“No religious test shall ever be required as a qualification to any office or public trust under the United States.” United States Constitution — Article VI, Paragraph 3
“No person who denies the existence of a Supreme Being shall hold any office in this state.” Mississippi State Constitution — Article 14, Section 265
INTRODUCTION
Mississippi is one of several states with statutes in their constitutions violating Article VI, Paragraph 3 of the United States Constitution. While such statutes would assumedly be overturned should an atheist gain legal standing to challenge them (i.e., by being elected to public office in such a state), the existence of such laws demonstrates that institutionalized discrimination against atheists continues to be endemic to the culture of the modern United States. Such discrimination, which occurs frequently in familial, social, educational, and work contexts (Cragun et al. Reference Cragun, Kosmin, Keysar, Hammer and Nielson2012; Hammer et al. Reference Hammer, Cragun, Hwang and Smith2012), should be of increasing concern as the non-religious have become a larger and more visible minority group of more than 55 million adults (Lipka Reference Lipka2015). While self-described “atheists” remain a minority within even the ranks of the non-religious (Lipka Reference Lipka2015), prejudice against the non-religious could affect any of the so-called “nones.” In one study, participants reported equally negative evaluations of “atheists” and people “without a belief in God” on a variety of dimensions (e.g., good-bad, immoral-moral; Swan and Heesacker Reference Swan and Heesacker2012). Another series of studies revealed atheists to be the least desirable marriage partners for respondents’ children out of a broad range of demographic groups as well the group least likely to share a respondent's vision for America (e.g., Edgell, Gerteis, and Hartmann Reference Edgell, Gerteis and Hartmann2006).
Strong and widely shared negative perceptions of atheists also make them less attractive as political candidates than individuals from other minority groups (e.g., Jones Reference Jones2012). Indeed, experimental research has demonstrated that the large deficits in political support for atheists are accompanied by elevated levels of negative feelings, especially distrust (Franks and Scherr Reference Franks and Scherr2014). During the 2016 Democratic Primary, leaked Democratic National Committee e-mails indicated that party insiders were considering a smear campaign against party outsider Bernie Sanders centered on his apparent lack of religiosity (East Reference East2016). While it has been demonstrated that atheists face a substantial degree of prejudice in the political arena, research has yet to examine factors that could increase their political acceptance. Identifying such factors has important implications for the non-religious and atheist communities, who may be able to make gains in political influence and representation. Accordingly, the overarching goal of the current research is to investigate potential influence strategies that could remediate anti-atheist prejudice in the high-stakes context of election politics.
FACTORS UNDERLYING ANTI-ATHEIST PREJUDICE
Perceived Lack of Morality
Christians have been shown to hold wildly inaccurate perceptions of the moral values of atheists (Simpson and Rios Reference Simpson and Rios2016a), and atheists are often perceived as less moral than religious believers (Hout and Fischer Reference Hout and Fischer2002), especially among those who are highly religious (Galen et al. Reference Galen, Smith, Knapp and Wyngarden2011). Consequently, immoral behavior is perceived as being consistent with atheism (Wright and Nichols Reference Wright and Nichols2014), which may be especially important in the political realm where various moral concerns have been shown to predict voting behavior (Franks and Scherr Reference Franks and Scherr2015).
Distrust
Related to a perceived lack of morality is the problem of a deficit in trust, which seems to be central to anti-atheist prejudice (Gervais, Shariff, and Norenzayan Reference Gervais, Shariff and Norenzayan2011). If atheists are not trustworthy and are a threat to cooperation and the social order, however, one would expect that chaos would reign in nations where atheism is very prevalent. This is not the case. The number of atheists worldwide has been estimated at over half a billion (Zuckerman Reference Zuckerman2008), and the highest prevalence of atheism per capita is in the modern secular democracies (e.g., The Netherlands, Norway; Gilani, Shahid, and Zuettel Reference Gilani, Shahid and Zuettel2012). These countries are the most cooperative, peaceful, and healthy nations in human history (Pinker Reference Pinker2012). Peaceful and prosperous conditions engender trust of atheists (Gervais and Norenzayan Reference Gervais and Norenzayan2012) and acceptance of their presence in political office. In Denmark, a country where secular institutions are adept at fighting poverty, crime, and disease, only 8% of respondents agree that atheists are unfit for public office compared to 95% in theocratic Pakistan, where poverty, crime, and disease are far more prevalent (Norris and Inglehart Reference Norris and Inglehart2004).
While religious affiliation and religious belief is on the decline in the United States (Lipka Reference Lipka2015), the vast majority of adults in the United States (including a majority of individuals who identify as religiously unaffiliated) believe in a god. As a disfavored out-group, atheists in the United States are likely to be viewed as less trustworthy than members of the religious majority (e.g., Guth, Levati, and Ploner Reference Guth, Levati and Ploner2006; Platow et al. Reference Platow, Foddy, Yamagishi, Lim and Chow2012; Tanis and Postmes Reference Tanis and Postmes2005). A lack of trust has repeatedly been implicated as a primary motivator of anti-atheist prejudice (e.g., Franks and Scherr Reference Franks and Scherr2014; Gervais, Shariff, and Norenzayan Reference Gervais, Shariff and Norenzayan2011; Giddings and Dunn Reference Giddings and Dunn2016), which is particularly problematic because trustworthiness is valued above all other personal characteristics (Cottrell, Neuberg, and Li Reference Cottrell, Neuberg and Li2007) and it is a predictor of actual election outcomes (e.g., Chen, Jing, and Lee Reference Chen, Jing and Lee2014).
