This is an open access article distributed under the terms of the Creative Commons
Attribution License (Understanding the complex manifestations of and contributions to sexual stigma is crucial in helping to prevent discrimination, harassment, and violence toward sexual minorities. In recent years, there has been a rapid and dramatic shift in policies and public opinion regarding sexual minorities in the United States (Herek, 2015). There has been a growing number of laws and policies prohibiting discrimination based on sexual orientation across the United States. Concurrently, Americans have shown increased support for sexual minorities on legal and policy issues including marriage equality, parenting, and employment discrimination and an increase in positive and accepting attitudes towards sexual minorities (Herek, 2015). However, these opinion changes and the decreasing moral righteousness associated with discriminatory policies belie the widespread discrimination that sexual minorities still face. The legitimacy of sexual stigma has been increasingly contested in recent years, but it remains robust in certain institutional and ideological systems. One such ideology is conservatism. In this research, we examine the contributions of political ideology to discrimination by examining the mediating role of prejudice toward sexual minorities. This research builds upon and extends previous work in important ways: we extend the scholarship linking political ideology to sexual stigma beyond attitudes to examine discrimination in an intern employment context; we broaden the research on employment bias to examine the role of political ideology in discrimination against gay men; and we empirically examine Herek’s (2009a) framework for conceptualizing reactions to sexual minorities by testing the mediating role of prejudice in linking political ideology and discrimination.
Sexual Stigma
When people possess characteristics that might lead others to consider them deviant, limited, or otherwise undesirable, they are said to carry a stigma. Goffman (1963) organized the nature of stigmatizing attributes into three general categories: blemishes of individual character, abominations of the body, and tribal stigma of race, nation, and religion. Whatever the source of the “undesired differentness” (p. 5), those who carry the “mark” belong to low status/power social categories and are more likely to encounter negative expectations and reactions from others (Major & O’Brien, 2005). Sexual stigma refers to “the stigma attached to any non-heterosexual behavior, identity, relationship, or community. In other words, it is socially shared knowledge about homosexuality’s devalued status relative to heterosexuality” (Herek, 2009a; p. 67). Sexual stigma is a social phenomenon and, like other stigmas, is not only manifest in individuals, but also in institutions (Herek, 2007); this can have important implications for discrimination against sexual minorities.
Individual Manifestations of Sexual Stigma: Prejudice and Discrimination
At the individual level, sexual minorities are often the target of negative stereotypes, prejudice, and discrimination. According to Herek’s (2007) framework, the individual manifestations of sexual stigma range from felt stigma (expectations regarding being the target of stigma enactments), to internalized stigma (acceptance of stigma), to enacted stigma (behavioral expression of negative stigma). Sexual prejudice, or negative attitudes toward sexual minorities, stems from internalizing sexual stigma and accepting it as part of one’s value system. This prejudice can motivate discrimination (Bernat, Calhoun, Adams, & Zeichner, 2001; Parrott & Zeichner, 2005). Sexual minorities regularly confront discrimination ranging from shunning to harassment and violence and experience inequities in many realms of their life from employment, to health care, to education (Robinson & Ferfolja, 2001; Sabin, Riskind, & Nosek, 2015; Tilcsik, 2011; Weichselbaumer, 2003). In addition to emanating from individual prejudice, discrimination against sexual minorities can stem from ideological manifestations of sexual stigma.
Ideological Manifestations of Sexual Stigma
Beyond individuals, sexual stigma also manifests in institutions, such as religious institutions and the law, and structural systems including ideological systems. The sexual stigma embedded within society’s institutions and structures is termed heterosexism (Herek, 2007). Importantly, heterosexism can serve as the foundation for the individual manifestations of prejudice and discrimination (Herek, 2009a). Like other structural or institutional manifestations of sexual stigma, ideological systems can legitimize the status and power disadvantages associated with sexual minority groups. Ideologies can provide people a way to justify the mistreatment and rejection of stigmatized individuals. Justification ideologies are frameworks of beliefs, values, and moral standards through which to view the world (Crandall, 2000). There are numerous ideologies that can enable people to justify discriminatory treatment ranging from the Protestant work ethic (a belief that links hard work with morality and success) to social dominance orientations (individual preferences for hierarchies amongst social groups; Crandall, 2000). These frameworks can enable justification by promoting attributions of responsibility and blame for the stigma or by promoting beliefs in social hierarchies. Some ideologies encourage both hierarchy beliefs and attributions that serve to justify stigmatization; one such ideology is political ideology.
Political Ideology
Political ideology plays an important role in guiding people’s responses to and justifications of social and political matters. Ideology encompasses both attributional processes and belief systems that can promote the justification of stigma. Conservatives are more likely than liberals to justify the way things are, to believe that existing social, economic, and political arrangements are fair and legitimate, and to accept inequality (Jost, Federico, & Napier, 2009; Jost & Hunyady, 2005; Jost, Nosek, & Gosling, 2008; Kerlinger, 1984; Rasinski, 1987; Skitka & Tetlock, 1993; Van der Toorn, Nail, Liviatan, & Jost, 2014). Through valuing hierarchies among groups of people as well as making greater internal attributions for individual outcomes, political conservatism can serve to justify stigmatization (Crandall, 2000).
Although both liberals and conservatives have been shown to demonstrate prejudice and discrimination (Nail, Harton, & Decker, 2003), research shows robust and reliable associations between conservatism and prejudice against stigmatized group members (Hodson & Busseri, 2012). In the United States, political conservatism has been shown to predict prejudice and discrimination against a range of stigmatized groups from racial and ethnic minorities (Hoyt & Goldin, 2016; Hurwitz & Peffley, 1992), to overweight individuals (Crandall, 1995), and to lesbians and gay men (Herek, 2009a; Van der Toorn, Jost, Packer, Noorbaloochi, & Van Bavel, 2017).
The Current Research
Discrimination against sexual minorities and the associated claims of moral righteousness has decreased in the United States (Herek, 2007). In this research, we aim to elucidate limitations associated with the recent decreased legitimacy of sexual stigma by demonstrating the robust legitimacy of heterosexism within conservative political ideology systems. Moreover, we aim to demonstrate the process through which these ideological systems promote discrimination by focusing on prejudice. We examine the critical role of sociocultural manifestations of stigma (ideology) in promoting individual sexual stigma (prejudice) and ultimately discrimination against sexual minorities. Thus, in the current research, we merge Herek’s framework for conceptualizing sexual stigma with political ideology theoretical perspectives to empirically test the claim that the sexual stigma that is legitimized in the conservative political ideology system can produce discrimination against sexual minorities largely through individual negative attitudes toward sexual minorities, or sexual prejudice.
In this work, we aim to extend the literature in important ways. Whereas extant work looking at the link between political ideology and sexual stigma has focused on the association between conservatism and both prejudicial and policy attitudes (Herek, 2009a; Hodson & Busseri, 2012), in this work we go beyond attitudes to examine discrimination in an intern employment context. Although significant research shows bias in employment contexts (Cunningham, Sartore, & McCullough, 2010; Tilcsik, 2011; Weichselbaumer, 2003), no studies have looked at the role of political ideology in these effects. Finally, working from Herek’s (2009a) framework arguing that heterosexism is the foundation for individual manifestations of sexual stigma, this is the first work to empirically examine the role of prejudice in mediating the link between political ideology and discrimination.
