Not just women for women: How gendered affinities affect candidate support

Abstract

Despite gains in women’s political representation, many studies continue to identify structural and attitudinal barriers that contribute to their persistent under-representation in elected office. One commonly cited explanation is a baseline reluctance among some voters to support female candidates. However, research on gender affinity suggests a more complex picture: voters often prefer candidates who resemble themselves, and especially women are more likely to support women. This identity-driven dynamic stands in contrast to a generalized bias against women and raises important questions about how gender identity, beyond binary gender, shapes voter behavior. To address these questions, we use a conjoint experiment to test whether gender affinity influences candidate preferences and how it is moderated by gender identity. Our findings reveal a consistent gender affinity effect with women respondents preferring women candidates, and men preferring men candidates. These patterns are stronger among individuals with more pronounced gender identities, that is, hyper-masculinity and hyper-femininity. Our study makes two contributions. First, by incorporating measures of gender identity, we move beyond a binary conception of gender to provide a more nuanced account of gender affinity. Second, our approach allows us to move beyond the traditional focus on affinity between women voters and women candidates to also consider how gender identity shapes political preferences among men.

Introduction

Women candidates are often expected to face an underlying reluctance from voters—a gender penalty—due to the mismatch between prevailing gender norms and culturally ingrained stereotypes of political leadership, which tend to align with masculine traits. While early studies reported modest gender gaps (e.g., Darcy and Schramm (1977)), most contemporary analyses find that women who run for office win at rates similar to men (Fox 2005; King and Leigh 2010; Thomsen and Sanders 2020).

A baseline reluctance to support women candidates is in contrast to gender affinity effects which hold that hostility is mostly toward a candidate of a different gender. Homophily, the tendency for people to associate with, build networks around, and identify with others who are similar to themselves, may draw individuals to political leaders who resemble them (McPherson et al., 2001). Political scientists refer to this dynamic as producing “affinity effects” (Dolan 2008; Plutzer and Zipp 1996; Rosenthal 1995). This social pattern can shape baseline preferences for political leaders, including along gender lines, as individuals may infer shared values or policy preferences from perceived similarity. This would lead to what Sanbonmatsu (2002) refers to as a “baseline preference” for a same gender candidate.

The gender affinity effect posits that women are more likely than men to support female candidates and rate them more favorably (Dolan 2008; Ortega et al., 2023; Sanbonmatsu 2003). Whether driven by a desire for descriptive or substantive representation, evidence for this phenomenon remains mixed (Blumenau et al., 2025; Schwarz and Coppock 2022). These mixed findings suggest that gender affinity effects are not universal but rather can be shaped by candidate characteristics (Schwarz and Coppock 2022), party dynamics (Lefkofridi et al., 2025), the broader electoral context (Badas and Stauffer 2019), and even voter characteristics (Ortega et al., 2025). To disentangle these dynamics, we move beyond binary conceptions of gender when measuring voters’ identities. At the same time, our experimental design retains binary categories of candidate gender (men and women) to isolate how respondents’ gender identities shape their evaluations. The analyses of gender affinity effects have largely neglected within-gender variation—especially among men—which may play a more substantial role than the commonly operationalized differences between men and women. If gender affinity effects are driven by social group identity, as the literature suggests, those with stronger group identities should be more prone to affinity.

To examine gender identity and affinity effects, we embedded a conjoint experiment in a post-election survey fielded after the July 2024 UK general election. A conjoint experiment asks voters to decide between two hypothetical candidates where attributes (e.g., gender and party) are randomized, reproducing the complexity of voter choice. Repeating the choice task for participants, we can estimate how changing a candidate attribute (e.g., from being a man to a woman) changes the probability of that candidate being chosen, taking into consideration the other attributes (Abramson et al., 2022; Hainmueller et al., 2014). We can also estimate the value and direction of these effects for different levels of respondent characteristics such as gender identity and the importance of gender in self-definition.

Drawing on this conjoint experiment with over 2,500 respondents rating randomized candidate profiles, we find strong gender affinity effects among hyper-masculine men and those for whom gender is central to their identity. However, contrary to expectations based on gender role theory that hyper-femininity represents more traditional views of gender, our results show that hyper-feminine women are more likely to support female candidates. These findings underscore the role of gender identity—among both women and men—in shaping voter preferences. By directly linking the intensity and salience of gender identity to candidate choice, this study offers a unique contribution to understanding how identity-based pre-dispositions can influence electoral behavior in ways not captured by traditional gender affinity explanations.

