Though women are still underrepresented at all levels of office, from local to statewide to the US Congress (CAWP 2023), in 2024, there is a record high number of women in state legislatures (2,430), serving as state governors (12), and serving in statewide elective executive office (99). Despite these strides toward gender parity, women still face various challenges spanning institutional and structural barriers, potential baseline gender bias from voters, and the existence of harmful societal stereotypes. This article focuses on the latter and asks if existing stereotypes about political leaders are still gendered and as biased against women as previous research suggests. Building on role congruity theory (Eagly and Karau Reference Eagly and Karau2002), a range of studies conducted among American respondents in previous decades demonstrate that women candidates, relative to men, are perceived as higher in so-called Horizontal traits like warmth, empathy, and compassion, but are also seen as lower in Vertical traits like agency, competence, and assertiveness (e.g., Alexander and Andersen Reference Alexander and Andersen1993; Eagly and Karau Reference Eagly and Karau2002; Huddy and Terkildsen Reference Huddy and Terkildsen1993a; Matland Reference Matland1994; Sanbonmatsu Reference Sanbonmatsu2002; Sanbonmatsu and Dolan Reference Sanbonmatsu and Dolan2009; for distinction between Horizontal and Vertical traits see below and Abele et al. Reference Abele, Ellemers, Fiske, Koch and Yzerbyt2021). Because Vertical traits are held as key personality characteristics for successful or “prototypical” leaders (Eagly and Karau Reference Eagly and Karau2002; Koenig et al. Reference Koenig, Eagly, Mitchell and Ristikari2011), it has been theorized that the existence of such stereotypes constitutes an important hindrance for women seeking political office (Eagly and Karau, Reference Eagly and Karau2002; but see also Wiezel et al. Reference Wiezel, Barlev, Martos and Kenrick2024). Relatedly, women are seen as more competent on “soft” policy issues like education and aid to the poor, whereas men are seen as better on “hard” policy issues like the economy and foreign policy (Cassese and Holman Reference Cassese and Holman2018; Holman Reference Holman2023; Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2016; Lawless Reference Lawless2004). In total, these studies highlight the existence of gendered trait and issue competence stereotypes that perpetuate the broader stereotypic notion of “think leader, think male,” representing important obstacles for women pursuing a political career.
However, despite the wealth of research supporting the existence of gendered candidate stereotypes in American society, recent meta-analytical work on broader leader stereotypes (i.e., outside the domain of politics) tells a different story. Evolving over time, leader stereotypes are now less strongly connected with masculinity (Eagly et al. Reference Eagly, Nater, Miller, Kaufmann and Sczesny2020; Koenig et al. Reference Koenig, Eagly, Mitchell and Ristikari2011). Consistent with these trends, studies of stereotypes about men and women political candidates from various countries also support the notion of shifting stereotype content (e.g., Devroe Reference Devroe2021; Lucciola Reference Lucciola2023). This article seeks to clarify the degree to which American citizens’ trait and issue competence stereotypes about political candidates are gendered, and if the content has changed in line with the evolution in broader leadership stereotypes. Moreover, we explore if the content of individuals’ candidate trait and issue competence stereotypes vary systematically as a function of gender, political orientation, and sexist attitudes. To answer these questions, we leverage two conjoint experiments embedded in a survey fielded among a large and diverse sample of American citizens (N = 1,200) in the spring of 2022. Results largely support that Americans’ candidate stereotypes have changed in line with the reported trends in broader leadership stereotypes: women candidates are still perceived as higher in Horizontal traits and corresponding policy issues like education and aid to the poor. Importantly, contrary to conventional wisdom, women candidates do not suffer any deficits on Vertical traits or corresponding policy issues like the economy and national security compared to men. Furthermore, moderation analyses reveal that liberals and respondents low in hostile sexism hold stereotypes that most strongly counter conventional wisdom. Consequently, trait and issue competence stereotypes likely constitute smaller obstacles for women candidates now than previously, and in line with other recent scholarly work (Schwarz and Coppock Reference Schwarz and Coppock2022), our results suggest that demand-side voter bias may not constitute the central reason for women’s underrepresentation; a point we return to and elaborate on in the discussion.
Gendered Trait and Issue Competence Stereotypes About Political Candidates
Stereotypes are general expectations or shared beliefs about members of particular social groups, typically defined by characteristics like race, age, or gender (Ellemers Reference Ellemers2018). In her broad review of gender stereotypes Ellemers (Reference Ellemers2018, 277) explains that “gender categorizations are immediately detected, are chronically salient, seem relatively fixed, and are easily polarized. This contributes to the formation and persistence of gender stereotypes and reinforces perceptions of differences between men and women.” In other words, gender stereotypes are omnipresent in society, and it is hardly surprising that citizens also use political candidates’ gender to form candidate stereotypes. Importantly, such candidate stereotypes come in two interrelated yet different versions: trait stereotypes and issue competence stereotypes.Footnote 1 Below, we review key findings about both trait and issue competence stereotypes and explain how such stereotypes potentially hinder women’s representation.
Trait Stereotypes
Scholars generally agree on the existence of gendered trait stereotypes about political candidates. Dolan (Reference Dolan2010, 71) states that “Numerous studies and surveys indicate that voters believe female politicians are warmer and more compassionate” [than men], and that “Male politicians are seen as strong and intelligent.” While we agree with this summary of the literature, we do, however, also want to stress the much longer lists of adjectives employed to capture and characterize the content of trait stereotypes for women and men candidates, respectively. Previous studies report that women candidates — compared to men — are perceived as more compassionate, collaborative, warm, emotional, affectionate, sensitive, gentle, caring, honest, trustworthy, family-oriented, talkative, and cautious. Conversely, men candidates — compared to women — are perceived as more intelligent, strong, competent, logical, assertive, self-confident, tough, arrogant, course, rational, experienced, decisive, possessing strong leadership skills, and ambitious (Bauer Reference Bauer2015; Devroe and Wauters Reference Devroe and Wauters2018; Dolan Reference Dolan2014a; Dolan and Lynch Reference Lynch and Dolan2014; Fridkin, Kenney, and Woodall Reference Fridkin, Kenney and Woodall2009; Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2016; Huddy and Terkildsen Reference Huddy and Terkildsen1993a, Reference Huddy and Terkildsen1993b; Lawless Reference Lawless2004; Lefkofridi, Giger, and Holli Reference Lefkofridi and Holli2019; Sanbonmatsu Reference Sanbonmatsu2002). Importantly, according to role congruity theory “a potential for prejudice exists when social perceivers hold a stereotype about a social group that is incongruent with the attributes that are thought to be required for success in certain classes of social roles” (Eagly and Karau Reference Eagly and Karau2002, 574). In connection to political representation of women, one clear hindrance to more equal representation is that the content of stereotypes about female candidates overlaps much less with impressions of “the ideal leader” (i.e., being intelligent, decisive, experienced, etc.) than does the content of stereotypes about male candidates. This discrepancy creates role incongruity (Eagly and Carli Reference Eagly and Carli2007; Glick and Fiske Reference Glick, Fiske, Crosby, Stockdale and Ropp2007).
