Sexual minorities: tolerance & change

Democratic societies are increasingly marked by increased self-reported support for sexual and gender minorities and LGBTQ+ rights (Adamczyk Reference Adamczyk2017). Indeed, and rather than viewing rising LGBTQ+ rights as an outcome of democracy, Velasco, Baral and Tang (Reference Velasco, Baral and Tang2024) argue that LGBTQ+ rights and the inclusion of LGBTQ+ persons are composite components of (liberal) democracy itself. Since the Netherlands’ entrepreneurial legalisation of equal marriage in 2001 (Kollman Reference Kollman2017), institutional recognition for same-sex couples has and is becoming, despite resistance (Ayoub Reference Ayoub2014; Ayoub and Stoeckl Reference Ayoub and Stoeckl2024; Mos Reference Mos2020; Velasco Reference Velasco2023), an increasingly normalised position among liberal regions across the globe (Kollman Reference Kollman2007). Although not devoid of prejudicial biases – which, as demonstrated by data on reported hate-crimes (Flores et al. Reference Flores, Stotzer, Meyer and Langton2022), instances of state-sponsored homophobia (López Ortega Reference López Ortega2024), and data on LGBTQ+ individuals’ lived experiences (Ammaturo and Slootmaeckers Reference Ammaturo and Slootmaeckers2024), remain – the social inclusion of sexual and gender minorities has transformed dramatically (Flores Reference Flores2019). The rate of change on the social inclusion of this minority population has been described as remarkable (Harrison and Michelson Reference Harrison and Michelson2017), and has far out-paced what might be expected from the natural forces of cohort replacement.

Several causal mechanisms have been identified to explain this transformation. In addition to generational replacement – which, according to Ekstam (Reference Ekstam2022), does indeed explain some of this change – evidence signals that individuals have also actively updated their views towards sexual and gender minorities (Ayoub et al. Reference Ayoub, Page and Whitt2021; Brookman and Kalla Reference Brookman and Kalla2016; Turnbull-Dugarte and López Ortega Reference Turnbull-Dugarte and López Ortega2024). A frequent-cited causal explanation can be found in increased inter-group contact (Ayoub and Garretson Reference Ayoub and Garretson2017; Flores Reference Flores2015; Garretson Reference Garretson2018; Jones et al. Reference Jones, Brewer, Young, Lambe and Hoffman2018; Lewis Reference Lewis2011). Whether it be social contact in psychical space or parasocial contact via a diverse media diet (Ayoub and Garretson Reference Ayoub and Garretson2017; Miller et al. Reference Miller, Flores, Haider-Markel, Lewis, Tadlock and Taylor2020), exposure to and meaningful interactions with LGBTQ+ individuals result in increased positive affect towards these individuals and support for policies that expand their welfare (Garretson Reference Garretson2018). Elites are often found to exhibit strong top-down effects on shaping inclusion (Flores and Barclay Reference Flores and Barclay2016). Top-down influence is observed not only in citizens responding to explicit elite-based cues (Flores and Barclay Reference Flores and Barclay2016; Jones and Bower Reference Jones and Bower2020), but also via state-sponsored inclusion. State recognition of LGBTQ+ rights can, via a positive policy feedback effect, engender a shift in societal norms that results in higher levels of inclusion for a minority that the state has moved to legitimise (Abou-Chadi and Finnigan Reference Abou-Chadi and Finnigan2019). Given social norms around the acceptability (or not) of anti-queer prejudices condition the prevalence of prejudicial behaviour (Masser and Phillips Reference Masser and Phillips2003), the norm-shaping role of state action and the ‘nationalisation’ of LGBTQ+ tolerance among western nations (Lægaard Reference Lægaard2007) has played an important role in transforming inclusion even if, in some instances, the inclusion of LGBTQ+ individuals is strategic as opposed to genuine (Turnbull-Dugarte and López Ortega Reference Turnbull-Dugarte and López Ortega2024).

Whether it be a consequence of this transformation in tolerance, or indeed a causal component behind the expansion of tolerance itself (Flores Reference Flores2015; Jones et al. Reference Jones, Brewer, Young, Lambe and Hoffman2018; Lewis Reference Lewis2011), more and more citizens in liberal democracies identify with diverse sexual and gender identities. According to comparative survey data from thirty countries as part of Ipsos’ (2023) Pride Study 2023 Global Survey, the average rate of LGBTQ+ identification sits at 9%. Notably, this proportion is even greater among young people – those born after 1997 (often termed ‘Gen Z’) – with 18% of this group identifying with an LGBTQ+ identity; the majority of which are bisexual (9%). In those societies where LGBTQ+ tolerance is highest, more individuals are comfortable living their authentic selves, as demonstrated by greater levels of non-heterosexual identification (Turnbull-Dugarte Reference Turnbull-Dugarte2020). As new emerging cohorts of the population are socialised in societies where not identifying with the sexual majority group or engaging in gender-inclusive, sexually diverse, and traditionally gender-incongruent activities are (more) normalised, young cohorts are notably more inclined to adopt these identities.