Anti-Atheist Prejudice in Politics
From 1978 to 2012, the percentage of United States respondents in Gallup polls who would support a “generally well-qualified” candidate from their preferred political party who happened to be atheist started low (40% in 1978) and increased to only 54% by 2012 (Jones Reference Jones2012). Consequently, there was only one openly non-religious Representative out of 535 members in the 114th United States Congress (January 2015 – January 2017).
While these polls provide a good preliminary glimpse at the obstacles American atheists face in the political arena, two important factors limit the conclusions that can be drawn from the above data. First, respondents could only answer “yes” or “no” when asked about their willingness to vote for each type of candidate. This is important because people who answer “no” may vary in the strength of that conviction and therefore in their ability to be influenced by other factors. Second, respondents were asked only to consider the religious group membership of the target candidates. This is an important issue because it has been shown that evaluations of an atheist target become less negative when additional information about the target is provided (although evaluations were still less positive than those of a similarly well-defined Christian target; Swan and Heesacker Reference Swan and Heesacker2012).
Experimental research has begun to extend the findings of these polling data. For example, one study demonstrated that an ostensibly real atheist political candidate suffered large deficits in voting intentions from Christian voters across the political spectrum (Franks and Scherr Reference Franks and Scherr2014). This lack of voter support for the atheist was accompanied by a wide-range of negative feelings, particularly distrust, which may explain why political candidates feel compelled to emphasize their religious values in order to elicit trust and enhance perceptions of their morality (Clifford and Gaskins Reference Clifford and Gaskins2015).
Research Overview
Thus far, research examining anti-atheist prejudice in the political domain has been focused on measuring deficits in political support for atheists. That is, research has examined factors that undermine atheists’ political aspirations, but it has yet to address factors that could increase political acceptance of atheists. The most important intended contribution of the current research is to identify conditions in which anti-atheist prejudice in politics may be reduced. Based on basic social psychology, this research focuses on the ability of several factors that may decrease deficits in political support for atheists.
Secular Morality (Study 1)
Secular values that do not directly reference atheism, such as protecting children from religious harms and removing religious bias from science education, have been shown to elicit strong support among religiously affiliated liberals (Franks Reference Franks2015). Because a perceived lack of morality may be motivating distrust and prejudicial attitudes toward atheists, one way to increase political acceptance of an atheist candidate among religiously affiliated voters may be for a candidate to express broadly popular secular moral values. The degree to which an individual perceives atheists to have a moral concern for the welfare of others was found to be especially predictive of anti-atheist prejudice (Simpson and Rios Reference Simpson and Rios2016b). Study 1 tests the hypothesis that an atheist political candidate would receive a boost in voting intentions, trustworthiness ratings, and morality ratings from participants when the candidate expressed policy positions reflecting popular secular moral values.
The Bandwagon Effect (Study 2)
The bandwagon effect refers to the increased probability than any single individual will adopt the majority opinion as their own as the majority opinion becomes increasingly popular. In effect, the bandwagon can be considered a form of social conformity (e.g., Asch Reference Asch1955) occurring on a very large scale. Exit poll survey data from three British general elections in the 1970's and 1980's indicate that election polling trends have an effect on voting behavior such that candidates leading in the polls receive an additional boost on election day from their pre-existing popularity (McAllister and Studler Reference McAllister and Studlar1991). Similar results have been found in the context of the 1992 United States Presidential general election (Morwitz and Pluzinski Reference Morwitz and Pluzinski1996) and the 1996 United States Republican Party primary election (Mehrabian Reference Mehrabian1998). Study 2 tests the hypothesis that an atheist candidate would receive increased voting intentions and ratings of trustworthiness and perceived morality from participants who receive information suggesting that the atheist candidate has a sizable lead in the polls.
Paradoxical Religiosity (Study 3)
According to the paradoxical religiosity effect (Sumaktoyo, Ottati, and Untoro Reference Sumaktoyo, Ottati and Untoro2016), voters differentiate between a candidate's personal religiosity and their political religiosity. Research germane to this effect has shown that while displaying personal religious convictions seems to bolster voter support for a candidate, candidates who espouse injecting their religion into public policy suffer overall losses in voter support (Sumaktoyo, Ottati, and Untoro Reference Sumaktoyo, Ottati and Untoro2016). Since prior research has indicated that some voters are less likely to vote for an Evangelical Christian than an atheist (Campbell, Green, and Layman Reference Campbell, Green and Layman2011), Study 3 examines the effect of an atheist running against a theocratic Christian (i.e., someone who wants public policy to be based on their religious doctrines). It is hypothesized that describing an atheist political candidate's opponent as a theocrat would increase intentions to vote for the atheist candidate, but because the manipulation would change attitudes about the opposing candidate rather than the atheist, the atheist candidate would not elicit increased trustworthiness or morality ratings.