Despite the richness of the insights produced across the cited articles, we believe that conclusions about trait stereotype content are less clear (than one might wish) because of the lengthy and varied lists of adjectives employed across studies. First, the many different adjectives complicate cross-study comparisons with heterogeneity in findings across studies, possibly caused by either different trait measurement questions, “real” differences, or some combination of the two. Second, from a purely theoretical perspective it also appears suboptimal to characterize the content of trait stereotypes about political candidates without building a larger theoretical framework for the main dimensionality of such stereotypes.
One solution to these challenges, we argue, is to reconsider the content of trait stereotypes about men and women candidates based on the dimensionality of validated and well-established models of social perception and social judgement from the psychological sciences (e.g., Abele and Wojciszke Reference Abele and Wojciszke2007; Fiske et al. Reference Fiske, Cuddy, Glick and Xu2002; Koch et al. Reference Koch, Imhoff, Dotsch, Unkelbach and Alves2016). This comes with at least two advantages. First, it explicitly grounds insights on trait stereotypes about political candidates in models aiming to capture social and group-based stereotypes broadly speaking (e.g., about ethnicity, race, age, class, and gender to name a few).Footnote 2 Given that these stereotype models have uncovered the automatic ways characteristics such as gender, age, and race are employed for forming impressions of new acquaintances outside the domain of politics (e.g., Fiske et al. Reference Fiske, Cuddy, Glick and Xu2002), it seems natural to also assume that the same processes are at play when citizens form impressions of unknown political candidates (or of candidates about whom only little relevant political information is available). Second, applying these broader stereotype models to political candidates also comes with the clear advantage of possibly comparing the content of stereotypes across different yet related objects such as political, business, and military leaders, as well as between politicians and ordinary citizens from the same group (for an interesting such comparison see Schneider and Bos Reference Schneider and Bos2014). A main takeaway from a recent theoretical synthesis of the models of social perception and judgment (Abele et al. Reference Abele, Ellemers, Fiske, Koch and Yzerbyt2021) is that social perceptions and stereotypes are formed along two fundamental dimensions labelled the Vertical and Horizontal dimension, respectively. While “The Vertical dimension relates to evaluation of status, prestige, esteem, power, and skill…The Horizontal dimension relates to benevolence, trust, cooperation, liking, belief compatibility/similarity, and value sharing” (Abele et al. Reference Abele, Ellemers, Fiske, Koch and Yzerbyt2021, 8). Moreover, the two fundamental dimensions are further split into two sub-dimensions each, with the Vertical sub-dimensions labeled “Ability” and “Assertiveness,” and the Horizontal sub-dimensions labelled “Morality” and “Friendliness.”
Applying this theoretical framework on the summarized work on gendered trait stereotypes about political candidates, it immediately becomes clear that the long lists of adjectives presented above neatly fall along the major Vertical and Horizontal trait dimensions. Specifically, stereotypes about women candidates center around the Horizontal sub-dimensions, Morality and Friendliness. In contrast, stereotypes about men candidates come much closer to “ideal leadership traits,” with content closely related to the Vertical sub-dimensions Ability and Assertiveness. We will refer to these trait dimensions as we progress.
Issue Competence Stereotypes
The second type of gendered candidate stereotypes relates to perceptions of candidates’ competence at handling various political issues. On the one hand, men are seen as more capable than women to deal with traditionally “harder” issues like national security, terrorism, the economy, and crime (e.g., Devroe and Wauters Reference Devroe and Wauters2018; Dolan, Reference Dolan2004; Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2016; Huddy and Terkildsen Reference Huddy and Terkildsen1993a). On the other hand, women candidates are often perceived as more competent than men when it comes to “softer” issues like education, aid to the poor, health care, childcare, reproductive health, and the arts (e.g., Dolan Reference Dolan2004; Dolan and Lynch Reference Dolan and Lynch2016; Fridkin, Kenney, and Woodall Reference Fridkin, Kenney and Woodall2009; Huddy and Terkildsen Reference Huddy and Terkildsen1993a; Kahn Reference Kahn1992). Relating trait stereotypes to these issue competence stereotypes, one easily sees how men’s stereotypic advantages on Vertical trait sub-dimensions — Ability and Assertiveness — lead voters to also form stereotypic impressions of men being more capable to handle national security and the economy. Likewise, women’s stereotypic advantages on Horizontal trait sub-dimensions — Morality and Friendliness — are linked to women being perceived as more competent on aiding the poor, sick, and elderly. Finally, because issues with stereotypes favoring men are generally more salient and decisive for voting behavior (Duch and Stevenson Reference Duch and Stevenson2008; Soroka and Wlezien Reference Soroka and Wlezien2010), issue competence stereotypes and trait stereotypes are seen as working in tandem to hinder women candidates’ electoral fortunes.
Expectations About Contemporary American Trait and Issue Competence Stereotypes
Based on the above summary of key findings from previous studies on trait and issue competence stereotypes about political candidates among Americans, it seems straight forward to form the following predictions:
H1a : Women candidates are seen as higher in Horizontal traits (i.e., Friendliness and Morality) than men, and more competent on “softer” policy issues. In contrast, male candidates are seen as higher in Vertical traits (i.e., Ability and Assertiveness) than women, and more competent on “harder” policy issues.
