Introduction

Operating through individuals and institutions, stigma represents a fundamental cause of population health inequalities, including inequalities related to sexual orientation (Bränström, Pachankis, Hatzenbuehler, & Link, 2016; Hatzenbuehler, Phelan, & Link, 2013). Accumulating research demonstrates that structural stigma, in particular, serves as a key driver of poor health among sexual minority populations—i.e., those who identify as lesbian, gay, or bisexual; engage in same-sex sexual behavior; or experience persistent same-sex attractions (Hatzenbuehler et al., 2014). Structural stigma refers to societal conditions, norms, or policies that compromise the opportunities, resources, and well-being of a socially marginalized group such as sexual minorities (Hatzenbuehler & Link, 2014). For example, sexual minorities living in U.S. states without legal protections against hate crimes and employment discrimination experience substantially more internalizing mental health disorders than sexual minorities living in states with such protections (Hatzenbuehler, Keyes, & Hasin, 2009). Sexual minority youth who live in highly stigmatizing communities experience a greater risk of suicidality than those who live in low-stigma communities (Hatzenbuehler, 2011). A growing body of research also demonstrates that among men who have sex with men (MSM), structural stigma predicts wide geographic variation in sexual health outcomes, including access to HIV-prevention education and associated services (Oldenburg et al., 2014; Pachankis et al., 2015b). Thus, the health of sexual minorities depends substantially on the degree of structural stigma present in the locations in which they live.

Despite consistent evidence that structural stigma is a major determinant of sexual minority health, few studies to date have investigated the possibility that manifestations of male sexual orientation are also shaped, in part, by structural stigma. On the one hand, men’s sexual orientation is assumed to be relatively impervious to sociocultural influences, especially compared to women’s sexual orientation (e.g., Baumeister, 2000; Savin-Williams, 1990; Savin-Williams & Diamond, 2000). In fact, compared to same-sex sexuality among women, same-sex sexuality among men is less likely to be described by men as chosen or under their personal control (Rosenbluth, 1997; Savin-Williams, 1990) and is less likely to depend on lifespan or historical context (Baumeister, 2000; Savin-Williams & Diamond, 2000; Twenge, Sherman, & Wells, 2015). On the other hand, a large body of literature points to variation across cultures and time in manifestations of male sexuality (Herdt, 1996; Risman & Schwartz, 1988; Troiden, 1988), suggesting the contextual nature of men’s sexuality. For example, across the globe, the reported prevalence of MSM’s bisexual behaviors and identities varies depending on country (Amirkhanian, Kelly, & Issayev, 2001; Colby, 2003; Choi, Gibson, Han, & Guo, 2004; Micheals & Lhomond, 2006; Wade et al., 2005).

Recent evidence suggests that, in addition to explaining the wide geographic variation in sexual minority health, structural stigma might also serve as one potential explanation for the wide variation in the proportion of MSM who openly disclose their gay/homosexual sexual orientation. Specifically, Pachankis et al. (2015b) found that MSM living in countries with homophobic laws, policies, and community attitudes (e.g., Russia, Ukraine) were significantly more likely to completely conceal their sexual orientation from all others than were MSM living in countries with low levels of structural stigma (e.g., Norway, Spain). Further, recent data collected from online searches and social networks indicate that many fewer adolescents and adult men openly disclose their sexual orientation (e.g., listing themselves as “interested in men” on Facebook) in U.S. states with high degrees of structural stigma toward sexual minorities (e.g., Mississippi) than in low-structural stigma states (e.g., California), despite a similar proportion of online search activity for both male pornography (e.g., searching Google for male pornographic websites) and casual male sex (e.g., searching for “casual encounters” on Craigslist.com) in both types of locales (Stephens-Davidowitz, 2014). There are several factors that might explain this relationship between structural stigma and disclosure of a gay/homosexual sexual orientation. MSM living in high-structural stigma locales might avoid discrimination by reporting attractions, behavior, or identity as heterosexual, as opposed to gay or homosexual. Additionally, MSM who report attractions as, behave as, and identify as bisexual or heterosexual might also expand sexual opportunities, with women in this case, in the face of environmental or attitudinal constraints against identifying and meeting same-sex partners. In high-stigma locales, therefore, MSM might perceive, and receive, more benefit from reporting attractions as, behaving as, or identifying as bisexual or heterosexual, as opposed to gay or homosexual.

