Will She Win? Gendered Media Coverage of the 2020 Democratic Party Presidential Primaries

Amount of Coverage

The amount of coverage female candidates received compared with men in U.S. electoral politics tends to vary. In the 2000 governor and senate races, women received proportionately more coverage than men did (Bystrom et al., 2001) and, in the subsequent election, nearly equal coverage (Lavery, 2013). Elizabeth Dole—up until that point, the woman with the “longest and most serious” bid for nomination by a woman in a dominant party—received less newspaper coverage than her male opponents did (Heldman et al., 2005). In the subsequent presidential election, the first-ever female vice presidential candidate, Sarah Palin, received 2.3 times more newspaper coverage and seven times more blog coverage than her opponent, Joe Biden, as well as more photo coverage in general (Bode & Hennings, 2012). The greater amount of photo coverage may reflect the commercial appeal of her image, which may have been damaging to her electability. In the 2008 election, Republican presidential candidate Michelle Bachmann received less coverage than her male opponents (Bystrom & Dimitrova, 2014).

Based on the proportions of coverage reported in these studies, no hard conclusions could be drawn about the effects of gender on the amount of coverage. However, the “novelty” of the female candidacy (especially for presidential campaigns)—the more acute focus on women’s physical appearance and the particular candidate’s perceived electability—can account for the case-to-case variance in coverage.

Gendered Policy Issues

One aspect of political candidates’ media coverage that has grown more equitable over the years, but still tends to surface in electoral rhetoric, is candidate association with “masculine” and “feminine” policy issues. Huddy and Terkildsen (1993) argue that “voters’ assumptions about a candidate’s gender-linked personality traits drive expectations that women and men have different areas of issue expertise.” The issues that tend to be characterized as “feminine” include child care, assisting the poor, solving problems of the elderly, reducing the gender wage gap, abortion, education (Huddy & Terkildsen, 1993), health care, women’s rights, the environment, and drug abuse (Lavery, 2013). In contrast, foreign policy, security, and economic issues tend to be constructed as “masculine” issues (Lavery, 2013). Within this policy schema, gender-equitable coverage is measured in terms of how each candidate gets covered across these issues in comparison with their counterparts of the opposite gender.

While some extant literature demonstrates only minor discrepancies between the coverage of issues between male and female candidates, others point out significant qualitative and quantitative differences that persist at various levels of political contest. In the 2000 gubernatorial and senate races, women were more likely to be covered on the topic of health care while men were mentioned more on the topics of budget, unemployment, and immigration (Banwart et al., 2003). Lavery (2013) found that television coverage of candidates in the 2002 midterm election offered roughly equitable coverage in terms of policy. In a more localized study—a 2003 gubernatorial race between Kathleen Blanco and Bobby Jindal in Louisiana—Major and Coleman (2008) found that news coverage tended to report more positively on the candidates referencing gender-congruent issues regardless of their experience with gender-incongruent ones, which the authors assert may deepen associations of candidates to gender-corresponding issues.

As for presidential politics, Heldman et al. (2005) found that the 2004 presidential election coverage of candidate Elizabeth Dole was generally equal to that of her male opponents when it came to policy topics. In the subsequent election, Sarah Palin became the second female vice president nominee for a dominant party (after Geraldine Ferraro in 1984). Across two different studies and a variety of media platforms, researchers found that Palin received limited policy coverage compared with her Democratic opponent, Joe Biden, despite receiving more coverage in general (Bode & Hennings, 2012; Wasburn & Wasburn, 2011). Furthermore, when her policies were discussed, the focus tended to be on “women’s issues” while Biden’s were biased toward “masculine” ones. In the 2012 Senate race, each gender was discussed in relation to opposite gender issues (Bystrom & Hennings, 2013) and Michelle Bachman did not receive more “feminine” issue coverage than her male opponents in the presidential primaries the same year (Bystrom & Dimitrova, 2014).

Viability and Electability

Previous literature has drawn a distinction between viability, which refers to the likelihood of the candidate receiving the presidential nomination, and electability, which refers to the likelihood of the nominated candidate winning the general election (Abramowitz, 1989). In Abromowitz’s (1989) electability model, the perceived viability of a candidate affects the degree to which they are perceived electable by a potential voter. In a more recent model, Bendle (2014) uses electability to account for both viability and electability as both are previously defined. This model accounts for electability by referencing a candidate in relation to other opponents within their party’s primary as well as the incumbent in the opposing party.