STUDY 1 (SECULAR MORALITY)
Method
Participants
Workers recruited from Amazon's Mechanical Turk online workforce (N = 219; 58% male) participated for a small cash payment. The sample was predominantly young (M = 31.2 years old, SD = 10.4) and white (78%) with a median income falling between $23,000 and $45,999. Half of the sample (50%) had at least an associate's degree. More than two-thirds (68%) were religiously affiliated. On a scale of 1 (very liberal) to 7 (very conservative), the mean self-reported political orientation was 3.28 (SD = 1.32). While the demographics of this and other mTurk samples reported in this article are substantially different that the demographics of the population of likely voters, research has demonstrated that effects found in mTurk convenience samples are similar to those found using representative samples (Mullinix et al. Reference Mullinix, Leeper, Druckman and Freese2015).
Experimental Design
Participants were randomly assigned to a 2 (target candidate faith status: Christian vs. atheist) × 2 (secular policy information: present or absent) experimental design, and participant religious status (affiliated vs. unaffiliated) was included as a non-experimental factor. The primary dependent variable for this study was voting intentions for the target candidate, while perceptions of the candidate as moral and trustworthy were secondary dependent variables. Candidate faith status was manipulated by labeling otherwise identical candidates as “Christian” or “atheist.” According to the anti-atheist prejudice hypothesis, Christian participants would indicate higher morality and trust ratings and greater likelihood of voting for the target candidate over the one that is described as Christian rather than as atheist. Secular policy information was manipulated by either including or not including a message from the candidate outlining a secular policy agenda that emphasized protection of children from religious harms, evidence based science education, and support for gay rights. According to the secular morality hypothesis, secular policy information would increase morality and trust ratings and vote confidence when for the target candidate regardless of the candidate's faith status.
Materials and Procedure
Participants responded to a post recruiting United States registered voters and read informed consent information on mTurk before agreeing to participate and following a web link to the experimental materials on surveymonkey.com.
Participants reported sex, age, ethnicity, religious affiliation, income, education, and sexual orientation. They also reported their political orientation on 1 (very liberal) to 7 (very conservative) scale. After completing these initial measures, participants read about Greg, a candidate for United States Congress. He was presented as an ostensibly real person and described as either a Christian or an atheist. His gender (male), ethnicity (white), sexual orientation (heterosexual), family situation (married with two kids), and other background information (e.g., hobbies and education) were always held constant. Further, participants were randomly assigned to either read a series of statements by the candidate espousing secular policies that have previously been shown to be amenable to religiously affiliated individuals (Franks Reference Franks2015; see Appendix A) or to not read such statements. Participants then indicated their perceptions of Greg as moral and trustworthy and reported their intentions to vote for Greg. All dependent measures were on 9-point Likert scales. Finally, a series of manipulation check questions were asked to ensure that participants understood information provided about the candidate.
Data Analytical Approach
The primary analyses for Study 1 were 2 (participant religious status: affiliated vs. unaffiliated) × 2 (candidate faith status: Christian vs. atheist) × 2 (secular policy information: present vs. absent) between subjects analyses of variance (ANOVAs) with morality ratings, trust ratings, and vote intentions as the dependent variables. Levene's test for homogeneity of variances failed for all dependent variables in Study 1, even when applying logarithmic transformations. No further attempts to normalize data were made as ANOVA is robust against violations of homogeneity, and raw scores are reported. In order to ensure the effectiveness of random assignment, I analyzed whether the treatment groups varied demographically and found no differences across treatment groups in terms of gender, race, religion, education, household income, or politics.
Results
Manipulation Checks
Participants responded to two multiple choice questions asking them to correctly identify the candidate's faith status and whether or not they received information on his policies. All participants answered these questions correctly.
ANOVAS
Morality Ratings
The main effect of secular policy information was significant indicating that participants perceived the candidate as more moral when the information was present (M = 6.46, SD = 1.91) than when it was absent (M = 5.72, SD = 1.87), F(1, 212) = 7.60, p < 0.01, η p 2 = 0.04. The results of this main effect for all Study 1 dependent variables can be viewed simultaneously in Table 1. There were no significant interactions involving the secular policy information variable (ps > 0.18), which indicates that the effect of introducing such information was equally effective regardless of the faith statuses of the candidate or participant. The main effect of participant faith status was significant, F(1, 212) = 8.48, p < 0.01, η p 2 = 0.04. However, that main effect was qualified by a significant Participant Faith Status × Candidate Faith Status interaction, F(1, 212) = 24.52, p < 0.001, η p 2 = 0.10. Exploring this interaction indicated that religiously affiliated participants perceived the candidate to be more moral when he was described as Christian (M = 6.72, SD = 1.35) than as atheist (M = 5.20, SD = 1.67), while religiously unaffiliated participants perceived the candidate as more moral when he was described as atheist (M = 7.35, SD = 1.20) than as Christian (M = 6.33, SD = 1.56). Religiously affiliated participants’ moral preference for the Christian candidate over the atheist candidate was large (D = 1.00), while religiously unaffiliated participants’ moral preference for the atheist candidate over the Christian candidate was medium-to-large (D = 0.72).
The main effect of secular policy information on Study 1 dependent variables
a (p < 0.01).
