Gender and Attack Advertising in U.S. Senate Campaigns

Abstract

We examine if candidate gender influences the issues opponents use in advertising attacks against them in U.S. Senate campaigns. We find evidence that candidates consider partisan issue ownership and issue salience when choosing the topics on which to engage opponents. However, the gender identity of the opponent does not have measurable effects on the selection of issues mentioned in contrast and attack advertising. While highlighting the dominance of partisanship as a key concern in the formulation of attacks against opponents, this study also illustrates the very different issue landscapes faced by women running as Republicans and Democrats, where Democratic women are most likely to face attacks on the issues where both party and gender stereotypes place them at a disadvantage while Republican women are most likely to be attacked where partisan issue ownership puts them at a disadvantage but their gender identity gives them a reputational edge.

Plain language summary

Gender and attack advertising in U.S. Senate campaigns

In this study we examine how political candidates in the United States may or may not attack their opponents differently based on the opponent’s gender. Using a comprehensive dataset of all advertising in U.S. Senate campaigns from 2010 through 2018, we find that the issues used in advertising that attacks opponents does not appear to be influenced by the gender of the target of the ad. Instead, the party of the opponent and the rise and fall of issue importance one campaign cycle to the next has the greatest impact. These findings indicate that women in each party face a different campaign landscape. Because gender and party both carry stereotypes of issue competency, Democratic women are most likely to face attacks on the issues where both party and gender stereotypes place them at a disadvantage, while Republican women are most likely to be attacked where partisan issue ownership puts them at a disadvantage but their gender identity gives them a reputational edge.

Introduction

The underrepresentation of women in Congress remains a persistent feature in American federal politics. Despite significant gains starting in the 1990s, women still only hold 28% of U.S. House seats and 25% of U.S. Senate seats in the 118th Congress (Center for American Women and Politics, 2022). Leading explanations for this gap have converged on the recruitment of women to run: barriers to entry by party gatekeepers (Sanbonmatsu, 2006), lower levels of political ambition where women are less likely to consider themselves qualified compared to male counterparts with similar experience (Lawless, 2012), and an undersupply of women in the pipeline of potential candidates (Thomsen & King, 2020).

Is greater reluctance of women to run and more sluggish recruitment of women by party elites in any way grounded in a realistic expectation of a more difficult campaign environment for female candidates? Recent scholarship on this question is mixed. A meta-analysis reviewing dozens of conjoint and vignette experiments found that women on average have a 2-percentage point advantage over men in candidate choice (Schwarz & Coppock, 2022), yet a recent observational study of campaign spending found that women need to spend more on campaign communications than men to achieve the same electoral result (Herrnson et al., 2022).

Some explanation for the varied findings on women’s electoral performance in the U.S. may lie in a complex messaging environment where social stereotypes impact perceptions of candidates’ personality traits, ideology, and policy competencies. Female candidates benefit from being associated with honesty and compassion, while male candidates are associated with stronger leadership and decision making (Alexander & Andersen, 1993; Huddy & Terkildsen, 1993; Koch, 1999). These trait stereotypes carry further into ideological expectations, as participants in experiments would infer that hypothetical female candidates were more liberal (Alexander & Andersen, 1993; Huddy & Terkildsen, 1993). Further, Koch (2000) found that individuals would perceive actual female candidates to be more liberal than they really were. Finally, these stereotypes impact perceptions of men and women’s competency to handle specific issues, where women are viewed as stronger on compassion issues and men are seen as more competent on leadership-oriented issues (Huddy & Terkildsen, 1993; Kahn, 1992; Sanbonmatsu, 2002).

Campaigns can attack their opponents by playing towards or away from traditionally masculine or feminine stereotypes in ads. We posit that candidates will choose to activate stereotypes by either attacking opponents on issues that are stereotypical to their own gender or candidates will attempt “trespassing” strategies that highlight messages that run counter to the gendered stereotypes of their opponent. Trespassing strategies have consistently yielded effects on voters’ candidate preferences in experimental settings (Bauer, 2015, 2020; Carpinella & Bauer, 2021; Holman et al., 2016). But observational studies of female candidates’ issue engagement have been mixed, in some cases concluding that women more frequently campaign on social issues (Kahn, 1993, 1996) and in other cases finding no difference by gender in the issues candidates emphasize when presenting themselves to voters (Dolan, 2005; Dolan & Lynch, 2017).