As previously mentioned, the 2020 election is unique in that all of the Democratic primary candidates who got media coverage appealed to their ability to beat the incumbent, Donald Trump. The general assumption underlying this operationalization is that “[k]nowledge of the opposition in the general election is critical to determining electability,” and therefore, a candidate’s electability will vary depending on who the candidate is discussed in reference to. In terms of media coverage, electability of a candidate tends to be discussed in comparison with his or her opponent(s), and therefore multiple candidates are likely to be mentioned (Bode & Hennings, 2012). Coverage of this sort is conceptualized as the “horse race” frame, in which the media constructs around candidates to assess their electability and determine who deserves coverage. This frame also justifies why a candidate might not deserve coverage (Lawrence & Rose, 2011). Such coverage tends to focus on personal traits and candidate characteristics, regardless of gender (Lavery, 2013). Furthermore, such frames are typically focused on the candidates’ likelihood of winning at the expense of issue-based coverage (Bystrom & Dimitrova, 2014). More recently, research has also focused on the in-group and out-group dynamics of coverage—where they found female journalists cover women’s political qualifications (Bauer, 2022). In terms of visual coverage, recent studies find women are frequently portrayed as less emotionally composed than men. This suggests the portrayals of women candidates may shift depending on modality, thus impacting framings of electability (Gruszczynski et al., 2022), and may be further exacerbated by the growing role of visual-based social media platforms in political processes (Chang et al., 2022).

For this study, we define electability as the likelihood of the candidates’ ability to win the general election, which assumes success in the primary.

Gendered Campaign Strategy: Policy Versus Personality

Gender stereotypes are well recognized by political candidates and their strategists, driving women candidates to strategically reinforce their association with “feminine” issues or emphasizing their acumen in tackling traditionally “masculine” issues. Candidates leverage or challenge gender stereotypes by emphasizing their issue priorities (Kahn, 1993) and generally prefer speaking to issues in which their party has perceived expertise in addressing (Petrocik, 1996). Windett (2014) also suggests that women are presumably more capable of securing a strong base of support from female voters, as a result of gender affinity, and strategically benefit from gearing their appeals toward male voters. Strategically, the author argues, there is a tendency for women candidates to avoid associating themselves too closely with feminine issues early on in their campaigns in an effort to boost their perceived electability. Ultimately, male candidates benefit from issue divergence, that is, keeping gender-stereotypical issues as a focus of their correspondingly gendered candidates. Conversely, female candidates strategically benefit from issue convergence, allowing them to subvert and downplay gendered assumptions about political strengths (Windett, 2014). A candidate’s campaign strategy may also notably change over time, adapting to such factors as polling numbers, the shifting campaign strategies of their opponents, and the political sentiments of the audiences they address on a regional basis.

Gender boundaries and their political implications are routinely negotiated between the media and candidates on the campaign trail (Ritchie, 2013). Huddy and Terkildsen (1993) suggest that “voters’ assumptions about a candidate’s gender-linked personality traits drive expectations that women and men have different areas of issue expertise.” More often than not, personality and policy stance is always in tension. Newspaper coverage of presidential candidate Elizabeth Dole focused more on her personality and appearance in comparison with her male opponents (Heldman et al., 2005). Likewise, vice presidential candidate Sarah Palin received significantly more coverage on “trivial” issues (“discussions of her childhood, family, physical appearance and personality”) than her rival, Biden (Wasburn & Wasburn, 2011). In that same election, Democratic primary candidate Hillary Clinton was “constructed at the antithesis of appropriate femininity—cold, calculating, emasculating and brutal,” which helped construct a more general image of her as a “preposterous” and even “dangerous” candidate (Ritchie, 2013). This characterization of Hillary as cold-blooded and bossy is aptly characterized as early as 1998 in Hating Hillary (Campbell, 1998).

In this regard, we propose the notion of transitive policy salience which captures the tension between personality and policy—If one’s novelty and personality are sufficiently strong, they can set the agenda against their opponents, which has significant downstream implications for candidates of the opposite gender.

Research Questions

In this study, we investigate how the media covered men and women candidates in the 2020 Democratic Party Presidential Primaries. The 2020 Primaries provides a critical comparison with prior elections, as one of the most diverse democratic primaries.

