Gendered Social Roles and Self-Presentation in Campaigns

Women remain numerically underrepresented in public office at all levels of government (Center for American Women and Politics 2022). Many attribute this underrepresentation to sexism against women candidates from the electorate. According to Hayes and Lawless (Reference Hayes and Lawless2016, 3), 47% of Americans agree that women “face bias from voters,” and 31% agree that women “don’t win as often as men.” However, scholarship since the 1990s has generally concluded that men and women perform equally well once they appear on the ballot (see Lawless Reference Lawless2015; Schwarz and Coppock Reference Schwarz and Coppock2022). Women’s underrepresentation may be better explained by gender differences in the decision to run for office. Women disproportionately decline to run compared to similarly situated men for a variety of reasons, including family care obligations (Bernhard, Shames, and Teele Reference Bernhard, Shames and Teele2021; Fulton et al. Reference Fulton, Maestas, Maisel and Stone2006), absent recruitment efforts from party leaders (Fox and Lawless Reference Fox and Lawless2010; Sanbonmatsu Reference Sanbonmatsu2002), and a relative lack of role models in high-level offices (Campbell and Wolbrecht Reference Campbell and Wolbrecht2006; Ladam, Harden, and Windett Reference Ladam, Harden and Windett2018).

Another potential reason women decline to run — one that we argue also relates to women candidates’ messaging choices — is that holding political office violates traditional gender norms. Social role theory holds that gender stereotypes and traits develop from a division of labor by sex in traditional societies — men provide by laboring in the public sphere, whereas women occupy caretaking roles (Eagly and Wood Reference Eagly, Wood, Kruglanski and Higgins2012). Building on that idea, role congruity theory explains that women will face social sanctions for defying gender norms by engaging in agentic behaviors, such as holding leadership positions or positions of power (Eagly and Karau Reference Eagly and Karau2002; Schneider and Bos Reference Schneider and Bos2019).Footnote ³

Together, these theories can help to explain several patterns in women’s self-selection into political candidacy. Men are more likely than women to express an interest in running for office (Fox and Lawless Reference Fox and Lawless2005; Schneider et al. Reference Schneider, Holman, Diekman and McAndrew2016). Resulting pressure from gendered expectations may lead potential women candidates to underrate their own qualifications for running (Fox and Lawless Reference Fox and Lawless2004) and make more risk-averse choices about when and where to run (Brown et al. Reference Brown, Mansour, O’Connell and Reeves2019; Fox and Oxley Reference Fox and Oxley2003; Fulton Reference Fulton2012; Ondercin Reference Ondercin2022; Pearson and McGhee Reference Pearson and McGhee2013). When explaining their interest in running for office, women are also more likely to express communal motivations (e.g., working with others, serving the community) than agentic motivations (e.g., gaining experience, changing the system) (Conroy and Green Reference Conroy and Green2020).

Gender role congruity should also inform how women candidates present themselves to voters. We focus on the presentation of experience. By experience, we mean a career that provided some skills or preparation to the candidate before running for office. Candidates discuss their pre-electoral experience to telegraph competence and build voters’ confidence in the candidate’s readiness for office. In our view, “experience” is closely related to a “qualification,” a frequently invoked term in the gender literature (e.g., Bauer Reference Bauer2020), since both point toward a candidate’s fitness to handle the responsibilities of an office.

Oftentimes, experience connotes prior service in elected office. In the political realm, prior officeholding experience signals that the candidate is already prepared for public service and can take on the responsibilities of elected office with lower starting costs. However, experience (and achievement) in a nonpolitical occupation can also build perceptions of competence. For example, in 2016, then-candidate Donald Trump stressed his experience in business as a qualification for the presidency. When respondents were provided information that Trump’s wealth was largely inherited from his father rather than built from his business, they perceived him as less competent in business and lowered their support for his presidency (McDonald, Karol, and Mason Reference McDonald, Karol and Mason2020).

Women candidates face a different set of challenges than men in communicating experience on the campaign trail because experience is a gendered characteristic, particularly in politics. Americans have long associated leadership with masculinity (Bauer Reference Bauer2020). Experience connotes competence, power, ambition, status, and leadership, all terms that take on an agentic and thus masculine tone (Conroy and Green Reference Conroy and Green2020). Given these often-unconscious expectations, women candidates face an uphill battle proving themselves to the public because voters do not automatically assume women are qualified or competent to hold office (Ditonto, Hamilton, and Redlawsk Reference Ditonto, Hamilton and Redlawsk2014; Mo Reference Mo2015).