In this research, we examine the role of political conservatism and sexual prejudice in predicting discriminatory evaluations of applicants that are presented with a sexual stigma or not. We test our predictions by employing a modified Goldberg (1968) experimental paradigm such that participants were presented nearly identical information in an intern applicant evaluation context, however, cues to sexual stigma were either present or absent. We signaled sexual stigma in different ways across studies with the applicant volunteering for an LGBTQ program, self-identifying as a gay athlete, or both. Across four studies, we test the role of political ideology and sexual prejudice in evaluating an applicant with sexual stigma. Data were collected from individuals living in the United States in the summer of 2016 (Studies 1 through 3) and the spring of 2017 (Study 4). Specifically, we hypothesize that those who score relatively higher (vs. lower) on conservative political ideology will more negatively evaluate the applicant with sexual stigma, those who score relatively higher (vs. lower) on sexual prejudice will more negatively evaluate the applicant with sexual stigma, and we predict that prejudice will mediate the link between ideology and evaluation (see Figure 1). More specifically, given the work showing that political conservatism predicts prejudice against a range of stigmatized groups, we propose that ideology will directly predict sexual prejudice, however, the link between prejudice and applicant evaluation will depend upon whether the applicant being evaluated has a sexual stigma or not. That is, we propose a second stage and direct effect moderated mediational model such that ideology will predict applicant evaluation indirectly and conditionally through prejudice, moderated by whether the applicant has a sexual stigma or not, controlling for the moderation of the direct effect of ideology on evaluation.
Figure 1
The conceptual second stage and direct effect moderated mediational model. (Process Model 15)
Study 1
Method
Participants
We recruited participants from Amazon’s Mechanical Turk (Buhrmester, Kwang, & Gosling, 2011; Weinberg, Freese, & McElhattan, 2014) to voluntarily participate in a study examining evaluations of athletic internship applicants. One hundred ninety-two participants completed the studyi with 27 participants failing either the attention check and/or the manipulation check leaving a final sample size of 165 (48.5% female; 51.5% male; median age = 33; 74.5% White; 7.3% African American; 0.6% Native American; 6.1% Asian American; 9.1% Latino/a; and 2.4% other)ii.
Procedure and Manipulations
Participants were asked to imagine that they are a staff member of a Division 1 university football program tasked with evaluating applications for an internship position (all intern evaluation materials can be found in Appendix A). They first read the position description for the university football player operations intern position. The description summarized the opportunity, essential duties and responsibilities of the position, required skills and abilities needed, and desired education and experience for the applicant. Next, participants read a cover letter from an applicant named John Peters, and read one of two resumes as part of his application for the player operations internship. The resumes were identical in every way except for volunteer experience with either Play 60 (control condition), an organization which encourages kids to get outside and play for at least 60 minutes per day, or Campus Pride’s “Out to Play Project” (stigma condition), an organization that promotes comfortable participation in sports for LGBTQ athletes with presentations meant to address anti-LGBTQ slurs, bias, and conduct in sports. After reading the cover letter and resume, participants responded to manipulation check items, evaluated the applicant, and then responded to the sexual prejudice, political ideology, and demographic measures.
Measures
Political ideology
Using a 7-point scale (strongly liberal to strongly conservative), participants responded to three questions indicating their overall political ideology, their political ideology on social issues and their political ideology on fiscal issues (α = .92). In assessing people’s beliefs systems regarding socio-economic systems, a useful distinction can be drawn between social and fiscal ideology. Although social and economic ideologies are distinct, they are reliably correlated within Western capitalist societies (Jost & Hunyady, 2005) and they similarly predict important measures such as system justification, resistance to change, and prejudice (Everett, 2013); thus, we combined the three items into a single measure of ideologyiii. Across the studies, the average political ideology scores were slightly liberal (see means in Table 1). However, the standard deviations reveal there was a combination of both liberal and conservative participants in the samples. Higher scores represent more conservative ideologies.
Applicant evaluation
Using 7-point scales, participants evaluated the applicants using a 10-item measure assessing the applicants in terms of their competence, hireability, fit, and willingness to mentor the applicant. Sample items included “How likely is it that the applicant has the necessary skills for this job?” and “How likely would you be to invite the applicant to interview for the player operations internship?” (see Appendix B for full measure; α = .92). Higher scores represent more positive evaluations.
Sexual prejudice
We assessed participants’ heterosexism with two scales: the 5-item Attitudes Toward Gay Men Scale (Herek, 1994) and the 15-item Homosexism Scale (Hansen, 1982; see Appendix B). Participants responded to items using a scale ranging from 1 (Strongly Disagree) to 7 (Strongly Agree). Both measures of sexual prejudice have been well-validated (Grey, Robinson, Coleman, & Bockting, 2013; Herek & Capitanio, 1999). The items were highly correlated thus we combined them into one measure of prejudice (α = .97). Higher scores represent greater prejudice.
Manipulation check
Participants were asked to indicate which organization the applicant volunteered for: Play 60, Campus Pride’s Out to Play, or Upward Sports. Additionally, they were asked to identify the applicant’s sexual orientation: Gay, Straight, or I don’t know.
Attention check. Participants were asked to give a particular response (e.g., “Please respond strongly agree”) at two points during the survey.
Results and Discussion
Twenty-seven participants either failed the manipulation check by failing to correctly identify whether the applicant volunteered for Out to Play or Play 60 (n = 18, 13 in the control condition and 5 in the stigma condition) or failed both attention checks (n = 9); these participants (13.64%) were excluded from analyses leaving a final sample size of 165iv. Both the prejudice and evaluation variables were skewed across all studies. A negative reciprocal (inverse) transformation was performed to reduce the positive skewness yet preserve order among values in the sexual prejudice variable and a log transformation after reflection was used to reduce the negative skewness of the evaluation variable. We report results from both (untransformed presented first, transformed presented in parentheses) and, for ease of interpretation, we report untransformed values in our descriptive data including graphs. Table 1 presents the scale means, standard deviations, and intercorrelations for the study variables.
Table 1
Scale Means, Standard Deviations, and Intercorrelations for Studies
| Dependent Variable | M | SD | 1 | 2 |
|---|---|---|---|---|
| Study 1 | ||||
| 1. Political ideology | 3.34 | 1.60 | ||
| 2. Applicant evaluation | 6.09 | .87 | -.25*** | |
| 3. Sexual prejudice | 1.93 | 1.18 | .51*** | -.37*** |
| Study 2 | ||||
| 1. Political ideology | 3.17 | 1.63 | ||
| 2. Applicant evaluation | 5.76 | 1.07 | -.25*** | |
| 3. Sexual prejudice | 2.02 | 1.24 | .50** | -.40*** |
| Study 3 | ||||
| 1. Political ideology | 3.37 | 1.61 | ||
| 2. Applicant evaluation | 6.02 | .83 | -.17* | |
| 3. Sexual prejudice | 1.97 | 1.20 | .58*** | -.39*** |
| Study 4 | ||||
| 1. Political ideology | 3.60 | 1.71 | ||
| 2. Applicant evaluation | 5.74 | 1.11 | -.23*** | |
| 3. Sexual prejudice | 2.68 | 1.68 | .55*** | -.41*** |
Note. Untransformed data reported.
*p ≤ .05. **p ≤ .01. ***p ≤ .001.
We conducted our analyses using Hayes’ (2013) PROCESS macro in SPSS. This macro uses an ordinary least squares or logistic regression-based path analytical framework to analyze statistical models involving moderation, mediation, and their combination, termed conditional process modeling. To test the hypothesis that political ideology will moderate the relationship between sexual stigma and applicant evaluation, we conducted a simple moderation analysis employing Model 1, mean centering the variables and regressing applicant evaluation on political ideologies, sexual stigma, and their interaction (1 = Stigma, -1 = Control). First, although political ideology significantly predicted applicant evaluation [B = -.13, p = .002; (B = .07, p < .001)] condition did not [B = .006, p = .922; (B = -.02, p = .490)]. Additionally, there was a significant interaction between political ideology and condition [B = -.10, p = .024; (B = .04, p = .052)]. The conditional effect of political ideology on applicant evaluation revealed that ideology significantly predicted evaluation in the sexual stigma condition [B = -.22, p < .001; (B = .10, p < .001)] and not in the control condition [B = -.03, p = .694; (B = .02, p = .421)]. Alternatively, liberal participants (-1 SD) trended toward rating the applicant in the stigma condition more positively than the applicant in the control condition [B = .16, p = .088; (B = -.08, p = .058)], and conservatives (+1 SD) showed the opposite pattern of rating the applicant in the stigma condition lower, though non-significantly so, than the applicant in the control condition [B = -.15, p = .128; (B = .04, p = .373)]. See Figure 2.
