The Effect of Counterstereotypic Gender Strategies on Candidate Evaluations in American Elections

Abstract

Women who ran for office in 2018 used a variety of strategies on the campaign trail, with some highlighting more masculine traits and others more feminine traits, but the latter was more common than in prior years. We ask how effective these strategies are for trait evaluations, perceptions of leadership and competence, likeability, and vote choice and how this effect varies based on respondent’s views about the role of women in society. To explore these relationships, we use data from a two-wave panel conducted in the winter of 2019. Results from our experiment show that female candidates who highlight more masculine traits are perceived as more agentic, less communal, and more competent, and, importantly, they do not appear to suffer from a backlash effect. Those higher in sexism also evaluate women who display these qualities as more competent, though as less warm and likable.

Perhaps more remarkable than the record number of women running for office in the 2018 election was the diversity of strategies used by them. There were women veterans running on both sides of the aisle who showcased their character and experience in a typically masculine arena. This emphasis on counterstereotypic traits is a strategy that scholars argue should be more prevalent among women running for office, since they need to counter the strong prototype of political leaders having agentic traits associated more with men (Bauer 2017; Fridkin and Kenney 2014). In fact, in their comprehensive content analysis of women senators running in the 2006 congressional elections, Fridkin and Kenney (2014) find that women are more likely to highlight agentic qualities in their campaign communications than communal qualities.

At the same time, other women seemed to throw that playbook out the window. Rather than run as men, a number of candidates ran unabashedly as women. According to an article in The New Yorker, “Some of those who have children are explicitly incorporating their identities as mothers into their appeals to voters” (Talbot 2018). Furthermore, on their campaign websites, female candidates emphasized communal traits such as the need to “care about others” and “bring people back together.” This shift toward embracing womanhood and communal traits did not start in 2018. Scholars note a similar evolution in Hillary Clinton’s strategy from 2008 to 2016, when she emphasized her role as a mother and grandmother (Sides 2015). While some analyses of campaign communications in 2018 find that women candidates are more likely to emphasize women’s issues in tweets (Mechkova and Wilson 2021), there has not been a systematic analysis of the presentation of agentic and communal traits in the 2018 election. Were women candidates highlighting communal traits more than agentic traits in 2018?

Furthermore, it is not entirely clear from the literature to what extent these different strategies help or harm women running for office. Are women more likely to be viewed as leaders by highlighting traits that are congruent with gender stereotypes or counter to them? How does an emphasis on stereotypic or counterstereotypic traits affect perceptions of women’s capabilities on masculine and feminine issues? The literature is somewhat mixed on these questions. Some scholarship suggests that by highlighting counterstereotypic traits, women are more likely to appear as leaders and as more competent on masculine issues (Huddy and Terkildsen 1993), though they may face some backlash for doing so (Krupnikov and Bauer 2014; Mechkova and Wilson 2021). Other work finds that counterstereotypic strategies neither help nor harm evaluations (Brooks 2013). And yet other scholarship suggests that women candidates may even benefit from highlighting communal traits in campaign communications (Burns, Eberhardt, and Merolla 2013; Dolan 1998).

Of course, one reason for these mixed findings, and for mixed strategies on the campaign trail, is that the answer to this question is “it depends.” Some scholars argue that it depends on the context of the election. For example, a context marked by security threat may make agentic qualities more relevant and applicable, so counterstereotypic strategies may be more effective (Holman, Merolla, and Zechmeister 2017; Lawless 2004). Another factor that may matter is the characteristics of voters themselves. For example, a counterstereotypic strategy may be more effective among in-partisans (Bauer 2017) and Republicans, while a stereotypic strategy may be more effective among Democrats (Burns, Eberhardt, and Merolla 2013). We argue that another individual-level factor that may influence voter reactions to these strategies are views about women in society. Those who do not hold sexist views may welcome stereotypic strategies and communal leadership; thus, they would evaluate women using stereotypic strategies more favorably. Meanwhile, those high in benevolent or hostile sexism may project leadership and competence onto women who use counterstereotypic strategies since they place more value on masculinity, though there may be some backlash on likeability and warmth.

In this article, we explore these questions by first conducting a content analysis of the use of stereotypic and counterstereotypic trait language on the websites of 242 women who ran for national-level legislative office in the 2018 election. This gives us a better picture of whether campaign strategy shifted in that election context compared to previous elections. In the second section, we look at the effectiveness of these strategies, adapting an experimental design used by Bauer (2017) in which she manipulates the use of stereotypic and counterstereotypic strategies. We build on and expand from this work by considering additional dependent variables and different moderators. Furthermore, we collect a large sample to have the power to detect main effects and conditioning effects. Finally, to explore moderators without priming participants on those dimensions, we use a panel design, collecting measures of gender, partisanship, and gender attitudes in the first wave and fielding the experiment on the second wave.

Our findings show that although women candidates mentioned stereotypic qualities more in 2018 compared to earlier elections, it does not appear that this strategy pays dividends. That is, the experimental findings show that women candidates receive more favorable evaluations from emphasizing counterstereotypic traits, even among those higher in sexist attitudes. Our research contributes to the broader literature by providing a systematic analysis of the ways in which women candidates highlighted traits in their campaign communications in the 2018 election, one in which gender was highly salient. It also builds on the literature that has assessed whether stereotypic or counterstereotypic trait strategies are more effective by considering another important factor that may condition reactions, views about women’s role in society.

Extant Understandings of the Effectiveness of Different Types of Trait-Based Strategies

Stereotypes are endemic to social and political life, and people rely on them, in part, to reduce information costs in a complex world (e.g., Allport 1954; Dovidio, Glick, and Rudman 2008). Some of the most pervasive stereotypes that cut across societies are gender stereotypes. Women are generally perceived as holding more communal traits, like compassion and caring (Bauer 2015; Dolan 2004; Dolan 2014a, 2014b), while men are generally perceived as having more agentic traits, such as being assertive and aggressive (e.g., Dolan 2004; Dolan 2014a, 2014b; Holman, Merolla, and Zechmeister 2011).