Gender bias and gender affinity effects

The concept of gender affinity in electoral behavior refers to the tendency for voters to be more likely to support same gender candidates but the focus has generally been on an affinity between women voters and women candidates (Dolan 2008; Goodyear-Grant and Croskill 2011; Sanbonmatsu 2003) in part because of their historically marginalized status in politics. Evidence suggests that women express a preference for candidates of their own gender (Holli and Wass 2010; Ono and Yamada 2020). Empirical studies have shown that women sometimes vote for female candidates even when those candidates represent rival political parties (Brians 2005; Cook et al., 1994; Plutzer and Zipp 1996), highlighting the role of shared gender identity in shaping vote choice (Giger et al., 2014; Kunovich 2012; Millard et al., 2013). Dolan (2008), for example, identified a 9 percentage point difference in U.S. House elections from 1990 to 2000, with women more likely than men to vote for female candidates. Research also highlights the potential for similar effects at the party leadership level (Campbell and Heath 2017; Goodyear- Grant and Croskill 2011; Karp and Banducci 2008; Ortega et al., 2023).

While some studies find a clear pattern of same-gender voting, with both men and women preferring candidates of their own gender, other studies find no significant effect (Goodyear-Grant and Croskill 2011; McElroy and Marsh 2010). Gender affinity effects appear contingent on context and individual characteristics: they are absent in partisan races (Badas and Stauffer 2019) and preferential systems (McElroy and Marsh 2010), but stronger among the less politically sophisticated and among men in preferential systems (Marien et al., 2017). Our study extends this line of research on individual-level conditioning factors by examining individual-level variation in gender identity.

Overall, while experimental evidence often shows modest support for female candidates, these effects vary in magnitude and are not universally consistent. This average advantage for women candidates in experiments shows that voters “do not seem to hold female candidates in disregard; all else equal, they prefer female to male candidates” (Teele et al., 2018, p. 537). A meta-analysis of experimental studies on gender bias in candidate evaluations finds that female candidates, on average, receive slightly greater support than male candidates—about a 2% increase in vote share (Schwarz and Coppock, 2022). However, this advantage is stronger among women respondents, who favor female candidates by about 3%, compared to 1% among men. Although both men and women respond positively to female candidates, the difference in the magnitude of their support is statistically significant (p = .004) but no similar gender affinity is reported for men.

Disentangling gender affinity effects

The literature on gender affinity voting presents a complex picture. Women, on average in experimental studies, may be more likely to support female candidates, but this tendency is mediated by ideological attitudes, candidate characteristics, and partisan loyalties (Campbell and Heath 2017; Dolan 2008). Similarly, men’s preference for male candidates appears to be habitual rather than driven by gender consciousness (Hellsten et al., 2007). To date, this literature has largely confined its focus to binary gender categories, with little attention to variations within genders. Our contribution to this debate highlights the role of gender identity, the variations that exist within gender categories among individuals, and their significance for understanding affinity effects.

Prior scholarship has advanced several theoretical frameworks to explain gender affinity effects, within which variations in gender identity may play a critical role. Anchored in the concept of representational congruence (Pitkin 1967), voters may prefer candidates who resemble them demographically. This logic holds that individuals are drawn to candidates who “look like them” and are presumed to share similar values or life experiences (Sanbonmatsu 2003). Gender, as one such marker, can inform a sense of descriptive representation. In a test of this representational explanation, Campbell and Heath (2017) show that women were more likely to vote for women candidates but only if they strongly supported the principle of descriptive representation with no similar impact for men. Because voters can rely on gender as a cognitive shortcut to infer whether candidate traits and policy positions reflect their own, particularly when lacking detailed information (Goodyear-Grant and Croskill 2011), gender affinity links also may be underpinned by stereotypes about candidates with, for example, women perceived as more compassionate, and competent in policy areas like education and health, while men are seen as stronger on defense and economic issues. Voters who place greater personal importance on gender or who hold more polarized gender identities may be especially inclined to view these gender similarities with candidates as politically meaningful. Another potential mechanism underlying gender affinity effects is group solidarity rooted in a sense of shared political struggle. Specifically, women with a strong awareness of systemic gender-based inequality may view support for women candidates as a form of collective action aimed at addressing under-representation (Tolleson-Rinehart 1992). This group solidarity draws on social identity theory which shows that people evaluate ingroup members more positively while discriminating against outgroups (Swan and Wyer Jr. 1997; Tajfel and Turner 2004). Social identities are thus an important basis for political attitudes and vote choice (Huddy 2013). Gender can readily structure an ingroup/outgroup divide that shapes candidate evaluations and electoral decisions (Dolan 2010; Jennings et al., 2025). In this view, voters may be more inclined to support candidates perceived as sharing their gender identity without implying women will “simply vote for any female candidate,” since identification instead reflects felt similarity and shared concerns (Winfrey and Schnoebelen 2019). Moreover, the stronger and more salient the identity, the larger its expected effect on vote choice, because the most salient identity should exert the greatest influence on political judgments (Huddy 2013; Tajfel and Turner 2004).