Figure 2
Study 1: Applicant evaluation as a function of political ideology and sexual stigma condition.
We conducted a similar analysis to test the hypothesis that sexual prejudice will moderate the relationship between sexual stigma and applicant evaluation; We employed Model 1, mean centering the variables and regressing applicant evaluation on sexual prejudice, sexual stigma, and their interaction (1 = Stigma, -1 = Control). First, although prejudice significantly predicted applicant evaluation [B = -.25, p < .001; (B = .47, p < .001)], condition did not [B = .007, p = .909; (B = -.02, p = .546)]. Additionally, the interaction between prejudice and condition was marginally significant [B = -.12, p = .033; (B = .19, p = .095)]. The conditional effect of sexual prejudice on applicant evaluation revealed that prejudice significantly predicted evaluation in the sexual stigma condition [B = -.36, p < .001; (B = .63, p < .001)] and not in the control condition [B = -.12, p = .175; (B = .26, p = .13)]. Alternatively, those low in prejudice (-1 SD) non-significantly trended toward rating the applicant in the stigma condition more positively than the applicant in the control condition [B = .12, p = .146; (B = -.07, p = .105)], and those high in prejudice (+1 SD) showed the opposite pattern of rating the applicant in the stigma condition lower, though non-significantly so, than the applicant in the control condition [B = -.13, p = .151; (B = .03, p = .447)]. See Figure 3.
Figure 3
Study 1: Applicant evaluation as a function of sexual prejudice and sexual stigma condition.
Next, we tested the proposed second stage and direct effect moderated mediational model; this model allows the effect of the mediator, prejudice, on the outcome, applicant evaluation, to be moderated by cues to sexual stigma and includes the moderation of condition on the direct effect of ideology on evaluation (PROCESS Model 15, Hayes, 2013; see Figure 1). We mean-centered variables and generated bootstrap-based confidence intervals (95%) for the indirect effect by taking 5,000 samples from the original data set. See Table 2a and 2b for results from the moderated mediational analyses. Results from the second stage and direct effect moderated mediational model suggested a significant indirect effect in the stigma condition but not in the control condition. However, the indices of moderated mediation were not significant suggesting that the indirect effects were not significantly different from one another (Hayes, 2015).
Finally, we conducted exploratory analyses examining a second stage moderated mediational analyses that does not control for the moderated direct effect (Process Model 14; see Figure 4). In Model 14, we fix the effect of ideology on evaluation, controlling for the mediator, to be independent of stigma condition. This assumption affects the estimation of the effect of prejudice on evaluation. In Model 15, we are not making such an assumption. Instead, we are assuming the effect of ideology on evaluation, independent of the mediator, to be a linear function of stigma condition. The estimation of the effect of prejudice on evaluation depends on what assumptions we make about the direct effect of ideology on evaluation. If the moderated mediation is significant in Model 14, it suggests that prejudice mediates the effect of ideology on evaluation differently across stigma conditions, and this occurs assuming that the direct effect of ideology on evaluation is not dependent on stigma condition. Once again, we mean-centered variables and generated bootstrap-based confidence intervals (95%) for the indirect effect by taking 5,000 samples from the original data set (see Table 2a and 2b). Tests of this model also showed a significant indirect effect in the stigma condition but not in the control condition. In this case the index of moderated mediation from the non-transformed data was significant with 95% confidence, but the confidence interval for the index from the transformed data included zero. Although the pattern was consistent with predictions, we cannot definitely say the indirect effect of prejudice depends on condition.
Interestingly, the majority of participants in the stigma (95%) and control (93%) conditions responded I don’t know when questioned about the applicant’s sexual orientation. Thus, it is important to note that these stigma effects were driven either by unconfirmed hunches about sexual orientation and/or by courtesy stigma effects, that is, stigma that occurs from being associated with stigmatized people (Goffman, 1963).
Figure 4
The second stage moderated mediational model without the moderated direct effect. (Process Model 14).
Table 2a
Results From Process Model 15 for Studies 1 Through 3
| Factor | Study 1
|
Study 2
|
Study 3
|
||||
|---|---|---|---|---|---|---|---|
| B | p | B | p | B | p | ||
| Ideology—Prejudice | UD | .37 | < .001 | .38 | < .001 | .43 | < .001 |
| TD | .09 | < .001 | .08 | < .001 | .09 | < .001 | |
| Prejudice—Evaluation | UD | -.22 | < .001 | -.28 | < .001 | -.31 | < .001 |
| TD | .36 | .006 | .54 | < .001 | .65 | .002 | |
| Ideology—Evaluation | UD | -.05 | .281 | -.06 | .227 | .06 | .146 |
| TD | .03 | .144 | .02 | .462 | -.03 | .164 | |
| Condition—Evaluation | UD | .01 | .867 | -.29 | < .001 | -.20 | < .001 |
| TD | -.02 | .487 | .12 | < .001 | .08 | .003 | |
| PrejudicexCond—Evaluation | UD | -.07 | .251 | -.19 | .006 | -.13 | .029 |
| TD | .09 | .498 | .14 | .330 | .13 | .279 | |
| IdeologyxCond—Evaluation | UD | -.07 | .148 | -.08 | .169 | -.02 | .721 |
| TD | .03 | .183 | .04 | .069 | .02 | .383 | |
| Conditional Direct effects | |||||||
| Stigma condition | UD | -.11 | .064 | -.15 | .098 | .05 | .471 |
| TD | .06 | .039 | .07 | .089 | -.01 | .774 | |
| Control condition | UD | .03 | .712 | -.00 | .970 | .08 | .194 |
| TD | -.00 | .965 | -.02 | .475 | -.04 | .113 | |
| Conditional Indirect effects | |||||||
| Ind eff | 95% CI | Ind eff | 95% CI | Ind eff | 95% CI | ||
| Stigma condition | UD | -.11 | [-.23, -.02] | -.19 | [-.29, -.11] | -.19 | [-.32, -.10] |
| TD | .04 | [.00, .08] | .06 | [.03, .09] | .07 | [.04, .12] | |
| Control Condition | UD | -.05 | [-.12, .01] | -.05 | [-.14, .02] | -.08 | [-.14, -.02] |
| TD | .02 | [-.01, .07] | .03 | [.00, .07] | .05 | [.02, .08] | |
| Index of Moderated Mediation | |||||||
| Index | 95% CI | Index | 95% CI | Index | 95% CI | ||
| UD | -.06 | [-.18, .05] | -.14 | [-.27, -.04] | -.11 | [-.24, -.00] | |
| TD | .02 | [-.03, .07] | .02 | [-.02, .07] | .02 | [-.02, .07] | |
Note. UD = Results from untransformed data; TD = Results from transformed data.
