Straight Identity and Same-Sex Desire: Conservatism, Homophobia, and Straight Culture

Abstract

This paper uses the 2013–2015 NSFG, a representative US-based dataset of individuals 15–44, to explore predictors of straight identification among all women and men and among subsets with substantial same-sex activity and/or attraction. After controlling for attractions and sexual practices, homophobia predicted straight identification in all groups. Among both groups of women, one femininity attitude and motherhood also predicted straight identification. One attitude reflecting alignment with normative masculinity significantly predicted straight identification among men with substantial same-sex activity and/or attraction. This paper also uses two waves of Add Health, a representative survey of young adults, to examine change to sexual identity over six years. Results show that among individuals who changed sexual identities between waves, heightened religiosity and political conservatism across waves were associated with increased odds of changing to a straight identity for women, but not for men. This suggests but does not prove a directional association between attitudes and identification for some individuals. Latent class analyses also found distinct groups of straight-identified men and women with substantial same-sex activity and/or attraction, indicating that it is a heterogeneous population in terms of attitudes, including homophobia. This suggests that straight identification is due partly to embeddedness in straight culture and enjoyment of straight privilege, not simply homophobia. While impossible to determine causality, the results show that straight identification is strongly related to non-sexual social factors, including religiosity and attitudes about sexuality and gender, in addition to attractions and sexual practices. The results also suggest that homophobia is related to identity formation for women, as well as men, but that there is substantial within-group variation.

Why do nationally representative surveys show that are there more men and women who acknowledge same-sex attractions and/or sexual practices than who identify as LGBQ (Copen, Chandra, and Febo-Vazquez 2016; Ward et al. 2014)? First, sexual orientation and sexual identity differ. Sexual orientation includes emotional attractions and sexual attractions, practices, desires, and fantasies (Savin-Williams 2014). Each element is related but distinct (Priebe and Svedin 2013). Aspects of orientation do not always cluster in ways suggested by common sexual identities. For instance, 3.6–4.1 percent of men and 7.6–9.5 percent of women have a small amount of same-sex sexual and/or romantic attraction (Savin-Williams and Vrangalova 2013). Sexual identities, on the other hand, refer to how individuals understand their sexuality and their relation to social groups (e.g., gay communities). They usually cannot capture the complex ways in which individuals experience their sexual orientation. Sexual behavior, attraction, and identity are distinct. Contrary to what the one-act rule of homosexuality suggests (Schilt and Westbrook 2009), same-sex sexual behavior does not necessitate a gay/lesbian or bisexual identity, nor does the presence of same-sex attraction.

A small base of qualitative research seeks to understand why some individuals with same-sex attractions and/or practices identify as LGBQ, whereas others identify as straight. Many women who have sex with women identify as straight, since they view their status as mothers or partners to men as precluding LGBQ identification (Budnick 2016; Walker 2014). Interviews with straight-identified men who have sex with men (MSM) show that many bolster their straight identities by framing sex with men as emotionless and by emphasizing exclusive or primary attractions to women (Carrillo and Hoffman 2016, 2018). Many also do so by prioritizing partnerships with women and by interpreting sex with men in ways that do not threaten their marriages, straightness, or masculinity (Silva 2017a, 2018). Content analyses of Craigslist personal ads suggest similar themes (Reynolds 2015; Robinson and Vidal-Ortiz 2013), and reinforce that many straight-identified men who hook up with other men enjoy straight privilege and culture, including heteronormativity (Ward 2015). Individuals who identify as straight also have considerable social advantages relative to LGBQ-identified individuals, such as protection from sexuality-based discrimination and prejudice. Structural heteronormativity encourages all people to identify as straight. Qualitative research, overall, shows that there are many factors other than attractions and sexual practices that influence straight sexual identification.

Two fairly separate literatures explore gendered differences in (1) sexual behavior and identification and (2) homophobia. A higher proportion of women report same-sex behavior than men, as well as non-heterosexual identities (England, Mishel, and Caudillo 2016; Silva 2017b). Additionally, among men and women with same-sex sexuality, a higher proportion of men identify as exclusively straight than women (Silva 2017b). The proportion of same-sex sexual behavior and identification as bisexual is increasing over time, but only for women (England, Mishel, and Caudillo 2016). This reflects in part how women are socially devalued relative to men, so they have somewhat more social flexibility to explore same-sex eroticism (England 2016). A separate literature shows that homophobia and normative masculinity are strongly related. One way many men “prove” their masculinity to other men is through displays of heterosexuality (Pascoe 2011), as well as homophobia. Men are especially likely to report homophobic attitudes when they feel that their masculinity is threatened (Falomir-Pichastor and Mugny 2009; Munsch and Gruys 2018; O’Connor, Ford, and Banos 2017; Weaver and Vescio 2015; Willer et al. 2013). Additionally, homophobia is significantly associated with straight identification among men with same-sex sexuality (Silva and Whaley 2018). Little research has merged these two literatures to examine how homophobia and attitudes reflecting gender normativity are related to straight identification for women.