Integrating these mechanisms with existing concepts of gender identity, we distinguish between two dimensions: self-expression and group belonging. The first refers to individuals’ internal sense of their own gender, captured through expressions of femininity and masculinity. Following Magliozzi et al. (2016), we conceptualize this as gendered self-expression—a construct that allows for varying degrees of femininity, masculinity, and androgyny. To empirically capture these gender expressions, we follow Gidengil and Stolle (2021b), using respondents’ self-reported feelings of masculinity and femininity. This approach captures the content of individuals’ gendered self-concepts whether they see themselves as more or less masculine, feminine, or both but not how important gender is to them. Our argument is that variation along this dimension may shape political behavior through processes of identity or perceived threats.

The second dimension reflects gender as group identity, measured by how important being a man or a woman is to one’s overall sense of self (Wood and Eagly 2002). In contrast to gendered self-expression, this measure does not capture how masculine or feminine individuals feel, but instead the centrality and importance of gender as a social identity. We capture this dimension using a direct measure of the subjective importance of gender in self-definition that reflects the salience of gender as a category of belonging and aligns with theories of group consciousness and linked fate (Dawson 1995; Stout et al., 2017). By examining both dimensions, we assess how gender identity operates as a mode of personal expression and as a basis for social identification in shaping political preferences.

Our first hypothesis specifies how gender affinity is conditioned by self-expressions of gender identity:

H1

Gender affinity is conditioned by the self-expressions of gender identity such that those who are hyper-masculine or hyper-feminine will be less likely to support women candidates and more likely to support men.

This hypothesis builds on research showing that masculine and feminine identities shape political behavior (McDermott, 2016). While masculinity tends to align with right-leaning politics and femininity with the left, Gidengil and Stolle (2021a) demonstrate that hyper-masculinity and hyper-femininity both predict Trump support among white voters driven by perceived threats to traditional gender identities that are particularly potent among men (see also Coffe et al., 2023). Extending this logic to gender affinity, we argue that rigid gender identities (Murnen and Byrne, 1991) condition responses to candidates. Hyper-masculine men may perceive women’s political leadership as a challenge to traditional gender hierarchies, strengthening affinity with male candidates. Hyper-feminine women may view women candidates as norm violators, dampening affinity with women and reducing support for them (Everitt et al., 2016; Strolovitch et al., 2017).

In this account, heightened gender identity intensifies perceived threat, amplifying affinity among men while weakening it among women. An alternative perspective, however, suggests that gender identity threat operates asymmetrically. Cassino and colleagues (Cassino 2018; Cassino and Besen-Cassino 2021) argue that heightened masculinity shifts men toward political identities perceived as congruent with masculinity but a similar effect is not expected for women and femininity. Applied to gender affinity, this view implies that hyper-masculinity should strengthen men’s support for male candidates, but hyper-femininity may have limited or no parallel effect among women.

H2

Gender affinity is conditioned by the sense of group identity, such that those who see gender as more important to who they are will be more likely to support same gender candidates.

On the other hand, this hypothesis draws on theoretical expectations about the strength of social identity and ingroup affinity. The strength of individuals’ attachment to social groups varies, and this variation carries important political implications. Those with more deeply felt group identities are more likely to develop political attitudes and behaviors aligned with their group interests compared to those with weaker attachments (Huddy 2013). Bittner and Goodyear- Grant (2017) show that considering the strength of gender identities allows one to capture variations in salience within genders finding that the gender gap on policy attitudes is amplified for those whom gender identity is important.

Similarly, individuals who view their gender group as central to their self-concept—such as women expressing a “linked fate” (Dawson 1995) with other women (Stout et al., 2017)—are more likely to prioritize gender-relevant issues and support candidates representing their group. Stronger gender identity not only informs values and issue positions but also increases the likelihood of voting in alignment with the perceived interests or symbolic representation of one’s gender group, making gender affinity voting more likely among those with a stronger sense of gender identity. Thus, our second hypothesis tests this effect and we expect that men for whom being a man is more important are more likely to support men. Likewise, women for whom being a women is more important are more likely to support women.