Drawing on emerging support for the context-dependent nature of sexual orientation disclosure (Pachankis et al., 2015b), we sought to investigate the degree to which structural stigma at the country level might also predict MSM’s sexual attraction, behavior, and self-identification. While the prevalence of MSM’s bisexual and heterosexual identification and behavior varies across countries (Amirkhanian et al., 2001; Colby, 2003; Choi et al., 2004; Wade et al., 2005), no study has examined country-level structural stigma as a correlate of this variation. Significant barriers have previously impeded testing this possibility. For example, few data structures capture wide variation in structural stigma, for example across supportive and homophobic countries. Further, while sexual orientation represents a multifaceted construct composed of at least sexual attraction, behavior, and identity (Savin-Williams, 2006), few large, multi-national datasets assess sexual orientation as a function of all three of these components.

Using a unique multi-national dataset with large variation in structural stigma as well as a multi-component assessment of sexual orientation allowed us to examine structural stigma as a key correlate of the geographic variation in MSM’s sexual attraction, behavior, and identity. Specifically, we used data from the European MSM Internet Survey (EMIS; EMIS Network, 2013), the largest study of MSM, spanning 38 countries. We hypothesized that MSM living in high-structural stigma countries in Europe would evidence higher odds of reporting bisexual and heterosexual attraction, behavior, and identity than MSM living in low-structural stigma countries, whom we expected would be comparatively more likely to identify as gay and to report exclusively same-sex attraction and behavior. We expected these associations to be robust when controlling for individual-level (e.g., immigration status) and country-level (e.g., income inequality) variables. We also expected that structural stigma would interact with the population size of one’s current city or town to predict sexual attraction, behavior, and identity among MSM. Given that sparsely populated areas often contain low community support for sexual minorities (Pachankis, Hatzenbuehler, & Starks, 2014), as well as numeric and attitudinal barriers to identifying and meeting same-sex partners, we expected that MSM who live outside of urban areas within high-structural stigma countries would be the most likely to report bisexual or heterosexual attractions, behaviors, and identities. Finally, given that high-stigma countries place constraints on adopting a same-sex sexual identity and engaging in same-sex sexual behavior, but not necessarily on experiencing same-sex attraction, we hypothesized that sexual identity and behavior would be more strongly influenced by country-level stigma than would sexual attraction, which is not necessarily inherently expressive (Savin-Williams, 2006; Stein, 2001). Thus, we also expected lower odds of concordance among attraction, behavior, and identity for MSM reporting any form of bisexuality or heterosexuality in high-stigma, compared to low-stigma, countries.

While the empirical evidence reviewed above supports our hypotheses that contextual factors can strongly influence manifestations of sexual orientation, this evidence does not necessarily suggest that contextual influences alone drive these manifestations or determine sexual orientation itself. In fact, our hypotheses are consistent with (but cannot fully test) hybrid sociocultural–biological theories of sexuality (Tolman & Diamond, 2001), which recognize that both sociocultural and biological, or innate, factors work in tandem to influence sexuality and its expression. In the present study, we focus on the largely overlooked influence of one potential sociocultural driver of MSM’s sexual attraction, behavior, and identity, namely structural stigma toward sexual minorities.