A reasonable response to low expectations from voters would be for women candidates to explain why they are qualified to hold office based on their experience. However, a claim of strong qualifications from prior officeholding would be a claim that one has engaged in power-seeking behavior in the past, itself a gender norm violation for women (Schneider and Bos Reference Schneider and Bos2019). Declaring one’s qualifications could invite backlash from voters for norm-violating behavior. As an alternative, women candidates might choose to downplay or omit their experience when campaigning in an effort to avoid anticipated backlash from voters. Although we generally expect discussion of experience to violate gender norms, we acknowledge that women candidates could evoke experience in ways that conform with gender norms, such as discussing how their experience helped them collaborate or achieve communal goals. Discussion of experience might also be aimed to demonstrate a candidate’s commitment to action on a specific issue or to loyalty to a group of voters.

In political science research, the extent to which voters truly penalize women candidates for discussing their experience remains an open question. When campaigning, women candidates face diminished prospects in electoral contexts when voters expect more agentic behavior from their leaders, such as in times of terrorism or when choosing leaders for high levels of office (Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2016; Huddy and Terkildsen Reference Huddy and Terkildsen1993; McDonald and Piatak Reference McDonald and Piatak2022). Messaging experience may produce a countervailing effect that helps women overcome stereotypes of incompetence (Bauer Reference Bauer2017, Reference Bauer2020; Ditonto, Hamilton, and Redlawsk Reference Ditonto, Hamilton and Redlawsk2014; Holman, Merolla, and Zechmeister Reference Holman, Merolla and Zechmeister2017; Mo Reference Mo2015).Footnote ⁴ Still, that messaging may come at a cost. When women promote their own experience, evaluations of their competence tend to improve, but evaluations of their social attractiveness or likeability simultaneously decline (Bauer Reference Bauer2017; Rudman Reference Rudman1998).

Regardless of the extent to which voters penalize women for messaging experience, candidates’ perceptions of what voters expect should determine candidates’ messaging choices more than voters’ true expectations. Candidates operate in environments of uncertainty about voters’ beliefs and expectations (Hershey Reference Hershey1974). Politicians’ sometimes-misplaced beliefs and assumptions about voters guide their behavior (Miler Reference Miler2010).

Therefore, even if voters’ gendered expectations do not ultimately affect the outcomes of elections, candidates often choose to message as if they might. Dittmar’s (Reference Dittmar2015) study of how campaign professionals address gender dynamics supports this claim. The campaign consultants and strategists she interviewed expected voters to apply gender stereotypes to candidates, even though they held mixed opinions on whether gender stereotypes ultimately affected election outcomes. Nonetheless, they accounted for gender in their messaging choices to counteract or forestall potential threats from violations of gender stereotypes. In Dittmar’s telling, recognizing and reacting to potential voter stereotypes often involved not a direct challenge to gender stereotypes but an adaptation of a message to provide maximum advantage to women candidates within the confines of a gendered political landscape. She writes, “While my interviews revealed little disruption in which credentials are expected of political candidates, they did demonstrate some potential shifts in the sites from which those credentials can be earned” (2015, 125, emphasis in original). Applied to messaging surrounding experience, we might expect women candidates not to evoke their prior officeholding experience directly but instead to discuss how forms of nontraditional experience qualify them for office. Pennsylvania State Rep. Tina Davis’ self-presentation as a “mom on a mission,” rather than a long-serving state legislator, would fit that mold.

To summarize the argument, candidates have incentives to present their experience to voters to establish their competence and qualifications for an elected office. However, the risk of violating gender role norms presents women candidates with opposing incentives to downplay or avoid discussion of their prior officeholding experience. On the campaign trail, public communications from women candidates should feature less information about their experience than similarly qualified men. We test the following hypothesis:

Although it is common for prior officeholders to emphasize their experience in ads, it is also common for candidates to emphasize their experience outside of elected office — for example, as business leaders or professionals. It is unclear whether women candidates would violate gender norms in advertising their work experience. A simplistic read of role congruity theory might yield the expectation that gender role violations would occur with any discussion of women’s agency, regardless of whether those pursuits are in political or professional realm. Yet, Schneider and Bos (Reference Schneider and Bos2019, 184) point out a distinction in agentic goals between dominance-oriented goals and independence or individual mastery goals. With the increased entry of women into the workforce over the last century, including greater representation in high-status professions, men and women alike share an interest in achievement. However, achievement does not necessarily indicate holding power-oriented goals; achievement may be instrumental to attaining communal goals. Women tend to endorse individual mastery goals and pursue careers that allow for achievement instead of dominance (Schneider et al. Reference Schneider, Holman, Diekman and McAndrew2016).