Some of these stereotypic traits are more privileged in the political sphere. More specifically, there is a general preference for agentic qualities in political leaders (Eagly and Karau 2002; Koenig et al. 2011), a trait that individuals are more likely to associate with men. Given the overlap in traits perceived in men and valued in political leaders, male politicians are more likely to be perceived as holding agentic qualities, since they are a subgroup of men (Schneider and Bos 2014). Women political leaders, however, are a subtype of women because they are expected to hold agentic traits as leaders, which are at odds with the communal traits perceived to be held by women (Schneider and Bos 2014). As a result, female politicians are not seen as particularly high on feminine characteristics (Bauer 2017; Brooks 2013; Dolan 2014a; Schneider and Bos 2014) or on masculine characteristics (Schneider and Bos 2014), and thus they are seen as “neither leaders nor ladies” (Bauer 2017, 279).

Given the preference for agentic traits in political leaders, some scholars argue that women candidates can benefit from using counterstereotypic strategies when running for office, especially with respect to traits (e.g., Bauer 2017; Fridkin and Kenney 2014; Schneider 2014b).¹ According to Bauer (2017), since women politicians do not fit into a clear category of being women or leaders, it is more difficult for voters to apply stereotypes. In this type of context, individuating information will be more relevant, so women may benefit from adopting counterstereotypic strategies. That is, by emphasizing agentic qualities, voters may be more likely to see women as leaders.

At the same time, some scholars contend that women may be punished for following such a strategy. That is, if women present themselves with more agentic characteristics, they may face a backlash for violating gender expectations (Bauer 2017; Krupnikov and Bauer 2014; Mechkova and Wilson 2021). This may put women in a double bind in which they may be more likely to be perceived as leaders, but they will not be viewed as likable. Other scholars argue that backlash may not be very pronounced for women candidates who adopt counterstereotypic strategies. For example, Bauer (2017) argues that when women emphasize positive counterstereotypic traits, they change the reference category used by voters to evaluate female candidates from ladies to leaders; thus, they will be less likely to face backlash. Furthermore, she argues that women candidates are not likely to face backlash from those within their own party. Cormack and Karl (2022) also find that voters are less likely to hold counterstereotypic trait signals against perceived copartisan women candidates. In a different vein, Deason, Greenlee, and Langner (2015) argue that certain women—those with children—may be able to avoid or at least limit the double bind, since they will be punished less for fulfilling the traditional role of motherhood.²

Research on the presentation of counterstereotypic strategies on traits in campaign communications shows that women candidates adopt this strategy. For example, as noted earlier, in their analysis of the 2006 congressional elections, Fridkin and Kenney (2014) find that women running for the Senate are more likely to emphasize agentic traits in their campaign communications than communal traits. Carpinella and Bauer (2021) find similar emphasis on agentic traits in their analysis of campaign ads by women senators in 2000, 2004, and 2008. This pattern also holds on Clinton’s website and tweets when she ran for president in 2016 (Lee and Lim 2016). And some scholars show that female candidates are more aggressive online (for example, attacking opponents) than male candidates (Evans and Clark 2016; Evans, Cordova, and Sipole 2014; but see Mechkova and Wilson 2021 for no difference in 2018).

The empirical evidence on the effect of these counterstereotypic strategies on public perceptions has been mixed. Some find that when women candidates are presented as having more agentic qualities, the public is more likely to perceive them as having the traits associated with leadership (Bauer 2018; Burns, Eberhardt, and Merolla 2013) and as more competent on masculine issues (Huddy and Terkildsen 1993). Several other studies observe a similar increase in overall evaluations and vote shares for women candidates when they are presented as having agentic traits (Banwart 2010; Bauer 2018; Burns, Eberhardt, and Merolla 2013; Huddy and Terkildsen 1993; Mechkova and Wilson 2021). Importantly, these findings are also present at the local level (Bauer 2020a). At the same time, others show evidence of a backlash in overall favorability when women are presented as having agentic qualities, presumably because they are violating gender norms (Bauer 2017; Eagly and Karau 2002; Krupnikov and Bauer 2014), though this double bind does not appear to apply to men who present themselves as having counterstereotypic traits (Ditonto 2017; Krupnikov and Bauer 2014). Finally, others find largely null effects of counterstereotypic trait strategies (Brooks 2013; Fridkin and Kenney 2014).

In terms of emphasizing communal traits in campaign communications, scholarship has been more limited. Fridkin and Kenney (2014) find that female senatorial candidates emphasize communal traits in their campaign communications but are less inclined to discuss their families or their role as mothers, though Schneider (2014b) shows that female congressional candidates in the 2002–08 cycles were more likely to mention details about family than male candidates. On the one hand, an emphasis on communal traits may increase perceptions that women leaders hold them, especially since scholarship shows that individuals do not transfer communal traits to women candidates (Bauer 2015; Brooks 2013; Dolan 2014a; Schneider and Bos 2014), and could translate into higher favorability (Burns, Eberhardt and Merolla 2013). On the other hand, since women candidates are generally perceived as more competent on feminine issues, it is not clear that more emphasis on them will necessarily increase perceptions even further (Bauer 2020b). However, there may be an expectation on the part of the public that women exhibit such traits, and they may be punished if they violate those expectations (Barnes, Beaulieu and Saxton 2020; Costa 2021; Ditonto 2017; Eagly and Karau 2002; Eggers, Vivyan, and Wagner 2018). As noted by Deason, Greenlee, and Langer (2015, 143), this may particularly be the case as motherhood has become increasingly politicized, whereby there is an expectation that women “must place family roles front and center in order to appear competent, well-balanced, or sufficiently feminine,” and research finds that voters prefer women with traditional profiles, such as being married and having children (Teele, Kalla, and Rosenbluth 2018).³

The findings for the effect of these strategies have been mixed. For example, Burns, Eberhardt, and Merolla (2013) find that a treatment emphasizing Sarah Palin’s more communal qualities boosts evaluations of those qualities, agentic qualities, as well as overall evaluations. However, Bauer (2017) observes null effects for a treatment emphasizing communal qualities on perceptions of the communal attributes of women candidates, and null effects on evaluations of leadership qualities and overall favorability. Similarly, Fridkin and Kenney (2014) find null effects using observational data. More recent empirical work by Bauer (2020b) shows that trait-based strategies tend to activate feminine stereotypes while reducing leadership evaluations of candidates.