Method

Participants

We used data from the EMIS, administered in 25 languages between June and August 2010 across the 38 European countries that yielded viable data (i.e., over 100 participants). Over 235 local, national, and international sexual minority websites recruited participants online through instant messages or banner advertising and national EMIS Network partners recruited offline through posters, cards, and face-to-face communication. Eligibility criteria included: male identification, European residence, at or above the age of homosexual consent in the country of residence (ranging from age 13 for Spain to age 18 for Bulgaria and Turkey), and sexual attraction to and/or sexual experience with men. Eligible participants had to indicate understanding of the study’s purpose and provide consent. Typical completion time was 21 min. No material inducement was offered. EMIS items were generated through consultation with NGOs, pilot testing for comprehension and length with MSM in 21 countries, and cognitive interviewing to ensure accurate interpretation. The survey development and methods are described in detail elsewhere (EMIS Network, 2013; Weatherburn et al., 2013).

The survey received 184,469 submissions. Three cases were lost to data corruption. Cases were removed for participants who: (a) did not specify a home country or indicated a country outside the study area (n = 2427); (b) were from a country that did not reach 100 qualifying cases (n = 291); (c) indicated being women, having no same-sex attraction or experience, or being under the age of homosexual consent in their country of residence or providing no age (n = 544); and (d) submitted a major inconsistent response (n = 6995), resulting in a final sample size of 174,209 MSM (Weatherburn et al., 2013). The per-country range was between 117 participants in both Macedonia and Moldova and 54,387 in Germany.

Measures

Explanatory Variables

Following previous analyses of EMIS data (Berg, Ross, Weatherburn, & Schmidt, 2013; Pachankis et al., 2015b; Ross et al., 2013), we assessed country-level stigma using a combination of national legislation and general population attitudes toward sexual minorities. We derived legislation from the International Lesbian, Gay, Bisexual, Trans, and Intersexual Association-Europe (ILGA-Europe) Rainbow Index 2010 (ILGA, 2010), an aggregate of the presence of ten supportive legislative policies (e.g., same-sex marriage, employment non-discrimination legislation), which were given positive scores, and four discriminatory practices and legislative policies (e.g., violation of freedom of assembly), which were given negative scores, creating a theoretical range from −4 to +10. The actual data range was from −2 (unsupportive) (i.e., Russia, Ukraine) to +10 (supportive) (i.e., Sweden) (M = 3.18, SD = 3.34). We derived country-level attitudes toward sexual minorities from the 2008 wave of the European Values Survey, a cross-national survey of social attitudes that randomly sampled approximately 1500 residents per European country. We included the proportion of respondents in each country who: (1) thought homosexuality could be justified; (2) agreed that homosexual couples should be able to adopt children; and (3) did not indicate not wanting to have homosexuals as neighbors. We calculated the standardized mean of these three items. We then combined this standardized mean with the standardized policy index to create a country-level index of support toward sexual minorities in each country. This index was the mean of these two variables—attitudes and policies. The inverse standardized score of this index was used in all analyses to facilitate interpretation as standard deviation units of stigma (i.e., higher scores indicated greater structural stigma).

To assess settlement size, participants were asked, “How would you describe the place you live in?” with response options: 1 (A very big city or town [a million or more people]), 2 (A big city or town [500,000–999,999 people]), 3 (A medium-sized city or town [100,000–499,999 people]), 4 (A small city or town [10,000–99,999 people]), and 5 (A village/the countryside [less than 10,000 people]). To facilitate interpretation of the results, responses were dichotomized at 100,000 and over to indicate living in a small town/city (= 1) compared to living in a large town/city (= 0).

Outcome Variables

Sexual attraction was assessed with the question, “Who are you sexually attracted to?” with response options 1 (Only to men), 2 (Mostly to men and sometimes to women), 3 (Both to men and women equally), 4 (Mostly to women and sometimes to men), and 5 (Only to women). Participants who selected any of the last three options were classified as reporting either bisexual or heterosexual attraction (= 1); participants who selected either of the first two options were classified as exclusively/mostly homosexual attraction (= 0).

Sexual behavior was assessed with the item, “When did you last have any kind of sex with a woman?” Participants who indicated having any sex with a woman in the past 12 months were classified as engaging in any heterosexual behavior (= 1) versus exclusively homosexual behavior (= 0). Participants who indicated having no sex with either men or women in the past 12 months (n = 8668) were excluded from analyses examining sexual behavior.