Building on these studies, this paper quantitatively explores the relationship between attitudes about sexuality and gender, religiosity, and conservatism, on one hand, and straight identification, on the other. Much of the previous research exploring straight identification has used non-generalizable samples, necessitating empirically generalizable data to further examine these relationships. This paper uses two datasets and two distinct analytical approaches to examine these topics. Utilizing the 2013–2015 National Survey of Family Growth (NSFG), a nationally representative US-based dataset of individuals 15–44, I examine men and women separately. Weighted logistic regressions compare straight- and LGBQ-identified individuals in the entire sample, as well as within subsets of women and men with same-sex sexual attractions and/or sexual practices. This paper pays particular attention to the relationship between straight identification and homophobia among women, given that little research has done so. It also uses two waves of the National Longitudinal Study of Adolescent to Adult Health (Add Health), a nationally representative US-based dataset of young adults. Analyses with Add Health examine how changes to certain social factors are significantly related to changing to a straight identity in wave IV from a previous non-heterosexual identity in wave III. Add Health’s longitudinal design does not prove causality, but suggests directional associations between attitudes and straight identification. Variable-centered approaches such as these identify what social factors are significantly associated with straight identification and changing to a straight identity.

Finally, this paper uses latent class analysis (LCA) to identify subgroupings among straight-identified women and men with same-sex sexuality. LCA is a person-centered form of analysis that identifies unmeasured groups in a population (Grzanka 2016). Few other studies have explored subgroupings of straight-identified women and men with same-sex sexuality. This paper shows that there are distinct groups that cluster by social attitudes and parenthood status. While homophobia as a variable is significantly related to straight identification, there are subgroups of straight-identified men and women with same-sex sexuality that are not overtly homophobic. Straight identification among individuals with same-sex sexuality is not entirely reducible to overt homophobia. Building on similar qualitative work, I argue that these results suggest that straight identification reflects, in part, alignment with straight culture, constraints associated with parenthood, and enjoyment of social advantages related to straight identification.

This paper does not make claims about the origins of sexual attractions or behavior, but rather explores sexual identity. Regardless of the causal mechanisms behind attitude-identity associations, this paper adds to the literature in four main ways. First, this is the first nationally representative research project to demonstrate that for both men and women some attitudes and non-sexual factors are associated with increased odds of heterosexual identification, even after controlling for personal sexual characteristics. Second, it considers this relationship among women, the experiences of whom have been underexamined on this topic—especially regarding homophobia. Third, it does both among individuals with and without substantial same-sex activity and/or attraction to better understand these relationships in different populations. Fourth, it utilizes LCA to identify distinct subgroups of straight-identified men and women, which has never before been done with nationally representative data. Doing so highlights that both variable- and person-centered approaches are important for understanding this population.

This paper focuses on straight identification because little research has examined it, and builds on the few other quantitative projects to examine straight identification. It examines both men and women, broadening the scope of Silva and Whaley (2018). It also uses the most recent data available, providing updated findings to Silva and Whaley (2018) and Silva (2017b).¹ Additionally, it uses LCA—a person-centered approach—to broaden those articles’ exclusive use of variable-centered approaches. Kuperberg and Walker (2018) use LCA to examine straight-identified college students whose last hookup partner was the same sex. Their analysis is the first to use LCA for straight-identified individuals with same-sex sexuality, and suggests that there are distinct subtypes characterized by differences in religiosity, homophobia, and circumstances of the hookup. The data they use are not representative, however, and come only from young college students, and only among those whose last hookup partner was of the same sex (and that hookup did not necessarily involve sex). I build on their important work by utilizing a sample representative of adults 15–44 in the United States.

Four key themes emerged about the relationship between straight identification and social factors. First, homophobia was associated with increased odds of straight identification for men and women, with or without substantial same-sex activity and/or attraction. This reinforces the need for researchers to study homophobia among women. Second, among women with or without substantial same-sex activity and/or attraction, motherhood and one attitude reflecting alignment with normative femininity were associated with increased odds of straight identification. Similar results emerged regarding one belief about masculinity among men with substantial same-sex activity and/or attraction, but not without. Regardless of the direction or nature of causality, the results suggest that heterosexual identification is strongly related to attractions, behaviors, and attitudes. Third, among women who changed sexual identities, increased religiosity and political conservatism were associated with heightened odds of changing to a straight identity. This was not the case for men, for whom only changes to attractions predicted changing to a straight identity. These analyses suggest that heightened religiosity and political conservatism may influence women to change to a straight sexual identity. That fewer variables were significant for men may reflect small sample sizes or fewer differences between men of different sexual identities, but more research is needed. Fourth, when utilizing variables that were significant in regression models, there were distinct clusters of straight-identified men and women with same-sex sexuality. Classes comprising about 46 percent of women and 56 percent of men were characterized by low homophobia and low alignment with conservative gender attitudes. This indicates that there are distinct subgroups of individuals with substantial same-sex activity and/or attraction who are tied together by their alignment with heterosexual identification, but not necessarily conservatism.

