CONCEPTUALIZING CANDIDATE SEX AND GENDERED SELF-PRESENTATION

Pioneering feminists and queer theorists developed “gender” as a distinct concept to signify the idea that society perceives and then enforces distinctions between men and women (“sex”) and this shapes social dynamics in a wide variety of ways (Scott Reference Scott1986), while others focused on the idea that gender is constructed via behaviors, such that when they are enacted this constitutes the creation of one’s gender identity (i.e., gender is “performative”) (Butler Reference Butler and Nicholson1990). At the root of this foundational work is the notion that gender is socially constructed and linked to hierarchy—in other words, it is intimately tied to power and politics. Although mainstream empirical political science has generally been slow to grapple with gender as a concept (Bittner and Goodyear-Grant Reference Bittner and Goodyear-Grant2017; Schneider and Bos Reference Schneider and Bos2019), there are key ways in which it ought to inform our analysis of various political phenomena.

Specifically, we argue conceptualizing gender as distinct from sex reveals significant voter contributions to the large partisan gap in women’s representation via a preference for masculinity. When we discuss “gender” we mean the culturally constructed “symbolism of masculinity and femininity that we connect to being male-bodied or female-bodied” (Wade and Marx Ferree Reference Wade and Ferree2019, 5). More precisely, the way political candidates “do gender” is a manifestation of what is variously called gender performance, gender expression, gender (self-) presentation, or gender strategy depending on intentionality and context. Language surrounding these issues can be tricky and the needs of the research question should drive how we approach conceptions of gender (Lindqvist, Sendén, and Renström Reference Lindqvist, Sendén and Renström2020).Footnote ³

We have chosen to use the relatively neutral term of gender self-presentation (or the umbrella term of “gender”) to capture ambiguity about the intentionality of a candidate’s performance of gender in the context of our work. We think agnosticism about motives is especially appropriate for our work for two reasons. First, our focus is on entry-level candidates with little or no political experience participating in neighborhood caucuses. Some of these candidates may be strategically interacting with cultural scripts about gender, but most likely are not. Second, and perhaps more importantly, our interest is less on the motivations of the candidates and more on the responses of the voters. For example, when candidates talk about being a nurse, voters may read that as a presentation of femininity whether the candidate intended to deploy feminine symbolism or not.

Operationalizing gender self-presentation has become an active discussion in the gender and politics field as scholars have pushed the field to think about gender and candidate evaluations. Schenider and Bos draw on classic sociological work and lay out one approach—social role theory—which argues that “gendered traits and stereotypes develop as a result of the differential roles that men and women occupy in society” (Reference Schneider and Bos2019, 175). This differentiation takes many forms, but two of the most observable are occupational segregation (176) and public opinion (178).

Our approach to designing the treatments in this project was to create short stump speeches that realistically reflect the way candidates for entry-level political positions present themselves. Our qualitative data reveal these kinds of speeches tend to focus on qualifications (work, prior political experience, life experience) and issue priorities. As Schneider and Bos (Reference Schneider and Bos2019) note, both of these dimensions are likely to have gendered components. Indeed, Carroll and Sanbonmatsu (Reference Carroll and Sanbonmatsu2013) show that men and women often come to politics with different backgrounds and life experiences. Men state legislators are more likely than women state legislators to come from law, business, and agriculture; women state legislators are more likely than men state legislators to come from education, nursing, the nonprofit sector, and homemaking (Reference Carroll and Sanbonmatsu2013, 23–4). These differences reflect widespread job segregation in the general population (Blau, Brummund, and Liu Reference Blau, Brummund and Liu2013) and are likely to have implications in elections. Men and women candidates—again, like men and women in the general population—also have somewhat different policy preferences and priorities (Crowder-Meyer Reference Crowder-Meyer2007; Lawless Reference Lawless2015). Through the lens of social role theory, we are able to see how occupations and issues that are dominated by men or women become the content of stereotypes regarding masculinity and femininity.

There are, of course, many other ways candidates may express gender in campaigns. Trait-based gender performance is also important, with masculinity focusing on “agentic” traits and femininity on “communal” traits (Conroy and Green Reference Conroy and Green2020; Schneider et al. Reference Schneider, Holman, Diekman and McAndrew2016) and can be operationalized via scales like the Bem Sex-Role Inventory that McDermott (Reference McDermott2016) employs in a survey of voters’ gendered personalities. Though the specific context of our research did not focus on personality or traits, we see all of this as important to understand the full terrain of how candidates do gender.

While conceptions of masculinity and femininity are derived from the types of things that male- and female-bodied individuals tend to do, they are also available for men and women to perform independent of sex. When applied to candidate self-presentation, this means two things are true: men and women candidates can self-present in masculine or feminine ways (or both), but it is more common that men candidates will present in masculine ways and women candidates in feminine ways. The latter point is key to our argument about voters’ roles in women’s representation. Discussions of representation tend to focus on the number of men and women in office—in other words, they focus on candidate sex. Empirical approaches that try to ascertain whether voters are to blame for women’s underrepresentation generally do so by trying to make all else equal comparisons between men and women who run. But the challenge with these approaches is that “all else equal” misses important real-life gender differences in pathways to power. In other words, sex and gender are confounded in the real world.