Trust Ratings
The main effect of secular policy information was significant indicating that participants perceived the candidate as more trustworthy when the information was present (M = 6.25, SD = 1.83) than when it was absent (M = 5.51, SD = 1.57), F(1, 212) = 8.75, p < 0.01, η p 2 = 0.04. There were no significant interactions involving the secular policy information variable (ps > 0.42), which indicates that the effect of introducing such information was equally effective regardless of the faith statuses of the candidate or participant. The main effect of participant faith status was significant, F(1, 212) = 19.28, p < 0.001, η p 2 = 0.08. However, that main effect was qualified by a significant Participant Faith Status × Candidate Faith Status interaction, F(1, 212) = 18.94, p < 0.001, η p 2 = 0.08. Exploring this interaction indicated that religiously affiliated participants perceived the candidate to be more trustworthy when he was described as Christian (M = 6.25, SD = 1.59) than as atheist (M = 5.04, SD = 1.83), while religiously unaffiliated participants perceived the candidate as more trustworthy when he was described as atheist (M = 7.13, SD = 1.26) than as Christian (M = 6.10, SD = 1.34). Religiously affiliated participants’ trust preference for the Christian candidate over the atheist candidate was medium-to-large (D = 0.71), while religiously unaffiliated participants’ trust preference for the atheist candidate over the Christian candidate was large (D = 0.80).
Vote Intentions
The main effect of secular policy information was significant indicating that participants reported higher voting intentions when the information was present (M = 6.57, SD = 1.82) than when it was absent (M = 5.75, SD = 1.40), F(1, 212) = 11.27, p < 0.01, η p 2 = 0.05. There were no significant interactions involving the secular policy information variable (ps > 0.34), which indicates that the effect of introducing such information was equally effective regardless of the faith statuses of the candidate or participant. The main effects of both participant faith status, F(1, 212) = 8.48, p < 0.01, η p 2 = 0.04, and candidate faith status, F(1, 212) = 5.71, p < 0.05, η p 2 = 0.03, were also significant. However, these main effects were qualified by a significant Participant Faith Status × Candidate Faith Status interaction, F(1, 212) = 25.99, p < 0.001, η p 2 = 0.11. Exploring this interaction indicated that religiously affiliated participants reported higher voting intentions for the candidate when he was described as Christian (M = 6.92, SD = 1.35) than as atheist (M = 5.20, SD = 1.67), while religiously unaffiliated participants reported higher voting intentions when he was described as atheist (M = 7.09, SD = 1.38) than as Christian (M = 6.30, SD = 1.45). Religiously affiliated participants’ voting preference for the Christian candidate over the atheist candidate was large (D = 1.13), while religiously unaffiliated participants’ voting preference for the atheist candidate over the Christian candidate was medium (D = 0.56).
Discussion
Study 1 provides initial evidence that informational manipulations such as the inclusion of secular policy statements can increase voter support for atheist political candidates, perhaps in large part due to concurrent increases in perceptions that the target candidate was honest and trustworthy. This interpretation of the results is consistent with the secular morality hypothesis as well as the findings of Simpson and Rios (Reference Simpson and Rios2016b), who showed that anti-atheist prejudice was more pronounced among those perceiving atheists to lack a moral concern for the welfare of others. While Study 1 offers encouragement for groups and individuals who wish to increase support for atheist candidates, these results are not without qualification. In support of the anti-atheist prejudice hypothesis, results of Study 1 show a large amount of anti-atheist prejudice among the Christians in this sample, who were recruited from a subject pool that is overall younger and more liberal than the typical voter. Religiously affiliated participants strongly preferred the Christian candidate, while unaffiliated participants strongly preferred the atheist candidate, and this pattern of results is consistent with prior research findings in this same area (Franks and Scherr Reference Franks and Scherr2014). Additionally, a Christian candidate who espoused the same broadly popular secular values received a similar boost in morality ratings, trustworthiness ratings, and voting intentions as an atheist, suggesting that a Christian candidate with popular secular values would maintain an advantage over an atheist candidate with popular secular values.
STUDY 2 (BANDWAGON EFFECT)
Method
Participants
Workers recruited from Amazon's Mechanical Turk online workforce (N = 115; 54% male) participated for a small cash payment. The sample was predominantly young (M = 33.3 years old, SD = 9.9) and white (80%) with a median income falling between $23,000 and $45,999. The majority of the sample (57%) had at least an associate's degree. Most participants (72%) were religiously affiliated. On a scale of 1 (very liberal) to 7 (very conservative), the mean self-reported political orientation was 3.28 (SD = 1.32).
Experimental Design
Participants were randomly assigned to a 2 (candidate faith status: Christian vs. atheist) × 2 (bandwagon: target candidate trails in the polls vs. target candidate leads in the polls) experimental design with participant faith status (religiously affiliated vs. religiously unaffiliated) added as a third non-experimental factor. The primary dependent variable, faith status manipulation, and operationalization of the anti-atheist prejudice hypothesis were all the same as in Study 1. Bandwagon was manipulated by including information that either describes the candidate as leading in the polls by a large margin (75% to 25%) or as trailing in the polls by a large margin (25% to 75%). According to the bandwagon hypothesis, participants would report increased voting intentions, morality ratings, and trust ratings for a candidate described as leading in the polls versus a candidate described as trailing in the polls.