Drawing from this scholarship, we test whether women using counterstereotypic trait strategies are perceived as holding more agentic traits ( H _1a ), while those using stereotypic trait strategies may be perceived as higher on communal traits ( H _1b ). We also test whether counterstereotypic trait strategies lead individuals to perceive the candidate more as a leader ( H _2a ) and as more competent at handling masculine issues ( H _2b ). Expectations are less clear-cut when examining perceptions of the candidates’ ability to handle feminine issues. On the one hand, we might imagine that an individual who is perceived as a capable political leader will also be perceived as better able to handle any issue area, including those more associated with women ( H _3a ). On the other hand, some may perceive a counterstereotypic woman as less able to handle women’s issues ( H _3b ). Finally, we test the competing hypotheses in the literature that emphasizing counterstereotypic strategies will ( H _4a ) or will not (H _4b ) lead to backlash, and explore whether stereotypic strategies improve overall evaluations ( H ₅ ).

Individual Differences in Reactions to Candidate Strategies

As noted in the introduction, one reason for the mixed findings in the literature on these questions may be that the extent to which these strategies are effective depends on the electoral context as well as the characteristics of voters. Our study design holds the context constant, so we focus on how the effectiveness of these strategies varies across individuals. Since we focus on in-partisans, we may not expect to see much by way of backlash (Bauer 2017), and existing work has not found many differences by respondent gender. We consider another individual difference that may also moderate reactions to the types of strategies that women candidates use on the campaign trail: the extent to which individuals hold benevolent or hostile views toward women.

Patriarchy is universal among human societies, meaning that men have structural control over political, legal, economic, and religious institutions (Goldberg 1993). While men are therefore the dominant group, they rely on women for reproduction. This reliance leads to women possessing a dyadic power, in which men cannot procreate without them. The simultaneous existence of male structural power and female dyadic power creates ambivalent sexist ideologies that are composed of both hostile and benevolent sexism (Glick and Fiske 1996). Hostile sexists seek to justify male power, traditional gender roles, and men’s exploitation of women as sexual objects through derogatory characterizations of women, while benevolent sexism relies on kinder and gentler justifications of male dominance and prescribed gender roles, while recognizing men’s dependency upon women and embracing a romanticized view of sexual relationships with women (Glick and Fiske 1996, 1997).

We might expect to find higher levels of hostile and benevolent sexism among men than women, since women are the subordinate group within the patriarchy. However, in order to preserve their status in other relevant hierarchies, such as the racial hierarchy, white women often reinforce the white male patriarchy (Junn 2017). In the 2016 presidential election, those high in hostile sexism were more likely to support Donald Trump (Cassese and Barnes 2019). Scholars have also begun to explore how sexism might moderate reactions to the behavior of women in office (Barnes, Beaulieu and Saxton 2020; Costa 2021) and how gender-based attacks in campaigns might activate such attitudes (Cassese and Holman 2019).⁴

Important for our purposes, it is possible that reactions to female candidates who present themselves in stereotypic or counterstereotypic ways may vary depending on whether someone is low or high on hostile or benevolent sexism.⁵ For example, given the more negative and derogatory views about women among hostile sexists, they may not see women who use counterstereotypic strategies as leaders or ladies, and instead have even more negative evaluations of them. Alternatively, since such individuals privilege agentic qualities in political leaders even more than others in the population, even if they have low evaluations of women who run for office on average, they may look more favorably on women who highlight agentic traits compared to those who do not. Given the kinder and gentler justification for male dominance among benevolent sexists, we might find that counterstereotypic strategies cause such individuals to view women as more masculine and as leaders, though they may show some evidence of lower likability given violations of gender expectations.

Those low on both dimensions may reject the patriarchy altogether and be drawn to a variety of strategies that women candidates employ, paying more attention to the particular traits being emphasized to draw inferences about agentic and communal qualities and about competency. However, it is also possible that such individuals have higher gendered expectations and may punish women who exhibit counterstereotypic behavior (Costa 2021). Therefore, we test whether the effect of the treatments for female candidates vary by levels of hostile sexism ( H _6a ) and benevolent sexism ( H _6b ).

The Use of Stereotypic and Counterstereotypic Strategies on the Campaign Trail

Before we turn to an experiment on voter reactions, we want to get a sense of whether the ways in which women candidates represent themselves in their campaigns have shifted over time, in particular in the context of the 2018 election. While the introduction presented anecdotes of women highlighting their unique experiences as women as well as stereotypic traits, it is not clear whether a broader cross section of women emphasized communal traits in that election. This would be in contrast with past research that finds women are more likely to emphasize agentic traits than communal traits in their campaign communications (Carpinella and Bauer 2021; Fridkin and Kenney 2014). There is already some evidence that women were more likely to mention women’s issues and less likely to mention male issues in 2018 (Mechkova and Wilson 2021),⁶ but does this extend to candidate traits?