We will refer to this as our “conventional wisdom expectation.” Importantly, recent work in political science and related fields suggests that such a “conventional wisdom expectation” does not come unchallenged. First, meta-analyses on leader stereotypes and gender stereotypes broadly speaking (i.e., outside the political domain) have uncovered how the content of stereotypes has evolved over time. Specifically, Koenig et al. (Reference Koenig, Eagly, Mitchell and Ristikari2011, 616) report that the “masculine construal of leadership has decreased over time,” and Eagly et al. (Reference Eagly, Nater, Miller, Kaufmann and Sczesny2020) conclude that women have maintained an advantage in communion (i.e., Horizontal) traits, but have closed or diminished the gender gap in perceptions of Ability and Assertiveness (i.e., Vertical traits). This is likely due to women’s changing role in society, as well as accompanying shifts in traditional gender attitudes (Eagly et. al. Reference Eagly, Nater, Miller, Kaufmann and Sczesny2020). Women are now fully integrated into the labor force and have made significant education gains (Kim and Sakamoto Reference Kim and Sakamoto2017). Furthermore, there have been documented shifts toward more egalitarian attitudes toward women’s work and family roles in the United States (Donnelly et al. Reference Donnelly, Twenge, Clark, Shaikh, Beiler-May and Carter2016).
Second, work in political science also gives credence to the notion that gender stereotypes and gender bias more broadly are shifting. Dolan (Reference Dolan2014b), as well as Hayes and Lawless (Reference Hayes and Lawless2016), find no support for the notion that women face a deficit on stereotypically “masculine” issues and traits in US politics. Even more recently, Devroe (Reference Devroe2021) finds that stereotypes about political candidates in Flanders (Belgium) are less gendered than conventional wisdom would assume, and Hargrave and Blumenau (Reference Hargrave and Blumenau2022) show how the rhetorical style of British MPs has changed to become less gendered now than in the past. Lastly, Lucciola (Reference Lucciola2023, 507) concludes that Brazilian voters even “hold positive stereotypes of women and a broad pro-female bias.” In total, these different findings give rise to a “rival expectation”:
H1b : Trait and issue competence stereotypes about American political candidates are less gendered now than previously, and women candidates are possibly evaluated similarly to (or even more positively than) men on the Vertical trait sub-dimensions, Ability and Assertiveness, and on competence on “hard” policy issues.
Below, we seek to uncover the content of trait and issue competence stereotypes about men and women candidates in contemporary American society, clarifying if the “conventional wisdom expectation” or “the rival expectation” has more support.
Exploring Heterogeneity in Trait and Issue Competence Stereotypes
In connection to pitting “the conventional wisdom expectation” and “the rival expectation” against each other, we also explore if certain sub-groups in the population are driving changing stereotype content. First, meta-analyses on gender stereotypes and gender bias both broadly speaking (Eagly et al. Reference Eagly, Nater, Miller, Kaufmann and Sczesny2020; Koenig et al. Reference Koenig, Eagly, Mitchell and Ristikari2011) and in the domain of politics (Schwarz and Coppock Reference Schwarz and Coppock2022) find that men and women — despite following the same general trend — also tend to express more favorable stereotypes and biases toward their own gender. Therefore, we explore if contemporary citizens in American society hold different trait and issue competence stereotypes about political candidates based on their own gender. Second, liberalism and left wing orientations are more closely connected with gender egalitarian values than are conservatism and right wing orientations (e.g., Gothreau, Warren, and Schneider Reference Gothreau, Warren and Schneider2022; Johnson and Tamney Reference Johnson and Tamney2001; Jost Reference Jost2017; Kteily et al. Reference Kteily, Rocklage, McClanahan and Ho2019; Sevincer et al. Reference Sevincer, Galinsky, Martensen and Oettingen2023). Furthermore, recent work shows how right wing (left wing) members of German and Austrian parliaments perceive women ministers as less (more) competent compared to their male counterparts (Dingler and Kroeber Reference Dingler and Kroeber2022); that right wing and left wing party members show preferences for male and female politicians, respectively (van der Pas, Aaldering, and Steenvoorden Reference van der Pas, Aaldering and Steenvoorden2022); and that potential Republican and Democratic campaign donors support and favor men and women candidates, respectively (Ondercin and Dalton Reference Ondercin and Dalton2023). Thus, ideological orientations possibly also moderate content of trait and issue competence stereotypes about men and women candidates. Finally, given the obvious relevance of sexist predispositions for understanding the role of gender in politics (e.g., Kinder, Reynolds, and Burns Reference Kinder, Reynolds, Burns and Berinsky2020; Winter Reference Winter2023), we investigate how the gendered content of Americans’ trait and issue competence stereotypes about political candidates possibly vary with hostile and benevolent sexism (Bareket and Fiske Reference Bareket and Fiske2023; Glick and Fiske Reference Glick and Fiske1996; Gothreau, Arceneaux, and Friesen Reference Gothreau, Arceneaux and Friesen2022).