Table 2b
Results From Process Model 14 for Studies 1 Through 3
| Factor | Study 1
|
Study 2
|
Study 3
|
||||
|---|---|---|---|---|---|---|---|
| B | p | B | p | B | p | ||
| Ideology—Prejudice | UD | .37 | < .001 | .18 | < .001 | .44 | < .001 |
| TD | .09 | < .001 | .08 | < .001 | .09 | < .001 | |
| Prejudice—Evaluation | UD | -.21 | < .001 | -.29 | < .001 | -.30 | < .001 |
| TD | .35 | .007 | .59 | < .001 | .65 | < .001 | |
| Ideology—Evaluation | UD | -.06 | .223 | -.05 | .357 | .06 | .149 |
| TD | .03 | .115 | .01 | .764 | -.03 | .176 | |
| Condition—Evaluation | UD | .01 | .843 | -.29 | < .001 | -.20 | < .001 |
| TD | -.02 | .465 | .12 | < .001 | .08 | .003 | |
| PrejudicexCond—Evaluation | UD | -.12 | .029 | -.24 | < .001 | -.14 | .003 |
| TD | .19 | .091 | .29 | .018 | .19 | .058 | |
| Ideology—Evaluation (Direct effect) | UD | -.06 | .223 | -.05 | .357 | .06 | .149 |
| TD | .03 | .115 | .01 | .764 | -.03 | .176 | |
| Conditional Indirect effects | |||||||
| Ind eff | 95% CI | Ind eff | 95% CI | Ind eff | 95% CI | ||
| Stigma Condition | UD | -.12 | [-.23, -.03] | -.22 | [-.32, -.14] | -.20 | [-.32, -.11] |
| TD | .05 | [.02, .09] | .07 | [.05, .11] | .08 | [.05, .12] | |
| Control Condition | UD | -.03 | [-.08, .03] | -.04 | [-.12, .03] | .08 | [-.14, -.02] |
| TD | .01 | [-.02, .04] | .03 | [-.00, .06] | .04 | [.02, .07] | |
| Index of Moderated Mediation | |||||||
| Index | 95% CI | Index | 95% CI | Index | 95% CI | ||
| UD | -.09 | [-.20, -.00] | -.18 | [-.30, -.08] | -.12 | [-.25, -.02] | |
| TD | .03 | [-.01, .08] | .05 | [.01, .09] | .04 | [.00, .08] | |
Note. UD = Results from untransformed data; TD = Results from transformed data.
Study 2
Study 1 provided initial support for the predictions that those who score relatively higher (vs. lower) on conservative political ideology will more negatively evaluate the applicant with sexual stigma and that those who score relatively higher (vs. lower) on sexual prejudice will more negatively evaluate the applicant with sexual stigma. It also provided limited initial support for the mediation predictions. With Model 15, although the indirect effect was significant in the stigma condition and not significant in the control condition, the confidence interval of the index of moderated mediation encompassed zero. Similar results were found with Model 14, although the index was significant with the untransformed data. In Study 2, we sought to again test our hypotheses using a different signal of sexual stigma. In addition to volunteering for an LGBTQ program, the applicant self-identifies as a gay athlete.
Method
Participants
Once again, we recruited participants from Amazon’s Mechanical Turk to voluntarily participate in a study examining evaluations of athletic internship applicants. One hundred eighty-two participants completed the studyv with 17 participants failing the attention checks and/or the manipulation check leaving a final sample size of 165 (50.3% female; 49.1% male; 0.6% other gender; median age = 31; 78.2% White; 6.7% African American; 9.1% Asian American; 5.5% Latino/a; and 0.6% other).
Procedure and Manipulations
The procedure for this study was similar to Study 1. This time, however, resumes were not shown and the cover letters differed from Study 1 in that they included information about prior work experience, volunteer experience, and education (see Appendix A). Participants saw one of two cover letters: in the stigma condition the applicant identified as a gay student athlete who worked for the LGBTQ organization Campus Pride’s Out to Play Project and in the control condition he identified himself as a student athlete working for the neutral organization Play 60. After reading the cover letter, participants answered the same manipulation check, applicant evaluation (α = .93), sexual prejudice (α = .96), and political ideology (α = .89) measures used in Study 1.
Results and Discussion
Ten participants failed the manipulation check by either not identifying the applicant in the stigma condition as gay (n = 8) or identifying the control condition applicant as gay (n = 2), and another 7 failed both attention checks; removing these participants (9.34%) left a final sample size of 165vi. Once again, transformations were performed to reduce skewness in the sexual prejudice and evaluation variables; we report results from both and, for ease of interpretation, we report untransformed values in our descriptive data including graphs. Table 1 presents the scale means, standard deviations, and intercorrelations for the study variables.
Once again, to test the hypothesis that sexual stigma condition will moderate the relationship between political ideology and applicant evaluation, we used Hayes’ PROCESS macro Model 1 with mean centered variables, regressing applicant evaluation on political ideologies, sexual stigma, and their interaction (1 = Stigma, -1 = Control). First, both political ideology [B = -.19, p < .001; (B = .06, p = .001)] and condition [B = -.31, p < .001; (B = .12, p < .001)] significantly predicted applicant evaluation such that conservatives evaluated the applicant more negatively than liberals and those in the stigma condition evaluated the applicant more negatively than those in the control condition. Moreover, there was a significant interaction (see Figure 5) between political ideology and condition [B = -.19, p < .001; (B = .07, p < .001)]. The conditional effect of political ideology on applicant evaluation revealed that ideology significantly predicted evaluation in the sexual stigma condition [B = -.42, p < .001; (B = .15, p < .001)] and not in the control condition [B = -.04, p = .488; (B = .01, p = .707)]. Alternatively, liberal participants (-1 SD) rated the applicant in both conditions similarly [B = .01, p = .961; (B = .00, p = .927)], however, conservatives (+1 SD) rated applicants in the stigma condition significantly lower than the applicant in the control condition [B = -.62, p < .001; (B = .23, p < .001)].
Figure 5
Study 2: Applicant evaluation as a function of political ideology and sexual stigma condition.
Next, we tested the hypothesis that sexual prejudice will moderate the relationship between sexual stigma and applicant evaluation. First, both prejudice [B = -.32, p < .001; (B = .61, p < .001)] and condition [B = -.29, p < .001; (B = .12, p < .001)] predicted evaluation. Additionally, there was a significant interaction between prejudice and condition [B = -.24, p < .001; (B = .29, p = .016)]. The conditional effect of sexual prejudice on applicant evaluation revealed that prejudice significantly predicted evaluation more strongly in the sexual stigma condition [B = -.62, p < .001; (B = .96, p < .001)] relative to the control condition [B = -.13, p = .100; (B = .38, p = .013)]. Alternatively, those low in prejudice (-1 SD) didn’t differ in their ratings of applicants across condition [B = -.03, p = .709; (B = .04, p = .324)], whereas those high in prejudice (+1 SD) rated the applicant in the stigma condition lower, than the applicant in the control condition [B = -.59, p < .001; (B = .20, p < .001)]. See Figure 6.
Figure 6
Study 2: Applicant evaluation as a function of sexual prejudice and sexual stigma condition.
Next, we tested the proposed moderated mediational model with Process Model 15. When results from the transformed and untransformed data differ, we interpret the most conservative, relative to our predictions, result (see Table 2a and 2b for results from the moderated mediational analyses). Results from this model showed a significant indirect effect in the stigma condition for both the transformed and the untransformed data, and also a significant indirect effect in the control condition for the transformed data only. Additionally, the index of moderated mediation was significant for the untransformed data but not for the transformed data. Hence, we obtained mixed evidence for our hypotheses. These findings suggest that there may also be an indirect effect of ideology on evaluation in the control condition, and the mixed findings with regard to the indices of moderated mediation suggest that the difference between the conditional indirect effects is not robust.
As in Study 1, we then explored the second stage moderated mediational analyses that does not control for the moderated direct effect (Process Model 14; see Figure 4). Tests of this model revealed a significant indirect effect in the stigma condition but not in the control condition. Additionally, the indices of moderated mediation were significant such that prejudice mediated the link between ideology and evaluation in the stigma, but not control, condition.
Study 3
The results from Study 2 provide additional support for our predictions that those who score higher (vs. lower) on conservative political ideology and sexual prejudice more negatively evaluate the applicant with sexual stigma. It also provided support for the second stage moderated mediation model (Figure 4). In this model, conservative political ideology indirectly predicted lower applicant evaluation through prejudice for the applicant with the sexual stigma, assuming that the direct effect of ideology on evaluation does not differ across stigma condition. In Study 2 we manipulated stigma with both volunteering for an LGBTQ program and self-identifying as gay. In this third study, we investigate whether these effects hold when the only signal to sexual stigma is self-identification as a sexual minority.