Applied to the discussion of experience on the campaign trail, women candidates might be hesitant to discuss political experience — more closely aligned with holding power-oriented goals — but less hesitant to discuss professional experience — which could indicate either power-oriented goals or achievement-oriented goals. Candidates could also intend for a discussion of their career to reinforce a personal brand, to lead voters to evaluate the candidate favorably based on positive stereotypes they have of the candidate’s occupation, or to appeal to specific constituencies in a district. Relevantly, Bauer (Reference Bauer2020) finds that women candidates, relative to men, are more likely to emphasize professional experience on their campaign websites than their political experience.

We acknowledge too, however, that achievement could be taken as a gender norm violation if discussion of it is perceived as self-promoting or bragging. Yet we might reasonably expect a discussion of political experience to evoke power-seeking goals more than other types of experience given its more direct connection to decision-making authority and political ambition. We ultimately hold mixed expectations about the extent to which women candidates will emphasize their professional experience relative to men, but to allow for comparability of results, we test a hypothesis directionally in line with our expectations around political experience. We test the following hypothesis:

Though we expect women candidates to downplay their experience in campaign ads, our hypotheses are falsifiable. Relevant literature has offered evidence supporting competing expectations. Bauer (Reference Bauer2020) advances and finds support for a strategic emergence hypothesis in which women emphasize their experience no less than (and perhaps even more than) men. There are two reasons for this expectation. First, women candidates may anticipate a gendered landscape and assert their qualifications to compete on an even footing with men and head off voters’ stereotypes about incompetence (see Ditonto, Hamilton, and Redlawsk Reference Ditonto, Hamilton and Redlawsk2014). Second, because of unequal expectations surrounding women’s qualifications, better-qualified women with more agentic motivations would select into running for office and run stronger campaigns than their male counterparts (Conroy and Green Reference Conroy and Green2020; Fulton Reference Fulton2012; Pearson and McGhee Reference Pearson and McGhee2013). As a result, self-selection would produce women candidates willing to emphasize their credentials, and no gender gap in messaging surrounding experience would emerge. Such a null finding would also be consistent with a range of studies finding few or no differences between male and female candidates in the issues and traits they choose to emphasize (Dolan Reference Dolan2005; Evans, Cordova, and Sipole Reference Evans, Cordova and Sipole2014; McDonald, Porter, and Treul Reference McDonald, Porter and Treul2020; Niven and Zilber Reference Niven and Zilber2001; Sapiro et al. Reference Sapiro, Walsh, Strach and Hennings2011).

Data and Methods

We test these hypotheses using evidence from television advertisements. Scholars have recently used TV ads in the study of campaign messaging and gender in part because they offer some advantages to understanding candidate self-portrayals over other media (Bauer and Santia Reference Bauer and Santia2023; English, Branton, and Friesenhahn Reference English, Branton and Friesenhahn2024). Unlike news coverage, messaging in TV ads falls under the control of candidates. Although both TV ads and campaign websites are highly accessible to the public, TV ads are more likely to be viewed by more casual observers of politics and may be more relevant to how an average voter perceives a candidate. Finally, and most importantly for this study, the short format of TV ads forces candidates to choose their priorities in messaging to the electorate. Campaign websites are constrained by space but allow candidates to provide much more information about themselves than TV ads. Therefore, TV ads allow us to observe better what “bottom-line” message a candidate wants to convey to the public. A limitation of using TV ads as data is that their production cost limits our ability to observe messaging choices from poorly funded candidates. Any results may or may not generalize to candidates with limited campaign funds.

To evaluate how candidates of both genders presented their previous experience, we watched and coded 1,030 unique TV ads for state legislative campaigns aired during the 2018 election cycle. We obtained ad data from the Wesleyan Media Project, which partners with Kantar Media/CMAG to capture television ads in each electoral cycle (Fowler et al. Reference Fowler, Franz, Ridout and Baum2020). CMAG’s automated system tracks ads on local channels, national networks, and national cable channels. It monitors and records each unique ad that airs, including its date, time, media market, station, and television show.