Hence, comparing identical men and women candidates means that one is not comparing typical men and women candidates. A focus on sex bias (whether men or woman candidates are disadvantaged in the electoral process) rather than gender bias (whether masculinity or femininity is disadvantaged in the electoral process) can miss the mechanisms that indirectly lead to lower women’s representation.

The bulk of the existing research on the role voters play in women’s underrepresentation focuses on measuring sex bias either through experimental manipulation of the sex of a hypothetical candidate or observational work that seeks to make ceteris paribus comparisons about the success of men and women candidates via controls (Bauer Reference Bauer2020b; Brooks Reference Brooks2011, Reference Brooks2013; Burrell Reference Burrell1994; Darcy, Welch, and Clark Reference Darcy, Welch and Clark1994; Dolan Reference Dolan2004; Dolan and Lynch Reference Dolan and Lynch2013; Ekstrand and Eckert Reference Ekstrand and Eckert1981; Fox and Smith Reference Fox and Smith1998; Hayes Reference Hayes2011; King and Matland Reference King and Matland2003; Madsen Reference Madsen2019; Mo Reference Mo2014; Sanbonmatsu Reference Sanbonmatsu2002; Smith and Fox Reference Smith and Fox2001), with a recent meta-analysis suggesting that there is little or no direct bias against women candidates among voters (Schwarz, Hunt, and Coppock Reference Schwarz, Hunt and Coppock2018). Yet most scholars studying voter support for men and women candidates either explicitly or implicitly acknowledge that to the extent that voters discriminate, underlying gender stereotypes are likely to blame (Bauer Reference Bauer2015; Reference Bauer2020a; Dolan Reference Dolan2010; Dolan and Lynch Reference Dolan and Lynch2016; Hayes Reference Hayes2011; Lawless Reference Lawless2004; Mo Reference Mo2014; Sanbonmatsu Reference Sanbonmatsu2002; Schneider and Bos Reference Schneider and Bos2014).

Given this widespread understanding, it is puzzling that there has been relatively little direct testing of the way that candidate gender self-presentation shapes voter evaluations. Huddy and Terkildsen’s (Reference Huddy and Terkildsen1993a; Reference Huddy and Terkildsen1993b) foundational work is a notable exception, as is the handful of more recent studies from political psychology (Bauer Reference Bauer2020a; Reference Bauer2020c) and campaign strategy (Schneider Reference Schneider2014) perspectives that push gender back to the forefront. Our work draws on many of these same themes while also seeking to more fully interrogate the role that partisan stereotypes, the femininity/competence double bind, and the resulting selection effects have on the broader question of why there has been such divergence in recent decades between women’s representation within the Republican and Democratic parties.

PARTISAN STEREOTYPES, THE DOUBLE BIND, AND SELECTION EFFECTS

Beyond providing conceptual clarity, distinguishing between how voters’ responses to candidate sex and candidate gender self-presentation is important because it holds clues for the partisan patterns in women’s representation that shape American politics. Since the 1980s, the image of the Republican Party has become tightly linked to masculinity, both explicitly and implicitly. The development of modern partisan issue ownership leaves Republicans with perceived expertise on “masculine” topics like taxes and defense and Democrats with perceived expertise on “feminine” topics like healthcare and concern for the poor (Petrocik, Benoit, and Hansen Reference Petrocik, Benoit and Hansen2003). This also bleeds into partisan trait ownership—Democrats are more compassionate (i.e., feminine) and Republicans are better leaders (i.e., masculine) (Hayes Reference Hayes2005). Hence, “when people think about the Republican Party, they are likely to draw on ideas about masculinity” (Winter Reference Winter2010, 589). The nomination and election of Donald Trump in 2016 exacerbated this tie between Republican politics and the performance of masculinity (Boatright and Sperling Reference Boatright and Sperling2019).

Republican voters also tend to be more committed to traditional gender roles than Democratic voters (Horowitz, Parker, and Stepler Reference Horowitz, Parker and Stepler2017). And conservative gender attitudes among Republicans have reverberated through elections in important ways (Cassese and Barnes Reference Cassese and Barnes2019; McDermott Reference McDermott2016; Schaffner Reference Schaffner2022; Schaffner, MacWilliams, and Nteta Reference Schaffner, MacWilliams and Nteta2018; Sharrow et al. Reference Sharrow, Strolovitch, Heaney, Masket and Miller2016; Thomsen Reference Thomsen2015; Wineinger Reference Wineinger2019). These partisan differences have been brewing for decades, part of a growing cultural divide between the two parties in which Republicans focus more on moral traditionalism, including traditional gender roles (Layman Reference Layman2001; Wolbrecht Reference Wolbrecht2000). Of course, attitudinal and policy preference variation exists among Republican identifiers (Barnes and Cassese Reference Barnes and Cassese2017). But since intraparty contests tend to mobilize stronger and more ideologically extreme partisans (Karpowitz and Pope Reference Karpowitz and Pope2015), those with the most conservative views have an outsized influence on candidate selection and broader party culture (Layman and Weaver Reference Layman and Weaver2016). This may help to explain some of the findings about women’s disadvantages in primary elections (Lawless and Pearson Reference Lawless and Pearson2006).