Materials and Procedure
As in Study 1, Participants responded mTurk posting, provided informed consent, reported politically relevant demographic variables as well as their political orientation, and read the same consistent background information about our target candidate Greg, who was either described as Christian or atheist. Additionally, participants were randomly assigned to either read that Greg was leading by 50 percentage points in the polls (75% to 25%) or that he was trailing by 50 percentage points in the polls (25% to 75%) before indicating their perceptions of Greg as moral and trustworthy and reporting their intentions to vote for Greg in the same manner as Study 1. Finally, a series of manipulation check questions were asked to ensure that participants understood information provided about the candidate.
Data Analytical Approach
The primary analyses for Study 2 were 2 (participant religious status: affiliated vs. unaffiliated) × 2 (candidate faith status: Christian vs. atheist) × 2 (bandwagon: target candidate trails in the polls vs. target candidate leads in the polls) between subjects ANOVAs with morality ratings, trust ratings, and vote confidence as the dependent variables. Levene's test for homogeneity of variances failed for all dependent variables in Study 2, even when applying logarithmic transformations. No further attempts to normalize data were made as ANOVA is robust against violations of homogeneity, and raw scores are reported. In order to ensure the effectiveness of random assignment, I analyzed whether the treatment groups varied demographically and found no differences across treatment groups in terms of gender, race, religion, education, household income, or politics.
Results
Manipulation Checks
Participants responded to two multiple choice questions asking them to correctly identify the candidate's faith status and whether he was leading or trailing in the polls. All participants answered these questions correctly.
ANOVAs
Morality Ratings
The main effect of bandwagon was significant indicating that participants perceived the candidate as more moral when he was described as leading in the polls (M = 6.67, SD = 1.54) than when he was described as trailing in the polls (M = 5.74, SD = 2.02), F(1, 107) = 9.98, p < 0.01, η p 2 = 0.09. The results of this main effect for all Study 2 dependent variables can be viewed simultaneously in Table 2. There were no significant interactions involving the bandwagon variable (ps > 0.10), which indicates that the bandwagon effect was equally effective regardless of the faith statuses of the candidate or participant. The main effects of participant faith status and candidate faith status were both non-significant (ps = 0.16 and 0.34, respectively). However, the Participant Faith Status × Candidate Faith Status interaction was significant, F(1, 107) = 9.21, p < 0.01, η p 2 = 0.08. Exploring this interaction indicated that religiously affiliated participants perceived the candidate to be more moral when he was described as Christian (M = 7.05, SD = 1.29) than as atheist (M = 5.71, SD = 2.04), while religiously unaffiliated participants perceived the candidate as more moral when he was described as atheist (M = 6.25, SD = 1.24) than as Christian (M = 5.73, SD = 2.15). Religiously affiliated participants’ moral preference for the Christian candidate over the atheist candidate was large (D = 0.79), while religiously unaffiliated participants’ moral preference for the atheist candidate over the Christian candidate was small (D = 0.29).
The main effect of bandwagon on Study 2 dependent variables
a (p < 0.01);
b (p < 0.001).
Trust Ratings
The main effect of bandwagon was significant indicating that participants perceived the candidate as more trustworthy when he was described as leading in the polls (M = 7.10, SD = 1.38) than when he was described as trailing in the polls (M = 5.78, SD = 1.64), F(1, 107) = 22.94, p < 0.001, η p 2 = 0.18. There were no significant interactions involving the bandwagon variable (ps > 0.23), which indicates that the bandwagon effect was equally effective regardless of the faith statuses of the candidate or participant. The main effects of participant faith status and candidate faith status were both non-significant (ps = 0.41 and 0.96, respectively). However, the Participant Faith Status × Candidate Faith Status interaction was significant, F(1, 107) = 7.62, p < 0.01, η p 2 = 0.07. Exploring this interaction indicated that religiously affiliated participants perceived the candidate to be more trustworthy when he was described as Christian (M = 6.79, SD = 1.42) than as atheist (M = 5.97, SD = 1.79), while religiously unaffiliated participants perceived the candidate as more trustworthy when he was described as atheist (M = 7.00, SD = 1.37) than as Christian (M = 6.38, SD = 1.77). Religiously affiliated participants’ trust preference for the Christian candidate over the atheist candidate was medium (D = 0.51), while religiously unaffiliated participants’ trust preference for the atheist candidate over the Christian candidate was small-to-medium (D = 0.39).
Vote Intentions
The main effect of bandwagon was significant indicating that participants reported greater voting intentions for the target candidate when he was described as leading in the polls (M = 6.85, SD = 1.54) than when he was described as trailing in the polls (M = 5.20, SD = 1.43), F(1, 107) = 37.99, p < 0.001, η p 2 = 0.26. There were no significant interactions involving the bandwagon variable (ps > 0.27), which indicates that the bandwagon effect was equally effective regardless of the faith statuses of the candidate or participant. The main effects of participant faith status and candidate faith status were both non-significant (ps = 0.27 and 0.54, respectively). However, the Participant Faith Status × Candidate Faith Status interaction was significant, F(1,107) = 10.97, p < 0.01, η p 2 = 0.09. Exploring this interaction indicated that religiously affiliated participants reported higher vote intentions for the target candidate when he was described as Christian (M = 6.36, SD = 1.64) than as atheist (M = 5.57, SD = 1.67), while religiously unaffiliated participants reported higher vote intentions for the target candidate when he was described as atheist (M = 6.81, SD = 1.68) than as Christian (M = 5.85, SD = 1.69). Religiously affiliated participants’ moral preference for the Christian candidate over the atheist candidate was medium (D = 0.48), while religiously unaffiliated participants’ moral preference for the atheist candidate over the Christian candidate was also medium (D = 0.57).