Sexual identity was assessed with the item, “Which of the following options best describes how you think of yourself?” with response options 1 (Gay or homosexual), 2 (Bisexual), 3 (Straight or heterosexual), 4 (Any other term), and 5 (I don’t usually use a term). Participants were classified as bisexual/heterosexual (= 1) or gay/homosexual (= 0). Participants who selected “I don’t usually use a term” (n = 12,195; 7.0 %) or “any other term” (n = 1497; 0.8 %) in response to the sexual identity item were excluded from analyses examining sexual identity.

Sexual attraction, behavior, and identity concordance was calculated for the subsample of MSM reporting any form of bisexual or heterosexual attraction, behavior, or identity. Participants who reported bisexual/heterosexual attractions, behavior, and identity were coded as concordant (= 1). Participants who reported bisexual/heterosexual attractions, behavior, or identity, but not all three, were coded as non-concordant (= 0).

Covariates

Respondents who reported bisexual or heterosexual attraction, behavior, or identity were significantly older and less likely to be in a relationship, to have been diagnosed with HIV, to have attained a high educational status—but also more likely to be unemployed—and to be immigrants than were respondents who reported exclusive same-sex attractions or behaviors, or a gay/homosexual identity, respectively. Therefore, we controlled for these demographic covariates in all analyses. Individual-level covariates included age, relationship status, HIV status, education, employment status, and immigration status. To control for the possibility that general structural inequality, rather than sexual orientation inequality specifically, is responsible for associations between structural stigma and sexual orientation attractions, behavior, and identity, we also included each country’s 2009 Gini coefficient, an index of income inequality, as a country-level covariate. The association of the Gini index with country-level stigma was moderate (r = .34, p < .05), consistent with previous research showing associations between income inequality and stigmatizing attitudes toward homosexuality (Andersen & Fetner, 2008).

Analytic Strategy

Given the nested structure of our data (i.e., respondents within cities/towns within countries), we specified three-level random intercept models using HLM v. 6 (Raudenbush & Bryk, 2002). Given the binary nature of our sexual orientation outcomes, we specified a Bernoulli distribution. This yielded an odds ratio of heterosexual/bisexual attraction, behavior, and identity (vs. gay/homosexual attraction, behavior, and identity) conditional on the explanatory variables. Fixed effects were estimated for each explanatory variable by means of full information maximum likelihood. Individual-level factors (e.g., demographic covariates) were modeled at Level 1, settlement size was modeled at Level 2, and country-level factors (i.e., Gini coefficient, structural stigma) were modeled at Level 3.

Missing data for explanatory variables ranged from 0 (0.0 %) for age to 4307 (2.3 %) for settlement size. Missing data for outcomes ranged from 396 (0.2 %) for sexual identity to 583 (0.3 %) for sexual attraction to 1898 (1.1 %) for sexual behavior. Thus, the sample size for each model depended on the amount of missing data for the outcome variable examined in that model.

Separate models were examined for each of the three sexual orientation outcomes in the full sample (i.e., attraction, behavior, identity), as well as for concordance among attraction, behavior, and identity in the subsample of MSM reporting any form of bisexual or heterosexual attraction, behavior, or identity. For each outcome (i.e., attraction, behavior, identity, and concordance), we examined associations with structural stigma and settlement size. Then, we examined the interaction between structural stigma and settlement size in predicting each outcome. Finally, given that these analyses combined heterosexual and bisexual (versus homosexual/gay) attraction, behavior, and identity, we ran sensitivity analyses in which we separately predicted only bisexual and then only heterosexual (versus homosexual/gay) attraction, behavior, and identity from our set of explanatory variables.

Results

Table 1 shows descriptive statistics of the full sample. The majority of the sample identified as gay/homosexual had been sexually active only with men in the past 12 months and was primarily attracted to other men. About half of the respondents were in a relationship and had completed primary and secondary education; about one-third lived in a locale with fewer than 100,000 inhabitants. The majority reported an HIV-negative last test result (or not being tested), currently living in the country in which they were born, and being employed. About three-quarters lived in a low-stigma country, defined in this case as a country that fell below the median score of our country-level structural stigma index.