If Republican politics is tied to masculinity and Republican voters have more traditional views on gender roles, many women candidates may struggle to present themselves in ways that are appealing to Republican voters—especially the most conservative Republican voters. The tightrope of needing to present as both strong and competent to comply with masculine ideals of leadership while also refraining from offending traditional gendered expectations of femininity is likely to be most acute for Republican women candidates. Jamieson dubs this the “femininity/competence double bind” and notes that it is likely to be difficult for women politicians to be seen as competent without being seen as unlikeable (Jamieson Reference Jamieson1995). While some research disputes the existence of this kind of double bind for women politicians (Barbara Lee Family Foundation 2016; Brooks Reference Brooks2013; Hayes and Lawless Reference Hayes and Lawless2015), recent research finds that these factors do matter for how candidates both emerge (Oliver and Conroy Reference Oliver and Conroy2020) and are perceived (Conroy Reference Conroy2016; Reference Conroy2018).

There is some research specifically about how Republican women have attempted to navigate this double bind. For example, Sarah Palin’s “mama grizzly” persona was an attempt to repurpose frontier mythology to create a space for strong but feminine women (Gibson and Heyse Reference Gibson and Heyse2014). Of course, as Wineinger (Reference Wineinger2019) points out, access to this solution to the double bind is primarily available to white Republican women. Former Representative Mia Love (R-UT), the first Black Republican woman to serve in Congress, had to navigate the gendered terrain much more carefully to avoid triggering “angry Black woman” and “welfare queen” stereotypes and thus focused much of her self-presentation on motherhood. Republican women from other racial and ethnic backgrounds face their own raced-gendered intersectional tightropes. For example, recently elected Korean-American Members of Congress Young Kim and Michelle Steel of California leaned on their conservative Christian identities and legal immigration stories as they campaignedFootnote ⁴ and have identified as “tough cookies” and “tiger moms”Footnote ⁵ as they address anti-Asian discrimination. Further, research suggests that lesbian candidates may face a particularly tricky masculinity/femininity tightrope, with men voters preferring more feminine self-presentations and women voters preferring more masculine self-presentations, though “these findings may not directly apply to the few cases of openly gay Republican candidates” (Golebiowska Reference Golebiowska2001). Other research confirms that women voters are more accepting of lesbian candidates, but that conservative voters have much lower levels of acceptance overall (Doan and Haider-Markel Reference Doan and Haider-Markel2010). And voters perceive disabled candidates in highly gendered ways (Reher Reference Reher2021) that could complicate navigating the double bind.

At the elite level, candidates have access to professional pollsters, consultants, and advisers who can steer them toward messages and imaging that account for the challenge of having to appear strong and competent as well as feminine and likable (Dittmar Reference Dittmar2015). But these dilemmas could have an especially profound impact on the early stages of the political pipeline of women, where women do not typically have access to these resources.Footnote ⁶ Because of this, we argue that there is good reason to believe that there are selection effects at play regarding which types of women are able to successfully navigate these challenges.

The possibility of early selection effects based on masculinity is important for contextualizing the decades of electoral return analysis that consistently find little bias against women (Burrell Reference Burrell1994; Darcy, Welch, and Clark Reference Darcy, Welch and Clark1994; Dolan Reference Dolan2004; Dolan and Lynch Reference Dolan and Lynch2016; Ekstrand and Eckert Reference Ekstrand and Eckert1981; Hayes Reference Hayes2011; Smith and Fox Reference Smith and Fox2001). Most studies of real-world vote choice focus on elections for relatively high office, namely Congress and statewide executive office. The majority of candidates running for these offices already have electoral experience (Fowler Reference Fowler1996), so these studies generally measure the difference in outcomes between men and women who have already proven themselves to be electable. Indeed, analyses of congressional candidate self-presentation mostly find that men and women candidates present themselves similarly (Banwart and Winfrey Reference Banwart and Winfrey2013; Dolan Reference Dolan2005) and some even find that men are more likely to emphasize “feminine” aspects of their stories, like their families (Stalsburg and Kleinberg Reference Stalsburg and Kleinberg2016).

We theorize these races may be too far downstream to catch the most significant ways in which voter biases shape women’s representation. Voters are most likely to rely on stereotypes in low-information elections (Crowder-Meyer, Gadarian, and Trounstine Reference Crowder-Meyer, Gadarian and Trounstine2020; McDermott Reference McDermott1997), such as intraparty contests and entry-level elections. Disadvantages at the earliest stages of the political process can prevent women from gaining political experience that would allow them to be seen as part of the pool of well-qualified candidates for higher office. It can also prevent women from being part of the political networks that most party leaders turn to when thinking about candidate recruitment (Crowder-Meyer Reference Crowder-Meyer2013). In fact, the handful of studies that have looked at lower-level offices do show some signs of discrimination against women candidates (Brown, Heighberger, and Shocket Reference Brown, Heighberger and Shocket1993; Brown Reference Brown1994; Crowder-Meyer, Gadarian, and Trounstine Reference Crowder-Meyer, Gadarian and Trounstine2015). Even small biases among voters can compound across the stages of the political pipeline to office, let alone over the course of several elections. Who runs for state and federal office is likely to reflect these selection effects, so examining earlier contests is crucial.

In short, we make three claims about the relationship between voters and the large partisan gap in women’s representation. First, voter biases may not be against women per se; rather, they may reflect preferences for masculinity over femininity, and women disproportionately perform femininity. Second, dynamics in the Republican Party that strongly favor masculinity may make it particularly difficult for those whose gender self-presentation skews feminine, disproportionately women, to successfully navigate elections. Finally, studying voter biases in entry-level, intraparty political contests is essential to understanding the role voters play in limiting the pipeline of women candidates. The typical focus on federal and state-level elections is problematic because selection effects have already taken root by that stage. In other words, the partisan gap in women’s representation is the result of early stage selection effects in the Republican Party that are technically neutral between men and women candidates but strongly favor masculinity over femininity.

