| College attendance (Wave 2) | 0.24 | 0 | 1 | |
| Race/ethnicity (Wave 1) | ||||
| Non-Hispanic White | 0.73 | 0 | 1 | |
| Non-Hispanic Black | 0.12 | 0 | 1 | |
| Hispanic | 0.09 | 0 | 1 | |
| Non-Hispanic Asian | 0.01 | 0 | 1 | |
| Islander, Native American, and Other | 0.03 | 0 | 1 | |
| Mixed | 0.02 | 0 | 1 | |
| Region of Residence (Wave 2) | ||||
| Northeast | 0.17 | 0 | 1 | |
| Midwest | 0.22 | 0 | 1 | |
| South | 0.38 | 0 | 1 | |
| West | 0.23 | 0 | 1 | |
| Religious Service Attend (Wave 2) | 2.85 | 2.21 | 0 | 6 |
| Religious Tradition (Wave 2) | ||||
| Conservative Protestant | 0.28 | 0 | 1 | |
| Mainline Protestant | 0.09 | 0 | 1 | |
| Black Protestant | 0.05 | 0 | 1 | |
| Catholic | 0.22 | 0 | 1 | |
| Jewish | 0.01 | 0 | 1 | |
| Mormon/LDS | 0.04 | 0 | 1 | |
| No affiliation | 0.13 | 0 | 1 | |
| Other Religion | 0.02 | 0 | 1 | |
| Indeterminate | 0.15 | 0 | 1 |
Table 2 provides bivariate descriptive statistics showing the percentage of respondents in each category who adopted an atheistic worldview. Those for whom most (17.28%) or all (22.50%) of one’s friends are not religious have higher than average percentages of adopting an atheistic worldview. On the other hand, those who have a majority (5.23%) or entirety of one’s friends (2.19%) who are a part of a religious youth group have lower than average percentages of adopting an atheistic worldview. In terms of religious affiliation, mainline Protestants, Jewish, no affiliation, and other religion have higher than average percentages of adopting an atheistic worldview. Males have higher than average levels of adopting an atheistic worldview (11.41%) whereas females have lower than average levels (6.60%). Non-Hispanic White respondents and non-Hispanic Asian respondents have higher than average levels at 10.61% and 12.54% respectively. Roughly 11% of those with some college education adopted an atheistic worldview compared to about 8% with less than some college education. In terms of U.S. regions, the Midwest region has the highest percentage of respondents who adopted an atheistic worldview (12.12%).
Table 2
Percentage of Respondents Adopting an Atheistic Worldview, Overall and by Select Variables.
| Overall | 9.00% |
|---|---|
| Focal Predictors | |
| Majority of friends not religious | 17.28% |
| 100% of friends not religious | 22.50% |
| Majority of friends relig youth group | 5.23% |
| 100% of friends relig youth group | 2.19% |
| Religious Tradition | |
| Conservative Protestant | 4.41% |
| Mainline Protestant | 16.32% |
| Black Protestant | 4.10% |
| Catholic | 6.82% |
| Jewish | 19.67% |
| Mormon/LDS | 5.36% |
| No affiliation | 19.71% |
| Other religion | 16.78% |
| Indeterminate | 5.40% |
| Gender | |
| Male | 11.41% |
| Female | 6.60% |
| Race/ethnicity | |
| Non-Hispanic White | 10.61% |
| Non-Hispanic Black | 1.11% |
| Hispanic | 4.70% |
| Non-Hispanic Asian | 12.54% |
| Islander, Native American, and Other | 4.53% |
| Mixed | 9.07% |
| Education | |
| Less than some college education | 7.91% |
| Some college education | 11.37% |
| Region | |
| Northeast | 9.96% |
| Midwest | 12.12% |
| South | 5.63% |
| West | 9.79% |
Table 3 presents the logistic regression models predicting adopting an atheistic worldview. Model 1 starts with the two main predictors—percentage of friends who are not religious and percentage of friends who are in a religious youth group. A 1% increase in one’s friends who are not religious is associated with a 1% increase in the odds of adopting an atheistic worldview. The percent of one’s friends who are in a religious youth group is not significantly associated with adopting an atheistic worldview. Model 2 estimates the association between the control variables and adopting an atheistic worldview excluding the predictors. Females have significantly lower odds of adopting an atheistic worldview compared to males. Non-Hispanic Black respondents have significantly lower odds of adopting an atheistic worldview compared to non-Hispanic White respondents. Religious service attendance is negatively and significantly associated with the odds of adopting an atheistic worldview. Mainline Protestants and those who identify with other religions have a significantly higher odds of adopting an atheistic worldview compared to conservative Protestants. Household income, college attendance, number of friends nominated, and region are not significantly associated with adopting an atheistic worldview. Model 3 presents the full model with predictors and control variables. All control variables maintain their associations net of the predictors. The percent of one’s friends who are not religious maintains its significant and positive association with adopting an atheistic worldview. Table 5 presents the predicted probabilities for percentages of friends who are not religious for Model 3 in Table 2 holding all other variables at their means. The predicted probability of adopting an atheistic worldview is 3% when a respondent has 0 friends who are not religious and 14% when 100% of one’s friends are not religious.