The Social Construction of Sexual Identity

This section overviews the social construction of sexual identity, which is not to be confused with sexual orientation (as described in the introduction). Its purpose is to underscore that sexual identities are socially constructed, subject to change over time, and influenced by social context. The common sexual identities today used most often in the West (bisexual, gay/lesbian, straight) have their roots in several simultaneous and interrelated historical developments that occurred mostly in the mid- to late nineteenth century. Individuals engaged in same-sex sexual practices prior to this point, but these acts were not usually associated with a sexual identity. Subcultures of MSM in urban areas first began emerging in the eighteenth century (e.g., in London: see Norton [2006]). Later, around the middle of the nineteenth century, medical scientists created new categories, framing what many people had understood as acts that anyone could engage in as characteristics of particular types of people (Foucault 1978). This helped legitimate new middle-class sexual practices that emphasized other-sex marital pleasure outside of reproduction (Katz 1995). New heterosexual identities also signaled that men and women had purportedly normal masculinity and femininity, respectively, after historical changes in gender practices as women increasingly entered the formal paid workforce and men began engaging in intellectual and interpersonal labor more than physical (Chauncey 1994). Wage labor within capitalism (D’Emilio 1997), and especially corporate capitalism (Boag 2003), allowed for individual economic self-sufficiency. Urbanization allowed individuals with a common sense of sexual difference to form communities, facilitating the adoption of gay/lesbian identities (D’Emilio 1997). The combination of all these factors was especially important for women, who previously had far fewer opportunities to be economically self-sufficient and thus form a collective sense of sexual difference (Faderman 1991). Bureaucratic practices in US immigration, welfare, and military institutions in the twentieth century also helped transform disparate sexual and relational practices, gender expressions, and physical embodiments into a single homosexual status or identity (Canaday 2009). From the mid- to late nineteenth century to the middle of the twentieth, sexual identities slowly disseminated across the West.

Understandings of sexuality differ between time periods and cultures, as well as within cultures. In the early nineteenth century, for instance, Victorians considered sexual passion and love to be mostly unrelated, facilitating socially validated romantic friendships between women (Faderman 1991; Smith-Rosenberg 1975) and men (Katz 2003; Rotundo 1989) that contained various types and degrees of physical contact (e.g., sleeping in the same bed and cuddling). This changed as the current system of sexual classification—which links emotional and physical attractions to a single “sexual orientation”—became hegemonic and pathologized these relationships. Cross-culturally, migration between different countries can change how individuals perceive and experience sexuality, as with MSM who migrate from Mexico to the United States (Cantú 2009; Carrillo 2017; Carrillo and Fontdevila 2014). Sexual identification can also differ within countries, depending on the weight individuals give to sexual attractions versus current sexual partnering practices (Rust 1992); the presence or lack thereof of emotional attractions (Adam 2000); stereotypes about LGBQ identity labels (e.g., lesbian women are butch) or a lack of awareness of alternative identities (Kitzinger and Wilkinson 1995); or gendered interpretations of penetration, as with some (but not all) MSM in Mexico (Carrillo and Fontdevila 2014). It may also differ based on whether the government under which an individual lives legally recognizes same-sex couples (Charlton et al. 2016; see also Frank, Camp, and Boutcher [2010] for a cross-national examination of changes to legal regulations of sexuality).

The current organization of other-sex sexuality as “heterosexual” did not exist in other historical periods. In the US Northeast during the colonial era, for instance, sexuality was organized on a reproductive/non-reproductive basis, such that most non-reproductive acts were viewed as similar offenses (Katz 1995). There was no social validation of sexual pleasure within this system. Heterosexual identification, however, normalized other-sex sexual pleasure (Katz 1995), a shift that was in part facilitated by Freud’s ([1905] 2011) insistence that humans were predisposed toward seeking sexual pleasure, and especially other-sex genital pleasure. Heterosexuality is an institution that creates and reinforces inequalities between men and women (Jackson 1999; Rich [1980] 1998), and an identity and a set of social and sexual practices and experiences (Jackson 1999). Straight identification grants individuals power and privilege relative to LGBQ-identified individuals (Messner 1999). Heteronormativity is part of the fabric of US society, evident in how parents raise their children (Martin 2009; Solebello and Elliott 2011); gendered sexual socialization in preschools (Gansen 2017); peer interactions in elementary school (Myers and Raymond 2010) through high school (Pascoe 2011); formal and informal curricula in educational institutions (Pascoe and Silva forthcoming); and how most institutions operate (Hearn and Parkin 1987; Hearn et al. 1989). State intervention also encourages alignment with heterosexuality, such as through current marriage promotion campaigns (Heath 2009). Historically, too, the federal government spread a national discourse about “normal” gender and sexuality, which used idealized images of rurality—as well as federal intervention in rural spaces—to construct and disseminate ideas about heterosexuality (Rosenburg 2016). All of this combined encourages identification as straight. Nonetheless, many currently straight-identified men and women considered alternative identities in their past before ultimately adopting an exclusively straight identity (Morgan 2012), reinforcing both the social construction of heterosexual identification and the power of compulsory heterosexuality.

There are two main takeaways from this body of research. While there may be biological influences on sexual attractions and practices, sexual identification—how individuals and groups understand and label sexuality—is socially constructed and differs by culture and time period. As social or historical context changes, how individuals understand and identify their sexuality can also change.

Homophobia, Gender, and Sexuality

Research suggests that for many men homophobia, normative gender practices, and heterosexual identification are tightly connected. Little nationally representative research has explored this, however, especially among women. Pascoe (2011) finds that boys regulate the masculinity and heterosexuality of peers through gendered homophobic harassment. A key way to demonstrate masculinity to peers is through control over girls’ bodies, such as through touching, sex, and discussions of sexual dominance over women. This reflects how heterosexual identification is central to hegemonic masculinity (Connell 1987). For many men heterosexual identification, homophobia, and normative expressions of masculinity are intertwined (Keiller 2010). This is the case even among many men who seem progressive (Munsch and Gruys 2018). Further, there are multiple, often subtle types of homophobia (Doan, Loehr, and Miller 2014) that reinforce both masculinity and straight identification (Bridges 2014; Bridges and Pascoe 2016).