Table 1 Characteristics of study respondents in the European MSM Internet Survey (N = 174,209)
Full size table

Country-Level Stigma and Sexual Attraction, Behavior, and Identity: Main Effects

Country-level prevalence of bisexual/heterosexual attraction ranged from 1.68 % (Malta) to 22.29 % (Slovenia). Bisexual/heterosexual behavior prevalence ranged from 8.38 % (Belgium) to 30.37 % (Bosnia/Herzegovina). Bisexual/heterosexual identity ranged from 9.43 % (Netherlands) to 38.58 % (Bosnia/Herzegovina). Using a median split of our structural stigma index to illustrate the distinction between high-stigma and low-stigma countries, the mean country-level prevalence of bisexual/heterosexual attraction was similar in both high-stigma (11.18 %, SD = 5.19) and low-stigma (9.79 %, SD = 4.06) countries, t = 0.91, p = .367. However, the mean country-level prevalence of bisexual/heterosexual behavior was higher in high-stigma (19.53 %, SD = 5.68 %) compared to low-stigma (13.86 %, SD = 4.08 %) countries, t = 3.53, p < .001. The mean country-level prevalence of bisexual/heterosexual identity was also higher in high-stigma (24.23 %, SD = 8.69 %) compared to low-stigma (17.44 %, SD = 5.57 %) countries, t = 2.87, p < .01.

Country-level stigma, examined as a continuous variable in all models, predicted all measured manifestations of sexual orientation (Table 2). Specifically, MSM living in high-stigma countries were more likely to (1) report bisexual or heterosexual attraction compared to reporting mostly or exclusively homosexual attraction (AOR = 1.18, 95 % CI 1.04–1.35) (Fig. 1); (2) report having sex with women or both men and women, as compared to reporting sex only with men, in the past 12 months (AOR = 1.33, 95 % CI 1.20–1.47) (Fig. 2); and (3) identify as bisexual or heterosexual than as gay/homosexual (AOR = 1.35, 95 % CI 1.19–1.54) (Fig. 3). That is, for every one SD increase in our structural stigma index, MSM reported 18, 33, and 35 % higher odds of bisexual or heterosexual attraction, behavior, and identity, respectively, all small effect sizes.

Table 2 Associations between structural stigma, population size, and sexual minority men’s odds of bisexual or heterosexual attraction, behavior, and identity: European MSM Internet Survey
Full size table
Fig. 1
figure 1

Proportion of MSM identifying as heterosexual or bisexual by country-level structural stigma

Full size image
Fig. 2
figure 2

Proportion of MSM reporting sex with women by country-level structural stigma

Full size image
Fig. 3
figure 3

Proportion of MSM reporting attraction to women by country-level structural stigma

Full size image

Somewhat inconsistent with our hypothesis that same-sex sexual identity and behavior would be more strongly related to structural stigma than same-sex attraction, we found that although country-level stigma was more strongly related to sexual behavior and identity than sexual attraction, the effect sizes were not significantly different from each other (identity vs. attraction: z = .50, p = .62; behavior vs. attraction: z = .39, p = .66). Consistent with our hypothesis regarding concordance across the three components of sexual orientation, concordance was significantly lower in high-stigma countries than in low-stigma countries (AOR = 0.87, 95 % CI 0.79–0.97). That is, bisexually or heterosexually attracted, behavioral, or identified MSM in high-stigma countries were 13 % less likely to report concordance across their attraction, behavior, and identity than their peers in low-stigma countries, a small effect size.