Table 3
Logistic Regression Models Predicting Adopting an Atheistic Worldview.
| PREDICTORS | OUTCOME: ADOPTED AN ATHEISTIC WORLDVIEW BETWEEN WAVE 2 AND WAVE 4 | |||||
|---|---|---|---|---|---|---|
| MODEL 1 | MODEL 2 | MODEL 3 | ||||
| ODDS RATIO | 95% CONFIDENCE INTERVAL | ODDS RATIO | 95% CONFIDENCE INTERVAL | ODDS RATIO | 95% CONFIDENCE INTERVAL | |
| Plausibility Structure | ||||||
| Non-religious friends percentage | 1.01*** | [1.01, 1.02] | – | – | 1.01*** | [1.01, 1.02] |
| Youth group friends percentage | 1.0 | [0.99, 1.00] | – | – | 1.01 | [1.00, 1.02] |
| Controls | ||||||
| Number of friends nominated | 1.11 | [0.67, 1.85] | 1.22 | [0.72, 2.05] | ||
| Gender | ||||||
| Male (referent) | – | – | – | – | – | – |
| Female | – | – | 0.55** | [0.36, 0.84] | 0.57** | [0.37, 0.86] |
| Age | – | – | 0.97 | [0.78, 1.2] | 0.96 | [0.77, 1.2] |
| Household Income | – | – | 0.92 | [0.81, 1.04] | 0.91 | [0.8, 1.05] |
| College Attendance | – | – | 1.64 | [0.85, 3.15] | 1.91 | [0.98, 3.72] |
| Race/ethnicity | ||||||
| Non-Hispanic White (referent) | – | – | – | – | – | – |
| Non-Hispanic Black | – | – | 0.03* | [0, 0.79] | 0.04* | [0, 0.82] |
| Hispanic | – | – | 0.44 | [0.2, 1.00] | 0.49 | [0.22, 1.08] |
| Non-Hispanic Asian | – | – | 0.73 | [0.14, 3.85] | 0.9 | [0.2, 4.04] |
| Islander, Native American and Other | – | – | 0.33 | [0.08, 1.40] | 0.29 | [0.06, 1.39] |
| Mixed | – | – | 0.78 | [0.14, 4.39] | 0.84 | [0.18, 3.87] |
| Region of Residence | ||||||
| South (referent) | – | – | – | – | – | – |
| Northeast | – | – | 1.54 | [0.79, 3.00] | 1.34 | [0.7, 2.61] |
| Midwest | – | – | 1.78 | [1.00, 3.17] | 1.73 | [0.97, 3.10] |
| West | – | – | 1.69 | [0.87, 3.29] | 1.58 | [0.81, 3.1] |
| Religious Service Attendance | – | – | 0.75*** | [0.64, 0.88] | 0.74*** | [0.63, 0.87] |
| Religious Tradition | ||||||
| Conservative Protestant (referent) | – | – | – | – | – | – |
| Mainline Protestant | – | – | 3.24** | [1.45, 7.26] | 3.9** | [1.72, 8.73] |
| Black Protestant | – | – | 11.48 | [0.59, 224.23] | 10.59 | [0.54, 207.32] |
| Catholic | 0.98 | [0.44, 2.20] | 1.18 | [0.52, 2.67] | ||
| Jewish | 2.02 | [0.52, 7.88] | 1.83 | [0.48, 6.97] | ||
| Mormon/LDS | – | – | 1.00 | [0.21, 4.74] | 0.87 | [0.21, 3.59] |
| No affiliation | – | – | 1.60 | [0.69, 3.7] | 1.2 | [0.5, 2.87] |
| Other religion | – | – | 3.75* | [1.03, 13.67] | 3.96* | [1.09, 14.35] |
| Indeterminate | – | – | 0.69 | [0.25, 1.91] | 0.65 | [0.23, 1.83] |
| N 1,554 | ||||||
To determine the robustness of the findings to alternative ways of measuring plausibility structures, we also estimated models using the binary variables for the majority of one’s friends are not religious or in a religious youth group as well as all of one’s friends are not religious or in a religious youth group. As seen in Table 4, the findings are consistent. Respondents for whom a majority of their friends are not religious have 2 times higher odds of adopting an atheistic worldview compared to those who do not. Additionally, respondents for whom all of their friends are not religious have a nearly 3 times higher odds of adopting an atheistic worldview compared to respondents with less than 100% of their friends being not religious. Table 5 presents the predicted probabilities for having a majority or entirety of one’s friends who are not religious for Models 1 and 2 in Table 4 holding all other variables at their means. The predicted probability of adopting an atheistic worldview is 4% when less than 60% of one’s friends are not religious and 9% when 60% or more of one’s are not religious. For respondents for whom 100% of their friends are not religious, their predicted probability of adopting an atheistic worldview is 12%, whereas those for whom less than 100% of their friends are not religious have a predicted probability of adopting an atheistic worldview of 5%.