Empirical Approach: Conjoint Experiment
To test the outlined predictions, we conducted two conjoint experiments: The “Trait Experiment” (n = 650) and “Issue Competence Experiment” (n = 550). Both experiments were part of the same larger survey and respondents were randomly assigned to participate in one or the other (see details below). The conjoint experimental approach has become increasingly common in political science (Atkeson and Hamel Reference Atkeson and Hamel2020; Kirkland and Coppock Reference Kirkland and Coppock2018; Teele, Kalla, and Rosenbluth Reference Teele, Kalla and Rosenbluth2018) as it comes with several desirable features. First, the conjoint experiment allows us to test if respondents freely and voluntarily apply traits and issue competences differentially to men and women candidates. Most existing studies either directly ask respondents to indicate if certain traits or issues are more tied to men or women candidates in an abstract manner, or assign respondents randomly to experimental conditions varying the degree to which certain traits describe a male or a female candidate (e.g., Huddy and Terkildsen Reference Huddy and Terkildsen1993a, Reference Huddy and Terkildsen1993b; Sanbonmatsu Reference Sanbonmatsu2002). Employing the conjoint experimental design, we can test if respondents voluntarily (i.e., without any instructional encouragement) assign certain traits or issues to men or women candidates when candidates also vary on a range of other relevant attributes besides their gender. Second, via the ability to vary many different candidate attributes simultaneously, the conjoint design disguises our interest in candidate gender. In the study, respondents are exposed to a range of candidates, all — besides gender — varying on attributes like age, experience, and marital status. Consequently, it is (almost) impossible for respondents to guess the purpose of our study, which alleviates social desirability concerns and improves internal validity. Finally, the conjoint design constitutes a close approximation of real election environments in which voters evaluate and make judgements about candidates based on various considerations and types of information. Thus, the conjoint experimental design, we argue, comes closer to imitating real electoral experiences compared to more traditional survey-based experiments (see also Ono and Burden Reference Ono and Burden2019).
Sample
Our study was fielded to a diverse and relatively large sample of American respondents via the survey platform Prolific in May 2022. The sample is broadly representative of the adult population in the United States with respect to gender, age, and income when compared to census data (United States Census Bureau 2022), and is large enough to uncover differences in respondents’ ascription of trait and issue competences to men and women candidates in the conjoint experiment (if they exist) (Schuessler and Freitag Reference Schuessler and Freitag2022; Stefanelli and Lukac Reference Stefanelli and Lukac2022). Our sample, like most online samples, is more educated than the general population in the US, though likely more representative than an in-person convenience sample (Berinsky, Huber, and Lenz Reference Berinsky, Huber and Lenz2012; Coppock Reference Coppock2019). Sample details and demographics are provided in Table A1 in the Appendix. Respondents who completed the 10-minute-long survey were compensated $1.30.
Procedure and Design of Conjoint Experiments
After consenting to participateFootnote 3 and completing demographic information, respondents answered various attitudinal questions, after which they were randomly assigned to either the Trait Experiment or the Issue Competence Experiment. In both experiments, respondents were instructed that they were about to see a “…series of candidates running for the U.S. House of Representatives”, after which they saw a total of seven randomly generated candidate profiles (one at the time) with the following instruction: “Imagine that during the next election the following candidate is running for the U.S. House of Representatives. Please read the candidate profile carefully and tell us to what extent you agree that the candidate possesses the traits below/has the competency to deal with the following policy issues.”Footnote 4 Respondents then answered either the trait or issue competence questions presented below. We primarily chose to conduct our studies in the context of American politics because the vast majority of previous work on candidate trait and issue competence stereotypes is based on American data. Thus, the American setting directly permits comparing our results to previously obtained conclusions regarding stereotype content about candidates. Moreover, situating our studies in the context of elections for the US House of Representatives serves the purpose of increasing realism and familiarity in the sense that respondents are used to being exposed to multiple candidates during an election.
Following procedures in existing studies, candidates displayed to respondents in the experiments varied across seven different attributes (Ono and Yamada Reference Ono and Yamada2020; Teele, Kalla, and Rosenbluth Reference Teele, Kalla and Rosenbluth2018), which appeared in random order across respondents (but remained fixed for each individual respondent): Gender (male, female), age (30–2, 45–7, 65–7 years), number of children (0, 1, 3), occupation (teacher, lawyer, business/finance, farmer), previous experience (0, 1–2, 4–5, 9–10 years in politics), race (white, Black), and marital status (unmarried, married).Footnote 5 Because we were interested in the direct effect of candidate gender on ascription of traits and issue competences to candidates, we did not experimentally vary candidates’ party affiliation. Importantly, for this reason, effects of gender could possibly be affected by gendered partisan stereotypes (Hayes Reference Hayes2005; Winter Reference Winter2010) — that is, perceiving women (men) candidates as Democrats (Republicans), a possibility we address in the discussion. Finally, after participating in either the Trait Experiment or the Issue Competence Experiment, respondents participated in another study included in the survey (unrelated to our interests in this article), before they were debriefed and thanked for their participation.
Dependent Variables: Trait and Issue Competence Application
Trait Experiment: Dependent variables tap how candidate profiles are perceived across the four central trait dimensions: Ability (measured using one item: “competence”), Assertiveness (measured using three items: “dominant,” “strong,” and “assertive”), Friendliness (measured using two items: “warmth” and “compassion”), and Morality (measured using one item: trustworthy). Respondents reported on five-point scales ranging from 1 “Strongly disagree” to 5 “Strongly agree” how they perceived each of the seven displayed candidate profiles on the items mentioned in parentheses above (displayed in random order). Subsequently, composite trait variables were created averaging across the relevant items and recoding each variable on a 0–1 scale, with 0 and 1 reflecting minimum and maximum values on a given trait dimension impression (Ability: M = 0.55, SD = 0.18; Assertiveness: M = 0.52, SD = 0.15, α = 0.82; Friendliness: M = 0.47, SD = 0.17, α = 0.88; Morality: M = 0.47, SD = 0.18).Footnote 6
Issue Competence Experiment: Dependent variables tap the extent to which candidate profiles are seen as having the competence to deal with various political issues. Responses were recorded on five-point scales and subsequently recoded on a 0–1 scale, with 0 and 1 reflecting minimum and maximum values on issue handling competence, respectively. Based on the extant work on gendered issue competence stereotypes, we included “aid to the poor” (M = 0.46, SD = 0.21), “reproductive rights” (M = 0.44, SD = 0.25), and “education” (M = 0.51, SD = 0.22) to represent issues typically “owned” by women, whereas “national security” (M = 0.39, SD = 0.22), “crime” (M = 0.45, SD = 0.22), “the economy” (M = 0.50, SD = 0.21), and “immigration” (M = 0.41, SD = 0.21) were chosen as issues typically “owned” by men.Footnote 7
Moderators: Gender, Ideology, and Sexism
Respondent gender was coded 1 and 0 for women and men, respectively. Political ideology was measured on a seven-point scale from “Extremely liberal” to Extremely conservative,” which we recode to a 0–1 scale, with 0 and 1 reflecting most liberal and most conservative positions, respectively (M = 0.38, SD = 0.30). Respondents’ sexist orientations were measured using eight items from the Ambivalent Sexism Inventory (Glick and Fiske Reference Glick and Fiske1996) using seven-point response scales; four items for both the hostile (example item: “women seek to gain power by getting control over men”) and the benevolent (“Women should be cherished and protected by men”) sexism sub-scales. Finally, we create composite scales for both sexism dimensions by averaging across the relevant items and recoding to a 0–1 scale, with 0 and 1 reflecting lowest and highest degrees of sexism (Mhostile = 0.30, SDhostile = 0.23, αhostile = 0.89; Mbenevolent = 0.46, SDbenevolent = 0.24, αbenevolent = 0.83).