To systematically explore candidates’ emphasis on communal (stereotypic) and agentic (counterstereotypic) traits, we conduct a content analysis of candidate websites. According to Fridkin and Kenney (2014), websites of candidates are an important vehicle through which candidates can present themselves to their constituents, explain their goals and priorities, discuss their backgrounds, and present their biographies. In short, campaign websites represent controlled messages by the candidate and the type of image they want to project to potential voters. Further, an advantage to coding websites is that every candidate has one, while not every candidate runs political advertisements. Several scholars use websites to capture various types of campaign content, including issue emphasis for men and women running for office (Dolan 2005, 2008, 2014b; Evans and Clark 2016; Herrick 2016; Kahn 1993; Lee and Lim 2016; Niven and Zilber 2001; Schneider 2014b); trait emphasis for men and women candidates (Bauer 2020b; Fridkin and Kenney 2014; Schneider 2014b); the impact of mixed-gender races compared to single-gender races on campaign appeals (Dolan 2005, 2008; Schneider 2014b); and mentions of gendered and racial identities (Brown and Gershon 2016).

A total of 266 women won their primaries and ran for congressional office in the 2018 midterm elections.⁷ Among them, 116 won their races (14 for the U.S. Senate and 102 for the U.S. House of Representatives). We coded all primary winners whose websites could be accessed, resulting in 242 candidates (52 Republicans and 190 Democrats). The candidates in our analysis were running for the House or the Senate. We coded the candidate’s official campaign website with the help of a research assistant.⁸ For the content analysis, in line with Schneider (2014b), we look at the main page, biography page, and issues page of each website. On the biography page, we examine the images (e.g., photographs of the candidates) as well as text. On both the main page and the biography page, we analyze descriptions of the candidates’ background, including facts about each candidate’s family and language used to describe themselves and their issue stances. On the issues page, we code for the way candidates talk about their traits with respect to different issues.

On each of these pages, we track the number of mentions of specific traits. Based on previous research by Fridkin and Kenney (2014) and Bauer (2017), we create unique codes for communal and agentic trait categories. Specifically, we code the following for communal candidate traits: advocate, champion, honest, trustworthy, integrity, gentle, sensitive, caring, and compassionate, active or good listener, and “work with.” An example of a code that is considered “work with,” a communal trait linked to being collaborative, comes from the website of Morgan Murtaugh, a Republican candidate in California running for the House of Representatives, who mentions “work with” nine times on her issues page. One example is “I will work with our military and national security agencies to keep our nation safe and ensure the Constitution is upheld in the process.” An emphasis on communal traits is considered a stereotypic strategy for a woman running for office. We also code mentions of agentic traits, such as ambitious, courageous, charismatic, strong leader, active, motivated, and passionate. An example of an agentic trait comes from Katie Porter, a Democrat running for the House of Representatives in California, who has 11 mentions of these traits throughout her website. Porter mentions she will “fight” for or against a multitude of issues in the interest of the American people—for example, “I will fight against the special money interest flooding into our nation’s capital.” The use of agentic trait language would be considered a counterstereotypic strategy.

For the analysis, we group together mentions of distinct communal traits and distinct agentic traits. On average, women candidates are slightly less likely to emphasize agentic traits, with an average of 5.95 mentions of agentic traits per candidate website compared to an average of 6.92 mentions of communal traits, though these differences are not statistically distinct (p = .19). The most common communal trait categories include “advocate,” “champion,” and “serve,” with an average of 5.18 mentions per candidate; followed by “talkative,” “good or active listener,” and “work with,” having an average of 1.16 mentions per candidate; and “care” and “compassionate,” with an average of 0.62 mentions per candidate. The most used agentic trait categories are “aggressive,” “tough,” and “fighter,” with an average of 3.75 mentions per candidate. The next most used agentic qualities include “ambitious,” “courageous,” “charismatic”, and “strong leader,” with an average of 2.19 mentions per candidate.

These results deviate from those of Fridkin and Kenney (2014), who find in their analysis of websites of U.S. senators in 2006 that mentions of agentic traits are more common than mentions of communal traits. Moreover, on average, agentic traits are much more likely to be mentioned on senator’s biography pages than communal traits. Fridkin and Kenney (2014) also examine how this breaks down by women and men senators, finding that women senators are significantly more likely to mention both agentic traits (e.g., leadership and experience) and communal traits (e.g., empathy and advocacy) on their biography pages than their male counterparts. Carpinella and Bauer (2021) observe a similar emphasis on agentic traits in their analysis of campaign ads by women senators in 2000, 2004, and 2008. It could be that the differences in findings are due to the period examined: perhaps the electorate was more open to communal qualities in candidates in 2018 compared to these earlier elections, though it could also be because we include both House and Senate candidates, and there are many more House candidates in our sample.

Given findings of partisan difference in the value placed on these different traits, we also explore whether there are differences in the use of agentic and communal trait language by the candidate’s partisanship (see Appendix A, Table 1 in the Supplementary Material). On the agentic dimension, the differences are large, with Democratic women candidates mentioning agentic traits 6.93 times on average and Republican women candidates mentioning agentic traits 3.33 times on average. There are few differences in mention of communal traits, with 7.05 mentions per candidate for Democrats and 6.57 for Republican candidates. In short, if anything, Democratic women are as likely to mention counterstereotypic traits as stereotypic traits. Republican women may be more hesitant to project counterstereotypic traits since Republicans, who are higher in hostile sexism, may react more negatively to women who violate gender norms.

In addition to coding mentions of agentic and communal traits, we also code whether candidates reference being a woman or a mother on their website. The candidates in our sample are much more likely to use the language of being a mother, 40%, than to use the language of being a woman, 25%. We also code for whether the candidate included family facts on the biography page and find that 63.57% of the candidates mention family facts. These findings suggest that women candidates are more implicitly cueing stereotypic strategies. They are not as likely to explicitly reference being a woman or a mother, though they implicitly cue these characteristics through mention of family facts and communal traits as they talk about their background and issues. These findings are more aligned with those of Schneider (2014b), while Fridkin and Kenney (2014) find that women are less likely than men to mention family facts.