Method
Participants
We recruited participants from Amazon’s Mechanical Turk to voluntarily participate in a study examining evaluations of athletic internship applicants. One hundred eighty-three participants completed the studyvii with 14 participants failing the attention checks and/or the manipulation check leaving a final sample size of 169 (53.8% female; 45.6% male; 0.6% other gender; median age = 32; 74.6% White; 5.3% African American; 1.2% Native American; 9.5% Asian American; 8.9% Latino/a; and 0.6% other).
Procedure and Manipulations
The procedure for this study was the same as Study 2. However, this time both of the cover letters described the applicant as having volunteered for Play 60. Thus, in this study the only signal to sexual stigma was the applicant’s self-identification of being gay (gay athlete) or not (student athlete). After reading the cover letter, participants once again responded to the manipulation check as well as the measures of applicant evaluation (α = .92), sexual prejudice (α = .97), and political ideology (α = .91).
Results and Discussion
Seven participants failed the manipulation check by either not identifying the applicant in the stigma condition as gay (n = 5), identifying the control condition applicant as gay (n = 1), or not responding (n = 1), and another 7 failed both attention checks. Removing these participants (7.49%) left a final sample size of 169viii. Once again, transformations were performed to reduce skewness in the sexual prejudice and evaluation variables and we report results from both the untransformed and transformed data. Table 1 presents the scale means, standard deviations, and intercorrelations for the study variables.
Using Hayes’ (2013) PROCESS macro Model 1 with mean-centered variables, we regressed applicant evaluation on political ideologies, sexual stigma, and their interaction (1 = Stigma, -1 = Control). First, political ideology marginally predicted evaluation [B = -.07, p = .071; (B = .03, p = .068)] and condition significantly predicted applicant evaluation [B = -.18, p = .004; (B = .08, p = .009)] such that those in the stigma condition evaluated the applicant more negatively than those in the control condition. The interaction between political ideology and condition was not significant [B = -.07, p = .080; (B = .03, p = .126)] but its pattern did resemble the findings obtained in Study 2 (see Figure 7). That is, the association between political ideology and applicant evaluation was stronger in the sexual stigma condition such that more conservative ideologies predicted more negative evaluations.
Figure 7
Study 3: Applicant evaluation as a function of political ideology and sexual stigma condition.
Next, Using Hayes’ (2013) PROCESS macro Model 1 with mean-centered variables, we regressed applicant evaluation on prejudice, sexual stigma, and their interaction (1 = Stigma, -1 = Control). First, both prejudice [B = -.26, p < .001; (B = .55, p < .001)] and condition [B = -.18, p < .001; (B = .08, p = .004)] predicted evaluation. Additionally, there was a marginally significant interaction between prejudice and condition [B = -.13, p = .003; (B = .19, p = .056)], indicating that prejudice significantly predicted evaluation in both the sexual stigma condition [B = -.41, p < .001; (B = .76, p < .001)] and the control condition [B = -.13, p = .037; (B = .38, p = .005)]. Alternatively, those low in prejudice (-1 SD) did not differ in their ratings of applicants across condition [B = -.03, p = .709; (B = -.05, p = .497)], whereas those high in prejudice (+1 SD) rated the applicant in the stigma condition lower than the applicant in the control condition [B = -.59, p < .001; (B = -.35, p < .001)]. See Figure 8.
Figure 8
Study 3: Applicant evaluation as a function of sexual prejudice and sexual stigma condition.
Next, we tested the proposed second stage and direct effect moderated mediational model with Process Model 15 (see Table 2a and 2b). Results from this model showed a significant indirect effect in both the stigma condition and the control condition. Additionally, the index of moderated mediation from the transformed data was not significant, although results from untransformed data suggested significance as the confidence interval does not encompass zero. Finally, we examined the second stage moderated mediational analyses that does not control for the moderated direct effect (Process Model 14; see Figure 4). Tests of this model also revealed significant indirect effects in both the stigma condition and the control condition. However, the indices of moderated mediation were significant such that the indirect effect of prejudice was stronger in the stigma, relative to control, condition.
Study 4
In this final study, we sought to replicate the past three studies by manipulating the nature of the signaling across the first three studies in a single study: volunteering for an LGBTQ program, self-identifying as a gay athlete, or both.
Method
Participants
Once again, we recruited participants from Amazon’s Mechanical Turk; three hundred ninety-one participants completed the studyix with 57 participants failing the attention checks and/or the manipulation check leaving a final sample size of 334 (53.3% female; 46.7% male; median age = 35.5; 77.8% White; 8.7% African American; 0.9% Native American; 7.5% Asian American; 3.3% Latino/a; and 1.8% other).
Procedure and Manipulations
The procedure for this study was similar to that of Studies 1-3. Participants were instructed to imagine that they were evaluating applicants for a Division I College Football Operations Intern position. After reading a position description, participants read one of four cover letters from an applicant. These cover letters were identical in every way except for whether the applicant self-identified as gay or not, and whether they volunteered for the LGBTQ organization or the neutral organization. We fully crossed self-identification of sexual stigma (identification stigma, IS) or not (identification control, IC) with volunteer work with stigma-related organization (volunteer stigma, VS) or not (volunteer control, VC). After reading the cover letter, participants answered the same manipulation check questions, applicant evaluation (α = .94), sexual prejudice (α = .96), and political ideology (α = .94) measures used in Studies 1-3.
Results and Discussion
Forty-four participants failed the manipulation check by either not identifying the self-identified gay applicant as gay (n = 15 in the IS,VS condition and n = 14 in the IS,VC condition), identifying the applicant in the control condition (IC,VC) as gay (n = 8), or not responding (n = 1), and another 13 failed both attention checks; removing these participants (13.41%) left a final sample size of 368x. Once again, transformations were performed to reduce skewness in the sexual prejudice and evaluation variables and we report results from both the untransformed and transformed data. Table 1 presents the scale means, standard deviations, and intercorrelations for the study variables.
First, we sought to replicate the conditions tested in the previous studies. We conducted similar analyses comparing the conditions akin to those tested in the first three studies. Results are reported in Table 3a and 3b. Results from both Hayes’ (2013) PROCESS macro Model 1 analyses with mean-centered variables, regressing applicant evaluation on either political ideologies or prejudice, sexual stigma, and their interaction (1 = Stigma, -1 = Control) are at the top of the table.
Table 3a
Study 4: Results From Process Model 1, Examining Ideology and Condition on Evaluation, Comparing Conditions to Replicate Studies 1 Through 3
| Factor | Replicating Study 1
|
Replicating Study 2
|
Replicating Study 3
|
||||
|---|---|---|---|---|---|---|---|
| B | p | B | p | B | p | ||
| Political Ideology | UD | -.08 | .070 | -.11 | .024 | -.13 | .010 |
| TD | .03 | .075 | .04 | .033 | .05 | .019 | |
| Condition | UD | -.04 | .550 | -.19 | .018 | -.14 | .106 |
| TD | .00 | .903 | .07 | .031 | .03 | .393 | |
| Interaction | UD | -.07 | .119 | -.10 | .039 | -.12 | .017 |
| TD | .03 | .088 | .04 | .033 | .05 | .019 | |
| Conditional effects | |||||||
| Stigma Condition | UD | -.14 | .022 | -.20 | .003 | -.24 | < .001 |
| TD | .06 | .012 | .08 | .004 | .09 | < .001 | |
| Control Condition | UD | -.01 | .936 | -.01 | .939 | -.01 | .945 |
| TD | -.00 | .948 | -.00 | .946 | -.00 | .951 | |
| Liberal Participants | UD | .07 | .494 | -.02 | .830 | .07 | .579 |
| TD | -.05 | .259 | .00 | .988 | -.05 | .282 | |
| Cons. Participants | UD | -.16 | .129 | -.36 | .002 | -.35 | .005 |
| TD | .06 | .197 | .14 | .003 | .11 | .024 | |
Note. UD = Results from untransformed data; TD = Results from transformed data. Replicating Study 1: IC,VS (n = 93) vs. IC,VC (n = 76); Replicating Study 2: IS,VS (n = 83) vs. IC,VC (n = 76); Replicating Study 3: IS,VC (n = 82) vs. IC,VC (n = 76).