Indeed, and as visualised in Figure 1, sexuality-based dynamics in the population are dramatically changing. According to British Election Study (BES) data (Fieldhouse et al. Reference Fieldhouse, Green, Evans, Mellon, Prosser and Bailey2023), which provides a very large and well-powered sample (N = 30,363), 5.6% of respondents aged 65 or over identify with a non-heterosexual identity. This proportion increases over successive cohorts, and among those who recently joined the adult population, the proportion is now in excess of 34%.Footnote ¹ Within the emerging cohort (those aged 18–24), there is also a sizeable gender gap in adopting diverse sexual identities. Thirty per cent of young men opt for one of these identities, but the proportion is even larger among young women at 38%. This asymmetry presents a puzzle. Assuming sexual diversity is an ascriptive characteristic that is randomly distributed across the population, these proportions should be as-good-as random. One, bi-erasing and sexist (Wurthmann and López Ortega Reference Wurthmann and López Ortega2024), argument presented to explain this variation is that women falsify their true heterosexual identities (Alarie and Gaudet Reference Alarie and Gaudet2013; Cipriano et al. Reference Cipriano, Nguyen and Holland2023) and claim to be bisexual because doing so increases their perceived desirability (Dodge et al. Reference Dodge, Herbenick, Friedman, Schick, Fu, Bostwick, Bartelt, Muñoz-Laboy, Pletta, Reece and Sandfort2016). An alternative explanation is that, rather than women falsifying their heterosexual preferences (biasing proportions of bisexual women upwards), men are more inclined to falsify their sexually modern preferences (biasing the proportion of bisexual men downwards). Should this latter explanation hold, one would expect there to be a larger cost on the dating market for men who express bisexual preferences and, importantly, for this penalty to be anticipated.

Figure 1. Generational shift in LGB+ identification.

The emerging population is increasingly diverse in terms of its sexual preferences. Importantly, the rising proportion of bisexual-identifying individuals and those who identify as neither heterosexual nor gay/lesbian presents those in the sexual majority group with a challenging test of social inclusion and tolerance. Straight individuals on the dating market, particularly those among the youngest in the population, are likely to engage with potential partners who have engaged with sexually inclusive and traditionally gender-incongruent activities. Do the sexual majority group (still) discriminate against this minority?

Despite rising (self-reported) tolerance towards sexual and gender minorities, alongside advances in LGBTQ+ visibility, there is evidence that prejudicial biases remain. In liberal democracies, lesbian, gay, and transgender political candidates are still subjected to a sizeable electoral penalty. Magni and Reynolds (Reference Magni and Reynolds2021), for example, present robust cross-national experimental evidence from Britain, New Zealand, and the USA to demonstrate that these candidates suffer electoral penalties vis-à-vis cis-heterosexual candidates. Importantly, the authors find that these discriminatory preferences are not observed by all voters and are notably smaller (if, however, still present) among younger citizens in the electorate (Magni and Reynolds Reference Magni and Reynolds2021).

Prejudicial biases against sexual and gender minorities are, of course, not limited to explicitly political processes like candidate selection. Individuals with diverse sexual identities often experience a penalty on the labour market (Baert Reference Baert2018; Drydakis Reference Drydakis2015; Flage Reference Flage2020), and similar discriminatory outcomes are observed for trans (Aksoy et al. Reference Aksoy, Carpenter and Sansone2024) and non-binary (Eames Reference Eames2024) individuals. Anti-LGBTQ+ biases have also been observed in distinct social scenarios including housing provision (Jansson and Fritzson Reference Jansson and Fritzson2022; Lauster and Easterbrook Reference Lauster and Easterbrook2014), healthcare access (Fumarco et al. Reference Fumarco, Harrell, Button, Schwegman and Dils2024), immigration processes (Hoffman and Velasco Reference Hoffman and Velasco2024), and parenting rights (Dotti Sani and Quaranta Reference Dotti Sani and Quaranta2020; López Ortega and Blanco Reference López Ortega and Blanco2025; Turnbull-Dugarte Reference Turnbull-Dugarte2024), as well as in general everyday social interactions (Hendren and Blank Reference Hendren and Blank2009). While self-reported tolerance may have increased and more LGBTQ+ individuals are willing to live openly as their authentic selves, this rich empirical evidence demonstrates that intolerance remains present, if, however, it is now simply more closeted. While developing social norms may make it less acceptable for individuals to publicly express their prejudicial biases (Masser and Phillips Reference Masser and Phillips2003), that does not mean these biases have disappeared but that they are simply harder to observe.

Building upon this literature, the first pre-registered hypothesis posits

Gender, LGBTQ+ tolerance, & gender-congruent norms

Research consistently demonstrates a gender gap in tolerance towards LGBTQ+ individuals, with women, on average, generally exhibiting greater (self-reported) support for LGBTQ+ rights and higher levels of social acceptance of sexual and gender minorities compared to men (Herek Reference Herek2002; Kuyper et al. Reference Kuyper, Sommer and Butt2018). This disparity is often attributed to a catalogue of causal processes including socialisation processes that encourage empathy and nurturing behaviours in women: women’s exposure to gendered prejudice fosters a more inclusive attitude towards other marginalised groups, women and LGBTQ+ individuals are more prone to engage in meaningful social interactions via overlapping social networks which result in more between-group contact.