In our sample of 242 women candidates running during the 2018 election cycle, we observe an increased use of stereotypic strategies compared to past research, whereby women are slightly more likely to mention communal qualities compared to agentic qualities on their websites. To what extent does emphasis on these traits influence electoral success? The correlations between use of traits and electoral results are generally weak (about 0.09 for agentic traits and 0.03 for communal traits). However, such weak correlations are mostly confined to Democrats, with the correlations of 0.02 and –0.07, respectively. Among Republicans, the correlations between traits and election results are more substantial (0.31 for agentic traits and 0.5 for communal traits). This descriptive analysis does not take into account possible confounders, and candidates are often using a mix of these strategies. To better assess how voters react to the use of stereotypic compared to counterstereotypic strategies, we turn to an experimental design.

An Experimental Test of the Effectiveness of Stereotypic and Counterstereotypic Strategies

Study Design

We conducted an experiment using Amazon’s Mechanical Turk (MTurk). Participants completed one survey and were invited to take another survey a few days later (N = 897). For each survey wave, MTurk participants received $1.00 compensation for participation ($2.00 total if they completed both). Sample descriptive statistics are presented in Appendix B, Table 1. As is typical with MTurk samples, our sample is younger, less diverse, more educated, and more left leaning than the general population (Levay, Freese, and Druckman 2016).

Our study is a panel design in which we collect both moderators and background measures in wave 1 and embed the experiment in wave 2. In wave 1, we ask participants to select traits associated with certain social groups, complete demographic questionnaires, reveal their partisan dispositions, and complete a sexism inventory. In wave 2, we adopt a modified version of the experimental design used in Bauer (2017) in order to directly manipulate candidate strategies. The experiment manipulates candidate gender (woman or man) and traits emphasized (agentic, communal, neither).

In wave 2, participants are presented with a primary election context that matches their own partisanship, as reported in the first wave. Specifically, Democratic participants receive the condition about Democratic candidates and Republican participants receive information about Republican candidates. Independent leaners are assigned to the party primary they lean closest to, while pure independents are randomly assigned to one of the two party primaries. The purpose of this partisan control is to preserve experimental power by only focusing on in-partisans (Bauer 2017 looked at both in-partisans and out-partisans in one of her studies).

The study was posted on the MTurk online forum with a title, a brief overview of the study, and length of the study (approximately 7 minutes). After agreeing to participate in the study, participants complete a demographic questionnaire, reveal their partisanship, and complete the sexism battery. Three days later, participants receive an invite with a direct link to the second wave.

In the second wave, participants are randomly assigned to one vignette on a hypothetical candidate. In each of the conditions, participants are asked to read about a hypothetical candidate whose gender varied between being a man or a woman and who either emphasized no traits (the control), agentic traits, or communal traits. The traits are embedded in the vignettes, which are the same as those used by Bauer (2017).

The control condition does not have any signals of candidate traits (see Appendix C). One condition highlights whether the candidate had their children present at a rally, presented themself as compassionate, and as a parent who understands the challenges families face, thus emphasizing communal traits (see Appendix C for the full text). In Table 1, we highlight the key elements of the treatment that signal these traits and cross it with the sex of the candidate. This condition is a stereotypic strategy for a woman running for office since it is consistent with gender stereotypes, and it is counterstereotypic for the man running. The other condition focuses on agentic traits, highlighting whether the candidate had business partners at the rally, presented an aggressive approach to politics, and focused on their role as a business owner who understands the challenges facing voters. This condition is a stereotypic strategy for men and a counterstereotypic strategy for women.

Table 1. Text of treatments, sex of candidate, and strategy type

Measures

In wave 1, in addition to asking about demographics and partisanship, we ask respondents to fill out a 22-item Ambivalent Sexism Inventory. Participants are asked the degree to which they agree (on a 6-point scale) with statements measuring “hostile” or “benevolent” prejudice toward women (Glick and Fiske 1996). For hostile sexism, sample items include “Most women interpret innocent remarks or acts as being sexist” and “Women are too easily offended.” For benevolent sexism, sample items are “Many women have a quality of purity that few men possess,” “Women should be cherished and protected by men,” and “Every man ought to have a woman whom he adores.” From these items, we create two variables to measure the level of hostile sexism and benevolent sexism. Reliability analyses of scales yield acceptable alpha coefficients (alpha = 0.93 and 0.94, respectively).

In wave 2, after the candidate vignette, we measure candidate categorization with a battery of questions asking participants to rate how well a specific trait describes the candidate they read about, with values ranging from 1 (not well at all) to 5 (very well). The batteries include a series of agentic and communal attributes. The agentic items correspond to masculine stereotypes and include the following traits: assertive, coarse, tough, aggressive, stern, masculine, active, rational, and self-confident. These items are adopted from Bauer (2017) and derived from Huddy and Terkildsen (1993), reflecting the characteristics that individuals associate with political leaders. Ratings that are more positive indicate higher perceptions of the candidate holding these agentic traits. The communal items correspond to feminine stereotypes and include the following traits: gentle, sensitive, feminine, emotional, talkative, and cautious. Individuals tend to associate these traits with women and women’s communal gender roles as mothers and caregivers (Eagly and Karau 2002). Ratings that are higher indicate perceptions of the candidate as being higher on these communal traits. We create two 5-point composite measures (one for agentic attributes, the other for communal traits) by taking the average of the corresponding batteries (alpha = 0.76 and 0.62, respectively).

In line with work by Bauer (2017), we then ask participants to rate the candidate on strong leadership and knowledge using the same 5-point scale. Following these questions, we ask participants to rate how well they think the candidate they read about is able to handle certain issues on a 5-point scale. We include the following issues: national security, education, crime, health care, taxes, unemployment, women’s health, and race relations. National security, crime, and taxes are typically considered better handled by male politicians, while women are generally perceived as better able to handle education, health care, unemployment, women’s health, and race relations (Huddy and Terkildsen 1993). These issues are sorted into two types, which range from 1 to 5 with national security, crime, and taxes categorized as masculine issues (alpha= 0.83), and the other issues categorized as feminine issues (alpha = 0.85).