Table 4
Logistic Regression Models Predicting Adopting an Atheistic Worldview.
| PREDICTORS | OUTCOME: ADOPTED AN ATHEISTIC WORLDVIEWBETWEEN WAVE 2 AND WAVE 4 | |||
|---|---|---|---|---|
| MODEL 1 | MODEL 2 | |||
| ODDS RATIO | 95% CONFIDENCE INTERVAL | ODDS RATIO | 95% CONFIDENCE INTERVAL | |
| Plausibility Structure | ||||
| Non-religious friends majority | 2.07** | [1.25, 3.44] | – | – |
| Youth group friends majority | 1.2 | [0.63, 2.31] | – | – |
| Non-religious friends 100% | – | – | 2.87** | [1.51, 5.42] |
| Youth group friends 100% | – | – | 0.49 | [0.08, 2.93] |
| Controls | ||||
| Number of friends nominated | 1.16 | [0.70, 1.93] | 1.23 | [0.72, 2.10] |
| Gender | ||||
| Male (referent) | – | – | – | – |
| Female | 0.57** | [0.37, 0.86] | 0.56** | [0.37, 0.85] |
| Age | 0.96 | [0.77, 1.19] | 0.93 | [0.75, 1.16] |
| Household Income | 0.91 | [0.80, 1.04] | 0.92 | [0.81, 1.04] |
| Grade completed | 1.82 | [0.94, 3.51] | 1.81 | [0.94, 3.48] |
| Race/ethnicity | ||||
| Non-Hispanic White (referent) | – | – | – | – |
| Non-Hispanic Black | 0.03* | [0, 0.91] | 0.03* | [0, 0.69] |
| Hispanic | 0.47 | [0.21, 1.06] | 0.45 | [0.2, 1.02] |
| Non-Hispanic Asian | 0.82 | [0.17, 4.08] | 0.72 | [0.15, 3.58] |
| Islander, Native American and Other | 0.31 | [0.07, 1.47] | 0.30 | [0.07, 1.33] |
| Mixed | 0.85 | [0.16, 4.38] | 0.80 | [0.15, 4.20] |
| Region of Residence | ||||
| South (referent) | – | – | – | – |
| Northeast | 1.37 | [0.70, 2.7] | 1.44 | [0.73, 2.83] |
| Midwest | 1.72 | [0.97, 3.05] | 1.71 | [0.94, 3.09] |
| West | 1.58 | [0.81, 3.08] | 1.63 | [0.82, 3.25] |
| Religious Service Attendance | 0.75*** | [0.64, 0.88] | 0.77** | [0.66, 0.89] |
| Religious Tradition | ||||
| Conservative Protestant (referent) | – | – | – | – |
| Mainline Protestant | 3.46** | [1.54, 7.80] | 3.33** | [1.49, 7.45] |
| Black Protestant | 12.40 | [0.50, 308.82] | 10.36 | [0.55, 195.65] |
| Catholic | 1.07 | [0.47, 2.43] | 0.97 | [0.44, 2.15] |
| Jewish | 1.95 | [0.51, 7.39] | 1.55 | [0.41, 5.82] |
| Mormon/LDS | 0.97 | [0.22, 4.24] | 0.90 | [0.21, 3.76] |
| No affiliation | 1.35 | [0.57, 3.21] | 1.38 | [0.59, 3.24] |
| Other religion | 3.77* | [1.03, 13.72] | 3.84* | [1.08, 13.68] |
| Indeterminate | 0.67 | [0.24, 1.88] | 0.68 | [0.24, 1.89] |
| N 1,554 | ||||
Table 5
Predicted Probabilities for Adopting an Atheistic Worldview.