Discussion
The results of Study 2 support the bandwagon hypothesis, and show a consistent bandwagon effect regardless of the target candidate's faith status. This lends further support to the notion that the political fortunes of atheists, while dire seeming based on public polling, can be improved upon when potential voters have more information about a candidate than just their faith status. Study 2 may suggest that popular incumbents who routinely win reelection by large margins can “come out of the closet” as atheists and still remain in office. These findings are not necessarily as optimistic for openly atheist candidates who have not yet established popularity with their constituents. And, yet again, the same manipulation that increased voting intentions for an atheist candidate was equally effective at increasing voting intentions for a Christian candidate, which indicates that an atheist would still be at a disadvantage to a similarly situated Christian. Additionally, Study 2 supports the anti-atheist prejudice hypothesis and is consistent with Study 1 and a large body of existing research finding a high degree of anti-atheist prejudice among Christians, and replicated this effect in a relatively young and liberal sample of Christian registered voters. The Participant Faith Status × Candidate Faith Status interaction is also consistent with Study 1 and previous research (e.g., Franks and Scherr Reference Franks and Scherr2014) demonstrating that religiously affiliated participants prefer a Christian candidate to an atheist one, while religiously unaffiliated participants prefer an atheist candidate to a Christian one.
STUDY 3 (PARADOXICAL RELIGIOSITY)
Method
Participants
Workers recruited from Amazon's Mechanical Turk online workforce (N = 124; 61% male) participated for a small cash payment. The sample was predominantly young (M = 33.9 years old, SD = 11.2), white (76%), and heterosexual (94%) with a median income falling between $23,000 and $45,999. The majority of the sample (56%) had at least an associate's degree. Most (71%) were religiously affiliated. On a scale of 1 (very liberal) to 7 (very conservative), the mean self-reported political orientation was 3.14 (SD = 1.49).
Experimental Design
Participants were randomly assigned to one of two levels of the paradoxical religiosity variable whereby our target candidate's political opponent was described as either a full-on theocrat (i.e., want public policy to be based on his religious views) or a relatively moderate Christian. Additionally, participants were again divided up by faith status to create a second non-experimental independent variable.
Materials and Procedure
As in Studies 1 and 2, Participants responded mTurk posting, provided informed consent, reported political relevant demographic variable as well as their political orientation, and read the same consistent background information about our target candidate Greg, who in Study 3 unlike the other studies was always described an atheist. The opposing candidate's views on religion and government were randomly manipulated to either be very theocratic (see Appendix B) or relatively moderate (see Appendix C) but still identifying him as Christian. Participants then indicated their perceptions of Greg as moral and trustworthy and reported their intentions to vote for Greg. All dependent measures were on 9-point Likert-type scales. Finally, a manipulation check question was asked to ensure that participants correctly recalled what was stated about the opponent's political/religious views.
Data Analytic Approach
The primary analyses for Study 3 were 2 (participant faith status: religiously affiliated vs. religiously unaffiliated) × 2 (paradoxical religiosity: moderate opponent vs. theocratic opponent) between subjects ANOVAs with morality ratings, trust ratings, and vote confidence as the dependent variables. Levene's test for homogeneity of variances failed for all dependent variables in Study 3, even when applying logarithmic transformations. No further attempts to normalize data were made as ANOVA is robust against violations of homogeneity, and raw scores are reported. In order to ensure the effectiveness of random assignment, I analyzed whether the treatment groups varied demographically and found no differences across treatment groups in terms of gender, race, religion, education, household income, or politics.
Results
Manipulation Checks
Participants responded to two multiple choice questions asking them to correctly recall the information given to them about the opponent candidate's political/religious views. All participants answered this question correctly.
ANOVAs
Morality
The main effect of paradoxical religiosity was non-significant indicating that participants’ moral perceptions of the target atheist candidate were not affected by whether or not the opponent supported religious based public policy, F(1, 174) = 0.05, p = 0.83, η p 2 = 0.00. The main effect of participant faith status was significant, indicating that religiously unaffiliated participants (M = 7.44, SD = 1.21) viewed the target atheist candidate as more moral than religiously affiliated participants did (M = 6.52, SD = 2.13), F(1, 174) = 8.49, p < 0.01, η p 2 = 0.05. The interaction effect was non-significant (p = 0.26).