However, campaigns’ advertising is often more oriented towards defining the opponent than in presenting the backed candidate to voters (Fowler & Ridout, 2012, 2014; Miller, 2019). So, in this study we turn our attention specifically to how candidates do or do not apply stereotypes of gender in the advertising they use attacking opponents. When women run, does their gender affect the choice of issues invoked by opponents in the advertising against them, or are they attacked purely as Democrats and Republicans?

In addition to offering the first examination of gender and issue emphasis specific to attack advertising, we offer a far more recent look at the interplay between candidate gender, party, and issues in campaigns ads. Prior observational studies on these questions have drawn from elections in 2010 and earlier, a less polarized period where party labels were less informative in legislative races, campaign spending was lower, and independent expenditure groups had a far smaller role in campaign advertising.

In our analyses we find that when holding party and year of the election constant, women are not attacked by their opponents on different issues than their male counterparts. Instead, partisanship and the rising and falling importance of issues in different election years are the dominant drivers of issue selection in advertising attacks. By analyzing the percentage of attack ads on gendered issues mentioned in attack advertising, we find that female Democrats and female Republicans both face different criticisms. This suggests that women in both parties face very different campaign experiences in terms of attack ads.

Gender Stereotypes and Issues in Campaigns

Parties have durable reputations for addressing some issues better than others (Riker, 1983). Issue ownership theory holds that because voters generally regard one party as more competent on some issues than the other party, campaign advertising should elevate considerations of the issues that the favored candidate’s party owns (Petrocik, 1996). Typically, Republicans enjoy a stronger reputation on issues of national defense, taxes, and crime, while Democrats have a better reputation for addressing health care, education, and the environment.

Issue ownership has been theorized to further extend to gender. In experimental settings individuals viewed female candidates on average as more competent in handling welfare, health care, education, and childcare, while viewing women as less competent in handling the issues of military, crime, and foreign policy (Huddy & Terkildsen, 1993; Kahn, 1992; Sanbonmatsu, 2002). Further, gender stereotypes are persistent within parties, suggesting that the presence of party cues does not fully overwhelm perceived differences by gender, but instead acts alongside party reputations (Sanbonmatsu & Dolan, 2009).

In addition, race acts alongside both party and gender when individuals make inferences about candidate positions (McDermott, 1998). These inferences by voters can essentially act as issue reputations for candidates based on their racial identity, where Black and Latino candidates are stereotyped by voters as having more liberal positions (Visalvanich, 2017). Furthermore, these stereotypes can intersect, where Black and Latino women running for office can be even more strongly stereotyped as more liberal than their white or male co-partisans (Philpot & Miller, 2020). Unfortunately, in the analyses that follow we are unable to address the potential overlapping effects of race and gender because women of color have so rarely been their party’s nominee for U.S. Senate seats (see page 5).

Given these expectations about identities and issue reputations, scholars could reasonably expect men and women running for office to pursue different issue strategies in their campaign communications. And candidates themselves may have had reason to expect that playing towards issues consistent with gender stereotypes would be a winning strategy. For example, an analysis based on survey responses from candidates for both federal and state legislative offices in 1996 and 1998 found that women running for office had greater electoral success when they reported that they focused on issues with positive feminine reputations, like education, welfare, or social security (Herrnson et al., 2003).

Studies on how these stereotypes can be activated in voters differentiate between implicit and explicit messaging strategies. Mendelberg (2001, 2008) finds that racialized implicit messages, for example, the “Willie Horton” ad, affects candidate evaluations. Explicit messaging, on the other hand, had the opposite effect since individuals more often reject cues that explicitly convey racial messages (Huber & Lapinski, 2006; Mendelberg, 2001, 2008). Furthermore, implicit attitudes about gender work in conjunction with explicit attitudes to amplify or mute gender stereotypes of female leadership qualifications, and these attitudes are responsive to campaign communications (Mo, 2015).

Whether campaigns attempt to prime gender stereotypes implicitly by attacking on issues with existing gendered reputations, or if campaigns explicitly attack an opponent’s suitability to address an issue based on her gender, both strategies would result in overt mention of these issues. The difference lies in the presence or absence of overt linkage of gender to claims of qualification or performance on an issue.