Country-Level Stigma and Sexual Attraction, Behavior, and Identity: Interaction with Settlement Size

Settlement size significantly predicted odds of bisexual/heterosexual attraction, behavior, and identity, such that in sparsely populated locales, MSM were 62, 49, and 50 % more likely to report bisexual/heterosexual attractions, behaviors, and identities, respectively, compared to MSM in densely populated locales, all small effect sizes (Table 2). Settlement size significantly interacted with country-level stigma to predict sexual attraction and sexual behavior, such that MSM living in sparsely populated locales within high-structural stigma countries were the most likely to report bisexual or heterosexual behaviors and attractions, although the slope of the association between country-level stigma and odds of bisexual or heterosexual attraction and behavior was steeper for large towns/cities than for small towns/cities (See Figs. 4 and 5). Country-level stigma did not significantly interact with settlement size to predict either sexual identity or concordance across attraction, behavior, and identity.

Fig. 4
figure 4

Interaction between country-level structural stigma and settlement size in predicting odds of bisexual or heterosexual (compared to exclusively homosexual) attraction

Full size image
Fig. 5
figure 5

Interaction between country-level structural stigma and settlement size in predicting odds of bisexual or heterosexual (compared to exclusively homosexual) behavior

Full size image

Sensitivity Analyses

Because our outcomes in the above analyses combined bisexual and heterosexual identity, behavior, and attraction, we conducted sensitivity analyses in which we limited analyses to predicting only bisexual—versus gay/homosexual—attraction, behavior, and identity. The direction and significance of main and interaction effects remained the same as in analyses that combined bisexual and heterosexual attraction, behavior, and identity. The one exception was for the interaction of country-level stigma and settlement size in predicting heterosexual (versus homosexual) behavior, which became nonsignificant (AOR = 0.96, 95 % CI 0.86–1.08), p = .50, unlike in the model that predicted bisexual and heterosexual behavior combined. We then limited analyses to predicting heterosexual—versus gay/homosexual—attraction, behavior, and identity; the direction and significance of main and interaction effects remained the same. The one exception was the main effect from country-level stigma to heterosexual (versus homosexual) attraction, which became nonsignificant (AOR = 1.05, 95 % CI 0.89–1.24), p = .56, unlike in the model that predicted bisexual and heterosexual attraction combined. This lack of association could potentially be explained by reduced power as only 236 participants reported exclusive attraction to women.

Discussion

Using a unique multi-national dataset of MSM spanning 38 countries containing wide variation in objectively measured structural stigma toward sexual minorities, as well as a comprehensive assessment of sexual orientation, we predicted variation in MSM’s sexual identity, attraction, and behavior as a function of country-level structural stigma. Consistent with our hypotheses, the wide variation in country-level structural stigma toward sexual minorities predicted significant variation in the odds of bisexual/heterosexual attraction, behavior, and identity among MSM across countries. Specifically, in more stigmatizing countries, MSM were significantly more likely to report bisexual/heterosexual attractions, behaviors, and identities than in less stigmatizing countries. Although country-level stigma was more strongly related to sexual behavior and identity than sexual attraction, the effect sizes were not significantly different from each other. Nonetheless, MSM who reported any bisexual/heterosexual attractions, identity, and behaviors were less likely to report concordance across these manifestations of sexual orientation in high-stigma, compared to low-stigma, countries. Finally, settlement size significantly predicted odds of bisexual/heterosexual attraction, identity, and behavior and moderated associations between country-level structural stigma and odds of bisexual/heterosexual attraction and behavior. The association between country-level structural stigma and odds of bisexual/heterosexual attraction and behavior was stronger for MSM residing in large, compared to small, towns/cities, while MSM residing in small towns/cities within high-structural stigma countries were the most likely to report bisexual or heterosexual attractions and behaviors.