Trust
The main effect of paradoxical religiosity was non-significant indicating that participants’ trust perceptions of the target atheist candidate were not affected by whether or not the opponent supported religious based public policy, F(1, 174) = 0.01, p = 0.93, η p 2 = 0.00. The main effect of participant faith status was significant, indicating that religiously unaffiliated participants (M = 7.23, SD = 1.46) viewed the target atheist candidate as more trustworthy than religiously affiliated participants did (M = 6.29, SD = 2.18, F(1, 174) = 8.11, p < 0.01, η p 2 = 0.05. The interaction effect was non-significant (p = 0.93).
Vote Intentions
The main effect of paradoxical religiosity was significant indicating that participants were more likely to vote for the target atheist candidate when his opponent was described as a theocrat (M = 6.98, SD = 1.34) than as a relatively moderate Christian (M = 6.24, SD = 1.61), F(1, 174) = 6.95, p < 0.01, η p 2 = 0.04. The results of this main effect for all Study 3 dependent variables can be viewed simultaneously in Table 3. The main effect of participant faith status was also significant, indicating that religiously unaffiliated participants (M = 7.25, SD = 1.07) were more likely to vote for the target atheist candidate than religiously affiliated participants were (M = 6.52, SD = 2.13), F(1, 174) = 13.83, p < 0.001, η p 2 = 0.07. The interaction effect was non-significant (p = 0.32).
The main effect of paradoxical religiosity on Study 3 dependent variables
a (p < 0.01).
Discussion
Whereas Studies 1 and 2 examine the effects of factors directly related to an atheist candidate (i.e., faith status, popularity, and policy positions) on vote intentions, Study 3 examines the paradoxical religiosity effect in regards to an atheist's political opposition. The results of Study 3 support the paradoxical religiosity hypothesis by demonstrating that voters were more supportive of an atheist candidate when his opponent held extreme theocratic views about the role of religion in government. Because the independent variable manipulated perceptions of the atheist candidate's opponent rather than the atheist candidate himself, changes in voting intentions were not accompanied by changes in perceptions of morality and trustworthiness for the target candidate. Study 3 and previous examinations of the paradoxical religiosity effect (Sumaktoyo, Ottati, and Untoro Reference Sumaktoyo, Ottati and Untoro2016) suggest that an atheist may garner more support by contrasting their secular views with an opponent's theocratic views. This is important because political candidates are often hesitant to criticize or negatively characterize the religious views of their opponents.
GENERAL DISCUSSION
The current series of research studies aim to discover possible conditions under which support for atheist candidates may be increased among samples of registered voters in the United States. Despite being a nation with a secular constitution and an increasingly large non-religious population, openly atheist and non-religious individuals are vastly under-represented in government at all levels. In the long term, research efforts that focus on increasing the proportion of under-represented groups (e.g., women, Lesbian, Gay, Bisexual, Transgender, Queer individuals, ethnic minorities, atheists) in our government serve to increase its ability to serve its constituents. Additionally, research efforts focusing on strengthening the secular basis of our government serve to decrease the possibility it trammels the interests of those who are most often harmed when religion encroaches on government (e.g., women, Lesbian, Gay, Bisexual, Transgender, Queer individuals, ethnic minorities, atheists).
These studies provide evidence for three factors that may improve the electability of atheists: secular morality (Study 1), the bandwagon effect (Study 2), and the paradoxical religiosity effect (Study 3). In an additional optimistic finding for non-religious Americans with political aspirations, religiously unaffiliated voters, who represent a quickly growing demographic, preferred an atheist candidate to a Christian one. While this finding may be consistent with a prior study using the same subject pool (mTurk workers in the United States; Franks and Scherr Reference Franks and Scherr2014), other studies have found anti-atheist prejudice to be robust among the non-religious and religious alike (Giddings and Dunn Reference Giddings and Dunn2016). Whether this is an effect of the particular sample or differing dependent measures is an issue for future research.
Studies 1 and 2, which manipulated the faith status of the target candidate, also share two findings that were not-so-optimistic for the political fortunes of atheists: (1) a similarly situated Christian candidate still outperformed the atheist candidate among religiously affiliated participants and (2) religiously unaffiliated participants preferred the atheist candidate to the Christian candidate.
Implications
The results of these studies provide several optimistic implications for atheists. First, it was found that a well-crafted, broadly appealing secular message can increase support for a political candidate among religious affiliated individuals. The secular values expressed by hypothetical candidates across these studies were found to have a large effect in terms of improving perceptions of atheist (and Christian) candidates, which demonstrates that there is a shared moral language that can allow moral coalitions to form between atheists and many members of the faith community.
While the experimental manipulations improved perceptions of atheists, Christian candidates received the same benefits from each of these manipulations. Thus, it would seem that, all other things being equal, an atheist candidate will receive less support than a Christian candidate from religiously affiliated voters. More optimistically, a popular incumbent who can intelligently and passionately communicate a secular message may now be able to confidently come out as an atheist and hope to be reelected, and this may be especially true as the nation becomes less religiously affiliated and the secular challenge to the religious right grows (e.g., Hansen Reference Hansen2011). Religiously unaffiliated participants in these studies favored an atheist candidate over a Christian one, which is another reason for atheists to be optimistic about their future political prospects unless this effect is an artifact of the non-religious individuals on mTurk being in some relevant way different from non-religious Americans in general, and the overall sample did report on average more liberal political orientation than Americans in general (see Twenge et al. Reference Twenge, Honeycutt, Prislin and Sherman2016).