An alternative explanation, however, suggests that the gender gap in LGBTQ+ tolerance emerges from homonegativity among men as opposed to social liberalism among women given that – at least in liberal societies – self-reported support for LGBTQ+ rights has lower social costs for women than it does for men (Morgan and Davis-Delano Reference Morgan and Davis-Delano2016; Wellman and McCoy Reference Wellman and McCoy2014) given heterosexuality reinforces men’s gender capital (Bridges Reference Bridges2009) and projects masculinity (Dean Reference Dean2014). Men perceive costs associated with being LGBTQ+ inclusive and they, therefore, disassociate from socially inclusive positions in order to avoid being misclassified as non-heterosexual (Bosson et al. Reference Bosson, Prewitt-Freilino and Taylor2005). When ‘being a man’ requires performative masculinity; supporting policy positions and being socially tolerant of those that defy the men-as-masculine trope, then men are socially incentivised to distance themselves from these issues (Bosson et al. Reference Bosson, Weaver, Caswell and Burnaford2012; Vandello et al. Reference Vandello, Upton, Wilkerson, Kubicki and Kosakowska-Berezecka2023; Wellman and McCoy Reference Wellman and McCoy2014).Footnote ² In short, rather than women being ‘over-supportive’, men are ‘under-supportive’.

Regardless of these alternative explanations, I find empirical confirmation of the gender gap in LGBTQ+ tolerance via data from the European Social Survey (ESS). Figure 2 reports the results of a meta-analysis that models the gender gap in LGBTQ+ tolerance across all participating country-years of the ESS using the three different measures of LGBTQ+ tolerance included in the survey.Footnote ³ The resulting coefficients (N = 425) demonstrate the widespread validity of the gender gap across different survey items, national contexts, and time.

Figure 2. European gender gap in LGBTQ+ tolerance.

A logical corollary of this widespread empirical finding, and regardless of the reason behind this gap, would be that women should be more accommodating to sexually modern and gender-inclusive dating partners compared to men. My theoretical proposition, and pre-registered empirical hypothesis, however, is that the gender gap in diffuse tolerance for LGBTQ+ rights and individuals will be reversed when it comes to both preferences and behaviours on the dating market.

The central rationale for this expectation is driven by the anticipated mechanism: gender incongruence is more socially norm defiant for men than it is for women, and that women’s private preferences contribute to enforcing this norm. First, there is variation in the social norm in liberal democratic societies related to the promotion of gender-incongruent behaviour of men and women. Gender incongruence is more norm-defiant when performed by men than by women (Bosson, Taylor and Prewitt-Freilino Reference Bosson, Taylor and Prewitt-Freilino2006; Feinman Reference Feinman1981; Heilman and Wallen Reference Heilman and Wallen2010; Rudman and Fairchild Reference Rudman and Fairchild2004). Across diverse experimental settings, a clear pattern has emerged: gender-defiant men are significantly and more negatively evaluated than gender-defiant women (Blakemore Reference Blakemore2003; Costrich et al. Reference Costrich, Feinstein, Kidder, Marecek and Pascale1975; Heilman and Wallen Reference Heilman and Wallen2010; Kane Reference Kane2012; Miller et al. Reference Miller, Bilimoria and Pattni2000; Magni and Reynolds Reference Magni and Reynolds2021; Motro and Ellis Reference Motro and Ellis2017; Sanborn-Overby and Powlishta Reference Sanborn-Overby and Powlishta2020; Sullivan et al. Reference Sullivan, Moss-Racusin, Lopez and Williams2018). Evolving and modernising social norms have increasingly encouraged and facilitated women to engage in activities traditionally associated with masculinity whereas the same is not true for men engaging in feminine behaviours (Eagly Reference Eagly1987; Eagly et al. Reference Eagly, Nater, Miller, Kaufmann and Sczesny2020; Kane Reference Kane2006; Miller et al. Reference Miller, Bilimoria and Pattni2000). As such, women who engage in gender-incongruent behaviour – although still rejected by individuals and institutions with misogynistic preferences (Ridgeway Reference Ridgeway2011; Rudman and Glick Reference Rudman and Glick2001; Rudman et al. Reference Rudman, Moss-Racusin, Phelan and Nauts2012; Shah Reference Shah2024) – are not defying an established social norm. Expectations around masculinity have remained remarkably entrenched and provide very little room for manoeuvre (Bosson et al. Reference Bosson, Prewitt-Freilino and Taylor2005; Eagly Reference Eagly1987; Wellman and McCoy Reference Wellman and McCoy2014).

This creates a double standard where men’s nonconformity to rigid heteronormative standards is seen as more transgressive and more socially undesirable than women’s.Footnote ⁴ In the case of sexually inclusive dating preferences, this is likely to lead heterosexual women to view sexually inclusive men as far less masculine (ie gender congruent). Bisexuality, by challenging conventional and rigid markers of masculinity (Bosson et al. Reference Bosson, Prewitt-Freilino and Taylor2005; Eagly Reference Eagly1987; Wellman and McCoy Reference Wellman and McCoy2014), disrupts a societal expectation about how men should and should not behave which is likely to produce a greater penalty for sexually inclusive (and therefore, norm-defiant) men on the dating market.