Our results add to a body of research showing contextual influences on sexual orientation (e.g., Baumeister, 2004; Diamond, 2008; Herdt, 1996; Risman & Schwartz, 1988; Troiden, 1988). While prior research has demonstrated cultural and historical determinants of sexual orientation (Herdt, 1996; Twenge et al., 2015) and life course fluctuations across sexual minority women’s sexual attraction, behavior, and identity (Diamond, 2008; Mock & Eibach, 2012; Savin-Williams & Ream, 2007), our study suggests that stigmatizing environments at the country level can also influence various manifestations of men’s sexuality. Country-level sociocultural contexts have previously demonstrated theoretically meaningful associations with other aspects of human sexuality (e.g., willingness to have casual sex; sexting behavior), explaining wide variation between country populations in these motivations and behaviors (Baumgartner, Sumter, Peter, Valkenburg, & Livingstone, 2014; Schmitt, 2005). The present study demonstrates the influence of country-level sociocultural contexts on manifestations of MSM’s sexual orientation, which previous research has found to be more impervious to social contexts than manifestations of women’s sexual orientation (e.g., Baumeister, 2000; Mock & Eibach, 2012; Twenge et al., 2015).

Our findings suggest that sociocultural factors might influence manifestations of MSM’s sexual orientation with implications for understanding the ontology of male homosexuality. On the one hand, our results suggest that stigmatizing environments might constrain MSM from publicly expressing their biologically determined same-sex attractions through their identities and sexual behaviors. Because sexual attraction often developmentally precedes, and shows less lifespan fluctuation, than sexual behavior and identity (Calzo, Antonucci, Mays, & Cochran, 2011; Diamond, 2008), sexual attraction could be argued to be more biologically determined than sexual identity and behavior, which are potentially more strongly influenced by sociocultural influences (DeLamater, 1981; Gagnon & Simon, 2011; McClintock & Herdt, 1996; Udry, 1988). In fact, biological factors are known to more weakly determine sexual behavior, such as the initiation of coitus, when social constraints are high (e.g., Udry & Billy, 1987). Our finding that MSM living in high-stigma, compared to low-stigma, countries experience lower odds of concordance among their sexual attraction, behavior, and identity potentially suggests that stigma might discourage MSM from enacting their attraction to men through congruent identities and behaviors.

On the other hand, structural stigma might not so much constrain the expression of an innate sexuality as it does shape the actual experience of one’s sexual attractions. Sociocultural factors can produce distinct experiences of sexuality inseparable from their local context (e.g., Brown-Saracino, 2015). For example, patriarchal gender hierarchies, perpetuated through cultural institutions such as sexual education, have been argued to fundamentally shape women’s experience of sexual desire, even outside their conscious awareness (e.g., Fine, 1988). Thus, rather than suggesting a dysfunctional incoherence across sexuality components, our findings regarding sexuality discordance might reflect local experiences of sexuality that adaptively fit a surrounding sociocultural milieu. However, without having measured sociocultural features other than stigma, we are unable to test the influence of other potential contextual factors.

Importantly, we caution against interpreting our results as implying that structural stigma is the predominant cause of bisexuality or that all bisexually or heterosexually attracted, behavioral, or identified MSM are hiding an innate homosexuality because of social pressure and personal shame. Instead, our results only suggest that geographic variation in structural stigma can be understood as a strong correlate of manifestations of men’s sexual orientation. The extent to which those manifestations are yoked to any innate sexual orientation or represent a locally derived and understood form of sexuality remains to be determined.

Consistent with our overall hypothesis that structural stigma is associated with geographic variation in MSM’s sexuality, results suggest that MSM living in areas with smaller populations are particularly likely to report attractions as, behave as, and identify as bisexual or heterosexual. Previous research shows inverse associations between local population size and structural stigma toward sexual minorities (Pachankis et al., 2014). Less densely populated areas might contain less visible gay communities or gay-related infrastructure, thereby perpetuating stigma toward MSM as well as providing fewer opportunities for MSM to locate and meet same-sex partners (Keene, Eldahan, White Hughto, & Pachankis, 2016; Weeks, 1985). In smaller towns/cities, therefore, identifying as heterosexual or bisexual and having sex with women might not only protect against discrimination but might also maximize sexual opportunities against a backdrop of constrained options. The finding that population size was a stronger predictor of bisexual or heterosexual attraction, behavior, and identity in low-stigma, as compared to high-stigma, countries suggests that MSM in low-stigma countries might perceive more freedom to maximize sexual opportunities when population size limits same-sex opportunities by reporting attractions as, behaving as, or identifying as bisexual or heterosexual. However, this possibility warrants future investigation.