After Elizabeth Warren, the last women candidate, announced the suspension of her campaign, media outlets were at a loss: How did primaries that began with the excitement of diversity end up with two White male candidates competing? Political commentators pinpointed gendered media coverage as a major reason for the loss of gender diversity. For instance, a CNN article headlined “Women were serious candidates for president. What happened?,” discussed the disproportionally high standard of electability for female candidates, saying “often, perhaps always . . . female candidates [seem] held to a higher standard when it came to being presidential, electable—and even likeable” (Hunt, 2020). Or as the New York Times asked, “Was It Always Going to Be the Last Men Standing?” citing “the narrative that somehow women are less electable than men” (Lerer, 2020).

Our principal aim is to then articulate how election discourse evolved around the gender of the candidate. This can occur either at the policy level or candidate level, specifically through direct comparisons with the incumbent. By measuring how these factors emerged in the media coverage of the 2020 Democratic Primaries, we hope to both characterize and theorize the media’s role in shaping election narratives, and potentially elections themselves.

Data Collection

Media Cloud, developed by the Berkman Klein Center at the Harvard Law School, provides a repository of all news articles published by major news sources. We used the 50 U.S. Top Newspapers collection compiled by Media Cloud based on the report published by the Pew Research Center in 2018. These 50 newspapers include sources from a wide array of political spectrum and include the New York Times, the New York Post, the Washington Post, and the Wall Street Journal. Furthermore, it comprises local newspapers such as the Seattle Times and Boston Globe, and thus was considered appropriate for our analysis that aims to study the larger media landscape of election coverage in the United States. This is to prevent the polarized media environment in the United States from effecting the analysis of gendered coverage if we were to choose only a few sources. A full list is provided in the Supplemental Appendix.

Using the Media Cloud API, we downloaded articles related to each of the Democratic Candidates. Specifically, Media Cloud exports a list of news article URLs given a list of Boolean search terms. Articles were then retrieved using the Newspaper3k package in Python 3. Articles were scraped from December 1, 2018, to March 12, 2020, the week after Super Tuesday, when Elizabeth Warren announced the suspension of her campaign. Basic statistics of the data set are shown in Figure 1.

OPEN IN VIEWER

Figure 1.

(A) Paragraphs about each candidate. (B) Articles about each candidate.

To ensure the quality of the search, we imposed a series of Boolean conditionals and keywords (such as the term “presidential”). A full list and description of the conditionals can be found in the Supplemental Appendix. For each candidate, a set of articles is produced. Each article is then tagged in multiple ways. First, the journalist and publishing news agency is recorded.

Topic Labeling

Next, we label paragraphs (or sentences) that contain the candidate and a specific policy issue. Specifically, suppose we have n policy issues. Let X be a string of text and F_i denote the filters for a specific category. The text string is either a paragraph or a sentence. As an example, the filter for electability is shown below:

F E ( X ) = { 1 i f X c o n t a i n s e l e c t i b i l i t y d i s c o u r s e 0 o t h e r w i s e

We search over 12 criteria: Trump, Clinton, electability discourse, and 9 policies—the environment, health care, national security, immigration, taxation, law enforcement, the economy, social issues, and education. The tagging process is repeated for all policies. We chose these topics in consultancy with the Comparative Agendas Project which lists 21 primary policy topics in the United States congress (Baumgartner et al., 2019). We selected the few that had the greatest salience (for instance, excluding culture) and chose keywords based on the codebook. For electability, we started with a seed list of “viable,” “electable,” and “electability,” then selected the top 10 co-occurring terms. This same approach was used for expanding the keyword set for the individual topics and is included in the Supplemental Appendix.

These tags are then stored in a vector of size 12, then timestamped by publication date. Note that some topics, such as health care, taxation, law enforcement, and education, are less likely to be isolated as paragraphs. There are likely several contributing factors to this variation, but the complexity of those topics likely compels journalists toward writing full articles rather than one-off paragraphs. Thus, we compare across both article and paragraph levels. Upon tagging each of the paragraphs, our data set consists of vectors that describe the policy and electability discourse, distributed over time. In addition to policy and electability tags, we analyzed the sentiment using word-based sentiment.