Analytical Strategy
We use OLS regression with respondent-clustered standard errors to estimate the average marginal component effect (AMCE) of candidate gender across our dependent variables. The AMCE represents the marginal causal effect of candidate gender over the joint distribution of all other candidate attributes (see Hainmueller, Hopkins, and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2014). Results are presented graphically with key coefficients reported in the text (see full regression outputs in Appendix Tables A.2–A.19).
Results
We start our analysis by addressing whether the “conventional wisdom expectation” or “the rival expectation” captures the content of contemporary American candidate trait stereotypes better. Recall that according to “conventional wisdom,” men candidates are expectedly stereotyped as higher than women on Ability and Assertiveness (Vertical trait sub-dimensions), while — conversely — women candidates are expectedly stereotyped as higher than men on Friendliness and Morality (Horizontal trait sub-dimensions). In contrast, “the rival expectation” suggests that contemporary trait stereotypes have changed such that Ability and Assertiveness stereotypes are no longer as favorable to men as they used to be, while women are still stereotyped as friendlier and more moral. Second, we repeat the exercise of pitting the “conventional wisdom expectation” against the “rival expectation” but this time testing the content of issue competence stereotypes. Finally, we explore if respondents’ gender, ideological orientations (from liberal to conservative), and their hostile and benevolent sexist orientations moderate the content of both trait and issue competence stereotypes.
Due to the way candidate gender is coded (1 and 0 reflecting “women” and “men,” respectively), any positive (negative) effects of candidate gender should be interpreted as women being perceived as higher (lower) on a given trait compared to men in the Trait Experiment, or as women being more (less) competent to handle an issue in the Issue Competence Experiment. Because all dependent variables are coded on a 0–1 scale, estimated effects of candidate gender can be interpreted as percentage point differences between women and men candidates.
Do Voters Hold Gendered Trait Stereotypes About American Political Candidates?
Yes, but mostly in line with “the rival expectation.” The upper panel (A) of Figure 1 displays the AMCEs of candidate gender on the four trait sub-dimensions (with Horizontal sub-dimensions placed to the right and Vertical to the left). In line with both the conventional wisdom expectation and the rival expectation, respondents, on average, attribute Horizontal trait sub-dimensions more to women candidates: Relative to men, women candidates receive about 6 percentage points higher on attributions of Friendliness (p < .001) and about 4 percentage points higher on attributions of Morality (p < .001). To contextualize these effects of candidate gender, estimated effects of the remaining six candidate attributes show that only occupation exerted stronger effects on ascription of Friendliness and Morality. For example, moving from a baseline of teacher to lawyer results in reduced attributions of 10 and 9 percentage points for Friendliness and Morality, respectively. In sum, these results stress that stereotypes about American political candidates on Horizontal traits are still significantly gendered.
Average marginal component effects (AMCEs) of candidate gender on trait (Panel A) and issue competence stereotypes (Panel B).
Turning to the Vertical trait sub-dimensions, results are clearly more in line with the rival expectation than conventional wisdom: For Assertiveness, we find a non-significant effect of candidate gender of −0.0 percentage points (p = .900), while for attributions of Ability, we find that women on average receive 3.2 percentage points (p < .001) higher compared to men. That is, these results suggest that on Vertical traits — often seen as core qualities in leaders and political candidates — women have closed the gender gap on Assertiveness and surpassed men on Ability. To contextualize these results, we briefly mention effects of the other candidate attributes. Occupation, previous political experience, and race all exerted significant effects on Assertiveness attributions, stressing that the non-significant difference between men and women candidates is not caused by Assertiveness generally not being affected by candidate attributes. Regarding Ability, experience unsurprisingly exerted a stronger effect than candidate gender (yielding a 6.8 percentage point difference between 0 and 4–5 years of experience), while occupation caused a sizably similar effect to candidate gender with a 4.4 percentage point difference (between candidates with a farming or teaching background, respectively) (see Appendix Tables A.2 and A.3 for estimated effects of all seven candidate attributes across all four trait sub-dimensions). In sum, the most striking result regarding Assertiveness is the “closed gap” in respondents’ attributions of this key leadership trait to men and women candidates. Furthermore, it is noteworthy that contemporary American respondents attribute Ability, also a key leadership trait, to women more so than men. With respect to Vertical traits, stereotype content about American political candidates has indeed changed compared to conventional wisdom expectations.
In summary, results from the Trait Experiment clearly support “the rival expectation” over the “conventional wisdom expectation.” Next, we investigate if similar patterns emerge for issue competence stereotypes.
Do Voters Hold Gendered Issue Competence Stereotypes About American Political Candidates?
Yes, but (again) mostly in line with “the rival expectation.” The lower panel (B) of Figure 1 displays AMCEs of candidate gender across the seven different issues included in the Trait Experiment. In line with both the conventional wisdom expectation and the rival expectation, women candidates are, on average, perceived as 23.5 (p < .001), 5.0 (p < .001), and 4.3 (p < .001) percentage points more competent than men on reproductive rights, aid to the poor, and education, respectively — issues traditionally “owned” by women. To contextualize the results, the estimated effect of candidate gender on perceived competence for education and aid to the poor matches the effect of having three (compared to no) children and having 4–5 or more years of experience (compared to no experience) (see full regression models in Appendix Table A.4). That is, candidate gender clearly exerts substantially important effects on inferred competence in these policy areas.