STUDIES 1 AND 2: EXPERIMENTS ON CANDIDATE GENDER PREFERENCES AMONG DEMOCRATS AND REPUBLICANS

To isolate the effect of candidate gender self-presentation on voter behavior, it is necessary to vary candidate gender self-presentation. In the experimental studies that follow, we draw on data about gendered pathways to office and gendered policy priorities to create candidate profiles that have stereotypically masculine and feminine qualifications and issue priorities. Our vignettes focus on precinct party office elections, as these are an entry-level opportunity for citizens to stand for election and require few qualifications and only a modest time commitment. We thus see this as a hard case for finding gender bias among voters, since the stakes are low and a very wide variety of candidates could adequately represent their neighborhood copartisans in this role.Footnote ⁷

With these vignettes, outlined in Table 1, we conducted a survey experiment on the 2016 and 2018 CCES to identify patterns in candidate gender preferences among Democratic and Republican voters.Footnote ⁸ Our experiment was included on the pre-election modules that included 1,000 respondents, with a total of 2,000 respondents completing the study over the two election years. Because the research designs of the two studies were identical and results were nearly so, we combine them here for purposes of analysis, a choice that has the added virtue of bolstering our statistical power.Footnote ⁹

Table 1. Experimental Profiles for Woman Candidates 2016–2018 CCES

Respondents were presented with a vignette election experiment with hypothetical candidates Julie Baker and Steven Hall using the following introduction: “Below are two hypothetical candidates for the position of delegate to the state [Republican/Democratic] convention. Imagine that they were running for state delegate in your precinct and gave the following speeches. Please read their speeches carefully, make a vote choice, and answer the additional questions on the pages that follow.”

In this iteration, Steven Hall always gave a “masculine” speech and Julie Baker gave one of the four speeches listed in Table 1: Masculine; Feminine; Law School Graduate Mom; and Mom. This allows us to ascertain the independent effect of a woman candidate’s gendered self-presentation on voter evaluations and vote choice. The first panel shows the language that was shown to Republicans, Republican leaders, and independents who said they would prefer to participate in a Republican primary. The second panel shows the language that was shown to Democrats, Democratic leaders, and independents who preferred to participate in a Democratic primary.

The Masculine condition pairs the occupation of bank vice president or software engineer with a brief conservative or liberal issue statement regarding the economy. The alternative not randomly assigned to Julie Baker was used as the masculine profile for Steven Hall. For the Feminine profile, we randomly assigned Julie Baker to be a teacher or a nurse and to talk about education. The Law Graduate Mom has a law degree to signal a middle ground of a more masculine set of qualifications paired with a more feminine issue like education. In the Mom Only condition, Julie Baker only discussed her role as a mother who volunteers at her children’s school (though she did not specifically identify herself as a stay-at-home-mom) and as in the other feminine profiles, emphasizes the issue of education. Both the occupation and issue content information, as well as the brevity of the speeches, are extremely realistic for these kind of precinct-level party elections (see Study 5 for more details).

We asked respondents to vote for one of the candidates and rate the likeability and competence of each one. Figure 1 outlines the ratings for likeability and competence that the CCES respondents gave the woman candidate compared to the baseline for the man candidate. The top panel shows the results for Democrats. Strikingly, Democrats always saw the woman candidate as significantly more likeable than the man candidate and feminine profiles received an exceptionally large boost on likeability. What’s more, Democrats’ competence ratings did not follow a “double bind” pattern, in which likeability is seen in opposition to competence and vice versa. The Masculine woman was seen as statistically indistinguishable from the Masculine man with regard to competence. Democratic respondents saw the Feminine woman and the Law Graduate Mom as more competent than the masculine man. Only in the Mom Only condition was the woman seen as less competent than the man. This is not to say that Democratic women candidates face no gender-related barriers, but Democratic voters tend to see a wide variety of woman candidates as both more likeable and more competent than a stereotypically masculine man candidate.

Figure 1. Gender Differences in Candidate Trait Evaluations, 2016 and 2018 CCES

Among the nationally representative sample of Republicans, however, a different pattern emerges. The woman candidate in the Masculine condition was seen as about as likeable as the masculine man, while the women candidates in the more feminine conditions were regarded as more likeable. However, the competence ratings followed the opposite pattern, yielding solid evidence of a likability/competence double bind for women candidates among Republican voters. These effects are modest but meaningful. Among Republicans, for example, Cohen’s d for the effect on likeability of the Mom Only condition (as opposed to the Masculine condition) was 0.26, and the effect on competence was 0.49.

Figure 2 shows the results of the respondents’ vote choice, and the dashed line is at 0.5—the point where the choice between the man and the woman is a coin flip. In every condition, Democratic voters significantly preferred the woman candidate, with the Feminine woman winning an astounding 75% of the time. Republican voters were less enthusiastic toward the woman candidate. In the Masculine condition, Republicans were indifferent between identical, masculine-coded men and women candidates. This replicates previous experimental work that finds little sex discrimination against identical women and men candidates. Republicans were also willing to vote for the Feminine candidate. However, the Law Grad Mom and Mom Only faced an electoral penalty, though only the Mom Only condition is statistically significant.