| TABLE 3, MODEL 3, NON-RELIGIOUS FRIENDS PERCENTAGE | ||
|---|---|---|
| Predicted Probability | 95% CI | |
| 0% | 0.03 | [0.02, 0.05] |
| 20% | 0.05 | [0.03, 0.06] |
| 40% | 0.06 | [0.04, 0.08] |
| 60% | 0.08 | [0.05, 0.11] |
| 80% | 0.11 | [0.06, 0.15] |
| 100% | 0.14 | [0.07, 0.21] |
| TABLE 4, MODEL 1, NON-RELIGIOUS FRIENDS MAJORITY | ||
| Predicted Probability | 95% CI | |
| 0 | 0.04 | [0.03, 0.06] |
| 1 | 0.09 | [0.05, 0.13] |
| TABLE 4, MODEL 2, NON-RELIGIOUS FRIENDS 100% | ||
| Predicted Probability | 95% CI | |
| 0 | 0.05 | [0.03, 0.06] |
| 1 | 0.12 | [0.05, 0.19] |
Discussion and Conclusion
The relationship between religious social ties and religious beliefs or worldviews is well-established in cross-sectional research. Fewer studies investigate non-religious ties and secular beliefs or worldviews particularly in the context of longitudinal research. Longitudinal research is important to separate selection versus influence effects. While Cheadle and Schwadel (2012) use longitudinal data to disentangle selection and influence effects, they examine religious participation, identification, and devotion, not belief. Smith (2011) and Zuckerman (2015) show the importance of social ties for becoming atheist or leaving religion, but because these studies are qualitative they cannot control for other factors. For example, Smith (2011) found that starting college was a key point in which participants began questioning theism and was also when they made new friendships further facilitating their questioning. Whether the friendships played a role in their becoming atheist independent of college cannot be determined with that data. The NSYR allows us to examine changes in belief longitudinally net of other relevant factors such as education, thereby contributing to the literature. The NSYR provides a unique opportunity to look at how the non-religiosity of one’s friends is associated with changing one’s worldview. In this study, we tested the association between non-religious and a form of religious (youth group participation) friendship ties on adopting an atheistic worldview. We found evidence that the percentage of non-religious friendships ties are associated with an increased odds of adopting an atheistic worldview net of a variety of controls.
Past research operationalizes strong plausibility structures in different ways. Facciani and Brashears (2019) argue that having even one friend who holds a different worldview operates as a hole in a person’s sacred umbrella that can undermine the plausibility of a worldview. Others suggest that the more ties that support one’s worldview the more plausible it is particularly when they outnumber the disconfirming ties (Cook et al., 2017; Corcoran, 2013). However, prior research has generally not adjudicated between these different propositions. We find that regardless of whether the plausibility structure related to a particular worldview is measured by percent, majority, or complete homogeneity of friendship ties, non-religious social ties are associated with an increased odds of adopting an atheistic worldview. Even having just one friend who is non-religious is associated with an increased odds of adopting an atheistic worldview. This appears consistent with Berger’s (1967) original argument that interaction with disconfirming others decreases the plausibility of beliefs (see also Facciani and Brashears, 2019). Having even one friend who is not religious may potentially serve to relativize one’s religious worldview thereby increasing the odds of adopting an atheistic worldview even when most of one’s friends are religious. Even still, those with friends who are non-religious, whether their whole friendship network or only a part, have greater odds of maintaining their worldview than adopting an atheistic worldview. Thus, worldviews may not be as fragile as some prior research suggests (Berger, 1967; Facciani and Brashears, 2019). The results indicate that once socialized into a worldview, it may be hard to change regardless of the worldview composition of one’s friends.