Results of the present study have implications for MSM’s physical and mental health. While previous research has found that structural stigma consistently predicts adverse mental and physical health among sexual minorities (e.g., Hatzenbuehler et al., 2014; Oldenburg et al., 2014; Pachankis et al., 2015b), the present study suggests that structural stigma is also a correlate of sexual orientation itself. Social structures differentially expose people to health-promoting or health-impairing resources (Diez-Roux, 2001). One way that social structures might compromise health is by shaping the identities that stigmatized populations adopt and the health-relevant knowledge, behaviors, and services associated with those identities (Oyserman, Smith, & Elmore, 2014). To the extent that MSM in high-stigma countries adopt identities and behaviors, including bisexual and heterosexual, that are not typically targeted by sexual health-promotion services and information, the health of MSM in these countries may be compromised (Pachankis et al., 2015b). Future research is needed to determine the exact mechanisms through which structural stigma might compromise the health of MSM across countries.

Despite this study’s strengths, results must be interpreted in light of several limitations. Given the cross-sectional design, we infer, but cannot test, causal relationships. Although a significant strength of utilizing an objective stigma index is that our outcomes cannot cause the predictor, it is possible that an unmeasured common factor, such as cultural variations in the form and meaning of men’s sexuality across countries (e.g., Lambevski, 1999), could still underlie the observed associations. Given that wide variations in the measured forms of sexuality nonetheless existed within low- and high-stigma countries even upon controlling for country-level income inequality, future research ought to consider additional country-level influences on these outcomes. Because of the infeasibility of creating an MSM sampling frame across countries, it was not possible to create a probability sample of MSM, which limits generalizability of the results. The extent to which MSM might have differentially selected into the EMIS based on sexual orientation across countries is unknown. For example, stigmatizing countries may have been less successful in recruiting exclusively homosexual men, perhaps because of a lack of recruitment organizations that specifically cater to exclusively homosexual men. At the same time, however, we note that the vast majority of participants (95.3 %) accessed the survey directly from a website or email invite, rather than typing the survey web address into a browser and that nearly three-quarters of the recruitment was conducted via the six largest dating sites via the internet. Thus, in-person recruitment via organizations was fairly minimal in this study. Further, because we relied on self-reported measures of sexuality, it is also possible that our results reflect systematic reporting biases across countries, although the nature of that bias is uncertain given that all eligible cases reported same-sex attraction or behavior. We assessed identity as how participants “think of” themselves, rather than how they publicly identify. Yet the proportion of individuals who think of themselves as sexual minorities is likely larger than the proportion who publicly identifies as a sexual minority (Pachankis et al., 2015a), and our estimates should be considered in light of the possibility that some participants who indicated a sexual minority identity (e.g., bisexual) in the present study might publicly identify as heterosexual. Finally, this study captures stigma at the country level, which may obscure important within-country variation. Because of this potential limitation, our results are particularly noteworthy, given that country-level factors are distal determinants of health; thus, our results are likely conservative estimates of the association between structural stigma and manifestations of sexual orientation.

The present study adds to a growing body of research highlighting the importance of structural stigma to the lives of sexual minority individuals. While previous research has demonstrated associations between structural stigma and adverse physical and mental health outcomes among sexual minorities (e.g., Hatzenbuehler et al., 2014), this study shows that structural stigma is also a key correlate of manifestations of MSM’s sexual orientation itself. The fact that the country-level variation found in MSM’s sexual attraction, behavior, and identity closely hews to the wide variation in structural stigma toward sexual minorities across those countries strongly suggests that manifestations of MSM’s sexual orientation are context-dependent. Specifically, the country across which an MSM lives affects the ways in which he views his sexual self, forms sexual motivations, and pursues sexual opportunities. These findings suggest that appropriate understanding of MSM’s sexuality and health will come from attending to the local contexts surrounding this important, although frequently overlooked, segment of the global population.