Limitations and Future Directions
The secular policy items, while found to be popular in research conducted to inform presentations given to non-profit groups (e.g., Franks Reference Franks2015) and useful in helping design studies such as those reported here, have not yet been published in a peer-reviewed format. Additionally, since several types of secular issues were addressed by these policies (e.g., gay rights, child protection) we cannot be sure to what degree each of the various statements affected outcomes in Study 1. Attempts should be made to validate and standardize a measure of secular public policy attitudes in a peer-reviewed format and determine the relative effects of declaring various secular values on perceptions of atheists.
The bandwagon manipulation used in Study 2 was not indicative of a competitive race. Future research should find the most realistic but effective method of manipulating poll results. Perhaps talking about odds of winning rather than percentage of votes would be more powerful with realistic numbers (e.g., a 10% lead in the polls may not sound as impressive as a 99% chance of victory, but those numbers may both reflect the same polling data).
Study 3 successfully divorced changes in voting intentions for an atheist candidate from changes in perceptions of morality and trustworthiness. It may be assumed that there would have been negative changes in the morality and trust ratings of the opposing candidate, but such measures were not taken. Future research may want to examine how paradoxical religiosity affects perceptions of morality and trustworthiness.
Additionally, the current research only looked at remedying one side of the non-religious under-representation equation. The fact that atheists are aware of the negative perceptions other people hold of them (Brewster et al. Reference Brewster, Hammer, Sawyer, Eklund and Palamar2016; Saroglou, Yzerbyt, and Kaschten Reference Saroglou, Yzerbyt and Kaschten2011; Sumerau and Cragun Reference Sumerau and Cragun2016) and that they perceive social distance between themselves and the religious majority (Guenther Reference Guenther2014) may dissuade them from seeking public office in the first place. Other interventions that may contribute to solving the problem of atheist under-representation may include identifying methods of encouraging atheists to openly pursue political careers.
Finally, this research lacks any investigation of intersectional identities. All candidates presented by the study were male, white, and heterosexual. Future research could explore how being atheist interacts with membership in other historically disenfranchised groups in determining other peoples’ perceptions.
CONCLUSION
The United States is a nation founded on secular principles such as the separation of religion and government. It is disconcerting that there is a de facto (and sometimes de jure) religious test to hold public office in most areas of the country. Research must continue to focus on understanding and reducing hostile feelings towards atheists, not only for the benefit of atheists but also for the benefit of non-atheists who fall into other groups harmed by religious bias in public policy. Women, persons of minority sexual orientations and gender identities, children, and religious and ethnic minorities are all negatively affected when governments lack strong secular voices. Increasing the representation of secular and atheist individuals will yield positive results for individuals in these other groups and reinvigorate the principle of church / state separation that is essential in a healthy democracy.
APPENDIX A
When pressed to elaborate on the relationship between religion and public policy, Greg stated, “I do not intend to impose my worldview on others. The government has no business telling people what to believe and no business telling people how to behave as long as their behavior doesn't harm others. On that note, I think it is important to prevent religiously motivated behaviors that bring harm to children and jeopardize the future of this country. Most people would agree that it is right for the government to prevent parents from performing female genital mutilation on preteen girls for religious purposes. We know that there are limits to what religiously motivated behaviors are protected by The Constitution. However, parents in 31 states are still allowed to engage in a practice known as “faith healing,” which is really just medical neglect of children for religious reasons resulting in suffering and death that is easily preventable. This, to me, is no different from human sacrifice, and it must stop. We can also no longer allow gay children and adolescents to be subjected to degrading and emotionally abusive conversion “therapy” that futilely attempts to change a healthy and natural part of their identities. Furthermore, the government funds adoption centers that deny children good homes with loving same-sex parents for religious reasons. Not only does this harm thousands of children waiting for adoption, it also uses the tax dollars of gay taxpayers to discriminate against same-sex couples. Finally, religiously motivated denial of science harms our children and our nation in the global economy. It is largely because we are increasingly disparaging science and devaluing science literacy that we have gone from the world leader in educational achievement in science and technology 40 years ago to falling behind many other Western democracies in these areas today. In order to restore American exceptionalism, our education policy must prevent religious ideologues from holding our future hostage with false controversies.”
APPENDIX B
When asked what he felt about Greg's position on religion and public policy, his opponent responded, “My opponent does not believe that The Bible should be used to make public policy. I say it is the only source of our morality, the only source of our salvation, and the only place we should look when making our laws. We need The Bible in the classroom so that evolutionists cannot trick our children into thinking their ancestors were monkeys. We need The Bible to protect marriage from gays and lesbians who want to change what this sacred concept means. We need God in our government because He is the foundation of all morality and authority.”
APPENDIX C
When asked what he felt about Greg's position on religion and public policy, his opponent responded, “My opponent does not believe that The Bible should be used to make public policy. I say that public policy should reflect the morality of the public, and Americans tend to get their morality from Christianity as do many lawmakers. I try not to force my religion on anyone. But it is not possible to completely divorce personal morality from public decision making, and we should not expect religious individuals to do so. We need to respect the freedom of religious communities and religious individuals to make their own decisions regarding whose moral values define public policy and education policy.”