Table 3b
Study 4: Results From Process Model 1, Examining Prejudice and Condition on Evaluation, Comparing Conditions to Replicate Studies 1 Through 3
| Factor | Replicating Study 1
|
Replicating Study 2
|
Replicating Study 3
|
||||
|---|---|---|---|---|---|---|---|
| B | p | B | p | B | p | ||
| Political Ideology | UD | -.38 | < .001 | -.36 | < .001 | -.45 | < .001 |
| TD | .74 | < .001 | .58 | < .001 | .71 | .008 | |
| Condition | UD | -.03 | .601 | -.34 | .021 | -.15 | .055 |
| TD | -.00 | .977 | .13 | .033 | .04 | .247 | |
| Interaction | UD | -.15 | .008 | -.28 | .028 | -.22 | < .001 |
| TD | .32 | .003 | .36 | .112 | .31 | .013 | |
| Conditional effects | |||||||
| Stigma Condition | UD | -.51 | < .001 | -.50 | < .001 | -.66 | < .001 |
| TD | 1.02 | < .001 | .75 | < .001 | 1.00 | < .001 | |
| Control Condition | UD | -.22 | .011 | -.22 | .022 | -.22 | .026 |
| TD | .39 | .015 | .39 | .020 | .39 | .029 | |
| Low Prej Participants | UD | .13 | .141 | -.02 | .936 | .09 | .364 |
| TD | -.08 | .033 | .03 | .700 | -.04 | .337 | |
| High Prej Participants | UD | -.20 | .027 | -.67 | .002 | -.41 | < .001 |
| TD | .08 | .039 | .23 | .009 | .12 | .010 | |
Note. UD = Results from untransformed data; TD = Results from transformed data. Replicating Study 1: IC,VS (n = 93) vs. IC,VC (n = 76); Replicating Study 2: IS,VS (n = 83) vs. IC,VC (n = 76); Replicating Study 3: IS,VC (n = 82) vs. IC,VC (n = 76).
Testing the hypothesis that sexual stigma condition will moderate the relationship between political ideology and applicant evaluation, two of the three comparisons revealed significant interactions (see Figure 9). Conditional effects analyses revealed that across all three comparisons (including the comparison with the interaction that did not reach statistical significance), ideology significantly predicted evaluation in the sexual stigma condition and not in the control condition. Similarly, in the two significant comparisons, liberal participants (-1 SD) rated the applicant in both conditions similarly, whereas conservatives (+1 SD) rated the applicant in the stigma condition significantly lower than the applicant in the control condition.
Figure 9
Study 4: Applicant evaluation as a function of political ideology and stigma condition comparing conditions to replicate Study 1 (top; stigma from volunteer work vs. control), Study 2 (middle; stigma from both volunteer work and self-identification vs. control), and Study 3 (bottom; stigma from self-identification vs. control).
Next, testing the hypothesis that sexual stigma condition will moderate the relationship between prejudice and applicant evaluation, two of the three comparisons revealed significant interactions (see Figure 10). Conditional effects analyses revealed that across all three comparisons (including the comparison with the interaction that did not reach statistical significance), prejudice significantly predicted evaluation in the sexual stigma condition and in the control condition. However, the effect was stronger in the stigma condition, significantly so in two of the three comparisons. Additionally, across all comparisons, low prejudice participants (-1 SD) rated the applicant in both conditions similarly, whereas high prejudice participants (+1 SD) rated the applicant in the stigma condition significantly lower than the applicant in the control condition.
Figure 10
Study 4: Applicant evaluation as a function of prejudice and stigma condition comparing conditions to replicate Study 1 (top; stigma from volunteer work vs. control), Study 2 (middle; stigma from both volunteer work and self-identification vs. control), and Study 3 (bottom; stigma from self-identification vs. control).
Next, we tested the proposed second stage and direct effect moderated mediational model with Process Model 15 (see Table 4a and 4b). Generally, across the comparisons, the indirect effect of prejudice was significant in both conditions (the only exception being the analysis in the first comparison, the 95% CI encompassed zero in the control condition with the untransformed data). Although there were indirect effects in both conditions, significant indices of moderated mediation would suggest that the indirect effects are statistically different (Hayes, 2015). Three of the 6 indices of moderated mediation that we tested did not encompass zero. Although the pattern of CIs was consistent with the indirect effect being stronger in the stigma condition, that some of the CIs contained zero suggests that these findings are not robust.
Finally, we examined the second stage moderated mediational analyses that does not control for the moderated direct effect (Process Model 14; see Figure 4). Generally, across the comparisons, the indirect effect of prejudice was significant in both conditions (the only exception being the analysis in the first comparison, the 95% CI encompassed zero in the control condition with the untransformed data). Once again, the indirect effect was larger in the stigma relative to control conditions; the replication of Studies 1 and 3 revealed significant indices of moderated mediation but the replication of Study 2 did not. However, in the Study 2 comparisons the pattern of findings is consistent with predictions; yet, that they contained zero suggests that these findings are not robust.
Table 4a
Study 4: Results From Process Model 15 Comparing Conditions to Replicate Studies 1 Through 3
| Factor | Replicating Study 1
|
Replicating Study 2
|
Replicating Study 3
|
||||
|---|---|---|---|---|---|---|---|
| B | p | B | p | B | p | ||
| Ideology—Prejudice | UD | .33 | < .001 | .35 | < .001 | .35 | < .001 |
| TD | .07 | <.001 | .08 | < .001 | .08 | < .001 | |
| Prejudice—Evaluation | UD | -.42 | < .001 | -.39 | < .001 | -.49 | < .001 |
| TD | .81 | < .001 | .61 | < .001 | .74 | < .001 | |
| Ideology—Evaluation | UD | .06 | .194 | .03 | .488 | .05 | .303 |
| TD | -.02 | .273 | -.01 | .629 | -.01 | .595 | |
| Condition—Evaluation | UD | -.04 | .546 | -.17 | .019 | -.15 | .053 |
| TD | .00 | .966 | .07 | .030 | .04 | .239 | |
| PrejudicexCond—Evaluation | UD | -.12 | .062 | -.10 | .193 | -.19 | .011 |
| TD | .26 | .029 | .09 | .494 | .22 | .128 | |
| IdeologyxCond—Evaluation | UD | -.03 | .476 | -.05 | .279 | -.04 | .447 |
| TD | .02 | .376 | .03 | .179 | .03 | .211 | |
| Conditional Direct effects | |||||||
| Stigma condition | UD | .03 | .632 | -.02 | .811 | .02 | .829 |
| TD | -.01 | .828 | .02 | .579 | .01 | .630 | |
| Control condition | UD | .09 | .152 | .09 | .195 | .09 | .209 |
| TD | -.04 | .153 | -.04 | .172 | -.04 | .200 | |
| Conditional Indirect effects | |||||||
| Ind eff | 95% CI | Ind eff | 95% CI | Ind eff | 95% CI | ||
| Stigma condition | UD | -.17 | [-.27, -.10] | -.17 | [-.28, -.00] | -.24 | [-.38, -.12] |
| TD | .07 | [.04, .11] | .06 | [.02, .10] | .07 | [.03, .12] | |
| Control Condition | UD | -.09 | [-.18, .03] | -.10 | [-.19, -.04] | -.10 | [-.19, -.04] |
| TD | .04 | [.01, .06] | .04 | [.02, .07] | .04 | [.01, .07] | |
| Index of Moderated Mediation | |||||||
| Index | 95% CI | Index | 95% CI | Index | 95% CI | ||
| UD | -.08 | [-.19, .01] | -.07 | [-.20, .05] | -.14 | [-.29, -.02] | |
| TD | .02 | [.01, .08] | .01 | [-.03, .06] | .03 | [-.01, .08] | |
Note. UD = Results from untransformed data; TD = Results from transformed data. Replicating Study 1: IC,VS (n = 93) vs. IC,VC (n = 76); Replicating Study 2: IS,VS (n = 83) vs. IC,VC (n = 76); Replicating Study 3: IS,VC (n = 82) vs. IC,VC (n = 76).