We chose the 2018 midterm because it produced a unique candidate pool for these state-level races. As a midterm election occurring under Donald Trump’s presidency, the 2018 election cycle saw a record number of women register to run for public office, allowing for greater variation in candidate gender and, in particular, women candidates’ background experiences. However, given the unique gender dynamics of that election cycle, it is possible that the sample of candidates in our data does not generalize to other election cycles.

Although many studies analyze the contents of congressional ads, we chose state legislative ads because the candidates they represent are usually not well known to the public. Candidates at that level must spend more time introducing themselves to voters than congressional or presidential candidates, including by discussing their biographies. Moreover, state legislative campaign spending has continued to increase over the last decade, creating an opportunity to study more widespread televised advertising among state legislative candidates.Footnote ⁵ Finally, thousands of Americans, both men and women, run for state legislature every year. Investigating campaign messaging at this level allows us to observe greater variation in messaging choices and to make inferences about lower-level campaigns, in contrast with studies of higher-salience but lower-frequency campaigns for statewide or national office.

To create a dataset for coding, we eliminated duplicates so that each observation is a unique ad.Footnote ⁶ We also limited our analysis to ads sponsored directly by the candidate’s campaign organization, removing ads sponsored by party organizations or outside groups.Footnote ⁷ Finally, to observe the relationship between a sponsor’s gender and message more cleanly, we removed ads jointly sponsored by multiple candidates, such as ads sponsored by copartisan candidates running in multimember districts. The resulting 1,030 ads represent all unique ads available from the Wesleyan Media Project that fit these criteria. Ads came from the campaigns of 529 different candidates, with each candidate recording between one and 20 unique ads.

We coded the ads to construct two binary dependent variables. The first, Political Experience, is intended to capture candidates’ prior service in elected office. We coded ads as mentioning political experience if the ad presented information that either states explicitly or logically implies the candidate had served in public office before. Sometimes, this came in the form of an explicit naming of a prior office they held. Other times, this came from candidates’ discussions of their votes or actions as a public official, even if they did not explicitly name the office they held.

The second dependent variable, Work Experience, is intended to capture whether the candidate discusses jobs or careers they held before running for public office. We recorded a “1” if a candidate mentions a job title or an industry in which they worked. Work experience would be counted regardless of how much focus was placed on it in the ad. For example, we counted an ad that centered the work experience in the message, as in the case of a Kentucky welder who used his work experience to frame his desire to work in a bipartisan manner in the state legislature. We also counted an ad in which the candidate primarily discussed issues but flashed a job title (e.g., “small business owner”) briefly on screen.

Two coders watched all 1,030 ads individually in their original video format in an initial round of coding. Comparing results, the coders agreed 87.5% of the time that an ad mentioned work experience (Cohen’s kappa = 0.70, signifying good reliability) and 78.7% of the time that an ad mentioned political experience (Cohen’s kappa = 0.58, signifying fair reliability). The coders then reviewed the ads on which they disagreed and made a joint final coding decision. Section A in the online appendix provides further details about the coding process, including a codebook, examples of coded ads, and results from the initial round of coding.

Overall, candidate mentions of experience were common but not omnipresent. Most ads (54.66%) mentioned either the candidate’s political or work experience. However, it was more common for ads to mention political (37.48%) than work (24.17%) experience. Few ads (only 6.99%) mentioned both political and work experience. Candidates messaged their experience quite consistently across their own ads. The average within-candidate standard deviation for Political Experience is about 0.28, and about 0.29 for Work Experience. For context, one would find a similar standard deviation with a binary variable among twelve observations if eleven observations took a value of “0” and one observation took a value of “1.”

We coded the primary independent variable, Woman Sponsor, based on the gender of the candidate sponsoring the ad. We caution that comparing differences between men and women alone has limitations, as it captures the gender identity of the candidate rather than their gender presentation. Conformity to normative social roles varies widely within genders (Diekman and Eagly Reference Diekman, Eagly, Shah and Gardner2008; Schneider and Bos Reference Schneider and Bos2019). Even within genders, gender presentation may have serious consequences for how voters interpret messages about experience and use those messages to decide their vote choice (see Teele, Kalla, and Rosenbluth Reference Teele, Kalla and Rosenbluth2018). To our knowledge, none of the candidates sponsoring ads in our dataset identified as trans, non-binary, or otherwise outside the traditional gender binary.