Figure 2. Women’s Electoral Success by Condition and Respondent Party, 2016 and 2018 CCES

What’s more, we find strong indications that voter’s self-identified ideology plays an important role in these evaluations among Republicans. Figure 3 presents these results. Among Democrats (Figure 3a), we find no strong difference between the most liberal respondents and voters with more moderate ideological commitments. The pattern is quite different in Panel B. Self-reported strongly conservative Republicans penalized women candidates in the more feminine conditions, specifically the Feminine and Mom Only conditions. This penalty is substantial: strong conservatives were about 20 percentage points less likely to vote for women who self-presented with these profiles. In the Feminine condition, the difference between strong conservatives and other Republicans was also especially pronounced. However, strong conservatives did not seem to penalize the Law Grad Mom, though the confidence intervals are large. One interpretation of these results is that once a woman disclosed a more “masculine” credential, the strongest conservatives were willing to vote for her at rates equal to the masculine man candidate.

Figure 3. Women’s Electoral Success by Condition and Respondent Party and Ideology, 2016 and 2018 CCES

These results are consistent with the idea that the parties have diverged, with femininity being culturally acceptable and even desirable in the Democratic Party but presenting challenges among Republican voters, especially the most conservative Republican voters. Nevertheless, the sample sizes from our CCES data limit the extent to which we can dive into what is happening among Republicans, so the next section presents data from two large surveys of politically active Republicans.

STUDIES 3 AND 4: EXPERIMENTAL EVIDENCE OF ELECTORAL PENALTY FOR FEMININITY AMONG REPUBLICANS

In this section, we use experimental data from two large original surveys of politically active Republicans in a conservative, Republican-dominated state to understand in more detail the dynamics of candidate gender preferences among Republicans. Because these surveys are drawn from party caucus participants, they give us an especially clear view into the ways in which the party selectorate views candidates. Politically active individuals like the ones we survey have an outsized influence on the party ethos (Layman et al. Reference Layman, Carsey, Green, Herrera and Cooperman2010), so understanding their candidate preferences may be especially important. Study 3 is very similar to the design of the CCES experiments, while Study 4 expands on these experiments to test the effects of masculinity and femininity for both men and women candidates.

Study 3: 2016 Survey of Republican Caucus Participants

Study 3 consists of an online survey of 2016 Republican presidential caucus participants in this conservative, Republican-dominated state. We administered this survey just a few days after the 2016 party caucus meeting. We had unexpectedly high cooperation rates and received more than 10,000 responses, about half of whom were asked to respond to the experiment described here.Footnote ¹⁰

After some warm-up questions, respondents were presented with a vignette election experiment very similar to the one in Study 1. Respondents were asked to evaluate candidates Steven Hall and Julie Baker to represent their precinct at a state party convention. As in the CCES experiment, for Steven Hall, the speech content always included masculine content about occupation and a brief issue statement. For Julie Baker, the speech content was randomized between five parallel profiles: Masculine, Masculine + Mom, Feminine, Feminine + Mom, and Mom Only.Footnote ¹¹ Table 2 contains the full question wording of the five possible candidate occupation and issue profiles randomly assigned to Julie Baker.Footnote ¹² Of particular note are the +Mom conditions, which added “…and a mom who volunteers at my kids’ school” to Julie Baker’s masculine or feminine occupation.

Table 2. Experimental Profiles for Woman Candidates 2016 Caucus Attender Study

In these studies, we again asked respondents for a vote choice between Steven Hall and Julie Baker, as well as their impressions of the candidates’ likeability and competence. Figure 4 presents the gender differences in judgments about the candidates, with positive scores indicating that respondents had a more positive view of the woman candidate than the (always masculine) man and negative scores indicating the opposite.

Figure 4. Gender Differences in Candidate Trait Evaluations by Condition, 2016 GOP Caucus Attender Study

As in Studies 1 and 2, when the woman candidate presented herself in a masculine way, her likeability rating was indistinguishable from the masculine man candidate. But when a women’s self-presentation included more feminine traits, she gained a significant “likeability advantage” over a masculine man. Though the full scale has a 7-point range, 95% of cases ranged between −1 and +1, so this advantage represents between 1/3 and 1/2 of a standard deviation (Cohen’s d ranges between 0.28 and 0.38). The most dramatic example of this is while women in the Masculine condition enjoyed no likeability advantage over men, women who simply added to their more masculine profile a mere mention that they volunteered at their children’s school were seen as substantially more likeable than the man candidate. This likeability advantage was present for each of the more feminine profiles (Feminine, Feminine + Mom, and Mom Only).

Ratings of competence again follow an opposite pattern. In the Masculine condition, women were seen as equally competent as men. But, in every other condition, all of which presented women in more stereotypically feminine ways, women were seen as substantially less competent than men (Cohen’s d ranges between 0.32 and 0.60). Together, the two trends in Figure 4 offer more clear evidence of a classic “double bind” for women candidates. When women described themselves as having more stereotypically feminine career paths and issue interests, respondents liked them more than the man candidate but simultaneously judged them to be less capable.