Second, despite a strong preference for gender equality, women’s support for traditional gender norms in their interpersonal relationships remains remarkably entrenched.Footnote ⁵ A robust empirical finding across observational and experimental designs signals that heterosexual women place a premium on demonstrations of masculinity when it comes to their preferences on the dating market (Alba et al. Reference Alba, Hammond and Cross2023; Ekrami et al. Reference Ekrami, Claes, Shriver, Weinberg, Marazita, Walsh and Van Dongen2021; Miller et al. Reference Miller, Bilimoria and Pattni2000; Ridgeway Reference Ridgeway2011; Sadalla, Kenrick and Vershure Reference Sadalla, Kenrick and Vershure1987; Snyder et al. Reference Snyder, Kirkpatrick and Barrett2008). Some instances of the masculinity premium are ascriptive – heterosexual women care more about their partner’s height than heterosexual men (Courtiol et al. Reference Courtiol, Picq, Godelle, Raymond and Ferdy2010; Gillis and Avis Reference Gillis and Avis1980; Pierce Reference Pierce1996; Stulp, Buunk and Pollet Reference Stulp, Buunk and Pollet2013) – whereas others are based on status – heterosexual women show a strong preference for potential partners that earn more than they do, whereas men neither reward nor penalise women based on their relative income (Ong and Wang Reference Ong and Wang2015; Rudman and Fairchild Reference Rudman and Fairchild2004; Whyte et al. Reference Whyte, Brooks, Chan and Torgler2021). In both instances, women reward gender-congruence and penalise gender incongruence more than men when it comes to mate selection. Importantly, men are aware of this asymmetry and engage in explicit acts of gender conforming behaviour when they anticipate a penalty from others in response to signs of incongruence (Rudman and Fairchild Reference Rudman and Fairchild2004).

Moreover, and independent of support for pro-feminist gender equality measures, women remain, on average, more supportive of traditional heteronormative dating rituals compared to men. Women are, according to Lamont (Reference Lamont2020), trapped between the simultaneous pursuit of new goals (equality in the public sphere) and old norms (gender roles in the private sphere). Women are more inclined than men to believe that women should take their spouse’s surname, that men should initiate dating contact, and pay for dates (Lever et al. Reference Lever, Frederick and Hertz2015; Ridgeway Reference Ridgeway2011; Serewicz and Gale Reference Serewicz and Gale2008). This is, of course, a product of women’s own socialisation into patriarchal society where their preferences and values have likely been shaped by what is expected of them. The stickiness of such preferences suggests, however, that, as empirically well-established (Glick and Fiske Reference Glick and Fiske1996, Reference Glick and Fiske2001), benevolent sexism not only remains widespread among society but is also prevalent among many women who have been socialised under norms of gender-congruent behaviour (Alba et al. Reference Alba, Hammond and Cross2023; Lamont Reference Lamont2014; Wu et al. Reference Wu, Luo, Klettner, White and Albritton2023).

H2 (gendered penalty thesis): Heterosexual women will be more inclined – vis-à-vis heterosexual men – to oppose potential partners that have engaged in sexual activity with men and women.

Study I – observational data & double-list experiment (UK & Spain)

Identifying prejudice is methodologically complex. As well-documented in diverse empirical work focused on different social groups, prejudicial attitudes are often hidden (Blinder, Ford and Ivarsflaten Reference Blinder, Ford and Ivarsflaten2013; Kuklinski, Cobb and Gilens Reference Kuklinski, Cobb and Gilens1997) because of social norms and socially desirable expectations that individuals anticipate to be expected of them. In a social context where a respondent expects that revealing their true preference (eg that they hate Beyoncé) is socially unacceptable, they are likely to mask their true preference and report what is expected (eg that they like Beyoncé). Self-reported measures are, therefore, potentially unreliable given they are often tainted by this sensitivity bias.

In Study I, and in addition to direct self-reported measures, I leverage an identification strategy based on a double-list (also known as dual-list) experiment. List experiments are uniquely placed to answer this research question given that the list experiment approach allows for the identification of population-level estimates of sensitive information while significantly minimising this sensitivity bias. Rather than directly soliciting an endorsement of belief on an individual sensitive issue – in this case, if dating someone who is romantically engaged with men and women – the list experiment approach invites respondents to report the total number of statements or items they agree with in a given list. Such an approach allows respondents to indicate their beliefs (more) truthfully than they would have had they been asked about their beliefs directly. By randomly assigning respondents to lists that vary with regard to the presence (treatment) or absence (control) of the sensitive item of interest, one can observe how the inclusion of the sensitive item among those in the treatment condition alters the total count and, as a result, one can identify the proportion of the population who share that belief.

The pre-registeredFootnote ⁶ experimental design used in Study I is summarised in Table 1. Following advancing developments in the design of list experiments (Diaz Reference Diaz2023; Glynn Reference Glynn2013; Turnbull-Dugarte, López Ortega and Hunklinger Reference Turnbull-Dugarte, López Ortega and Hunklinger2025), I apply the double-list experiment model. Essentially, the double-list experimental approach is equivalent to running two sequential list experiments and, therefore, allows the researcher to significantly increase statistical power. Maximising statistical power is important given, as demonstrated by Blair, Coppock and Moor (Reference Blair, Coppock and Moor2020), most list experiments are not sufficiently powered (Rosenfeld, Imai and Shapiro Reference Rosenfeld, Imai and Shapiro2016). In addition to randomising exposure to the treatment (long-list) condition, the double-list design also randomises i) under which baseline (eg List A vs. List B) individuals receive treatment, ii) the order in which each list was presented. This design matches the randomised-randomised variant in Diaz (Reference Diaz2023)’s four-point typology of variations in double-list experiment designs.

Table 1. Double-list experimental design

The sensitive item of interest in the double-list experiment relates to a partner engaging in same-sex behaviour. Study I manipulates sexual diversity based on behaviour (has sex with and dates both men and women) as opposed to an identity (is bisexual).Footnote ⁷ As displayed in Table 1, List A and List B include a different catalogue of non-key items. As recommended by Glynn (Reference Glynn2013), I include an equal baseline proportion (0.5) of negatively correlated items that respondents will agree and disagree upon in order to limit the potential for ceiling and floor effects. As per the pre-registered design, the baseline includes a strategically selected balance of values that respondents are assumed to, on average, view as undesirable in a partner. These include: living with their parents, admiring Vladmir Putin, having more than twenty sexual partnersFootnote ⁸ (List A); calling their opposite sex parent everyday, having an OnlyFans profile, and losing their virginity at fourteen (List B). Note that the order of items within a list, including the sensitive item, is also subject to randomisation.