Observational studies of the differences in male and female candidates’ issue emphases have had conflicting results. Kahn (1993, 1996) found that indeed men devoted more attention to economic issues and women focused more on social issues in Senate campaign advertising in the 1980s. However, Dolan (2005) examined the content of campaign websites for House and Senate candidates in the 2000 and 2002 elections and found that women did not focus on gender-stereotyped issues but campaigned on topics similar to their male opponents. Most recently, a study of the issue content of advertising in the 2010 House and Senate campaigns found that candidates more strongly emphasized issues where their party had an ownership advantage but found no evidence that male or female candidates sought to elevate feminine or masculine issues (Dolan & Lynch, 2017).

In addition, issue selection has been shown to be driven in part by the gender matchup in the race and not just the gender of the sponsoring candidate. That is, a male candidate may present himself differently when running against a male opponent versus when he is running against a female opponent. A study of House and Senate campaign advertising from 2000 through 2004 finds that when men faced a female opponent they were more likely to begin the campaign by focusing their attack ads on feminine issues, whereas female candidates began by avoiding references to feminine issues in their own advertising (Windett, 2014). However, as the campaign neared election day, women countered male opponents’ gendered attacks by increasing the amount of feminine issue references in their advertising.

There are good reasons for campaigns to emphasize issues other than those that the candidate owns by virtue of party identification or gender. Issues most important to voters deserve greater attention, and the importance of issues fluctuates from one election to the next (Budge & Farlie, 1983). Campaigns can respond to these ebbs and flows of public attention by altering the attention devoted to issues in an effort to “ride the wave” of public opinion (Ansolabehere & Iyengar, 1994; Kahn & Kenney, 1999). Campaigns may deviate from issues more favorable to their own candidate and “trespass” on opponent-owned issues based on the state of the economy and issue salience in a given election cycle (Damore, 2004). Specific to gender, there is experimental evidence that women running counter to stereotypes by emphasizing masculine traits can improve their standing with voters, suggesting that there could be electoral benefits for female candidates to run on masculine issues (Bauer, 2017).

Observational studies on gender differences in issues employed in campaign advertising have considered the question of how male and female candidates present themselves to voters. But campaigns use some advertising to define their own candidate, some advertising to define and attack the opponent, and some advertising to draw contrasts between the candidates. Here we ask a different question: do candidates emphasize different issues when characterizing male and female opponents to voters?

Recent studies specific to messages attacking candidates have been confined to experimental designs. Experiments exposing participants to short newspaper stories found that both male and female candidates were vulnerable to attacks on issues stereotypically associated with their party and gender (Cassese & Holman, 2018). However, Bauer (2020) found that messages reinforcing feminine issues did not activate feminine stereotypes among the experiment’s participants.

Given the conflicting conclusions on the effectiveness of different issue emphases in attacks on opponents from experimental studies and the competing evidence of campaigns’ emphasis of owned issues versus trespasses on opposition owned issues from observational studies, we use campaign advertising content from Senate races to evaluate two rival expectations. First, does advertising aimed at female opponents seek to remind voters of the issues where female candidates are presumed to be at a disadvantage, particularly when the attacking candidate is male and his gender does not give him the same reputational disadvantage? Or second, does advertising about opponents instead “trespass” and attempt to chip away at female candidates’ presumed positions of strength, especially when the attacking candidate is male and his gender does not impart the same reputational strength? More formally, the hypotheses we evaluate are as follows:

• Issue ownership hypothesis: Candidates will attack female candidates more frequently on masculine issues.

• Issue trespass hypothesis: Candidates will attack female candidates more frequently on feminine issues.

Or are neither of these hypotheses supported? Given the increasingly elevated partisan polarization in federal races and the integration of gender issue reputations into party images of “masculine Republicans” and “feminine Democrats” (Winter, 2010), does the party of the candidate overwhelm considerations of gender in campaign advertising against opponents?