Data and methodology

Following the July 2024 UK general election, we fielded a post-election online survey incorporating a conjoint experiment designed to assess voter preferences for female versus male candidates. While post-election timing allows us to examine gendered evaluations in a salient political context, it may also heighten the prominence of gender and campaign narratives surrounding women voters. Recent work on the 2024 UK election suggests such dynamics could amplify responses to candidate gender, making our estimates context-dependent (Campbell et al., 2024; Sanders and Gains 2024). We commissioned Deltapoll, an established British online survey agency, to conduct the survey of a representative sample of the British population (achieved n = 2511; fieldwork 6–12 July 2024) with quotas applied to the following variables: gender, age, government office region, and education. ¹

In the conjoint experiment, we present respondents with two candidate profiles each with 8 listed attributes that are assigned randomly to each candidate. The main attribute of interest is gender—a man or a woman. The other randomly varying attributes are previous political experience, occupation prior to politics, number of children, ethnicity, party affiliation, position on Brexit, and time spent in the constituency they are running in. ² Each respondent is shown five pairs of candidates, with one male and one female candidate consistently featured in each pair. Respondents are then asked “Which candidate would you prefer to vote for?” and then they select one of the candidates In our models, our estimate of interest is the probability of choosing a female candidate over a male candidate in the conjoint experiment.

To test our hypotheses about the conditioning effects of self-expressions of gender identity on affinity effects (H₁), we follow the classification of feminine and masculine identities proposed by Gidengil and Stolle (2021b)). Participants were asked two questions: “Regardless of whether they were born male or female, sometimes people differ in how masculine or feminine they feel. In general, how masculine/feminine would you say you feel?”, using a scale from 0 (“Not at all masculine/feminine”) to 6 (“Very masculine/feminine”). Based on their responses and self-reported gender, women were categorized as hyper, strong, weak feminine identities or counter-typical. For men, we coded similar categories of masculine identities. Women (men) who answered 6 in femininity (masculinity) and 0 in masculinity (femininity) were categorized as hyper-feminine (hyper-masculine). Women (men) reporting 6 or 5 in femininity (masculinity) and values between 1–3 in masculinity (femininity) were classified as strong feminine (masculine). Women (men) with femininity (masculinity) between 5–3 and masculinity (femininity) below 3 were considered weak feminine (masculine). All remaining combinations were classified as counter-typical. ³ To measure the strength of group identity, we capture the importance of gender in self-definition as an explanatory variable (H₂) using the question: “How important is your gender to how you define yourself?”.

Testing gender affinity effects in the UK 2024 elections

Given the mixed results in the literature regarding the gender affinity effect, we start by reporting the predicted probability of the interaction ⁴ between the respondent’s reported gender and the candidate’s gender in the conjoint analysis on the probability of choosing one candidate over another. To do this, we fit a generalized linear mixed-effects model with random effects for individuals to account for repeated measures (i.e., five tasks), controlling for individual factors such as age, education, parenthood, sexual orientation, political interest, and party voted for in the 2024 election. In this first model, we estimate without gender identity as a covariate to assess baseline effects.

We also acknowledge the methodological concerns raised in the recent literature. Abramson et al. (2022) argue that Average Marginal Component Effects (AMCEs), commonly used in conjoint experiments, conflate the intensity and direction of preferences and therefore may not reflect the true proportion of voters who prefer one attribute level over another. In response to this limitation, we report AMCEs in the text to show the relative effect of each attribute level compared to a baseline, and visualize predicted probabilities to convey the overall distribution of preferences across attribute levels. We also provide upper and lower bounds on the proportion of the respondents in our sample that prefer a same gender candidate for each subgroup. ⁵ Because we have a large number of candidate profiles, the bounds have a large range and can be inconclusive about majority preferences over a single attribute such as the gender of the candidate. Our main interest remains to show how these effects vary by gender subgroups.

Figure 1 shows significant differences between same gender pairs of candidates and respondents indicating initial support the gender affinity hypothesis, but mostly for women. Among women respondents, the predicted probability of choosing a female candidate is approximately 3.2 percentage points higher than that of choosing a male candidate. On the other hand, among men respondents, the probability of choosing a male candidate over a female one is reduced approximately 1.45 percentage points. However, confidence intervals overlap for men. Overall, these findings provide evidence of gender affinity effects, especially for women.

Figure 1.

The Role of Gender on Candidate Selection.