Random allocation of the sensitive item to different lists facilitates the identification of potentially socially undesirable preferences without introducing sensitivity bias that may result in preference falsification. Through the difference-in-means estimator, the proportion of respondents who view a potential partner that is and has been sexually involved with men and women can be estimated.Footnote ⁹ The sensitive item includes the word ‘frequently’ strategically to signal habitual behaviour as opposed to a onetime same-sex experience. While this might prime promiscuity (and stereotypes about bisexuals being promiscuous) beyond just the sexual diversity information, additional signals of promiscuity were strategically included to allow for informational equivalence (Dafoe, Zhang and Caughey Reference Dafoe, Zhang and Caughey2018).

Study I: Results

I begin by considering respondents’ self-reported views opting for romantic partners who have engaged in (traditional) gender-incongruent behaviour. In Figure 3, I visualise the proportion of men and women who signalled that bisexual behaviour would be a ‘red flag’ for them when considering a potential partner. Two things stand out from the responses to this question: not only is there a difference in the relative tolerance for a partner that engages in gender-inclusive sexual activity between men and women, but there is also a divergence in their absolute tolerance for this behaviour. According to this self-reported item, women are, on average, more likely than not to evaluate a bisexual partner negatively. The same is not true for men who report a mean probability of doing so comfortably below 50% of the time. The difference between men and women is significant (p < 0.001) and substantive: given a baseline probability of indicating that sexually diverse preferences are undesirable among men of 0.42, the sixteen-point gap between men and women means that the penalty among women is 38% larger.

Figure 3. Self-reported views on dating bisexuals among men and women.

As in the case of the direct observational measure, the double-list approach demonstrates sizeable and significant heterogeneity in tolerance of engagement in gender-incongruent sexual behaviour between men and women. Consider the combined sample of respondents from Britain and Spain (left-hand column). Compared to men, women are ten percentage-points (p< 0.05) more likely to consider a partner that has sex with men and women to be undesirable. Given the baseline prevalence of prejudice among men is 0.37, a ten percentage-point gap for women equates to a difference of 27%.

Soliciting the item-counts based on both the respondents’ own preferences as well as those of their peer group, facilitates the distinction between what respondents perceive to be a normalised preference from their own preference (Troyer and Younts Reference Troyer and Younts1997). Among the combined sample of respondents from Britain and Spain, the gender gap is symmetrical – ten percentage-points higher among women – regardless of whether one evaluates tolerance based on partners being negatively evaluated by peers or by the self. There is, however, variation between the two countries considered in the analysis. In Britain, the difference in tolerance towards a sexually diverse partner is significantly larger when personal preferences are considered over those of peers. British women are fifteen percentage-points more inclined to believe a partner that has sex with men and women would be considered undesirable by their peers and 23 percentage-points more likely to consider the same to be true for themselves. The difference in the difference (eight percentage-points) is also significant (p< 0.1) and signals a specific rejection for women personally that is beyond what can be explained by socialised expectations. The same is not true of Spain where the observed penalty is statistically symmetrical for men and women regardless of whether one considers individual or peer-based preferences.Footnote ¹⁰

Figure 2 demonstrated that men tend to express lower levels of tolerance and support for LGBTQ+ individuals than women. The empirical results presented in Figures 3 and 4 reveal, however, that the same gender gap does not travel to considerations of individuals’ own intimate relationships. In this case, the gender gap is actually reversed, with men exhibiting – by means of both self-reported and experimentally identified preferences – increased acceptability of a gender-inclusive partner compared to women.

Figure 4. Prevalence of prejudice against gender-incongruent behaviour.

Study II – visual conjoint experiment (UK & Spain)

The objective of Study II is to assess whether the estimated prevalence of prejudicial preferences towards sexually inclusive partners extends to a behaviour-based setting. To do so, I pre-registered a visual Footnote ¹¹ conjoint survey experiment (López Ortega and Radojevic Reference López Ortega and Radojevic2024; Turnbull-Dugarte and López Ortega Reference Turnbull-Dugarte and López Ortega2025) that randomly exposes respondents to visually manipulated fictional dating profiles. Respondents were presented with a single profile sequentially and asked to behave in a way that they normally would on a dating application (N = 39,488). There is no forced comparison between opposing profiles; ten individual profiles were presented to respondents successively. The outcome measure of interest is behavioural.

Rather than coercing a preference or asking respondents to self-report the likeability and/or desirability of a profile, respondents were asked to swipe on one of three emoticons presented below each profile, as customary on the dating app Tinder, in order to behaviourally evaluate the presented dating profile. The emoticons included a red cross (indicating reject), a blue star (indicating ‘superlike’), and a green heart (indicating like).Footnote ¹² The profiles presented rely on faces originally sourced from the Chicago Face Database (Ma, Correll and Wittenbrink Reference Ma, Correll and Wittenbrink2015), which were selected on their comparable scores on attractiveness. Using a sample of ten base faces that varied on gender and ethnicity, I then leveraged an AI-based facial editing algorithm to create eight variations of a single face. These iterations resulted in a total population of eighty faces that varied the profile’s gender, ethnicity, age, and emotion (smiling or neutral expression), as well as their gender-congruence.Footnote ¹³ The eight variations of a single face reproduced in Figure 5 represent an illustrative example. All of the facial manipulations for each of the ten base faces are available for consultation in the supplementary material.