In the following section presenting the results, we conduct two tests of each of the two hypotheses. We first conduct simple difference of means tests to establish a baseline, bivariate relationship between candidate gender and presentation of each type of experience. Then we proceed with more rigorous tests of the hypotheses by estimating multilevel logistic regression models with covariates. Logistic regression is an appropriate choice for our binary dependent variables of interest. Because ads are the unit of analysis and one candidate could sponsor several ads, the multilevel approach helps account for the clustered structure of the data and candidate effects. We nest ads within candidates and calculate robust clustered standard errors. We note that this modelling approach cannot identify causal effects of gender on discussion of experience. Readers should treat the analysis as descriptive in nature.

Confounding factors like the varying length of the ads, the partisanship of the candidates, and the districts in which they choose to run might also help to explain differences in candidate messaging. To account for potential confounding variables, we add controls for candidate characteristics (incumbency status, prior officeholding experience, party affiliation, race/ethnicity, and campaign contributions received), ad characteristics (ad length, tone, and estimated cost), and the type of election the ad was prepared for (general, primary, or special/runoff election).Footnote ⁸ Table B1 in the online appendix displays descriptive statistics for all variables in the models.

We evaluate each hypothesis using different subsets of the dataset. To evaluate the first hypothesis, we must reduce the sample only to candidates with experience in elected office who could be considered “at risk” of mentioning it. No comprehensive source lists the prior elected experience of state legislative candidates. Through an online search of candidates’ webpages, state legislators’ official websites, local media reports, and election information sites (e.g., Ballotpedia), a research assistant recorded the actual officeholding background of each of the ad sponsors in our dataset. Examining the messaging only in the 614 ads sponsored by candidates who had held elected office before running for state legislature, 384 (62.54%) alluded to the sponsor’s political experience.Footnote ⁹ We assume all candidates have held some prior work experience and so use the full sample of ads to evaluate the second hypothesis.

Political Experience

We begin by evaluating the first hypothesis that men are more likely to mention prior political experience in ads than women. We take a first step by conducting a bivariate analysis of the frequency of experience mentions by candidate gender. Figure 1 displays the results. Among ads from prior officeholders, a statistically significant gender difference appears — 66.81% of ads from politically experienced men mention that experience compared to 48.61% ads from politically experienced women.Footnote ¹⁰ The bivariate results lend tentative support to our expectations.

Figure 1.

Ads mentioning political experience by candidate gender.

Note: Data from the Wesleyan Media Project and the authors.

We proceed with a more rigorous test of the first hypothesis. We specify a multilevel logistic regression model with covariates as described in the previous section. Supporting evidence would be present if we find a negative statistically significant coefficient for the independent variable of interest, Woman Sponsor. We present the full table of regression results in the first column of Table A2 in the online appendix. In line with the descriptive findings in Figure 1, we continue to find that male-sponsored ads are more likely to discuss political experience than female-sponsored ads among experienced candidates. The difference is statistically significant at the 0.05 level of confidence.

Because it is difficult to interpret the size of a relationship from coefficients in a multilevel logistic regression alone, we illustrate the results in Figure 2 by plotting the predicted probabilities derived from the fully specified model in Table B2 in the appendix following the observed value approach (see Hanmer and Kalkan Reference Hanmer and Kalkan2013). Among men-sponsored ads, the probability of mentioning experience is 0.65, whereas for women-sponsored ads, it is 0.53. The size of the difference is meaningful — it translates to about one more ad discussing experience out of every eight.

Figure 2.

Predicted probability of mentioning political experience by candidate gender.

Note: Data from the Wesleyan Media Project and the authors.

Turning to the controls, listed in the first column of Table A2 in the online appendix, we see few meaningful relationships between them and the outcome variable. The results indicate incumbents are significantly more likely to discuss experience than experienced challengers. We also find that promotional ads discuss the sponsor’s political experience more than attack ads. Finally, we see that ads are significantly less likely to mention political experience when run during special or runoff elections than when run during general elections. However, we do not see evidence that a candidate’s party, race, or fundraising level is associated with mentions of experience, nor do we see evidence that the length, cost, or volume of ads is associated with it either. Overall, results of these tests provide empirical support for the first hypothesis.

Work Experience

Next, we turn to evaluating the second hypothesis, that ads sponsored by men will be more likely to mention their work than ads sponsored by women. As before, we begin by observing the raw differences in mentions of work experience between ads sponsored by men and those sponsored by women. Figure 3 displays the findings. We see that women-sponsored ads are slightly more likely than men-sponsored ads to mention the candidate’s prior work experience (26.51% vs. 23.22%, respectively). However, the difference is small — 3.29 percentage points — and not statistically distinct from zero.