Finally, to measure overall evaluations, we ask participants to rate the candidate on both likability and warmth with a 100-point feeling thermometer and to indicate their likelihood of voting for the candidate. If the counterstereotypic woman candidate faces a backlash effect, participants will rate them higher on leadership and knowledge but less positively on warmth and likeability, and they will be less likely to vote for the candidate.

Findings: Average Treatment Effects

We first test how stereotypic and counterstereotypic strategies influence ratings of candidates on agentic and communal traits. To refresh, we expect to find that women who employ counterstereotypic strategies will be perceived as holding more agentic attributes ( H _1a ), while those emphasizing stereotypic strategies should be perceived as holding more communal attributes ( H _1b ). To test these arguments, in Table 2, we present t-tests between the stereotypic and counterstereotypic conditions compared to the control group by the candidate’s sex. The values in the table therefore reflect the difference between the relevant treatment group and the control group, matched to the candidate’s sex. Positive values indicate that participants view the candidate as being higher on the given attributes compared to the control group.

Table 2. The effect of strategies on evaluation of traits, leadership, and knowledge

Notes: Each entry contains the difference from the given treatment to the control group that matches the sex of that candidate, with standard errors in parentheses. * p < .1; ** p < .05; *** p < .01. All dependent variables are coded on a 5-point scale.

As expected, we find that women who employ counterstereotypic campaign strategies receive a significant boost in public perceptions of their agentic traits ( H _1a ). This finding is in line with Bauer (2017). This effect is also substantively meaningful in that it results in half a unit shift on the agentic attributes scale (control mean = 2.77). We also find that women using stereotypic strategies leads to a 0.41-unit boost in evaluations of communal traits, in support of H _1b (control mean = 3.04). This finding is in line with extant literature that suggests emphasizing female candidates’ stereotypic qualities can lead to higher evaluations of their communal traits (Burns, Eberhardt, and Merolla 2013). However, the employment of counterstereotypic strategies causes a 0.28-unit decrease in evaluations of communal traits. Women therefore appear to face some cost to perceptions of communal attributes when they employ this strategy (Bauer 2017; Eagly and Karau 2002; Krupnikov and Bauer 2014).

As for men, our findings indicate that sticking to traditional masculine campaign strategies (stereotypic condition) increases perceptions of their agentic traits by three-quarters of a unit relative to the control group (control mean = 2.73) and decreases perceptions of their communal traits by over half a unit (control mean = 2.76). Employing counterstereotypic strategies leads to an increase in evaluation of their communal traits by about one-quarter of a unit, but does not harm evaluations of their agentic qualities. It should be noted that using the same type of campaign strategy can have differential effects for both genders. Furthermore, men do not face a cost on perceptions of their agentic attributes for using counterstereotypic strategies, but women face a cost on perceptions of their communal attributes from using such strategies.⁹

There is clear evidence that candidates can alter perceptions of their agentic and communal attributes by using different types of strategies. The next important question is whether these strategies then influence other evaluations. In Table 3, we explore whether the use of stereotypic and counterstereotypic strategies translates into higher evaluations of leadership and knowledge. As expected, women who employ counterstereotypic strategies are significantly more likely to be perceived as strong leaders and as more knowledgeable ( H _2a ). They are half a unit higher in perceptions of leadership and one-quarter of a unit higher in perceptions of knowledge relative to the control group woman condition. Thus, while counterstereotypic strategies do lead to some backlash against women in public evaluations of their communal traits, they can be conducive for evaluations of other traits (Bauer 2017; Burns, Eberhardt, and Merolla 2013). Men who use strategies that emphasize their agentic attributes are also more likely to be perceived as strong leaders, by one-third of a unit relative to the control male candidate. However, women and men who emphasize communal attributes in their campaign communications receive no boost in leadership evaluations.

Table 3. The effect of strategies on leadership and issue competence evaluations

Notes: Each entry contains the difference from the given treatment to the control group that matches the sex of that candidate, with standard errors in parentheses. * p < .1; ** p < .05; *** p < .01. All variables are measured on a 5-point scale.

Next, we expand on Bauer’s work (2017) to consider whether the use of these strategies affects perceptions of the candidate’s ability to handle masculine and feminine issues. The findings are presented in the last two columns of Table 3. Our findings show that women who adopt counterstereotypic strategies are indeed perceived as more capable of handling masculine issues (control mean = 3.11) such as crime and security ( H _2b ), and this does not lead to any penalty or increase in perceptions of their ability to handle feminine issues (control mean = 3.56) (thus, we do not find support for H _3a or H _3b ). Women who adopt more stereotypic strategies do not see an increase in perceptions of their ability to handle feminine or masculine issues. These results provide further evidence of the potential benefits of breaching gender expectations in campaigns and the glass-ceiling effect of emphasizing feminine traits (Burns, Eberhardt, and Merolla 2013). In addition, male candidates adopting counterstereotypic strategies are perceived as more capable of handling feminine issues like education and healthcare (control mean = 3.28), and they do not face a penalty in perceptions of their ability to handle masculine issues (control mean = 3.24).⁵ Unlike women, men emphasizing areas of stereotypic strength see a slight advantage in perceptions of their issue competency on masculine issues.

Finally, we explore the effect of these strategies on overall favorability in Table 4. We do not find any effect of the use of counterstereotypic strategies on overall favorability for women candidates (control means = 71.11, 60.13, and 3.37, respectively). Therefore, women do not appear to face a backlash from using such strategies, in support of H _4b , and consistent with Bauer (2017). Men using counterstereotypic strategies do, however, appear to face a penalty on likability and warmth, but not on vote choice (control means = 72.84, 73.95, and 3.32, respectively). Finally, we do not find that using stereotypic strategies helps with overall favorability for women. Therefore, even though we witnessed greater use of stereotypic strategies by women in the 2018 election, it does not appear that the strategy improves overall favorability ratings. Therefore, we find little support for H ₅ in the full sample. Men get a slight bump in warmth when using stereotypic strategies, but not on the other two indicators.