Figure 5. Example eight variations in base face.

To ensure real-world external validity, respondents are never exposed to a variation of a single face (or the randomly assigned accompanying name) more than once. Respondents were exposed to a total of ten profiles. In real terms, this means that respondents with single-sex preferences were exposed to a single variation, of the eight potential variations, from each of the initial population of ten faces. Individuals who expressed being attracted to men and women were exposed to a random subsample of five men and five women. Relying on experimental manipulations of a smaller pool of base faces, and including fixed-effects for these, allows us to control for those features in faces that I cannot easily observe (eg mandibular symmetry or biocular width) but which may well influence a profile’s probability of success during initial evaluations on the dating market (Valentine et al. Reference Valentine, Li, Penke and Perrett2014).

The core explanatory attribute of interest in Study II is identification as bisexual on presented profiles. This information was manipulated in the accompanying profile description, and profiles were randomly tagged as bisexual with a probability of 0.25. In addition to the facially manipulated characteristics detailed above, the visually presented dating profiles also manipulated an individual’s occupation, education, distance from respondent, dating preferences (eg hookup or serious partner), the presence of nationalist symbols, the presence of information on pronouns (Britain) or diverse (vegan) diet (Spain), facial hair (men only), hair length (women only), political partisanship, as well as three hobby tags. Simultaneous randomisation, a core feature of the value-added from conjoint designs (Hainmueller et al. Reference Hainmueller, Hopkins and Yamamoto2014), facilitates the identification of the average effect of a concrete attribute value (eg bisexuality) independently of all other attribute values included in the manipulation. The inclusion of hobbies and dating preferences is of theoretical importance given the potential for identification of sexual diversity to result in statistical discrimination (Phelps Reference Phelps1972). In the online Appendix Table B.1, I summarise the full list of attributes and their potential values.

Typical examples of a visually manipulated profile are reproduced in Figure 6, which are presented to a respondent attracted to men (a) or women (b). The analysis I report in the main paper focuses on respondents who identify as heterosexual only: I am interested in how the majority sexuality group responds to the minority sexually diverse population (bisexuals). The comparison between heterosexual and non-heterosexual respondents is, however, explored in the discussion as part of the robustness checks (see Appendix Figure B.1).

Figure 6. Examples of visually manipulated dating profiles.

Study II: Results

I begin by discussing the overall results among the full sample, combining men and women in Britain and Spain. Figure 7 reports two estimands: the marginal mean (MM) and the average marginal component effect (AMCE). The MM indicates the favourability toward dating profiles that express bisexual identification or not, marginalising across all other attributes and their corresponding values (Hainmueller, Hopkins and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2014). The AMCE indicates the change in the favourability for profiles caused by their being bisexual.

Figure 7. Overall prejudice in online dating market.

Despite likely increasing exposure to more sexually diverse individuals on the dating market, individuals discriminate significantly based on signals of this behaviour. Marginalising across the influential effects of all other attribute values, profiles identified as bisexual are subjected to an eight percentage-point (p < 0.001) penalty compared to heterosexual profiles. The probability of positive selection on the online dating market for a heterosexual profile is 0.46. As a result, an eight percentage-point penalty for being bisexual is not small and equates to a 17% change in dating market success.

Figure 8 compares the MM and the AMCE among men and women. As demonstrated by the left panel, which reports the MM – and consistent with observational evidence from online dating applications – women are more selective than men (Berkowitz et al. Reference Berkowitz, Tinkler, Peck and Coto2021). The baseline probability of selection for women is 0.35, and for men, it is 0.56. Both men and women apply a penalty on potential partners when their online dating profile signals they engage in sexual activity with men and women. While bisexual women are subjected to a five percentage-point penalty from men, bisexual men are subjected to a nine percentage-point penalty from women. The difference in the penalty between the two is significant (p< 0.01). Moreover, given the divergence in the baseline levels of positive selection for heterosexuals among women and men already, the substantive penalty exhibited by the former is also three times larger at -24% than that exhibited by the latter (-8.6%). As in the case of Study I, which manipulates sexual modernity via expressed behaviour, I show that the majority sexuality group insulates itself and penalises those with more gender-inclusive preferences. The results also show – as demonstrated by Study I and the observational evidence that supports it – that there is a sizeable gender component to this penalty with men exposed to notably higher costs of being sexually inclusive compared to women.

Figure 8. Gendered penalty in online dating market.

As part of an additional exploratory analysis – that is, analysis not included in the pre-registered analysis plan – I also consider if the presence of an overall binegativity penalty and the sizeable gendered nature of this penalty is conditioned by political leanings. To do so, I report the results of a subgroup analysis that stratifies respondents into four groups based on gender and left-right partisanship. The results (visualised in Figure B.3) demonstrate that the penalty exhibited against men by left-wing women and right-wing women is symmetrical. Regardless of the political leanings, women apply a penalty on bisexual men equal to nine percentage-points. Political preferences matter more for men: the penalty exhibited by left-wing men is significantly lower than left-wing women (four percentage-points), whereas for right-wing men, the penalty is statistically indistinguishable from women. This additional test speaks to the robustness of the main findings – even among those respondents we would identify as the least likely to penalise sexually inclusive individuals, we still observe a substantive penalty. That the penalty applied by left-wing women is more than 100% larger than that applied by left-wing men speaks to the sizeable penalty sexually inclusive men experience, even within a sub-population that we would assume to be more accommodating of sexually inclusive behaviours.