Table 4b
Study 4: Results From Process Model 14 Comparing Conditions to Replicate Studies 1 Through 3
| Factor | Replicating Study 1
|
Replicating Study 2
|
Replicating Study 3
|
||||
|---|---|---|---|---|---|---|---|
| B | p | B | p | B | p | ||
| Ideology—Prejudice | UD | .33 | < .001 | .35 | < .001 | .35 | < .001 |
| TD | .07 | <.001 | .08 | < .001 | .08 | < .001 | |
| Prejudice—Evaluation | UD | -.42 | < .001 | -.39 | < .001 | -.49 | < .001 |
| TD | .80 | < .001 | .62 | < .001 | .75 | < .001 | |
| Ideology—Evaluation | UD | .06 | .182 | .04 | .449 | .05 | .304 |
| TD | -.02 | .252 | -.01 | .544 | -.01 | .579 | |
| Condition—Evaluation | UD | -.04 | .548 | -.17 | .018 | -.15 | .053 |
| TD | .00 | .970 | .07 | .030 | .04 | .241 | |
| PrejudicexCond—Evaluation | UD | -.14 | .012 | -.14 | .030 | -.22 | < .001 |
| TD | .31 | .004 | .18 | .113 | .31 | .012 | |
| Ideology—Evaluation (Direct effect) | UD | .06 | .182 | .04 | .449 | .05 | .304 |
| TD | -.02 | .252 | -.01 | .544 | -.01 | .579 | |
| Conditional Indirect effects | |||||||
| Ind eff | 95% CI | Ind eff | 95% CI | Ind eff | 95% CI | ||
| Stigma condition | UD | -.18 | [-.27, -.11] | -.19 | [-.29, -.09] | -.25 | [-.38, -.15] |
| TD | .07 | [.05, .11] | .06 | [.03, .10] | .08 | [.05, .12] | |
| Control Condition | UD | -.09 | [-.16, -.03] | -.09 | [-.16, -.02] | -.09 | [-.18, -.02] |
| TD | .03 | [.01, .06] | .03 | [.01, .06] | .03 | [.01, .06] | |
| Index of Moderated Mediation | |||||||
| Index | 95% CI | Index | 95% CI | Index | 95% CI | ||
| UD | -.09 | [-.18, -.01] | -.10 | [-.21, .01] | -.16 | [-.28, -.05] | |
| TD | .04 | [.02, .08] | .03 | [-.01, .07] | .05 | [-.01, .09] | |
Note. UD = Results from untransformed data; TD = Results from transformed data. Replicating Study 1: IC,VS (n = 93) vs. IC,VC (n = 76); Replicating Study 2: IS,VS (n = 83) vs. IC,VC (n = 76); Replicating Study 3: IS,VC (n = 82) vs. IC,VC (n = 76).
In sum, in this study we tested all three approaches to signaling sexual stigma employed across the first three studies: self-identifying as a gay athlete, volunteering for an LGBTQ program, and both. In our analyses, we tested each stigma condition compared to the control condition (mimicking the previous studies). Conditional effects analyses revealed that political ideology predicted internship applicant evaluation, but only when evaluating an applicant who had a sexual stigma; the interaction reached significance in two of the three comparisons. Also, in two of the comparisons prejudice predicted applicant evaluation more strongly in the sexual stigma relative to control condition. Moreover, the moderated mediation analyses using both Model 15 and Model 14 revealed that prejudice mediated both conditions. Although the indirect effect was larger in the stigma condition, the indices of moderated mediation reveal that these indirect effects are not significantly different across all comparisons. Finally, like in Study 1, the majority of participants in both conditions where the applicant does not self-identify as gay, the volunteer stigma condition (78%) and the control (96%) condition, responded I don’t know when questioned about the applicant’s sexual orientation. Thus, once again the discrimination found in the volunteer stigma condition occurred with the majority of participants indicating they didn’t know the applicant’s sexual orientation.
Mini Meta-Analysis
Finally, we conducted a mini-meta-analysis across all four of our studies (see Goh, Hall, & Rosenthal, 2016 for a primer on conducting meta-analyses on one’s own studies). All effect sizes must be independent, thus in Study 4 we included a single effect size for each hypothesis that tests the combined sexual stigma conditions to the control condition rather than testing each of the three comparisons to avoid the control participants being repeated across different effect sizes. Following the procedures of Hall and Goh (2017), we examined the interaction effects by conducting two meta-analyses looking at simple effects split by condition. We meta-analyzed the studies using fixed effects in which the mean effect size (i.e., mean correlation) was weighted by sample size. We first converted our t-values into Pearson’s correlations, which were Fisher’s z transformed for analyses and then converted back to correlations. Results from untransformed data are reported first.
Overall, political ideology predicted applicant evaluation in the stigma condition such that more conservative ideologies predicted lower evaluations [M r = -.35, Z = -8.02, p < .001 (M r = .34, Z = 7.64, p < .001)] but not in the control condition [M r = -.04, Z = -.68, p = .496 (M r = .04, Z = .72, p = .472)]. The average effect size of -.35 in the stigma condition across the four studies indicates a medium effect. Next, prejudice predicted applicant evaluation such that greater levels of prejudice predicted lower evaluations in both the stigma condition [M r = -.58, Z = -14.50, p < .001 (M r = .51, Z = 12.43, p < .001)] and the control condition [M r = -.20, Z = -3.65, p < .001 (M r = .24, Z = 4.54, p < .001)]. The average effect sizes indicate a large effect in the stigma condition and a small to medium effect in the control condition.
Finally, we meta-analyzed the indirect effects using the same approach of examining the indirect effect by condition. Given that we were meta-analyzing the same indirect models using the same measures of variables across datasets, we computed our mini meta-analysis using the estimates of the ab path and used inverse variance weighting from the standard errors of the indirect effects. For Model 15, the effect was significant across both condition, albeit stronger in the stigma condition [M ES = -.18, Z = -8.25, p < .001 (M ES = .06, Z = 8.37, p < .001)] relative to the control condition [M ES = -.07, Z = -4.13, p < .001 (M ES = .04, Z = 5.04, p < .001)]. Similar effects were found for Model 14 with the indirect effect being significant in both conditions, but stronger in the stigma condition [M ES = -.19, Z = -8.97, p < .001 (M ES = .07, Z = 9.46, p < .001)] relative to the control condition [M ES = -.06, Z = -3.45, p < .001 (M ES = .03, Z = 4.39, p < .001)]. Finally, we computed a mini meta-analysis on the indices of moderated mediation, once again using inverse variance weightings from the standard errors of the indices. The analyses revealed significant effects for both Model 15 [M ES = -.10, Z = -3.61, p < .001 (M ES = .02, Z = 2.43, p = .0153)] and Model 14 [M ES = -.11, Z = -3.94, p < .001 (M ES = .04, Z = 4.45, p < .001)] such that the indirect effect was stronger in the stigma relative to the control condition.
General Discussion
In this research, we sought to examine the role of heterosexism within political ideology systems and demonstrate the process through which these ideological systems promote discrimination by focusing on sexual prejudice. Across four studies, we tested the predictions that political ideology and sexual prejudice will predict the evaluation of an applicant with a sexual stigma and we tested a moderated mediational model such that conservative ideology negatively predicts the evaluation of those with a sexual stigma indirectly through prejudice. We varied the nature of the sexual stigma signal across the studies, from volunteering for an LGBTQ program, self-identifying as a gay athlete, or both. In the first three studies, we tested one of these stigma signals, and in the final study we examined all three. Although there were differences in results across studiesxi, the fact that consistent patterns were obtained bolsters our confidence in our theorizing. We followed up with a mini meta-analysis to determine the overall effects across the studies. Overall, the moderation hypotheses were supported. Political ideology predicted applicant evaluation in the stigma, but not control, condition such that more conservative ideologies predicted lower evaluations. Additionally, prejudice predicted applicant evaluation such that greater levels of prejudice predicted lower evaluations in both conditions, but the effect was stronger in the stigma condition. This suggests that our measure of sexual prejudice is also predicting a general tendency to make more negative, or more positive, evaluations. Finally, there was limited support for the moderated mediation predictions. In general, across both the second stage and direct effect moderated mediational model and the second stage model not including the moderated direct effect (the results were substantively similar across models), ideology indirectly predicted applicant evaluation through prejudice in both conditions but the effect of conservative political ideology indirectly predicting lower evaluations was stronger for the applicant with the sexual stigma.