In addition to offering the first evaluation of the association between issues and gender in campaigns specific to opposing advertising, this article also seeks to confirm the temporal validity of prior scholarship more generally on the issue landscape when women run for office in the U.S. by examining the issue content of attacks on candidates from 2010 through 2018. The current state of knowledge on this question largely rests on studies based on observational data from campaigns in 1982 to 1986 (Kahn, 1996), 2000 to 2004 (Dolan, 2005; Windett, 2014), and the most recent from the 2010 election cycle (Dolan & Lynch, 2017). The campaign advertising landscape since 2010 is different in ways that could influence the content of attacks against opposing candidates. First, polarization in the public has only increased over time. Second, the sheer volume of advertising in campaigns has dramatically increased since the Citizens United and Speechnow.org court decisions in 2010 opened floodgates of new independent expenditures from super PACs and 501(c) “dark money groups” (Fowler et al., 2021).

Data and Method

To examine the issue content of recent campaign attack advertising we turn to U.S. Senate general election campaigns from 2010 through 2018. We choose to study attack and contrast advertising in Senate campaigns for two reasons: First, even Senate campaigns that are not particularly competitive feature higher levels of advertising than House elections, and Senate elections typically include a high volume of attacks against opponents. Second, while governors’ contests are also often well-funded affairs with robust advertising like Senate races, issue selection in gubernatorial races can be affected by idiosyncrasies of local and regional issue concerns that could obscure our results.

The unit of analysis for this study is the total general election campaign, including the total advertising volume by the candidate as well as all party and independent group advertising supporting that candidate. We confine our study to general election advertising for comparability across cases. Senate primaries only occasionally feature any advertising and any attack advertising within primaries would have very different motivations and intended audiences.

To measure issue content in advertising targeting opponents, our study draws on data from the Wesleyan Media Project (WMP) for the 2010 through 2018 elections (Fowler et al., 2014, 2015, 2017, 2019, 2020). The WMP captures and codes all airings of political advertisements on broadcast television in all 210 media markets in the United States for federal and gubernatorial campaigns. The WMP indicates general election advertising with a dummy variable in the data for 2012 to 2018. For 2010 advertising we code ads that ran after the state’s primary as general election ads and ads that ran before or on the date of the primary as primary election ads.

The WMP classifies advertising as positive about the supported candidate, contrast that mentions both the sponsoring candidate (or candidate supported by the ad creator) and the opposing candidate, or negative advertising that mentions only the opposed candidate. Because our interest is in the issues employed when candidates talk about their opponents, our analyses are restricted to contrast and attack advertising. In total, our analyses include the issue mentions in 2,626,262 individual airings of campaign spots about 214 Senate opponents. Of the 214 Senate candidates who were the subject of any television advertising from the opposing campaign in the 2010 through 2018 elections, 28% were women. Consistent with overall partisan differences in candidate gender, in the sample of our study female general election candidates targeted by opposing advertising were more often Democrats (33%) than Republicans (22%).

We are unable to quantitatively address the potential overlapping effects of race and gender due to data limitations because in Senate contests over the period studied the major parties put forward only a handful of non-white women in the general elections: 3 Black women—Joyce Dickerson in South Carolina in 2014, Connie Johnson in Oklahoma in 2014, and Kamala Harris in California in 2016; and only 2 Latina women—Catherine Cortez Masto in Nevada in 2016, Loretta Sanchez in California in 2016. Only one of these, Cortez Masto, was the target of any attack advertising.

WMP data includes indicator variables for the mention of 86 issue categories in each ad airing. In a process similar to that in Franz et al. (2016) and Sides (2006) the issue mentions in each ad are weighted according to the number of issues mentioned in each ad, for example, for a spot that mentions three issues each issue mention is weighted 0.33. The total mentions of each issue in each ad are then summed for the span of the general election campaign. The 86 issues were collapsed into 20 categories to provide a more compact view of the general themes of advertising targeting opponents in Senate campaigns (the complete list of issue codes is provided in Table A1 of the Appendix). This process yields summary measures of the proportion attacks on various issues. Summary statistics for all variables are provided in Table A7 of the Appendix.