Table 4. The effect of gender strategies on overall evaluations

Notes: Each entry contains the difference from the given treatment to the control group that matches the sex of that candidate, with standard errors in parentheses. * p < .1; ** p < .05; *** p < .01. Voting variable is measured on a 5-point scale, whereas likability and warmth are on a 100-point scale.

Findings: Moderating Effects by Hostile and Benevolent Sexism

The aggregate results may be masking heterogeneous reactions to the use of these strategies, so next we evaluate how individual differences in attitudes about and toward women influence the effect of gender strategies employed by candidates. We utilize a series of interaction models to explore such potential moderation effects. Specifically, we consider how hostile ( H _6a ) and benevolent sexism ( H _6b ) moderate the effect of the treatment conditions. Thus, for each outcome variable, we run a model with dummy variables for each treatment condition (the female control serves as the baseline), the two subscales of sexism, and interactions between the two subscales and each treatment.

Given that this entails a large set of analyses, we summarize the significant results here and show more detailed results in the Supplementary Material (see Appendix C, Tables 1–18).¹⁰ We begin by looking at hostile sexism. We do not find any significant interaction effects between hostile sexism and the experimental conditions for perceptions of the candidate’s agentic qualities, communal qualities, competence on feminine issues, leadership, or knowledge. However, we do find a meaningful interaction for competence on masculine issues, likeability, warmth, and likely vote, but only for the counterstereotypic woman candidate condition. In each of these models, those higher on hostile sexism have more negative evaluations of women leaders. However, the key question is whether hostile sexism moderates individual reactions to the treatments.

In Figure 1, we plot the marginal effect of the counterstereotypic woman candidate condition relative to the control condition across levels of hostile sexism. In the top left hand panel of the figure (a), we find that those higher in hostile sexism who are exposed to the counterstereotypic woman treatment view the candidate as more competent on masculine issues relative to their counterparts in the control condition with a woman candidate. However, as we can see in the top right of the figure, those at the highest levels of hostile sexism also view such a candidate as less likable (panel b). We observe a similar effect when we examine perceptions of warmth in the lower left panel (c). However, those low in hostile sexism perceive the counterstereotypic woman candidate as higher in warmth.

Figure 1. Marginal effect of the counterstereotypic woman candidate condition relative to the woman control condition, by levels of hostile sexism. The x-axis stands for the level of hostile sexism (low to high), with bars representing the sample distribution on that scale. The y-axis stands for the treatment effect on the corresponding dependent variable relative to the control condition.

We might expect that these cooler evaluations on likeability and warmth then translate into a lower likelihood of voting for the counterstereotypic woman candidate. However, as can be seen in the lower right panel (d), those higher in hostile sexism become more likely to vote for the counterstereotypic woman candidate relative to their counterparts in the control woman candidate condition, while there is no effect among those low in hostile sexism. These findings are therefore more consistent with a pattern by which those higher in sexism evaluate women who display agentic qualities on the campaign trail as more competent on masculine issues, and are more likely to support the candidate, even if they have more negative feelings toward her. Another important implication is that the mixed results for the effect of these strategies on warmth and likeability for women candidates may relate to prior work not considering individual differences in sexist views.

If we turn to moderating effects by benevolent sexism, few of the interaction terms are statistically meaningful. Benevolent sexism does not moderate the effect of the treatments on perceptions of agentic qualities, knowledge, competence on feminine issues, likeability, warmth, or vote likelihood. We do find a significant moderating effect with the counterstereotypic woman condition for perceptions of communal qualities. As illustrated in the left panel of Figure 2(a), those low in benevolent sexism to the middle value rate the counterstereotypic woman candidate as lower on communal qualities than their counterparts in the woman control candidate condition. We also observe a significant moderating effect for this candidate type on perceptions of leadership. As can be seen in the right panel of Figure 2(b), even though the counterstereotypic woman candidate is perceived as lower on communal qualities, those with low to middle values on benevolent sexism view her as a stronger leader. We only find one other significant effect for another candidate type. Those higher in benevolent sexism perceive the stereotypic candidate who is a man as more competent on masculine issues compared to the baseline control candidate. It is quite interesting that out of all the analyses save this last one, all of the cases in which we find a significant moderating effect by hostile or benevolent sexism are for the counterstereotypic woman candidate. Therefore, it is largely this candidate, who violates gender expectations, that is evaluated distinctly on some dimensions among those at varying levels of hostile sexism or benevolent sexism, with more consistent findings for the former.

Figure 2. Marginal effect of the counterstereotypic woman candidate condition relative to the woman control condition, by levels of benevolent sexism. The x-axis stands for the level of benevolent sexism (low to high), with bars representing the sample distribution on that scale. The y-axis stands for the treatment effect on the corresponding dependent variable relative to the control group.

What may be somewhat puzzling is that the stereotypic woman candidate does not garner more positive evaluations, even among those low in hostile or benevolent sexism, especially when more women who ran for office in 2018 highlighted stereotypic traits in their campaign communications. However, the lack of much effect for highlighting stereotypic qualities is consistent with existing research (Bauer 2017) and our main treatment effects. It is also possible that other measures of individual difference might reveal more positive reactions to women who use stereotypic strategies. For example, it is possible that Democratic women react more positively to such strategies (Burns, Eberhardt, and Merolla 2013), as well as mothers and younger women. These are valuable questions for future research to address.

Discussion and Conclusion

The year 2018 has been dubbed the “year of the woman” since a record number of women ran for and won elected office. In a context in which gender issues were salient, many women running for office highlighted communal qualities in their campaign communications. Our content analysis demonstrates that women in that election cycle were more likely to emphasize a mixture of agentic and communal qualities on their campaign websites. This is in contrast with past work showing a much greater emphasis on agentic qualities than communal qualities by women on campaign websites and campaign ads in prior election cycles (Carpinella and Bauer 2021; Fridkin and Kenney 2014). Presumably candidates felt empowered to emphasize not only agentic qualities but also communal qualities because they thought doing so would appeal to voters.