How do the experimental conditions affect respondents’ vote choices? Figure 5 presents the proportion voting for the woman candidate, and several key results emerge from the figure. First, women are neither advantaged nor disadvantaged electorally when they present themselves with a more masculine profile. The proportion voting for the woman candidate was almost exactly 0.5. This result again replicates previous experimental work that has shown little disadvantage for women running for office when they have identical profiles to men.

Figure 5. Women’s Electoral Success by Condition, 2016 GOP Caucus Attender Study

But our results also qualify previous work in important ways. For example, women who self-presented with a masculine profile and also mentioned their identities as mothers received a significant electoral advantage, with voters choosing the woman candidate much more than half of the time. This “supermom” profile is the only experimental condition in which women had an electoral advantage among the respondents in this sample of Republican caucus attenders. Conversely, women who self-presented in more stereotypically feminine ways and did not mention more masculine attributes were electorally disadvantaged, with the proportion voting for a woman falling well short of 0.5 in the Feminine, Feminine + Mom, and Mom Only conditions. The greatest disadvantage for women was in the Mom Only condition, where the proportion voting for a woman was only 0.426 (95% CI ranges from 0.395 to 0.457). Relative to the Masculine condition, this is the difference between a coin flip election and a nearly 15-point loss.

As in the CCES data, self-identified ideology plays a key role in driving these results. We analyzed the data by self-reported degree of conservativeness. Figure 6 shows these trends. “Strong conservatives” were generally much less likely to support women candidates than other kinds of Republicans across the board.Footnote ¹³ Feminine women faced especially stiff electoral penalties among strong conservatives, a fact that is especially important given evidence that those who self-identify as most conservative are also more likely to participate in primary elections and caucuses designed to winnow the candidate field (Brady, Han, and Pope Reference Brady, Han and Pope2007; Karpowitz and Pope Reference Karpowitz and Pope2015). We estimate that among strong Republicans, women who self-presented with a Feminine or Mom Only profile were likely to face a more than 20-point electoral rout. However, among other Republicans, the results were more promising for women. These more moderate Republicans were especially enthusiastic about the Masculine + Mom profile, and Mom Only was the only woman candidate who faced a clear electoral penalty.

Figure 6. Women’s Electoral Success by Respondent Ideology, 2016 GOP Caucus Attender Study

In short, women do better with Republican voters—especially more conservative ones—when they focus on more stereotypically masculine qualifications and issues like taxes and the budget deficit rather than more stereotypically feminine issues like education. Women who self-present by emphasizing comparatively more feminine qualifications and issue interests faced a classic double bind; while they are seen as more likeable, they are also judged as less competent and were ultimately less likely to receive the votes of Republican caucus attenders, with the strongest conservatives driving this pattern.

Study 4: 2018 Survey of Republican Caucus Participants

Thus far, we have made the case that Republican voters do not necessarily show a bias against women candidates, but rather show a bias against feminine candidates. In other words, the underrepresentation of women in the Republican Party is the result of gender bias more than sex bias, especially among the most conservative voters. However, to fully make the case that femininity is an electoral liability independent of sex, we need to examine what happens when respondents are presented with masculine and feminine men candidates as well as women candidates.

To do so, in 2018 we recontacted the pool of Republican voters we surveyed in 2016 with a new survey experiment. Once again, we received more than 10,000 responses.Footnote ¹⁴ In this experiment, we again measured support for Julie Baker and Steven Hall with four profiles: Masculine, Masculine+Parent, Feminine, and Feminine+Parent (see Supplementary Table A6 for full text). However, instead of always assigning Steven Hall the masculine profile, we allowed for all 16 possible permutations between the candidates. This fully crossed design allows us to compare the effect of a masculine or feminine profile on Julie to the effect of a masculine or feminine profile on Steven, which is important for untangling the influence of sex and gender.

Figure 7 reveals the effect of the experimental treatments on the gap in likeability and competence. The left panel of the figure displays the results when the man candidate took on a Masculine profile across each of the four profiles to which the woman candidate could have been randomly assigned.Footnote ¹⁵ These results essentially replicate our findings from the two previous survey experiments. When Steven had a Masculine profile, Julie experienced no deficit in likeability or competence if she also had a Masculine profile. Her likeability increased and competence decreased when she emphasized her role as a parent, and the competence-likeability double bind was largest in the conditions where Julie’s was described with a Feminine profile or when she combined her Feminine profile with references to motherhood.

Figure 7. Gender Differences in Candidate Trait Evaluations, 2018 GOP Caucus Attender Study

The right panel of Figure 7 shows how trait evaluations are affected by giving the man candidate the Feminine profile. The third set of bars, for example, demonstrates the effect of giving the man candidate a feminine profile and the woman candidate a masculine profile. In this condition, the pattern was the inverse of the double bind the feminine woman candidate experienced in our previous studies and in the left panel of Figure 7. When given a Masculine profile and running against a Feminine man, Julie was seen as more competent and less likeable than Steven. In other words, when we reversed candidate sex and gender, it was the man, not the woman, who experienced the double bind.

However, the right panel of Figure 7 is not a perfect mirror reverse of the left panel. When compared to Steven with a feminine profile, Julie experienced an advantage in both competence and likeability in nearly every other experimental treatment. Thus, our results suggest that Republican voters especially penalize feminine men. In other words, gendered expectations interact with and cannot be completely disentangled from sex because men who violate those expectations by taking on a feminine profile are regarded in less positive ways across multiple evaluative dimensions.