Next, estimating AMCEs of candidate gender on issues traditionally “owned” by men, we see that contemporary American issue competence stereotypes do indeed depart from conventional wisdom: women candidates are neither perceived as less competent on national security (p = .281), crime (p = .174), nor on the economy (p = .121). Importantly, both candidates’ occupational background, experience, and race significantly affect perceived issue competence on the mentioned issues, which underlines that the null effects of candidate gender are not caused by competence perceptions on these issues generally being unmovable (see full regression models in Appendix Tables A.5). Furthermore, women candidates are rated a significant 3.4 percentage points (p < .001) more competent than men on immigration. These findings contrast with conventional wisdom and previous studies concluding that women candidates were typically stereotyped as less competent than men on key issues like the economy, national security, and crime (Huddy and Terkildsen Reference Huddy and Terkildsen1993a; Lawless Reference Lawless2004). This is clearly not the case based on the results reported in Panel B of Figure 1. Rather, women candidates have closed the gap on salient issues traditionally “owned” by men while still holding an advantage on issues typically “owned” by women. We return to the broader theoretical and normative consequences of the support for “the rival expectation” over “the conventional wisdom expectation,” and the revealed change in trait and issue competence stereotypes about American political candidates in the discussion.
Does Respondents’ Gender, Ideology, and Sexism Moderate Stereotype Content?
Yes, liberals and individuals low in hostile sexism hold stereotypes least in line with conventional wisdom. Below, we test if content of trait and issue competence stereotypes about male and female political candidates is moderated by respondents’ gender, ideological orientations, and/or sexist dispositions. We report estimated interaction terms and illustrate our results plotting the marginal effect of candidate gender across values on the moderating variable of interest.
Our first moderation test concerns respondents’ own gender. Based on existing work, one might expect potential gender affinity effects such that women respondents rate women candidates in more favorable ways (e.g., less in line with conventional wisdom on Ability and Assertiveness traits and on issues traditionally “owned” by men). Interestingly, we only find limited support for gender affinity effects. For trait stereotypes, women compared to men perceive women candidates as significantly higher on Friendliness, Morality, and Ability (p < .05), but these differences are modest, hovering between 2.6 and 3.8 percentage points. No significant interaction between respondent gender and candidate gender was observed for the Assertiveness subdimension (p = .200). Regarding issue competence stereotypes, women respondents see women candidates as more competent than men in handling reproductive rights and crime (p < .05) (between 3.6 and 5.6 percentage points), while respondent gender did not moderate remaining issue stereotypes. Thus, some weak signs of gender affinity effects emerge, but overall women and men respondents tend to hold trait and issue competence stereotypes that are closely related (see Appendix Tables A.15–A.19 for full regression models).
In contrast, we find much more variation in stereotype application by respondent ideology and hostile (but not benevolent) sexism. Figure 2 displays the estimated effect of candidate gender on trait sub-dimensions across different levels of hostile sexism (red lines) and ideology (blue lines), with higher values indicating more conservative orientations and higher levels of hostile sexism, respectively. Results are from regression models including either the candidate gender X ideology moderation or the candidate gender X sexism moderation. Across all four trait sub-dimensions, ideology significantly moderates the effect of candidate gender (p < .01) such that liberal respondents tend to hold more favorable trait stereotypes for women compared to men candidates. The ideology moderation is similarly strong for Friendliness, Morality, and Ability, while slightly weaker for Assertiveness. Interestingly, content of trait stereotypes among the most conservative respondents is generally non-gendered; that is, men and women candidates are rated as equally high on all traits except Assertiveness, with men getting a slight edge over women. Turning to hostile sexism (red lines), similar patterns emerge: hostile sexism significantly moderates the effect of candidate gender on all four trait sub-dimensions, with the weakest moderation for Assertiveness (p < .05), and respondents low in hostile sexism hold trait stereotypes most favorable to women. Respondents highest in hostile sexism see men candidates as higher than women in Ability, while candidate gender does not affect remaining traits among respondents high in hostile sexism. Interestingly, benevolent sexism did not moderate the effect of candidate gender for any trait sub-dimensions (see full regression models with candidate gender and ideology/sexism interactions in Appendix Tables A.6–A.9 and Figure A.5).
Marginal effect of candidate gender on trait stereotype content across ideology (blue lines) and hostile sexism (red lines).
Finally, we test if ideology (blue lines) and hostile sexism (red lines) moderate content of issue competence stereotypes with results displayed in Figure 3 for the two issues aid to the poor (left-hand panel) and national security (right-hand panel). We focus on these two issues as they represent issues traditionally “owned” by men (national security) and women (aid to the poor), and because findings on these issues represent the overall findings across the remaining issues (see Appendix Tables A.10–A.14 and Figure A.2). Ideology significantly moderates the effect of candidate gender on both issues (p < .01), such that stereotypes are more favorable to men (and less favorable to women candidates) among more conservative respondents. Similarly, hostile sexism also significantly moderates the effect of candidate gender on both issues (p < .05), with stereotype content becoming more (less) favorable to men (women) with increasing hostile sexism. An important difference between the two issues is that for aid to the poor, the least conservative and least hostile sexist respondents express stereotype content strongly favoring women, which then decreases across both moderators. In contrast, for “national security,” the most liberal and least hostile sexist respondents see women candidates as slightly more competent than men, though this female advantage shifts as conservatism and sexism increase, with the most conservative/sexist respondents advantaging men in this policy area. In sum, despite the parallel way in which ideology and hostile sexism moderate the content of stereotypes for both aid to the poor and national security, the absolute levels of these stereotypes still reflect that the two issues were historically “owned” by women and men candidates, respectively.
Marginal effect of candidate gender on issue competence stereotype content across ideology (green lines) and hostile sexism (red lines).
In summary, the reported moderation results show that the evolving content of trait and issue competence stereotypes about men and women candidates in the United States seems driven by liberals and individuals low in hostile sexism.