Figure 3.

Ads mentioning work experience by candidate gender.

Note: Data from the Wesleyan Media Project and the authors.

We proceed with a more formal test of the second hypothesis by again specifying a multilevel logistic regression model with full controls. We add the control variable for Prior Experience to this model, omitted from the prior model because the sample was limited to ads from politically experienced candidates. Evidence supporting the second hypothesis would be present if we find a negative statistically significant coefficient for the independent variable of interest, Woman Sponsor. The second column in Table B2 in the online appendix presents the full regression results. Contrary to expectations, but consistent with the descriptive findings in Figure 3, we find no differences in discussion of work experience between men-sponsored ads and women-sponsored ads.

Figure 4 illustrates the finding. It shows the predicted probability that an ad mentions the candidate’s work experience by the sponsoring candidate’s gender, holding controls at observed values in the data. The predicted probabilities fall very close to the proportions illustrated in Figure 3. Ads sponsored by male candidates have a 0.26 likelihood of mentioning work experience, whereas ads sponsored by female candidates have a 0.23 likelihood. The difference is not statistically significant. Therefore, we fail to find support for the second hypothesis.

Figure 4.

Predicted probability of mentioning work experience by candidate gender.

Notes: Data from the Wesleyan Media Project and the authors. The figure shows the predicted probability of a mention of political experience with controls held at observed values.

Among the controls listed in the second column in Table B2 in the online appendix, we find that promotional ads are significantly more likely to mention candidates’ work experience. White candidates are significantly more likely to mention their work experience than nonwhite candidates, perhaps reflecting a difference in messaging strategies from candidates of color. Incumbents and prior officeholders are significantly less likely to mention their work experience, in line with the idea that politically experienced candidates advertise their backgrounds in public office instead of their prior careers.Footnote ¹¹ We find no statistically significant relationship between the dependent variable and any of the other controls.

The results with regard to work experience may differ depending on whether candidates hold prior political experience. The descriptive results above show that most politically experienced candidates choose to mention that experience in TV ads, but most candidates do not mention their work experience. Such results suggest that the choice to mention work experience may be conditional on whether the candidate has prior political experience. Politically experienced candidates choose to evoke that background; inexperienced candidates may opt to evoke work experience as a second-best option.

Table B3 in the online appendix shows the results when a mention of work experience is regressed on the sponsoring candidate’s gender and the full set of controls, but with results broken out among politically experienced and inexperienced candidates. In both cases, ads sponsored by women are estimated to be less likely to mention work experience than ads sponsored by men. However, the coefficient estimates are not significantly different from zero in either case. Therefore, we continue to infer that there are no meaningful differences between ads sponsored by men and women in terms of mentioning work experience among either politically experienced or inexperienced candidates.

Supplementary Analyses

In this section, we summarize a series of supplementary analyses exploring additional possibilities in the data. First, we consider whether Democratic and Republican women candidates messaged differently in their ads. We present expectations for the analysis and the findings in Section C of the online appendix. Overall, we find few differences between the two parties. Women’s ads appear less likely to mention political experience than men’s ads within both parties. The difference is more pronounced among Democrats than among Republicans, but Republican women’s ads are not significantly more likely to mention political experience than Democratic women’s ads. We continue to find no notable partisan or gender differences for ad mentions of work experience.

Second, we consider whether candidates message their experience when their opponents do so as well. Section D in the online appendix provides details and results. Here too, we find little noteworthy. Analyzing only general election ads, we find that ads are no more or less likely to mention a candidate’s political experience when at least one of their opponent’s ads mentions the opponent’s political experience. The same finding applies to mentions of candidate’s work experience.

Finally, we consider whether women with high levels of actual political experience undersell their qualifications relative to men with less actual political experience. It is difficult to measure actual experience on a single scale given the wide variety of candidates’ backgrounds. However, we can compare incumbents to nonincumbent prior officeholders. Although not universally true, incumbents usually have more experience in the offices they currently hold than challengers do. Section E in the online appendix provides the findings. We find that gender differences persist for both groups. Women’s ads discuss the candidate’s political experience less than men’s ads among both incumbents and nonincumbents. However, incumbent women’s ads emphasize experience at a higher rate than nonincumbent men’s ads, suggesting that the actual level of experience is related to a candidate’s likelihood of advertising their experience despite gender differences.