Little research addresses homophobia among straight women, though some finds that homophobia is more central to masculinity than to femininity (Falomir-Pichastor and Mugny 2009; Nagoshi et al. 2018; Willer et al. 2013). Many men express homophobia as a strategy to “prove” masculinity and heterosexuality and thus maintain privilege, especially in response to perceived threats to their masculinity (O’Connor, Ford, and Banos 2017; Weaver and Vescio 2015; Willer et al. 2013). Women do so for different reasons, however. Heteronormativity and gender inequality both contribute to straight women’s homophobia. This is why some college-aged straight women who kiss other women in social settings distance themselves from lesbians, who they perceive as threats to their attempts to appeal to men and thus gain status they would otherwise lack (Hamilton 2007). It is also why some girls use homophobic harassment against other girls as a resource to construct their own sexual subjectivity when it is otherwise difficult for them to do so (Miller 2016). More research is needed about the relationship between homophobia, femininity, and straightness, however, given that few studies have explored it.

As the visibility and social acceptance of LGBQ individuals increase, many straight-identified individuals consciously express their straight identity through normative gender practices (Dean 2014). Previous research using non-representative samples demonstrates that many straight-identified MSM internalize higher levels of homophobia and align themselves more strongly with hegemonic masculinity, on average, than gay and bisexual men (Amola and Grimmett 2015; Baunach and Burgess 2013; Miller, Serner, and Wagner 2005; Rowen and Malcom 2002; Wolitski et al. 2006). This demonstrates the links between homophobia and hegemonic masculinity among straight-identified MSM. No similar research exists for straight-identified women who have sex with women. Previous research, overall, suggests that non-sexual social factors such as homophobia, normative gender practices, gender attitudes, and enjoyment of straight culture shape sexual identification for many straight-identified MSM (Carrillo and Hoffman 2016, 2018; Silva 2017a, 2018; Ward 2015). This framework is useful for considering straight identification in other populations as well, even as the direction of causality remains unclear.

Combined, previous research suggests that straight identification, homophobia, gender attitudes, and normative gender practices are connected for many people. These relationships have been underexplored in nationally representative samples, however, especially among individuals with substantial same-sex activity and/or attraction. Even in non-representative samples, little research examines these relationships among women with or without substantial same-sex activity and/or attraction. This research aims to help fill these gaps in the literature. Readers should be mindful of the analytical distinction between individuals that secretly identify as LGBQ, and are thus “closeted,” and those that identify in ways counter to the dominant system of sexual classification. While internalized homophobia may affect both groups (see Szymanski, Kashubeck-West, and Meyer [2008] and Newcomb and Mustanski [2010] for a review of internalized homophobia), they are distinct—hence this paper’s examination of straight-identified individuals with substantial same-sex activity and/or attraction.

Methods

All data were gathered from two datasets: the NSFG and Add Health.² The 2013–2015 NSFG is a nationally representative, probability-based sample of US residents aged 15–44 residing in non-institutional households. Interviews lasted approximately 74 minutes for women and 51 for men, and respondents were paid $40. The total response rate was 71.2 percent for women and 67.1 percent for men. In total, 5,699 women and 4,506 men participated. The survey was designed and administered by the National Center for Health Statistics, within the Centers for Disease Control (US Department of Health and Human Services 2016). Questions about same-sex sexuality were self-administered, to encourage honest reporting. The NSFG is one of the few large-scale, nationally representative surveys to ask questions about both same-sex sexuality and social attitudes.

As a supplemental source of data, this study uses waves III and IV of Add Health. This is a nationally representative, US-based panel study of young adults who were in grades 7–12 in the 1994–1995 school year (Harris 2013). Wave III was collected in 2001–2002 and had a 77.4 percent response rate; its 15,197 respondents were 18–26 at the time of the interview. Wave IV was collected in 2008 and had a response rate of 80.3 percent, and its 15,701 respondents were 24–32 during data collection. Add Health is a supplemental dataset, rather than a primary one, for several reasons. First, Add Health has a large sample but does not ask any attitudinal items. Second, while Add Health is longitudinal, it is dated: its last two waves were collected in 2001–2002 and 2008. Third, Add Health examines only one cohort of Americans. Fourth, Add Health does not measure degree of attraction to men and women—only attraction to either as yes/no questions—undercutting the ability to comprehensively control for attractions in regressions. Despite these shortcomings, Add Health does provide stronger (but inconclusive) evidence for the direction of association between variables.

Other major surveys that researchers have often used in sexuality research are designed in ways that make them difficult to use for this project. The General Social Survey (GSS), for instance, utilizes a ballot design such that not all respondents answer the same attitudinal questions. This makes the already small subset of respondents who report same-sex sexuality difficult to analyze when including more than one attitudinal variable. Further, in the GSS, respondents report lifetime same-sex sexual practices, but not attractions, making it impossible to control for attractions in regressions. Other surveys do not ask attitudinal questions at all. Given these factors, the NSFG is the most appropriate primary dataset, with Add Health as a supplementary dataset.

Same-Sex Sexuality Subsets

Respondents in the NSFG were coded as 1 for having substantial same-sex activity and/or attraction if they indicated substantial same-sex attraction and/or two or more lifetime same-sex sexual partners.³ Given that respondents may have included non-consensual acts when reporting sexual partners, men who reported two or more male sexual partners were excluded from the subset if they reported exclusive attractions to women and non-consensual sex with men. The NSFG asked no questions to women about non-consensual sex with other women, so this strategy was impossible for them. Men and women were also included if they reported equal attractions to men and women or primary or exclusive same-sex attraction. The subsets, consequently, include only respondents with substantial same-sex activity and/or attraction.