We did not find any significant associations between friendship ties to those who are members of a religious youth group and adopting an atheistic worldview. Madge and Hemming (2017) found that friends rank higher as an influence on one’s non-religious/religious identity for non-religious youth compared to religious youth. It may be the case that non-religious friends matter more for adopting an atheistic worldview than religious friends matter for retaining one’s religious worldview. In supplemental models (not shown), percentage of friends who are members of a religious youth group is significantly and negatively associated with adopting an atheistic worldview in a bivariate model. However, net of non-religious friendships or the religion controls the association is attenuated suggesting that non-religious friendships and measures of religiosity are more important for predicting the odds of adopting an atheistic worldview. It may also be the case that the inverse of non-religious friendships is religious friendships, which would account for why the religious youth group association is attenuated net of non-religious friendships. We discuss the limitations of the religious youth group measure below.
We do not find evidence of a significant association between attending college and adopting an atheistic worldview net of other factors. While Pew Research Center (2024) found that atheists are more likely to have graduated college than religious individuals, they used cross-sectional data, In this study, we exclude those with an atheistic worldview in wave 2 because they are not at risk of adopting an atheistic worldview in wave 4. Thus, we focus on correlates of adopting an atheistic worldview versus holding an atheistic worldview, which may explain the difference in results. We also do not find a significant association between age and adopting an atheistic worldview although this is likely because the age range of the respondents is constrained and has little variation.
The longitudinal panel nature of the NSYR allowed for the opportunity to test predictors of adopting an atheistic worldview. However, the measures were not perfect. Having direct measures of the percentage of one’s friends who held an atheistic worldview, rather than using non-religious friendship ties, would have better captured the strength of the plausibility structure for adopting an atheistic worldview. Yet, we still found the expected associations. We imagine those associations would be stronger had a direct measure been used. We did not have a corollary measure of religious ties. Being involved in a religious youth group likely represents those with stronger religious commitments, so we’d expect the measure to capture an even stronger religious plausibility structure compared to a measure that only asked about religious ties. However, it may be a poor measure of general religious ties particularly for our sample of those aged 16–21. Additionally, our dependent variable is limited to three response options and thus has less variation than measures in other surveys capturing certainty or doubt in God. It is possible that those with more doubts prior to Wave 2 are more likely to have non-religious friends and those with fewer doubts are more likely to have religious friends. Thus, it is possible that those with weaker worldviews initially may be more open to friendships with those who don’t share their worldview and also be more likely to change their worldview. While our study cannot adjudicate between these mechanisms, what we can say is that several prior studies have found that friendships and social interactions are important prior to and for adopting a new worldview (Lofland and Stark, 1965; Snow and Phillips, 1980; Stark and Bainbridge, 1980; Stark and Finke, 2000). In terms of atheism specifically, Smith’s (2011) qualitative study identified that, for some, new friendships as well as interactions with professors in college “prompted and drove” questioning their belief in God. It may be best to think about weakening theistic worldviews, burgeoning atheistic worldviews, and friendships with others who don’t believe in God as dynamic processes that may mutually reinforce each other and potentially lead to the adoption of an atheistic worldview. Further research is needed in this area. This study draws on only two waves of the NSYR to maximize the exposure period between waves and the number of adoptions of an atheistic worldview to allow sufficient power to estimate our models. As such, we were unable to incorporate additional variables from other waves or to examine the potential liminality of nonreligious worldviews (Lim et al., 2010), which would require multiple waves and is beyond the scope of this study. Lastly, this study is unable to describe the content of the atheistic worldviews of the respondents beyond that they do not believe in God. Such research is important and may be best captured by rich qualitative studies. Despite these limitations, this is one of only a handful of studies that examines on the individual-level how non-religious plausibility structures are associated with adopting an atheistic worldview over time. It would be beneficial for future research to measure religious and secular ties using a variety of measures and for a diverse range of social ties (e.g., friendship, familial, work).
Data Accessibility Statement
The data used in this paper is publicly available at www.thearda.com.
Notes
[1] Including those who said ‘no’ in Wave 2, weighting the data, and using the analytical sample, in Wave 2, 78.03% of respondents said yes, 4.59% said no, and 17.9% said don’t know/not sure. In Wave 4, 74.05% said yes, 11.40% said no, and 14.55% said don’t know/unsure.
[2] We also estimated the models with the number of respondents’ friends who are not religious and also the number of respondent’s friends who are involved in a religious youth group net of the number of friends the respondent nominated (results not shown). The results were consistent with those from the other operationalizations.
Competing Interests
The authors have no competing interests to declare.
Author Contributions
All authors (KEC and CPS) contributed to the study’s design, development, and implementation. KEC wrote the first draft of the manuscript. All authors contributed to drafting and editing the manuscript. All authors read and approved of the final manuscript.