Study III: Testing the gender incongruence mechanism (UK)

So far, the observational and multi-experimental evidence has evinced two clear patterns. First, individuals who engage in sexual activity with both men and women are penalised. The sexual majority group penalises those who operate in both the sexually heteronormative and sexually inclusive arena. This is true regardless of whether this sexual inclusivity is signalled via behaviour (Study I) or via sexual identity (Study II). Second, this penalty is notably greater for men than it is for women. Theoretically, and despite women’s increased support for LGBTQ+ rights vis-à-vis men, the proposed mechanism to explain the gendered nature of this penalty is that there exists a stricter social norm around gender-congruent behaviour for men than is the case for women, and that simultaneously, women contribute to maintaining this norm (Kane Reference Kane2012; Ridgeway Reference Ridgeway2011).

Should this mechanism be at play, one would expect to observe two things. First, individuals anticipate the gendered nature of the penalty. If indeed gender incongruence is more socially costly for men than women, we should observe that individuals hold an empirical expectation of this penalty (Bicchieri Reference Bicchieri2017). Second, potential partners who express sexual attraction towards multiple genders – in addition to being less desirable on the dating market – should also be perceived as less gender-congruent and this incongruence signal should be greater for women considering a male partner, than for men considering a female partner. The value-added of Study III is to test these expectations and assess if this gender-norm violation mechanism explains the gendered nature of the significant prejudicial biases identified against men.

To assess this, I fielded a new survey among individuals of dating-market age in Britain. The survey included both an instrument that would allow us to observe empirical expectations on the gendered penalty, as well as an experiment that assesses if men with sexually inclusive dating preferences are viewed by heterosexual women to be less gender-congruent than comparable women are by heterosexual men.

The experimental design randomly exposed respondents to a simple vignette with a fictitious profile that manipulated sexuality. As in the case of Study I, I manipulated sexuality based on behaviour rather than identity. Treatment was manipulated in expressions of the type of romantic partner the profile was looking for: ‘An openminded [girl/girl or guy] who likes jumping on a plane for a weekend away as much as I do!’ (see Appendix C.1 for examples).

After being presented with the profile, respondents were asked how likely it is that they would date the presented individual and how likely it is that they would recommend a friend to date the individual. These outcomes serve as a replication of the distinctive preferences from Study I: those for oneself and those for others. Building on this, and as a test of the core theorised mechanism, I also asked respondents to evaluate the profile against a number of traits: femininity, masculinity, attractiveness, whether the individual appears to be trustworthy, would be a faithful partner, or is sexually adventurous. The pre-registered hypothesis anticipates that respondents randomly assigned to see the sexually inclusive profile will view the profile as less gender-congruent and that this perceived gender incongruence will be greater for women evaluating men than vice versa. I operationalise gender incongruence as men being less masculine and women being less feminine. I also report effects on all other traits (Table 2) – there are no gender differences in the effect of treatment on these additional traits. As in the case of the previous studies, I limit the analysis to those who identify as heterosexual; tests including non-heterosexuals are reported in the Appendix.

Table 2. Treatment effects on profile traits conditioned by gender

p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001.

Study III: Results

Do respondents anticipate a gendered penalty against men with gender-inclusive sexual preferences? Should a norm exist whereby gender incongruence among men is more socially costly than it is for women, individuals should believe that sexually inclusive men will be subjected to this penalty (empirical expectations). They do.

In Figure 9, I demonstrate that men and women harbour the same empirical expectations. Citizens anticipate that women will punish gender-inclusive preferences among men. Simultaneously, citizens do not anticipate the same penalty to be exercised by men against women. There is a social norm that penalises gender incongruence in sexual preferences, that norm is gendered, and that norm is anticipated equally by men and women. As demonstrated in Study I and Study II – and I will now show below in Study III – the empirical expectation that heterosexual men reward bisexual women in the dating market is not observed in any of the experimental manipulations. Instead, and like bisexual men, bisexual women are penalised, if however, to a lesser extent than men.

Figure 9. Empirical expectations: do individuals anticipate a gendered penalty?.

Figure 10 reports the core findings from Study III. The left-hand panel reports the overall average treatment effect (ATE) as well as the conditional average treatmrnt effect (CATE) among men and women on two core outcomes: whether they would encourage a friend to date the individual in the profile, or if they would personally date them. The former evaluates second-order preferences (respondent believes the profile will be considered desirable by others) and the latter evaluates first-order preferences (respondent personally believes the profile is desirable). The results visualised corroborate the empirical evidence presented so far: sexually inclusive profiles are significantly penalised; the penalty exhibited by heterosexual women against bisexual men is significantly larger than the reverse when considering personal (first-order) preferences but not those of others (second-order). In terms of the magnitude, the penalty among women (0.14) is 100% greater than that among men (0.07).

Figure 10. Mechanism test: Gendered nature of the gender incongruence penalty.