Several items determined the subsets within the NSFG. Questions included: People are different in their sexual attraction to other people. Which best describes your feelings? Are you… Categories included only attracted to males, mostly attracted to males, equally attracted to males and females, mostly attracted to females, only attracted to females, not sure, and don’t know. Women were asked if they ever gave or received oral sex with another woman. Women who did not report oral sex with a woman were asked, Have you ever had any sexual experience of any kind with another female? Men were asked if they ever gave or received oral or anal sex with a man.⁴ If participants answered yes to any of these questions, they estimated their total number of lifetime same-sex sexual partners. Overall, 704 women and 212 men were included in the same-sex sexuality subsets.

The Add Health analyses examine the 256 men and 998 women who reported a different sexual identity at wave IV than wave III. The question gauging sexual identity was Please choose the description that best fits how you think about yourself. Options included 100 percent heterosexual, mostly heterosexual, bisexual, mostly homosexual, 100 percent homosexual, asexual, and don’t know.

Dependent Variable

In all regressions with NSFG data, the dependent variable was whether or not the respondent identified as straight: Do you think of yourself as… heterosexual or straight (female frequency = 5,098; male = 4,248); homosexual, gay, or lesbian (female f = 131; male = 101); bisexual (female f = 410; male = 103); and don’t know (female f = 22; male = 16). Refusals were coded as missing. In the subsets, 253 women identified as straight and 451 did not; 49 men identified as straight and 163 did not. For Add Health data, the dependent variable was whether or not respondents changed to an exclusively straight identity between waves III and IV. In total, 337 women and 113 men did so. The longitudinal design allows researchers to examine factors associated with identity change at different stages in life. I use straight versus not-straight as the dependent variable because this paper is expressly concerned with markers of difference between individuals who align themselves with sexual normativity (straightness) or non-normativity (bisexuality or lesbianism/gayness).

NSFG Independent Variables

The NSFG offered respondents four answer categories from strongly agree to strongly disagree; all attitudinal variables were recoded such that more conservative responses were given higher values.⁵ Non-attitudinal independent variables and attitudes about gays/lesbians were included in regressions for men and women, but attitudes about childrearing were included only for women and masculinity only for men.⁶ For LGBQ attitudes, an additive scale combined two items: Gay or lesbian adults should have the right to adopt children, and Sexual relations between two adults of the same sex are all right. For both men and women in all subsets, the Cronbach’s alpha was above 0.75.

The items in the NSFG most related to femininity included three questions about childrearing: People can’t be really happy unless they have children. It is okay to have and raise children when the parents are living together but not married. It is okay for a young, unmarried woman to have and raise a child. Given the social expectation that women have children and the tight link between femininity and childrearing for many women (McQuillan et al. 2008), I examine these attitudes (the second two reverse-coded) as indicators of alignment with normative femininity. While imperfect, these are the closest items measuring women’s gender conservatism in the NSFG. Hegemonic masculinity attitudes for men included: Men only need to see a doctor when they are hurt or sick. When a man is feeling pain he should not let it show. Men have greater sexual needs than women. The Cronbach’s alphas for the three masculinity items and the three femininity items were too low to combine them into additive scales.

Regressions included several other independent variables. Age ranged from 15–44, and marital status options included never married, currently married, and divorced/widowed/separated.⁷ A yes/no item assessed whether respondents had biological children or raised non-biological children. Options for race/ethnicity included Latina/o any race, non-Latina/o white, non-Latina/o black, and non-Latina/o other race. For highest educational degree, options were recoded into less than a bachelor’s and bachelor’s or above. A measure of religious attendance was recoded into never, once or twice a year, 3–11 times a year, once a month, 2–3 times per month, once a week, and more than once a week.⁸ While often over-reported, religious attendance is a good reflection of religiosity (Brenner 2011). Region included US West, Northeast, Midwest, and South. For attraction, options were recoded into (1) exclusively or mostly attracted to another sex; (2) equally attracted to women and men or mostly or exclusively same-sex attracted; and (3) unsure or missing. A variable assessing last-year sexual partners was calculated from two questions about the exact number of sexual partners in the last year. Categories included (1) no partners; (2) other-sex only; and (3) same-sex partners (including, perhaps, both men and women).⁹ The coding of the attraction and sexual behavior variables reflect the analytic interest in heterosexual identification. Respondents were more likely to identify as straight if they were only or mostly other-sex attracted, or if they had only other-sex partners in the last year.

Add Health Independent Variables

Fixed-effects models by default control for all respondent characteristics that do not change over time. Consequently, only variables that change over time were included in analyses. Political ideology was measured using In terms of politics, do you consider yourself very conservative, conservative, middle-of-the-road, liberal, or very liberal? “Don’t know” answers were assigned to “middle-of-the-road.” Religious attendance was measured using How often have you attended church, synagogue, temple, mosque, or religious services in the past 12 months? Options in wave IV included never, a few times, once a month, 2 or 3 times a month, once a week, and more than once a week. Wave III contained an additional option, “several times,” which I recoded into “a few times.” Attractions were measured using Have you ever had a romantic attraction to a [female/male]? Measuring change to last-year sexual partners was not possible because of a lack of necessary variables in wave III.