Beyond baseline gender affinity, our core objective is to test whether the strength and salience varies across categories of gender identity. In Figure 2, we test whether gender identity rigidity shapes candidate preferences—specifically, whether individuals with strongly gendered self-identifications (hyper-masculine or hyper-feminine) are more likely to support male candidates. Here we report the predicted probabilities for supporting a same-gender candidate by subgroups of gender of the respondent and their level of gender identity. The findings offer partial support for Hypothesis 1 (H₁). Among men (left panel), those identified as hyper-masculine show a preference for male candidates (Dif = 7.8%). In contrast, the pattern for women (right panel) runs counter to our expectations: Women with hyper-feminine identities are more likely to favor female candidates (Dif = 4.7%). This result challenges prior findings from Gidengil and Stolle (2021a), suggesting that gender affinity may, in some cases, override the influence of political values typically associated with rigid gender identities. While H₁ is only partially confirmed, the results nonetheless demonstrate that gender identity exerts a significant and measurable influence on vote choice within the controlled conditions of a conjoint experiment.

Figure 2.

The Interaction of Gender and Gender Identity on Candidate Selection.

The bounds estimate shown in Table A6 in the Appendix confirms our results and supports our conclusions regarding H₁. Among women, the strongest evidence of gender affinity is found among voters with hyper-feminine and strongly feminine identities, suggesting that a substantial proportion of these voters actually prefer female candidates. Even women with weak gender identity show a moderate effect. Only in the counter-typical group, that is, those whose gender expression deviates from conventional feminine norms, is the effect ambiguous. Among men, the results reconfirm that only hyper-masculine respondents show a clear preference for male candidates, which is consistent with our earlier probability prediction results. In contrast, other male subgroups show ambiguous or null effects, including the “Strong” and “Weak” masculine groups, whose bounds range from near zero to near one.

Drawing on theories of social identity and ingroup favoritism, Hypothesis 2 (H₂) proposes that individuals who attach greater importance to gender are more inclined to support candidates of the same gender. Figure 3 offers partial support for this claim. Among men, those who view gender as a central aspect of their identity are significantly more likely to prefer male candidates. This pattern, however, does not extend to women: regardless of how important gender is to their self-perception, their candidate preferences remain unchanged ⁶ . These findings suggest that the salience of gender moderates the gender affinity effect, but only for men.

Figure 3.

The Interaction of Gender and Importance of Gender on Candidate Selection.

Conclusions

Our findings show that gender affinity is neither uniform nor symmetrical. Women consistently prefer female candidates, especially those individuals with strongly feminine identities. Among men, affinity for male candidates appears only under specific identity conditions, most notably among hyper-masculine respondents and those for whom gender is central to their self-concept. Gender affinity is not simply a function of shared demographic traits: we show that affinity effects can be shaped by self-expressions of gender identity and the salience of gender as a social identity (Huddy 2013).

These results challenge binary framings of the gender affinity debate. Instead of asking whether voters penalize women or whether women “vote for women,” our evidence indicates that gender-based voting can be moderated by gender identity. The critical division exists not only between men and women, but also within each group. Consistent with recent work on gender identity (Gidengil and Stolle 2021a), our findings show that treating gender as a fixed demographic variable obscures significant heterogeneity in political behavior. Furthermore, our findings were not contingent on interest or education suggesting that gender can serve as more than a shortcut to voting decisions.

The asymmetry observed among men has broader implications for contemporary identity politics. Support for male candidates is concentrated among men with rigid or highly salient masculine identities, suggesting that candidate choice may serve as a means of identity affirmation. This pattern resonates with scholarship linking masculinity, perceived status threat, and right-wing mobilization (Coffe et al., 2023; Strolovitch et al., 2017). Under these conditions, resistance to women candidates may reflect not generalized bias, but identity-based reactions among specific constituencies. However, hyper-feminine women’s greater affinity for women candidates is counter to a threats-based explanation and suggests instead those with hyper-feminine identities have preferences of aligned with progressive women.

These findings highlight the need to further clarify what “hyper” gender identities capture. Self-reported hyper-femininity or hyper-masculinity may reflect adherence to traditional gender roles, progressive gender consciousness, or perceived threats to established hierarchies. Future research could distinguish these pathways more directly, as they may operate differently for masculinities and femininities and carry distinct implications for democratic representation.

Supplemental material

Supplemental material - Not just women for women: How gendered affinities affect candidate support

Supplemental material for Not just women for women: How gendered affinities affect candidate support by Alejandro Tirado Castro, Susan Banducci in Research & Politics

Supplemental material

Supplemental material - Not just women for women: How gendered affinities affect candidate support

Supplemental material for Not just women for women: How gendered affinities affect candidate support by Alejandro Tirado Castro, Susan Banducci in Research & Politics