The right-hand panel of Figure 10 provides the core value-added of Study III and allows for the identification of the underlying mechanisms behind the gendered component of the sexually inclusive penalty. Women are significantly prone to perceive men who date both men and women as less gender-congruent. The same is not true for men. On average, straight women perceive the straight man profile to be gender-congruent at 0.76 on a 0-1 scale. The -0.086 CATE for the profile presented as sexually inclusive equates to a penalty of 11%. For heterosexual men, the straight woman condition is perceived to have a gender-congruent score of 0.80 and manipulating gender-inclusive sexual preferences does not significantly alter this. The CATE (-0.025) is indistinguishable from zero at conventional levels (p <0.05) and represents only a 3% change from the heterosexual baseline. The difference in the CATE observed among men and women is also significant (p < 0.01) and sizeable (0.06). In real terms, manipulating sexual preferences among men and women results in a penalty to the perceived gender-congruence among heterosexual dating partners that is 400% greater for women evaluators than it is for men evaluators. As theorised, the significantly greater penalty for sexually inclusive men compared to sexually inclusive women is the result of heterosexual women perceiving men who date both men and women as less ‘manly’.

Threats to inference: Alternative explanations

Theoretically, one might make the case that the rejection of potential partners that engage in gender-incongruent sexual behaviour may be a function of alternative mechanisms other than prejudicial preferences and discriminatory penalties against gender incongruence.

First, rejecting individuals who select sexual and romantic partners from a wider population might be considered rational, given that doing so limits dating market competition. A heterosexual opting for an individual sexually attracted to men and women increases market competition by competing not only with individuals of their own gender but also those of the opposing gender. Should competition-based rationality drive these effects, however, a similar significant rejection should be observed among lesbian women and gay men who, if willing to date individuals attracted to men and women, open themselves up to a shift in market competition that is notably greater in magnitude than would be observed among heterosexuals. I rule out this possibility by comparing the effects conditioned on respondent sexuality (see Figures B.1 and B.2). There is no observed penalty among lesbian women or gay men in Study II, and, in the case of Study III, I observe a premium for bisexual profiles among non-heterosexual respondents (see Appendix C). Dating market competition is, therefore, not an alternative explanation supported by the data.

Second, perhaps the negative gender gap emerges because of a double standard towards bisexuals: gender-inclusive sexual preferences among men may be stigmatised by heterosexual women because it undermines gender norms around masculinity, while, simultaneously, gender-inclusive sexual preferences among women may be eroticised and fetishised by heterosexual men. This double standard may result in sexual modernity providing women with a premium as opposed to a penalty, which, consequently, produces a divergence between the median preference among heterosexual men and women. While this is a potential theoretical explanation, as shown across the three experiments, it is not a behaviour identified in the data. Men do not reward sexually inclusive women with a premium on the dating market but are significantly and negatively biased against them, even if, however, to a lesser extent than women. Empirically, I also demonstrate that experimentally manipulating a profile’s sexually inclusivity does not result in any significant variation in men’s and women’s perceptions of the profile as sexually adventurous (see Table 2).

Third, heterosexuals’ rejection of potential partners who have sex with men and women may be based on statistical discrimination (Phelps Reference Phelps1972); a process via which individuals rely on information they hold about an underlying association between sexually diverse behaviour and adjacent items. For example, an individual may assume that those who have been engaged in sexual activities with men and women might have different values around romantic partners (eg monogamy), relationship stability, or sexual adventurism. Likewise, individuals might assume that those who engage in gender-incongruent sexual activity may also be inclined to engage in other gender-incongruent hobbies or social activities. Both the double-list experiment and the visual conjoint experiment were designed in a way to rule out this possibility by maximising informational equivalence (Dafoe, Zhang and Caughey Reference Dafoe, Zhang and Caughey2018). In the latter double-list, sexual adventurism was included as one of the non-sensitive items (number of sexual partners). In the visual conjoint experiment, the desired relationship type (eg stable partner) and social hobbies were simultaneously randomised specifically to allow us to identify the effect of being bisexual, independent of these adjacent factors. Moreover, in Study III – which explicitly tests opposing mechanisms – I find no evidence of gendered asymmetry across various different traits, with the exclusion of gender incongruence, including: attractiveness, trustworthiness, faithfulness, and sexual adventurousness. Both heterosexual men and heterosexual women consider sexually inclusive partners to be less faithful, but so do to a symmetrical degree. This view is not shared by LGB men and women.

Fourth, the gender distinctions observed across the observational and the three pre-registered experiments may be a result of heterosexual women being more likely to misclassify men engaged in gender-incongruent sexual behaviour as gay, and heterosexual men misclassifying women engaged in gender-incongruent sexual behaviour as straight. Such an expectation is based on evidence of bisexual erasure (and biphobia) which assumes that bisexuality does not exist: bisexual men are really gay men falsifying their homosexuality (Dodge et al. Reference Dodge, Herbenick, Friedman, Schick, Fu, Bostwick, Bartelt, Muñoz-Laboy, Pletta, Reece and Sandfort2016; Morgenroth et al. Reference Morgenroth, Kirby, Cuthbert, Evje and Anderson2022) and bisexual women are really straight women engaging in attention-seeking behaviour and masking their heterosexuality (Alarie and Gaudet Reference Alarie and Gaudet2013; Cipriano, Nguyen and Holland Reference Cipriano, Nguyen and Holland2023). The original data I leverage above does not include an instrument to assess gendered differences in bi-erasure. Publicly accessible survey data, however, suggests this is not the case. In Appendix Figure D.3, I show that according to data from the YouGov’s online tracker, women are less inclined, compared to men, to believe that sexuality is dichotomous rather than a spectrum. Moreover, as highlighted by Wurthmann and López Ortega Reference Wurthmann and López Ortega2024) using representative survey data from Germany, observational evidence shows that i) the prevalence of bi-erasure is low, and ii) women are significantly less likely to harbour bi-erasing views compared to men.