Conclusion and Discussion
Long-standing conventional wisdoms suggest that American citizens hold gendered trait and issue competence stereotypes about political candidates: women candidates are seen as higher than men in Horizontal traits like Friendliness and Morality, but as lower in Vertical traits like Ability and Assertiveness, which in turn colors how voters perceive the competence of women in key policy areas (Huddy and Terkildsen Reference Huddy and Terkildsen1993a, Reference Huddy and Terkildsen1993b; Matland Reference Matland1994; Sanbonmatsu Reference Sanbonmatsu2002; Sanbonmatsu and Dolan Reference Sanbonmatsu and Dolan2009). Because these gendered stereotypes run counter to ideals of what constitutes the prototypical leader or politician, their existence presents a hindrance to women seeking political office and thus also contributes to the overall underrepresentation of women in public life. However, leadership stereotypes are evolving in a less masculinized direction (Eagly et al. Reference Eagly, Nater, Miller, Kaufmann and Sczesny2020; Koenig et al. Reference Koenig, Eagly, Mitchell and Ristikari2011), and recent meta-analytical work on gendered candidate preferences suggest that, particularly among Americans, women candidates are favored above men (Schwarz and Coppock Reference Schwarz and Coppock2022). Thus, rivaling conventional wisdom, these findings suggest that Americans’ candidate stereotypes are also evolving. Above, we reported results from two well-powered conjoint experiments designed to illuminate the possibly gendered content of contemporary trait and issue competence stereotypes about American political candidates. Results largely support that Americans’ stereotypes about men and women candidates have changed. Regarding trait stereotypes, Americans still tend to perceive women candidates as higher in Friendliness and Morality, but also as higher in Ability (i.e., competence) and equally as Assertive as men. We find the fact that voters seem to see women as higher in Ability as particularly important because women, on average, are higher quality political candidates (Bauer Reference Bauer2020b; Milyo and Schlosberg Reference Milyo and Schlosberg2000) and more effective legislators (Anzia and Berry Reference Anzia and Berry2011; Volden, Wiseman, and Wittmer Reference Volden, Wiseman and Wittmer2013). With respect to issue competence stereotypes, our American respondents perceive women candidates as more competent to handle “soft” issues like aid to the poor, education, reproductive rights, but also immigration. However, most indicative of changing stereotypes was the finding that respondents on average see women and men as equally competent to handle key political issues like the economy, crime, and even national security. Importantly, these aggregate patterns were moderated by gender — though effect sizes were substantively small — and more strongly moderated by respondents’ political ideology and levels of hostile sexism: liberals and individuals low in hostile sexism generally hold stereotypes that are more favorable to women candidates. Below, we first elaborate on the of robustness and generalizability of our key results, discuss the theoretical and methodological implications and contributions, and point out limitations of our studies, while also suggesting directions for future research where we find appropriate.
Robustness and Generalizability
Our results about aggregate trait and issue competence stereotypes are generally robust across a series of alternative modeling strategies and specifications. When we disaggregate the trait composite scales (see Appendix Tables A.22 and A.23) into the underlying single-trait items, we arrive at similar conclusions.Footnote 8 Further, one might question how generalizable our findings are to the wider public given that, after all, they are not based on a probability sample of the American population, and moreover, given the reported moderation by ideological and sexist orientations one might suspect that support for “the rival expectation” is driven by an overly liberal sample. We report two sets of results to alleviate this concern: i) based on our models interacting candidate gender with respondents’ ideological orientations we can estimate the content of trait and issue competence stereotypes for respondents self-reporting as ideological moderates (i.e., 0.5 on our recoded 0–1 scale);Footnote 9 ii) we can restrict our estimation of aggregate trait and issue competence stereotypes to exclude the most ideologically extreme respondents who possibly bias the results reported above (i.e., analyze only respondents self-placing from 0.17 to 0.83 on ideology). Importantly, following both strategies we still reach the similar conclusions: women candidates are perceived as higher than men on Friendliness, Morality, and Ability, and equally as Assertive as male candidates. Additionally, they are seen as more competent than men on reproductive rights, aid to the poor, education, and immigration, while men and women are perceived as equally competent on “the economy,” crime, and national security (see Appendix Tables A.24–A.28). Finally, the moderations of trait and issue competence stereotype content by respondents’ self-reported political ideology replicate using two psychological constructs found to underlie differences between self-reported liberals and conservatives, social dominance orientation, and right wing authoritarianism (e.g., Duckitt and Sibley Reference Duckitt and Sibley2010) (see Appendix Figures A.3 and A.4). This suggests that the tendency for liberal and conservative individuals to hold different candidate stereotypes is rooted in a wider catalogue of psychological differences, a possibility we see as worth pursuing in future research.
Theoretical Implications and Contributions
Our main conclusion about the evolving content of trait and issue competence stereotypes about political candidates has various theoretical implications and also raises new questions. First, the parallel between changing content of broader societal leadership stereotypes (see Eagly et al. Reference Eagly, Nater, Miller, Kaufmann and Sczesny2020; Koenig et al. Reference Koenig, Eagly, Mitchell and Ristikari2011) and our findings highlights the importance of seeing and analyzing stereotypes about political candidates in connection with the broader phenomenon of leadership stereotypes, while still paying attention to obvious and important differences between leaders in politics and other domains (e.g., business or military). Broadly, these results, in tandem with other work on leadership stereotypes, suggest that societal leadership stereotypes, whether in the domain of business, the military, or politics, are not fixed. Instead, they are responsive to changes in the roles that men and women take on in society. Furthermore, we speculate that if women are seen as more competent than men in politics — a complex and highly competitive domain — it is likely that this applies to women in leadership positions in work, civic organizations, and other domains more broadly. Importantly, we acknowledge that although many voters may see women as baseline more competent than men, there is also evidence that they hold female candidates to higher qualifications standards, particularly Republican women (Bauer Reference Bauer2020a).