One goal of our study is to assess whether voters react more positively to women highlighting communal traits compared to agentic traits in campaign communications in today’s political environment. The results from our experiment suggest that the answer is no. Those in our study exposed to a woman emphasizing communal qualities did see that candidate as possessing such qualities, but this did not translate into more favorable assessments of the candidate’s knowledge, competence, warmth, or ability to handle feminine issues (or masculine issues). Individuals are also unmoved in their likelihood of supporting the candidate when she emphasizes communal traits. Instead, women who present agentic qualities in their campaign communications are perceived as possessing those traits, as more competent on masculine issues, and as stronger leaders and more knowledgeable. These findings are consistent with work by Bauer (2017). And, women emphasizing counterstereotypic traits do not suffer penalties with respect to warmth and likeability, consistent with Bauer (2017) but counter to other findings (Eagly and Karau 2002; Krupnikov and Bauer 2014).

Our findings suggest that, at least on average, women get more mileage from emphasizing agentic traits in their campaign communications, even in a political environment in which gender is particularly salient. In terms of implications for women candidates in actual elections, they may want to put slightly more emphasis on these counterstereotypic traits. This is not particularly surprising given how highly valued agentic traits are in political leaders (Eagly and Karau 2002; Koenig et al. 2011). The findings do not suggest that women should avoid talking about stereotypic traits—there were no negative effects of such strategies on evaluations, just very limited positive effects.

A second goal of our work is to examine whether there are differences in how individuals react to such strategies, particularly depending on their views about the role of women in society. To do so, we use a panel design, so that we would not prime gender attitudes before the experiment. Again do not find any evidence that emphasizing communal qualities benefits women candidates, not even among those who are low in benevolent or hostile sexism. Instead, some of the advantages women receive in emphasizing agentic traits are among those higher in hostile sexism (on competence in handling masculine issues). At the same time, those higher in hostile sexism do perceive the candidate as less warm and likeable—thus there was some penalty and evidence of backlash. However, and importantly, those higher in hostile sexism are still more likely to say they would vote for the women who emphasizes agentic traits compared to the control candidate. One important reminder is that participants in our study only evaluate copartisan candidates. Those higher in hostile sexism might exhibit more negative reactions toward an out-partisan woman. The effects are different for benevolent sexism, where those low in benevolent sexism penalize the counterstereotypic candidate on communal qualities, but perceive of the candidate as a stronger leader. These findings are thus more nuanced than what the average treatment effects reveal and suggest that an important next step in this line of research is considering measures of individual difference that may condition how people react to the strategies that women employ on the campaign trail.

The results from our study indicate that men do not experience exactly the same consequences as women candidates. The men in our study similarly benefit from emphasizing agentic traits in their campaign communications, including in warmth, though for men, these are stereotype consistent. Employing counterstereotypic strategies for men leads to an increase in evaluation of their communal traits and ability to handle feminine issues; however, unlike women who use counterstereotypic strategies, men suffer a penalty in warmth and likeability. This is somewhat different from the null findings on warmth and likeability for men employing these strategies in Bauer’s (2017) study. Importantly, we do not find that reactions to the strategies that men employ systematically vary by benevolent or hostile sexism. It is only for the counterstereotypic woman candidate that we these measures of individual differences matter for reactions.

One important lingering question for future work is whether there are other measures of individual differences that lead to more favorable evaluations of women emphasizing their communal qualities in campaign communications. While scholars have considered partisanship, gender, and now types of sexism, there may be other factors that moderate how people react to the traits women emphasize to voters, such as parental status, age, and race/ethnicity.

The findings also raise some interesting puzzles. On the one hand, existing work on women’s representation finds that there is an expectation that women fulfill traditional roles. For example, women candidates with children are preferred to those without children (Teele, Kalla, and Rosenbluth 2018), and motherhood has become more politicized and valued in the political sphere (Deason, Greenlee, and Langner 2015). At the same time, motherhood is linked to communal traits, and our findings show that highlighting communal traits does not really result in any benefit for women candidates, while highlighting agentic traits does (which is consistent with Bauer 2017). One important avenue for future research, then, is to untangle motherhood from strategic emphasis on different traits. In the design we use, the communal treatment mentions children, thus highlighting parenthood, as well as communal traits. Such an approach has been used in the literature given the strong linkages between motherhood and expectations of communal traits (see, e.g., the discussion in Deason, Greenlee, and Langner 2015). However, it is possible for a candidate to highlight being a mother and agentic traits, such as Palin’s emphasis on being a “pitbull with lipstick” (Burns, Eberhardt, and Merolla 2013).

Our focus on campaign strategies emphasizes candidate traits, given that a more extensive body of work focuses on issues. And, our primary interest is understanding how the use of stereotypic and counterstereotypic strategies might impact evaluations of women running for office. Our findings have implications for the quality of representation. For example, the use of communal traits on websites might be a type of symbolic representation (Brown and Gershon 2016; Mansbridge 1999; Pitkin 1967), showing that traits stereotypically associated with women are valued in the political sphere. And, to the extent that certain strategies enhance women’s electoral prospects, there will be implications for increased descriptive representation in government. At the same time, the use of these strategies in campaigns does not in and of itself guarantee substantive representation. It is possible, however, that the traits candidates highlight when they are running for office are a signal of the approaches they will take to solving important substantive issues while in office. For example, research shows that women legislators are more collaborative than men (e.g., Barnes 2016; Holman and Mahoney 2018; Swift and VanderMolen 2021). It is possible that a collaborative approach is used even more among women who emphasize stereotypic (communal) traits when they run for office. It may also be that women who highlight counterstereotypic traits take a more assertive approach when engaging in things like floor speeches or engaging with bureaucratic actors (Lowande, Ritchie, and Lauterbach 2019). An exploration of linkages between campaign strategies and behavior in office is a fruitful avenue for future research.