Figure 8 illustrates how the various permutations affected vote choice decisions, with each of the 16 boxes indicating one experimental condition. The estimated probability of voting for the woman candidate in each matchup is highlighted by the color coding, with darker reds indicating a greater likelihood of a victory for her. In this sample, unlike the 2016 sample, voters somewhat preferred Julie over Steven on average. But we do see considerable variation across the different matchups, with both Julie and Steven performing much better when they have masculine profiles than when they have feminine profiles. The worst scenario for Julie (38%) is when she has the Feminine profile and Steven has the Masculine Dad profile, and the best scenario for her (67%) is the reverse—when she has the Masculine Mom and he has the Feminine profile.

Figure 8. Probability of Voting for Woman Candidate by Condition, 2018 GOP Caucus Attender Study

As in other analyses, these results become even more dramatic when we subset our sample on self-identified “strongly conservative” respondents. Figure 9 shows an extremely clear pattern of masculinity mattering more than sex for these respondents. When both candidates present as masculine or both present as feminine, the outcomes depend on who talks about parenthood but generally show a weak preference for Julie (in keeping with this sample’s overall weak preference for her). However, when Julie presents as feminine and Steven as masculine (the upper left-hand quadrant of the figure), she faces a deficit of between 18 and 34 percentage points.Footnote ¹⁶ When Steven presents as feminine and Julie as masculine (the lower righthand quadrant), he faces a deficit between 16 and 32 percentage points. In other words, both candidates are extremely unlikely to win the support of conservative voters when they have a feminine profile and their opponent has a masculine profile.

Figure 9. Probability of Voting for Woman Candidate by Condition, 2018 GOP Caucus Attender Study, Strong Conservatives Only

Together, these results provide evidence that among Republicans, gender presentations profoundly shape voters’ judgments about candidates and their leadership traits, and they do so independent of candidate sex. This influences vote choice, especially among the most conservative Republicans.

STUDY 5: OBSERVATIONAL DATA ON REPUBLICAN MEN’S AND WOMEN’S SELF-PRESENTATION

In the real world, is variation in candidate gender actually correlated with candidate sex? Existing research suggests that women and men do have somewhat different pathways to power (Carroll and Sanbonmatsu Reference Carroll and Sanbonmatsu2013), but there is no evidence that men and women candidates for offices like gubernatorial or congressional races present themselves in dramatically different ways with regard to masculinity and femininity (Banwart and Winfrey Reference Banwart and Winfrey2013; Dolan Reference Dolan2005; Stalsburg and Kleinberg Reference Stalsburg and Kleinberg2016). However, we suspect this may be the results of selection effects and that data from entry-level elections may show gender differences.

In this section, we report data we gathered from neighborhood Republican Party precinct caucus meetings in 2014 in the same conservative state as Studies 3 and 4.Footnote ¹⁷ We find solid support for the idea that among these candidates for entry-level party offices, there are distinct differences in how men and women present themselves with regard to their qualifications and issue priorities, and these correspond with stereotypical notions of masculinity and femininity. Of course, there are also many similarities between men and women in these races, but we believe the evidence we present here provides an important plausibility check for our argument that voters have ample opportunity in entry-level races to shape the candidate pipeline according to their gender biases and that this disproportionately hurts women candidates.

We sent dozens of trained student observers to a random sample of locations within a four-county geographic area that contains about 75% of this Republican-dominated state’s population. They gathered information about who ran for delegate and other precinct positions, as well as what was said during those contests. The neighborhood caucus meetings examined here are much like the well-known Iowa presidential caucuses in structure and form. Redlawsk, Tolbert, and Donovan (Reference Redlawsk, Tolbert and Donovan2011, 24) succinctly describe the Iowa caucuses as “lengthy local party meetings used to conduct party business and select delegates.” Neighbors in the precinct gather as partisans to nominate and elect precinct leaders, collect funds for the local party organization, discuss issues, and select delegates to county and state nominating conventions.

We focus on candidate speechmaking and self-presentation in campaigns for the position of delegate to the state party convention (“state delegates”). State delegates have historically played an important role in nominating candidates for statewide and federal offices and determining the party platform. These are the most important and competitive races held at the caucus meetings, with over 97% of state delegate contests having more than one candidate. However, barriers to election as a state delegate are low, even for political novices. About 3,500 delegates are elected statewide every two years at more than 2,000 precinct caucus meetings that are attended by as few as a half-dozen or as many as a couple of hundred voters per meeting. Getting nominated requires no advance preparation; campaigning is typically limited to a short speech of a minute or two at the neighborhood meeting. Turnover from year to year is high. Hence, these caucuses are an ideal setting to study and observe low-stakes, grassroots candidate emergence. Women typically comprise about 20–25% of elected delegates, despite the fact that caucus meetings are, on average, roughly evenly split between men and women.