In these Senate campaigns we see examples of candidates attacking women both on issues where they should enjoy a reputational advantage and on issues where women are stereotyped as having less credibility. For example, in 2018 Dean Heller (R) attacked his female opponent Jacky Rosen (D) in the Nevada Senate race repeatedly on education and on healthcare, both issues considered an advantage for women. The most prominent Heller campaign advertisements were two versions where grainy black and white images of Rosen are shown while a female narrator accused the congresswoman of passing no legislation to improve healthcare or education during her time in the U.S. House, for example, “Zero. That’s the number of bills Jacky Rosen passed while in Congress before announcing she was running for the Senate. Zero bills to fix education, zero bills to improve healthcare…”

Conversely, candidates may attempt to attack female opponents more frequently on stereotypically masculine issues. In 2016 in the Illinois Senate contest a top theme of attack ads from the Mark Kirk (R) campaign against Tammy Duckworth (D) was security and foreign policy—an interesting choice given Duckworth’s distinguished military service. The most prominent attack line was typified by an ad in which the female narrator attacks Duckworth’s position on the nation’s response to the Syrian civil war and allowing the entry of Syrian refugees “even though the FBI says there’s no way to guarantee they aren’t terrorists.”

More broadly, we examine the patterns of issues raised in advertising opposing Democratic Senate candidates in Figure 1, separately for male and female targets of the advertising. From 2010 through 2018 the top subjects of attack and contrast advertising were healthcare, taxes, the economy, and good government (the last category is a set of issues that includes references to campaign finance reform, government ethics/scandal, term limits, government regulations, government shutdown, and emergency preparedness and response). Female Democratic candidates were attacked about 3 percentage points more often on healthcare, taxes, good government and other social issues such as LBGTQ rights and gun control, while male Democratic candidates were attacked 4 percentage points more often on spending and 3 percentage points more often on Social Security.

Figure 1.

Issues in advertising attacking democratic senate candidates.

Figure 2 displays the frequency of issues appearing in advertising opposing Republican Senate candidates from 2010 through 2018, again separately for male and female targets of the advertising. The top subjects of attack and contrast advertising were the economy, taxes, spending, and healthcare. Female Republican candidates were attacked about 4 percentage points more often than their male Republican counterparts on average on the economy and taxes. Republican men were attacked 6 percentage points more often on average on healthcare.

Figure 2.

Issues in advertising attacking republican senate candidates.

Does the greater emphasis on health care in advertising targeting male Republican and female Democratic opponents indicate attack strategies that factor in the gender of the opponent when selecting issues for attacks? Individually, these differences alternately point towards attacks on issues where the candidate is presumed to have the greatest reputational advantage based on party and gender on one hand and attacks on the other hand on issues where the party and gender identities of the opponent combine to give them the weakest issue reputation. So, we turn in the following section to multivariate assessments of the associations between the topics of attacks and candidate gender, party, and issue salience in the public.

Analysis

Our analysis proceeds in two parts. First, we measure the change in frequency of specific issues appearing in advertising mentioning female versus male opponents. Second, we assess the degree to which female candidates are attacked in campaign advertising more or less frequently on feminine or masculine issues overall.

Individual Issues

Our first step is to evaluate the effect of the targeted opponent’s gender on the selection of issues appearing in advertising mentioning that opponent. We estimate linear models where the dependent variable is the percentage of overall contrast and attack advertising from campaign referring to the given issue. Our principal independent variable of interest in each model is the indicator variable for a female opponent.

We account for the effect of partisan issue reputations with an indicator variable for a Democratic Senate candidate. In addition we include a set of dummy variables for year to account for the rising and falling salience of different issues in each election cycle.

We estimate linear models for each of 20 issues. Because the models are linked, that is, each is an estimate of the advertising proportion on a given issue within the same pool of campaigns where greater emphasis on one issue decreases the potential emphasis on another issue, we use a seemingly unrelated regression (SURE) where the error terms remain independent across observations but these errors are correlated across equations. The functional form for the set of equations where i represents each issue and r represents each observation is as follows:

$$\begin{array}{c}\multicolumn{1}{c}{\%\mathrm{issu}{e}_{\mathrm{ir}}={\beta }_{0i}+{\beta }_{1i}\mathrm{femal}{e}_{\mathrm{ir}}+{\beta }_{2i}\mathrm{Democra}{t}_{\mathrm{ir}}}\\ \multicolumn{1}{c}{+{\mathrm{\Sigma }}_{j=3}^6{\beta }_{\mathrm{ij}}\mathrm{yea}{r}_{\mathrm{ir}}+e_{\mathrm{ir}}}\end{array}$$

(1)

Figure 3 displays the point estimates and 95% confidence intervals for the estimated effects of a female opponent for a campaign’s percentage of mentions of each issue in contrast or attack advertising, while holding party and election year constant (complete results from each of the models are provided in the Appendix Table A2). Only crime reaches the standard threshold of statistical significance, estimated to be mentioned 2 percentage points more frequently in attack and contrast advertising when the opponent is female. No other issue is mentioned more or less frequently in campaign advertising that targets female candidates.