Analytic Strategy

All analyses were run in Stata 14.2 and 15.1. Person, strata, and cluster weights were applied to NSFG analyses to adjust for the complex survey design, making them generalizable to all Americans aged 15–44 living in non-institutional households. Logistic regressions using the NSFG compared (1) straight- versus LGBQ-identified women in the entire sample and (2) in the subset, as well as (3) straight- versus gay/bisexual-identified men in the entire sample and (4) in the subset. The SUBPOP function in Stata produced accurate estimates of standard errors in the same-sex sexuality NSFG subsets. Fixed-effects models using Add Health data controlled for characteristics that usually do not change over time. Models utilizing Add Health data examined associations between changes to key IVs and the DV of changing to a straight identity.

LCA analyses examined subgroupings of straight-identified men (n = 49) and women (n = 253) with substantial same-sex activity and/or attraction in the NSFG. Variables were included if they were significantly associated with straight identification in the weighted logistic regression models for subgroups of men and women with substantial same-sex activity and/or attraction. Attraction and sexual partnering variables were excluded in part because of small sample sizes, and in part to focus on other personal characteristics that characterize these subsets. Thus, homophobia was examined for both women and men; motherhood status and attitudes about single mothers were included for women, and attitudes about men showing pain were included for men.¹⁰ This paper was interested to examine whether there (1) were subgroups of straight-identified men and women, or if this population was fairly homogeneous. It was also (2) interested to examine whether variables that were markers of distinction between straight- and LGBQ-identified respondents further distinguished subgroups of straight-identified men and women.

Results

Population Proportions using the NSFG

Table 1 shows weighted sexuality estimates for all women and men aged 15–44. An estimated 10.6 percent of women and 3.9 percent of men report substantial same-sex activity and/or attraction. This is likely an underestimate, given that enduring stigma encourages under-reporting of same-sex sexuality, even on anonymous surveys (Coffman, Coffman, and Ericson 2017; Robertson et al. 2018). Differences in the proportions of women and men with same-sex sexuality partially reflect the distinct ways in which heteronormativity affects women and men. This is in part because women’s same-sex eroticism is fetishized when directed at a male audience in certain contexts, as when straight women kiss at parties to gain the attention of men (Hamilton 2007; Rupp and Taylor 2010). It is also in part because women are devalued relative to men, so there are fewer social sanctions for women who transgress certain expectations related to gender and sexuality—including same-sex sexual practices—than for men who transgress similar conventions (England 2016).

Table 2 shows weighted sexuality estimates for the subpopulation of women and men with substantial same-sex sexuality, revealing extensive variation in identity, attraction, and behavior. Within this population, 35.9 percent of women and 24.6 percent of men identify as straight. About 12.8 percent of women and 5.4 percent of men within this subpopulation have exclusive other-sex attractions, indicating that many have, or had, same-sex sex due to other elements of their sexuality not captured in the survey—such as sexual desires or romantic attractions, or past sexual attractions—or some other reason, such as experimentation or general sexual needs regardless of the partner. About 84.7 percent of women in this population have had same-sex sex, compared to 75.4 percent of men. Within the last year, 18.9 percent of women had same-sex sex and 18.6 percent had sex with both men and women, whereas 37.3 percent of men had only same-sex sex and 14.4 percent had sex with both men and women.

Weighted Mean Differences using the NSFG

Table 3 shows weighted mean differences for attitudes and religious attendance (1) among straight versus non-straight individuals with substantial same-sex activity and/or attraction; and (2) among straight individuals with and without substantial same-sex activity and/or attraction. Not shown are weighted full-population means for homophobia, which are 4.06 for women and 4.60 for men: men are more homophobic on average than women. Straight-identified women with substantial same-sex activity and/or attraction were significantly more conservative than LGBQ-identified women in terms of attitudes toward gays and lesbians (p < 0.001) and single mothers (p < 0.05), and attended religious services significantly more (p < 0.05). Similarly, straight-identified men with substantial same-sex activity and/or attraction were significantly more conservative than gay or bisexual men in terms of homophobia (p < 0.001) and attitudes about going to the doctor (p < 0.001) or showing pain (p < 0.01). Thus, straight-identified women and men with substantial same-sex activity and/or attraction were significantly more conservative and religious than LGBQ-identified respondents.

Next we turn to straight people with and without substantial same-sex activity and/or attraction. Among straight women, those with substantial same-sex activity and/or attraction were significantly less conservative about attitudes toward LGBQ people (p < 0.001), cohabitating with children (p < 0.05), and happiness without children (p < 0.01), and attended religious services significantly less frequently (p < 0.001), than straight women without substantial same-sex activity and/or attraction. In contrast, the only significant difference between straight men with and without substantial same-sex activity and/or attraction is that the former attended religious services significantly less frequently (p < 0.01). Straight women with substantial same-sex activity and/or attraction are significantly different from straight women without, but straight men with substantial same-sex activity and/or attraction are not significantly different than straight men without—excepting frequency of religious attendance.