Second, if women candidates are now both seen as higher in Friendliness, Morality, and Ability, while having also closed the perceptual gaps on Assertiveness and competence on key political issues such as national security and the economy, it seems straight forward to conclude that stereotypes as a hindrance to women’s political representation have been reduced. However, according to role congruity theory, it could be the case that being perceived as high on the Friendliness trait sub-dimension — that is, as very kind, warm, and compassionate — might constitute a misfit with existing leadership stereotypes such that women candidates are still facing challenges rooted in the public’s stereotypes. Nonetheless, we stress the point that leadership stereotypes have become less focused on stereotypically masculine traits over time (Koenig et al. Reference Koenig, Eagly, Mitchell and Ristikari2011), and it is no longer totally clear whether being perceived as high in Horizontal traits, particularly Friendliness, constitutes an advantage or a hindrance to political candidates. In addition, recent work has uncovered other more subtle and implicit demand-side hindrances to female representation such as “shifting standards” (Bauer Reference Bauer2020a) or “the double bind” (Teele, Kalla, and Rosenbluth Reference Teele, Kalla and Rosenbluth2018).Footnote 10 We agree with these studies, that serious stereotype-related challenges are still facing women candidates. Yet, we also find it worth stressing that the changed nature of trait and issue competence stereotypes reflected in our results constitutes a reduction in direct obstacles facing women seeking a political career, at least on the demand side.
Third, our results also raise the question of why have trait and policy stereotype content changed? One option is that over time, women have increasingly taken on executive political offices and entered political issues traditionally “owned” by men (e.g., Armstrong et al. Reference Armstrong, Barnes, Chiba and O’Brienforthcoming; O’Brien et al. Reference O’Brien, Mendez, Peterson and Shin2015), but the flip side of this argument is that women politicians are potentially seeking these offices as a consequence of a cultural change in existing stereotypes about political candidates. Consequently, future research could seek to clarify the causal relationship between stereotype content and the types of offices held by women politicians through carefully mapping both phenomena in a given country of interest over time, or by conducting comparative and cross-national studies.
Finally, it is worth considering what these results tell us about how women in politics can appeal to voters. Political pundits often deliberate about whether women in politics have to “act like men.” In other words, do they need to make a special effort to appear tough, assertive, and competent in “masculine” policy domains to seem like viable leaders? As we have already noted, the qualities that citizens see as essential for leadership have become less masculinized over time (Koenig et al. Reference Koenig, Eagly, Mitchell and Ristikari2011), and traits like empathy and the ability to compromise are valued more highly than ever (Brooks Reference Brooks2013; Powell, Butterfield, and Jiang Reference Powell, Butterfield and Jiang2021). This suggests that not only do women in American politics benefit from being perceived “like men” in terms of Vertical traits, but they may also benefit from being perceived as higher in Horizontal traits as well; though again, we see this as an open empirical question that future research should address more directly.
Besides our empirical contributions presented and discussed above, we also want to stress two additional contributions. First, we hope that the larger theoretical framework on social perceptions and judgements from the psychological sciences on which we built will be seen as beneficial by other researchers in the field. Explicitly conceptualizing candidate traits along the major dimensions of Horizontal and Vertical traits is an important theoretical integration that could facilitate novel cross-disciplinary insights. Second, we believe that the conjoint experiment is particularly attractive for studying content of candidate stereotypes as it allows for testing whether respondents voluntarily assign a certain trait to specific candidates (based on gender, race, social class, etc.). In this way, we answer calls to consider stereotype application opposed to stereotype knowledge (Bauer Reference Bauer2013).
Limitations
Despite the mentioned contributions, we also want to stress several limitations of our studies. First, we chose to study the content of gendered trait and issue competence stereotypes in the absence of clear information about candidates’ party affiliation or ideology. Yet, existing work stresses “party ownerships” of certain traits and that partisan information about candidates color trait impressions and stereotypes (e.g., Hayes Reference Hayes2005; Winter Reference Winter2010). Hayes (Reference Hayes2011, 133) even concludes that “gender stereotyping is limited by the relevance of party stereotypes.” Indeed, all things held constant, we also find that female candidates are inferred to be slightly more liberal than male candidates (
$ \mathrm{p}<.001 $
; see Appendix Table A.20 and Figure A.6). This could mean that the relationship between gender and stereotype application is partially mediated by party or ideology stereotypes. For example, if women are seen as more liberal or more likely to belong to the Democratic party, they are perceived as warmer and more likely to be competent in “soft” policy areas. Despite this possibility, we still believe that the reported studies tell us something important about whether contemporary American candidate stereotypes are gendered. However, we also urge future research to clarify if and to what degree our conclusions hold for other kinds of elections, also introducing partisan and opinion-based differences between candidates. Second, our studies presented one candidate at a time to respondents. Consequently, one might wonder how stereotypes become activated in competitive election settings. Relatedly, it is possible that the electoral context, and scenarios such as a race in which a man is running against a woman, could ignite things like gender resentment and sexism that cause voters to rely on stereotypes more (e.g., Kalaf-Hughes and Leiter Reference Kalaf-Hughes and Leiter2020). Third, we also want to acknowledge that in our conjoint experiment, the distribution of the other candidate attributes (experience, occupation, etc.) was not constrained by gender and therefore did not reflect the distribution of those attribute levels in the real world. In other words, men are more likely to be farmers, lawyers, and business owners. Furthermore, because of incumbency advantage and women’s underrepresentation in every political office throughout history, men, on average, have more political experience. Thus, women candidates would not benefit from the same bump that these additional traits give men. Finally, despite the relevance of the United States as one of the world’s central and biggest democracies, our results are only obtained from one study with American respondents. Thus, we strongly urge future research to initiate truly comparative research projects uncovering how stereotype content vary across institutional settings, levels of gender egalitarianism, macro-cultural differences, and historical legacies for political representation of women. Such cross-national research projects, we believe, hold important potential for both uncovering deeper truths about causes underlying women’s political underrepresentation, and also for potentially improving it.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S1743923X25000042.
Acknowledgments
This study was funded by The Danish National Research Foundation (DNRF144). The authors would like to thank Kira Sanbonmatsu for her helpful comments on an early version of this project, as well as the three anonymous reviewers and the editor for helping us improve this manuscript.
Competing interest
The author(s) declare no competing interests.