Figure 3.

Effect of female opponent on issue mentions in attack advertising.

Feminine and Masculine Issues

Advertising mentions of any single issue often occur at low frequencies, so statistically significant differences by gender can be difficult to detect. In a second analysis we evaluate our hypotheses with a similar approach used in previous studies by grouping issues into those expected to have feminine or masculine reputational associations (see Carpinella & Bauer, 2021; Dolan & Lynch, 2017; Huddy & Terkildsen, 1993; Windett, 2014). The issues included in each broad category of feminine or masculine issues are listed in Appendix Table A1.

We estimate sets of linear models each for the percent of attack and contrast advertising mentioning masculine (Table 1) and feminine (Table 2) issues. OLS models are employed for ease of interpretation. Linear models where the dependent variable is bounded (in this case at zero and 1) can produce biased estimates, particularly when large numbers of observations occur at boundary values of the dependent variable. To verify that the results presented here are not a consequence of the choice of model form, regression models are also estimated with fractional probit specifications in Tables A3 and A4 in the Appendix.

Table 1.

Attacks on Masculine Issues.

	No controls	All candidates	Male candidates	All candidates	Male candidates
	[1]	[2]	[3]	[4]	[5]
Female opponent	−0.016 (0.038)	−0.027 (0.042)	−0.036 (0.048)	−0.032 (0.042)	−0.046 (0.048)
Republican opponent	—	−0.111** (0.034)	−0.124** (0.041)	−0.112** (0.034)	−0.108** (0.041)
Female opponent × Republican opponent	—	0.022 (0.066)	0.066 (0.084)	0.022 (0.066)	0.075 (0.085)
% opposing advertising on masculine issues	—	0.120 (0.075)	0.222* (0.090)	—	—
% opposing advertising on feminine issues	—	—	—	−0.130 (0.089)	−0.099 (0.103)
2012	—	−0.188*** (0.044)	−0.174** (0.052)	−0.191*** (0.044)	−0.186** (0.053)
2014	—	−0.335*** (0.047)	−0.327*** (0.054)	−0.345*** (0.045)	−0.365*** (0.053)
2016	—	−0.202*** (0.048)	−0.184** (0.057)	−0.205*** (0.048)	−0.199** (0.058)
2018	—	−0.290*** (0.049)	−0.244*** (0.059)	−0.301*** (0.047)	−0.291*** (0.057)
Intercept	0.394*** (0.020)	0.599*** (0.052)	0.539** (0.061)	0.689*** (0.041)	0.673*** (0.048)
Adj. R2	<0.01	0.30	0.30	0.30	0.27
N	214	214	156	214	156

Note. Dependent variable in each model is the percentage of attack advertising mentioning masculine issues, scaled from 0 to 1. Standard errors in parentheses below OLS regression coefficients.

*p < .05. **p < .01. ***p < .001, two-tailed tests.

Table 2.

Attacks on Feminine Issues.

	No controls	All candidates	Male candidates	All candidates	Male candidates
	[1]	[2]	[3]	[4]	[5]
Female opponent	−0.006 (0.034)	0.017 (0.042)	0.041 (0.048)	0.011 (0.041)	0.032 (0.047)
Republican opponent	—	0.162*** (0.034)	0.174*** (0.040)	0.167** (0.033)	0.190*** (0.040)
Female opponent × Republican opponent	—	−0.020 (0.065)	−0.074 (0.084)	−0.018 (0.065)	−0.065 (0.083)
% opposing advertising on feminine issues	—	0.049 (0.087)	0.071 (0.102)	—	—
% opposing advertising on masculine issues	—	—	—	−0.136 (0.073)	−0.201* (0.088)
2012	—	0.103* (0.043)	0.058 (0.052)	0.092* (0.043)	0.045 (0.051)
2014	—	0.147** (0.044)	0.127* (0.052)	0.123** (0.045)	0.092 (0.053)
2016	—	0.076 (0.047)	0.050 (0.057)	0.063 (0.047)	0.035 (0.056)
2018	—	0.159** (0.046)	0.107 (0.057)	0.130* (0.048)	0.055 (0.058)
Intercept	0.302** (0.018)	0.112** (0.040)	0.135** (0.047)	0.192*** (0.051)	0.251*** (0.060)
Adj. R2	<0.01	0.16	0.12	0.17	0.15
N	214	214	156	214	156

Note. Dependent variable in each model is the percentage of attack advertising mentioning feminine issues, scaled from 0 to 1. Standard errors in parentheses below OLS regression coefficients.