To collect detailed data about the proceedings of the precinct meetings, we sent trained observers who were blind to the purposes of the study to a sample (n = 130) of meetings held in 2014.Footnote ¹⁸ Observers kept systematic notes about the candidates who ran for office, including which candidates gave speeches and which did not, as well as the content of these speeches. (See Supplementary Appendix for additional details about the observers.) After observers completed their work, a separate set of coders, also blind to the overall purpose of the study, were trained to read and code the observers’ notes. The observer’s notes on speeches were not especially extensive. At the same time, the norm for these speeches is that they are quite short. Thus, while coding for the content of speeches is a blunt instrument for identifying patterns because of their brevity, it is also likely that student observers could accurately capture the gist of what was said.Footnote ¹⁹

Overall, we have speech summaries for 339 men and 120 women candidates. Table 3 captures the gender differences in the general topic areas mentioned by the candidates (see Supplementary Appendix for details of the coding instructions). Large percentages of both men and women candidates talked about their qualifications for office, raised issues they cared about, and made reference to ideology. In this sense, these speeches are similar in structure to stump speeches candidates at higher electoral levels might give.

Table 3. Gender Differences in Topics Mentioned 2014 Neighborhood Caucus Observations

Note: Cell entries in the women and men columns are average values for each gender; standard errors are in parentheses. The difference column, z, and p-values are generated from a difference-of-proportions test. N = 339 men, 120 women.

Men and women both mentioned qualifications, issues, ideology, and neighborhood most frequently. But, men were somewhat more likely than women to talk about the party platform, while women were substantially more likely than men to talk about gender and to reference family in some way. This is the opposite of what analysis of congressional candidates finds (Stalsburg and Kleinberg Reference Stalsburg and Kleinberg2016), but it is consistent with studies about the politicization of motherhood (Deason, Greenlee, and Langner Reference Deason, Greenlee and Langner2015) and how conservative women politicians frame their family experiences (Schreiber Reference Schreiber2016; Wineinger Reference Wineinger2019).

More specifically, Table 4 presents the types of qualifications men and women highlighted in their speeches, with results limited to the subset of candidates who mentioned any qualifications at all. Among this group, many candidates talked about their professional backgrounds (work as lawyers, doctors, engineers, nurses, or teachers, for example), their previous experience as a state delegate, previous experience in other party offices, or previous campaign experience. For these qualifications, we find no meaningful gender differences. However, men referenced executive or business experience (work as business managers, small business owners, and other similar management positions) much more often than women and were marginally more likely to invoke past military service. Women were substantially more likely than men to describe themselves as parents or homemakers.

Table 4 also shows the gender differences in specific issues raised among the subset of candidates for state delegate offices who mentioned any issues at all. The most common issues referenced by candidates of both sexes were education, the Constitution, marriage and marriage equality, government spending and the deficit, and the neighborhood caucus system itself—a list that is not surprising for a conservative gathering.

Table 4. What Sorts of Qualifications Are Mentioned?

Note: Cell entries in the women and men columns are average values for each gender; standard errors are in parentheses. The difference column, z, and p-values are generated from a difference-of-proportions test. Analysis limited to candidates who mentioned qualifications (N = 199 men, 78 women) or issues (N = 175 men, 60 women) in any way.

Men were somewhat more likely than women to talk about the economy, government spending and the deficit, taxes, and healthcare,Footnote ²⁰ though these differences do not reach statistical significance in a difference-of-proportions test. In a regression model with controls, however, men candidates’ increased likelihood of mentioning government spending and the deficit (p = 0.09) and taxes (p = 0.09) come closer to standard levels of statistical significance. On the other hand, women were dramatically more likely to talk about education than men, with more than 43% mentioning it, compared to 25% of men. This gender difference is statistically robust (p = 0.008) in a difference-of-proportions test and remains so in regression models with controls (p = 0.005). These results are largely consistent with previous work (Crowder-Meyer Reference Crowder-Meyer2007; Karpowitz and Mendelberg Reference Karpowitz and Mendelberg2014), which finds that in the general population, men tend to be somewhat more focused on financial issues like taxes, government spending, and the economy, while women often prioritize “care issues” like education or the needs of families and children.

Overall, our content analysis of the speeches shows many similarities between men and women candidates. But it also reveals meaningful differences between men and women that reinforce and reproduce stereotypically masculine and feminine identities and concerns. While candidates constructed their speeches in broadly similar ways, men and women often emphasized different sorts of social roles, personal experiences, and political issues. Men were more likely to highlight their executive experience and military service, label themselves as conservatives, and emphasize concerns about government spending and taxes. Women were more likely to invoke the role of homemaker or stay-at-home parent, to describe themselves in terms of their familial relationships, to raise the issue of education, and to make gender-based appeals to the voters.

These gendered differences in self-presentation also seem to be related to candidates’ electoral prospects. While our findings here are strictly observational and constrained by the limited information we have about each candidate, regression analysis shows that women who presented in distinctively feminine ways—mentioning being a homemaker or parent, invoking family as a topic, or raising the issue of education—were less likely to be elected as a state delegate (see Supplementary Table A9).Footnote ²¹ The size of the effect for women candidates was substantial, reducing the predicted probability of being elected by more than 20 percentage points (Supplementary Figure A2). For men, the effect of feminine self-presentation was also negative, but much smaller and not statistically robust. In addition, the effect of male self-presentation, defined as mentioning business or executive experience, military background, conservative ideology, taxes and government spending, or deficits, was positive for both men and women, but not statistically significant.

In sum, our observations of real-world precincts show that men and women candidates for at least this type of Republican entry-level partisan office look different from each other. The previous sections show that Republican voters are predisposed to select for masculinity among candidates, and this section shows that they have the opportunity to do so in real-world elections in a way that disproportionately disadvantages women candidates. Theories of representation that fail to account for this miss an important on-the-ground reality.