Weighted Logistic Regressions Using the NSFG

Table 4 shows results from weighted cross-sectional logistic regressions. Attractions and last-year sexual practices were significant in all regressions in the expected direction. Otherwise, five variables emerged as significant for women and two for men. In the entire sample of women, each one-unit increase in homophobia was associated with increased odds of straight identification by 1.46 (p < 0.001). Additionally, each one-unit increase in belief that single women should not raise children was associated with increased odds of straight identification by 1.33 (p < 0.05). As compared to women without a bachelor’s degree, women with a bachelor’s degree or higher had increased odds of straight identification by 3.14 (p < 0.001). Women with children, as compared to those without, had increased odds of straight identification by 1.73 (p < 0.05). Finally, non-Latina women of “other” races had decreased odds of straight identification as compared to white women by 0.48 (p < 0.05). Results were substantively identical in the women’s same-sex sexuality subset, except that education level was just shy of significance (p < 0.055) and no categories of race/ethnicity differed from non-Latina white women. Findings about motherhood mirror what Budnick (2016) found in her interview study of straight-identified women who had sex with women. Overall, homophobic attitudes, conservative beliefs about single mothers raising children, and motherhood predict straight identification among all women and among women with substantial same-sex activity and/or attraction.

Other than attractions and last-year sexual practices, two variables were significant for men. Among all men, each one-unit increase in homophobia was associated with increased odds of straight identification by 1.50 (p < 0.001). Among men with substantial same-sex activity and/or attraction, each one-unit increase in homophobia was even more strongly associated with increased odds of straight identification: by 2.25 (p < 0.001). Belief that men should not show pain was also significantly associated with straight identification among men with same-sex sexuality by 2.44 (p < 0.05). Location in the West as compared to the South was just shy of significance (OR = 0.21, p < 0.058). Overall, homophobia was significantly associated with straight identification in both regressions, and one attitude about masculinity was significantly associated with straight identification in the same-sex sexuality subset.

Fixed-Effects Regressions Using Add Health

Table 5 shows fixed-effects models examining the relationship between social factors and changing to a straight identity. Among women, each one-unit increase to political conservatism and religious attendance was associated with increased odds of changing to a straight identity by 1.18 (p < 0.05) and 1.20 (p < 0.001), respectively. While gaining attraction to men was not significant, gaining attraction to women was associated with reduced odds of changing to a straight identity, by 0.1 (p < 0.001). For men, neither changes to political ideology nor religious attendance were significantly associated with changing to a straight identity. Gaining attraction to women was associated with increased odds of changing to a straight identity, by 3.42 (p < 0.01). Similarly, gaining attraction to men was associated with reduced odds of changing to a straight identity, by 0.16 (p < 0.001). The findings for men may reflect small sample sizes, but more research is needed. Religiosity and conservatism appear to be key for women who change to a straight identity, though causality is impossible to determine.

LCA Models Using the NSFG

The variable-centered approaches utilizing the NSFG showed that homophobia, gender conservatism, and (for women) motherhood were significantly associated with straight identification. Were there distinct classes of straight-identified women and men with same-sex sexuality along these axes, or were these populations fairly homogeneous? LCA models in table 6 show that there were, in fact, distinct classes. Among straight-identified women with substantial same-sex activity and/or attraction, there were seven main classes. The most deeply homophobic—classes 5, 6, and 7—together comprised only a moderate portion of the subset: about 25 percent. They differed in degree of homophobia and attitudes about single mothers, but all were characterized by substantial conservatism on these measures and high likelihood of being a mother. Classes 1, 2, and 4—about 16 percent, 31 percent, and 25 percent of the subset, respectively—were not very conservative and had similar chances of motherhood. Class 2 was the most liberal of all, being as accepting of LGBQ people as possible (mean: 2.00, the lowest possible) and almost as accepting of single mothers (1.71). About 65 percent were mothers themselves. This class comprised about 31 percent of the subset, showing that a substantial portion of straight women with same-sex sexuality was actually quite liberal. Of the seven classes, six—representing about 96 percent of the subset—had motherhood rates of 57 percent or higher. Putting this in conversation with Budnick’s (2016) findings, it appears that motherhood is central to straightness for many women with substantial same-sex activity and/or attraction. In contrast, there was considerable variability in terms of attitudes.

Straight-identified men with substantial same-sex activity and/or attraction were in four main classes. Class 2 comprised about 42 percent of this group, and they were neither very homophobic (2.98) nor did they strongly believe that men should not show pain (2.00). Class 1, about 14 percent of the subset, was slightly more homophobic (3.14) but also less wedded to the masculinity attitude (1.00). In contrast, classes 3 and 4 were deeply homophobic, with estimated means of 6.19 and 5.41, respectively. They differed based on alignment with hegemonic masculinity: class 3 had a moderate mean (2.00), whereas class 4 was deeply conservative (3.25). Thus, about half of straight-identified men with same-sex sexuality are neither very homophobic nor wedded to hegemonic masculinity, whereas the other half are.

Discussion and Conclusion

This paper is the first to use nationally representative data to examine variables associated with straight identification among (1) all women and men and (2) women and men who reported substantial same-sex activity and/or attraction. It also examines predictors of changing to a straight identity over a six-year period, using a supplemental dataset. Finally, it compares these variable-centered approaches to a person-centered approach—LCA—to examine variability among straight-identified women and men with substantial same-sex activity and/or attraction. It is the first to do so with generalizable data. Several key results emerged in cross-sectional regressions after controlling for attractions and sexual practices. First, homophobia was associated with increased odds of straight identification in all groups. Second, endorsement of one attitude reflecting alignment with normative femininity was associated with straight identification among women with and without substantial same-sex activity and/or attraction. One attitudinal variable about masculinity was significantly associated with straight identification among men with substantial same-sex activity and/or attraction. Finally, among women—but not men—who changed sexual identities increases in religious attendance and conservatism were associated with changing to a straight identity.