*p < .05. **p < .01. ***p < .001, two-tailed tests.

We again control for party with a dummy variable for a Republican opponent (or Democratic attacking campaign), and we again account for the shifting salience of different issues across campaign cycles with year dummy variables. The effects of party and gender may be conditional on one another, that is, the degree to which women are targeted on feminine or masculine issues may be conditional on their party affiliation separate from the main effect of their partisanship. Thus, we also include interaction terms for the binary variables of gender and party.

Because campaigns may also alter the issues on which they attack an opponent in response to that opponent’s own decisions of which issues to emphasize (Windett, 2014), in these models we control for the opposing advertising issue content. The percent of opposing advertising on feminine or masculine issues represents the total issue mentions of either feminine or masculine issues in all advertising from the opposing campaign.

Table 1 presents the results for masculine issue mentions in attack and contrast advertising. Column [1] reports the zero-order relationship between opponent gender and the use of masculine issues in attack and contrast advertising. Columns [2] through [5] include controls for the issue content of the opposing side, control for a Republican opponent, and the election year. In Columns [2] and [4] these effects are estimated across all Senate candidates, while in Columns [3] and [5] the sample is restricted to male candidates.

The gender of the opponent is not associated with the frequency of masculine issue mentions in contrast or attack advertising about that candidate. Whether in an un-controlled comparison in Column [1], with additional controls for party, year, and the opposing campaign issue content in Columns [2] and [4], candidates are no more or less likely to attack using masculine issues whether either candidate identifies as a woman or man. Furthermore, the interaction of gender and party shows no significant association. In other words, the strong effect of party observed in these models is not conditioned by the gender of the opponent. Likewise, the overall null effect of gender on issue selection in attack advertising does not hide significant differences within one or both partisan groups.

Previous scholarship has highlighted that issue emphases could be different in male-female campaign matchups versus campaigns where both major party candidates are the same gender (Dolan & Lynch, 2017; Windett, 2014). With this in mind, we restrict the sample to the campaign advertising from male Senate candidates in Columns [3] and [5] and still find that women running in Senate general election campaigns are no more or less likely to be the subject of opposing advertising that employs masculine issues such as taxes, trade, military, and foreign policy even when the opponent is male.

In these models we see a negative association between Republican opponents and the mention of masculine issues in advertising against those candidates. This relationship holds across each of the specifications in Columns [2] through [5] where masculine issues are mentioned about 8 to 10 percentage points less in advertising that targets Republicans.

The percentage of the opponent’s advertising mentioning masculine issues is positively associated with a candidate’s mention of masculine issues in attack and contrast advertising, but only among the subsample of male candidates (Column[3]) and not in the full sample (Column [2]). Model estimates in Columns [4] and [5] show that the frequency of opponent advertising mentioning feminine issues was not associated with the frequency of attack or contrast advertising mentioning masculine issues.

Each of the year dummy variables 2012 through 2018 were strongly negative, reflecting the tremendous weight of contrast and attack advertising on masculine issues in 2010. The 2010 election was an unusual year where the Tea Party Movement crystallized and narrowed the Republican campaign platform onto a handful of issues around reducing the size of government. Shown in Appendix Table A6, Republican candidates in this Tea Party election year overwhelmingly attacked their opponents on government spending (24%), the economy (17%) and jobs (14%). By contrast, Republican candidates attacked on these issues much less 2012 through 2018, in the range of 7% to 9%.