The current research makes theoretical and practical contributions to our understanding of discrimination against sexual minorities. This is the first paper, to our knowledge, to directly explore the process through which political ideology predicts discrimination against those with a sexual stigma. This work shows the critical interplay of both sociocultural (ideology) and individual (prejudice) manifestations of sexual stigma in discrimination against sexual minorities. According to the theoretical work of Herek (2009a), “heterosexism serves as the foundation and backdrop for individual manifestations of sexual stigma” (p. 67). In this work, we examined the role of heterosexism, specifically, the stigma that is anchored in conservative ideological systems, in promoting internalized stigma in the form of sexual prejudice. Across all studies, political conservatism was strongly correlated with higher levels of sexual prejudice. In turn, this sexual prejudice predicted discrimination in the form of negative evaluation in the employment context. Thus, the conservative political ideological system in the United States serves in part to legitimize sexual stigma and promote discrimination against sexual minorities through individual internalization of sexual stigma (Herek, 2009a). This research points to the importance of integrating sociocultural and individual perspectives when investigating processes involved in sexual stigma. Moreover, by showing the critical role of ideologies, this research suggests that without also taking into account sociocultural manifestations of sexual stigma it can be difficult to get an accurate understanding of the prevalence of discrimination (Bertrand & Duflo, 2017).
Beyond theoretical contributions, this research has important implications for approaches to promoting social justice by lessening prejudice and discrimination against sexual minorities. Sexual minorities face significant discrimination-based barriers in many contexts including employment (Cunningham, Sartore, & McCullough, 2010; Tilcsik, 2011; Weichselbaumer, 2003). We sought to gain a better understanding of bias in employment contexts because sexual minorities experience great inequities in this domain (Tilcsik, 2011) and there are many places where LGBT-based employment discrimination remains legal in the United States. In Tilcsik’s (2011) large-scale audit study, sexual minorities were 40% less likely to be invited for an interview than similarly qualified sexual majority applicants. Research focusing on understanding factors that exacerbate or attenuate this bias has typically focused on factors such as geography, job requirements, and evaluator’s gender (Tilcsik, 2011; Weichselbaumer, 2003). By focusing on political ideology and thus taking both sociocultural and individual perspectives to understanding sexual stigma, this work offers insights into both reducing discrimination and promoting equitable and non-discriminatory public policies.
Our work suggests that political ideology serves as an important foundation for internalized sexual prejudice and in turn discrimination against minorities, thus, pointing to the important role of intervening at both the ideological and individual levels. Attempts to reduce sexual prejudice often, unsurprisingly, focus on the individual level. For example, one dominant approach is to promote personal contact with sexual minorities (Herek & Capitanio, 1996; Pettigrew, 1998). These approaches are critical, but likely not sufficient. Ideologies are powerful purveyors of stigma in part because they provide the necessary justifications for devaluing others; in the case of political ideology, it endorses a hierarchical understanding of social relations and that people are responsible for the outcomes they receive (Pratto, Sidanius, Stallworth, & Malle, 1994; Weiner, 1995). Our work suggests that one way to counter the stigma justification processes associated with political ideology is to promote beliefs and ideologies that serve to quell prejudice (Crandall, 2000). For example, an ideology that works to suppress stigma is the humanitarianism-egalitarianism belief system that promotes the equal worth and value of all humans and the concern for others’ wellbeing (Katz & Hass, 1988). Another sociocultural force that can work to suppress stigma is pressure from social norms (Crandall, 2000).
Despite the theoretical advances and practical implications of this work, there are limitations and significant opportunities for future research. First, there are other potential explanations for why political ideology might predict discrimination; for example, people might have assumed the applicant with the stigma has more politically liberal attitudes and this assumption may have led more conservative individuals to evaluate the applicant less favorably. However, the results showing modest support for the role of sexual prejudice in mediating the link between ideology and discrimination lends support to our explanatory framework that ideological systems promote discrimination at least in part by legitimizing anti-gay prejudice. Second, although we relied on a non-student sample to help generalize beyond a college population, it is not clear whether these findings will generalize to those who make internship decisions in athletic contexts. Additionally, this work focused on the evaluation of a gay man and it is not clear whether the findings will generalize to other sexual minorities. Research shows that the likelihood of experiencing discrimination based on sexual orientation, from violence to employment or housing discrimination, is not uniform among sexual minorities and that gay men report experiencing the most discrimination (Herek, 2009b)
Future research should also investigate these processes in contexts other than athletic internships. For example, the research showing that employment discrimination against gay men is strongest amongst employers searching for applicants with stereotypically masculine traits (Tilcsik, 2011), suggests that discrimination will vary across occupations that are highly gender segregated. Moreover, the research showing that conservatives evaluate counterstereotypical gay men more negatively than stereotypical gay men (Stern, West, & Rule, 2015) suggests that the discrimination observed in this counterstereotypical context (football internship) might not generalize to gay men applying to intern in more stereotype consistent, or stereotype neutral, positions. Future work might also explore more and varied signals to sexual stigma.
In examining our questions about the critical interplay of sociocultural and individual manifestations of sexual stigma, we focused on political ideologies in the United States largely because this ideology blends the two components of justification ideologies (Crandall, 2000): an endorsement of hierarchies and greater internal attributions for the stigma. Future work should further examine the generalizability of and limits to this framework. For example, researchers should investigate ideologies beyond American conservatism that justify stigma and should test political ideologies in other political contexts that are not associated with the same hierarchy beliefs and attributions as in the United States.
Finally, our use of Mechanical Turk workers, or Turkers, is not without its limitations. For example, there are concerns that many Turkers participate in multiple similar studies which may lead to participant nonnaïveté (Chandler, Mueller, & Paolacci, 2014). There are also concerns about the attentiveness of Turkers; however, empirical support of this concern is scant and recent work has shown that Turkers might be more attentive to instructions than traditional samples (Hauser & Schwarz, 2016). Most of the debate about Turkers, however, concerns the external validity of the samples. Although the sample is more representative than typical undergraduate samples for experimental survey work, Turkers in the U.S. are not representative of the U.S. For example, they tend to be younger, better educated, and less racially/ethnically diverse than the general population (Hitlin, 2016). More relevant for this research is the consistent finding that Turkers tend to be more politically liberal; however, recent research examining the personality and value-based motivations of ideology across various samples lends support to the validity of MTurk samples for research related to political ideology (Clifford, Jewell, & Waggoner, 2015).
Conclusion
In this research, we examined the role of both sociocultural and individual manifestations of sexual stigma in discrimination against sexual minorities. Specifically, we examined political ideology as the backdrop to sexual prejudice and ultimately discrimination. The pattern of results from four experimental studies show that conservative political ideology negatively predicted the evaluation of an internship applicant with a sexual stigma but not a similar applicant without the stigma. Additionally, higher, relative to lower, levels of sexual prejudice more strongly negatively predicted the evaluation of the applicant with the stigma relative to the control applicant. Finally, whereas ideology indirectly predicted applicant evaluation through prejudice generally, the effect was stronger for the applicant with the sexual stigma. Understanding the role of both political ideology as well as individual sexual prejudice in discrimination may facilitate efforts to dismantle discrimination and promote equitable public policies.