That homophobia is associated with straight identification reinforces that while overt homophobia may be declining in some contexts, it nonetheless remains related to straight identification. Regardless of the differences in women’s and men’s expressions of homophobia, the centrality of it to femininity and masculinity, and the reasons for its manifestation in women and men, homophobia is related to straight identification for both women and men. These findings indicate the need for researchers to continue exploring the relationship between homophobia and straight identification among women, since previous research has mostly emphasized men.

While results from cross-sectional (NSFG) and longitudinal (Add Health) analyses are suggestive, the direction of causality between attitudes, religiosity, and sexual identity remains unclear. Due in part to experiences with marginalization, LGBQ-identified individuals are more socially liberal than straight-identified individuals (Schnabel 2018). Women and men with same-sex sexuality may identify as straight in part because it grants them considerable privilege relative to LGBQ-identified individuals. Building on qualitative critical heterosexuality research, conservative attitudes about sexuality and gender may also influence the ways in which individuals interpret their identity, sexual practices, and other aspects of their lives. Longitudinal Add Health analyses suggest this. Undoubtedly, however, the relationships between social factors and sexual identification are highly complex.

Notable findings emerged with regard to childrearing. That motherhood was significantly related to straight identification among both groups of women suggests that many women interpret it as one part of their straight identity, reinforcing Budnick’s (2016) interview study. Further, parenthood, and especially motherhood, constrains individuals’ ability to identify as LGBQ. Parenting takes considerable time and energy—especially for women, who on average do more childcare than men in other-sex partnerships. Spending so much time and energy on children, particularly in the context of a partnership with a man, may make it seem to many women as though anything but straight identification is not possible or reasonable. This is especially the case given that many institutions with which individuals become involved after having children (e.g., schools) are often deeply heteronormative. Similarly, connecting with networks of other parents often means associating mostly with straight people. Additionally, women who have sex with men are more likely to have children than women who exclusively have sex with women. This is both because it is easier (e.g., no need for donor sperm) and because the risk of unintended pregnancy is high. Unintended pregnancies may help embed many women in straight culture, for the reasons outlined above. Many LGBQ people are parents, of course, and some of them became parents through a prior other-sex relationship (sometimes before identifying as LGBQ). There are nonetheless many constraints to identifying as LGBQ in the context of childrearing, given the way society is currently structured. That fatherhood was not significantly associated with straight identification among men may reflect that childrearing is more central to normative femininity than to normative masculinity.

Finally, latent class analyses showed major variation within subsets of straight-identified women and men with substantial same-sex activity and/or attraction. Homophobia, motherhood, and gender attitudes were significantly related to straight identification in variable-centered models. This does not mean that all straight-identified individuals are homophobic or conservative about gender attitudes, however: person-centered analyses showed that there was considerable variation on those axes. Indeed, about 56 percent of straight-identified men and 46 percent of straight-identified women with substantial same-sex activity and/or attraction endorsed neither overt homophobia nor conservative beliefs about masculinity or femininity, respectively. Most—66 percent—of straight-identified women in this subset were mothers, suggesting that motherhood is a key aspect of straightness for many of them. What tied all the classes of straight-identified women and men together was straight identification, rather than uniformity of attitude. Putting this in conversation with related interview research (Budnick 2016; Carrillo and Hoffman 2016, 2018; Silva 2017a, 2018; Walker 2014) suggests that straight-identified individuals with same-sex sexuality identify as straight in part because of their embeddedness in straight culture, including enjoyment of social benefits related to straight identification (i.e., freedom from sexuality-based discrimination and prejudice). It also reflects constraints associated with motherhood, which may make LGBQ identification more difficult for many women. Individuals’ identification as straight, as LCA analyses suggest, is not entirely due to overt homophobia or gender conservatism (see also Kuperberg and Walker 2018).

Future research could expand on this study. Longitudinal studies are needed to chart differences in sexual identity and attitudes over the life course. Using more nuanced measures of attitudes will also help match historical changes in gender and sexuality, especially given that many men adopt purportedly progressive masculinities that nonetheless leave intact sexual and gender inequalities (Bridges 2014; Bridges and Pascoe 2014, 2016). Larger sample sizes are also necessary, especially for men. In the NSFG and Add Health, subsets of men with substantial same-sex sexuality or who changed sexual identities, respectively, are small enough to warrant caution when interpreting the results.

The findings reflect how attitudes, religiosity, and identities reinforce one another in complex ways. Putting them in dialogue with those from critical heterosexuality studies highlights the utility of analyzing heterosexual identification not just as a collection of attractions and sexual practices, but also as a matrix of beliefs, interpretations, and alignment with cultural groups.

Footnotes

About the Author

Tony J. Silva is a 2018–2020 postdoctoral scholar at Northwestern University. His research interests include gender, sexuality, rural sociology, qualitative and quantitative methods, and race/ethnicity. His research examines how individuals negotiate and understand social categories that—while often perceived as natural—are actually affected by social context. As a mixed-methods researcher, he conducts research using interviews and secondary data analysis of nationally representative surveys.

References

Author notes

The author would like to thank C. J. Pascoe and David John Frank for their comments and suggestions on this manuscript.