Table 2 presents these results for attacks on feminine issues. Again Column [1] provides the zero-order relationship between the opponent gender and the use of feminine issues in attack and contrast advertising. Columns [2] through [5] include controls for the issue content of the opposing side, control for a Republican opponent, and the election year. In Columns [2] and [4] these effects are estimated across all Senate candidates, while in Columns [3] and [5] the sample is restricted to just male candidates. As is the case with masculine issues, across all specifications the gender identity of the opponent has no association with the frequency of feminine issues mentioned in advertising targeting that opponent. In addition, any association between opponent gender and the issues mentioned in advertising attacks are not conditioned by the party of the candidates. In our restricted sample of campaign advertising only from male Senate candidates in Columns [3] and [5] we again find that women running in Senate general election campaigns are no more or less likely to be the subject of opposing advertising that employs feminine issues such as healthcare, environment, welfare, or education.

In each of the models with control variables included in Columns [2] through [5] a Republican opponent is positively associated with a greater proportion of contrast and attack advertising on feminine issues. In other words, holding gender, year, and a candidate’s own proportion of advertising referencing masculine or feminine issues constant, advertising aimed against Republican candidates mentions feminine issues about 16 percentage points more frequently. Furthermore, several positive associations in 2012, 2014, and 2018 with feminine issue attacks compare to the omitted year of 2010 when feminine issues received little attention in attack and contrast advertising. Columns [2] and [3] provide coefficients from models that include control variables for the proportion of the opponent’s total advertising that mentions feminine issues. Columns [4] and [5] include controls for the opponent’s total advertising that mentions masculine issues. Just one of these coefficients reaches the p < .05 threshold of statistical significance: male Senate candidates were less likely to attack or contrast opponents on feminine issues when those opponents dedicated more of their own advertising to masculine issues.

Finally, we account for the potential effects of the increased volume of outside advertising from the national parties and independent expenditure groups in Senate campaigns. Within our sample 38% of the total attack and contrast advertising airings were produced by campaign actors other than the candidates. We estimate models interacting the percentage of attack and contrast advertising from a campaign with the female opponent dummy variable to evaluate if outside group involvement attenuates gendered issue selection strategies from campaigns. In line with other results in this study, we find no significant differences in the frequency of masculine or feminine attacks on women as the percentage of attacks from outside groups changes (see Appendix Table A5 for full results from these models).

Discussion

Overall, results suggest that campaigns consider partisan issue ownership and public issue salience when choosing issues to feature in advertising that refers to opponents. However, the gender identity of the opponent did not have measurable effects on the issues Senate campaigns raised in attack advertising in the 2010 through 2018 election cycles. Instead, Democrats were attacked more frequently than Republicans on masculine issues and Republicans were attacked more frequently than Democrats on feminine issues.

While this work focuses specifically on the issues that campaigns invoke when characterizing opponents, these findings are descriptively consistent with the most recent work on campaigns’ overall issue engagement and gender. Our findings slightly depart from studies that draw on observational data from earlier decades, but in our view this does not challenge the results of these earlier works. Instead, our findings document an evolution in campaign communications towards an issue landscape even more dominated by partisan reputations over other candidate attributes.

Do these patterns of issue selection by Senate campaigns indicate that they believe that gender does not impact issue reputations and thus is not factored into messaging decisions? Or has the connection between Republican and Democratic party images and gender stereotypes described in Winter (2010) fully extended to issue reputations? This is an important difference.

If women and men still carry issue reputations that are highly collinear but still distinct from party, this has different implications for Democratic and Republican women running for office. Given that campaigns have demonstrated a preference to attack candidates on the issues where the target candidate has a partisan weakness, women from each party face differing circumstances: Democratic women drawing attacks on Republican/masculine owned issues are doubly vulnerable by virtue of both party and gender. But Republican women drawing attacks more frequently on Democratic/feminine issues will have a slight counteracting issue reputation from gender identity that may make these attacks less effective. If, on the other hand, gendered issue reputations are fully displaced by partisan issue ownership, then gender is irrelevant to the effect of issues on candidate evaluations.

The absence of any impact of candidate gender on the topics of attacks candidates face places an even harsher light on the sluggish recruitment of women to run for legislative office in the United States. Reluctance by party elites to encourage women to run could be explained as a cold but rational choice if women contend with a more challenging issue landscape than men and the party’s singular interest is in winning the maximum number of seats. However, our findings demonstrate that the topics of attack are determined by party affiliation and the salience of topics in that election year. Identical issue landscapes for men and women of the same party in a given election year undermines arguments about the electability of women and provides evidence that more frequent nomination of female candidates for public office is not a risky choice for the parties.