Data Aggregation, Analysis, and Pollsters

The resulting data table is thus defined on every article (or paragraph) as rows, with boolean columns for topic (policy, electability, incumbent comparisons), boolean columns for every candidate, boolean columns for gender, and timestamps. This allows us to conduct simple but highly effective cross-sectional analysis across men and women candidates, as well as focus on specific candidates. Time series were aggregated across 3-day windows. The exception is Figure 8 when comparing journalism with poll results, which use 7-day windows.

We integrated polling data from 144 national polls—such as YouGov and SSRS—sourced from fivethirtyeight.com, the polling aggregation site by Nate Silver and licensed feature of the New York Times. In our final analysis, we consider the standardized difference between the press and polls, first standardizing each time series, before taking their difference. A full list of the pollsters is included in the Supplemental Appendix.

Discrepancies Between Men and Women

We begin with RQ1, which considers how our topics vary by the gender of the candidates. Figure 2 shows the co-occurrence of policy issues, electability discourse, and individuals (Trump and Clinton) within news articles. We include Clinton to see whether women candidates were similarly compared with the prior election. Note, having confirmed the topic modeling criteria does not skew results to a particular gender, these bar graphs make direct comparisons between men and women, for each topic. As articles were scraped by Media Cloud, we can treat this as the population level.

OPEN IN VIEWER

Figure 2.

Co-occurrence of policy issues, individuals, and electability discourse with male and female candidates.

In the articles (Figure 2A), we observe most discourse topics being mentioned in relatively equal proportions. The same holds for the number of paragraphs, save co-occurrences with Trump. Male candidates are mentioned almost 4% more in conjunction with Trump, which is around 60,000 paragraphs. This is significant. Although women are mentioned in parity with men, for every article, they are under-mentioned at the paragraph level. Reuning and Dietrich (2019) contend that media coverage, defined as “the extent to which a candidate is covered in the news,” influences public interest and may act as a mode of “status conferral” for the candidates, offering voters cues as to who has the best chance of winning. However, it remains unclear the degree to which mentions of candidates may affect their public interest because of where they appear in text (i.e., in the first few paragraphs), how they are discussed in each instance, and which articles are the most read.

Note, the explicit mention of gender issues—men and women, male and female—was almost non-existent (less than 10 articles total). This may be a sign of progress toward gender equality, or in any case a marked move away from explicit mention of gender as a casually addressed aspect of a particular candidate’s viability or comparison with another candidate.

Furthermore, we conducted statistical tests to see whether or not such differences are statistically significant. As the data showed a distribution with severe right skew means, we conducted the Mann–Whitney U test instead of t tests which assume a normal distribution. The results are shown in Figure 3, which demonstrates the distribution of data for men and women for Trump. The test results are summarized in Table 1. The results show that as we predicted from Figure 2, male candidates and female candidates showed a significantly different distribution for the issues relating to Trump, Education, National Security, Economy, Environment, and Social Issues. For issues such as Health Care, Clinton, Electability, Immigration, Law Enforcement, and Taxation, there was no statistically meaningful disparity between candidate gender.

OPEN IN VIEWER

Figure 3.

Independent-samples Mann–Whitney U test results comparing the data distribution between men and women for the issue Trump.

OPEN IN VIEWER

Table 1.

Summary of Independent-Sample Mann–Whitney U Tests Comparing the Distribution of Data Among Male and Female Candidates in 12 Topic Areas.

Topic	p	Topic	p
Education	.0006*	Electability	.066
Trump	.00001*	Environment	.0005*
National security	.0003*	Immigration	.324
Health care	.229	Law enforcement	.845
Economy	.00008*	Social issues	.0005*
Clinton	.057	Taxation	.927

*

p < .05.

Before we analyze why significant distributions appeared in reference to some topics and not others, we first consider additional differences between the men and women, by conducting a time series analysis. Figure 4 shows the relative proportion of paragraphs among six select discourse topics: Trump, health care, education, the economy, immigration, and national security.

OPEN IN VIEWER

Figure 4.

Time series of discourse topics, between men and women.

Immediately, we observe a large spike in Trump-related discourse with men in September 2019. This continues throughout the primary season, peaking on Super Tuesday. This likely points to a spike in articles backfilling with explanations of how the impeachment effort developed, especially in discussions between Biden and Trump and Biden and Sanders, and results in a vertical translation within the graph. For other discourse topics, men and women candidates appear highly coupled. However, there are a few exceptions. Women have more coverage in mentioning education, health care, and the economy. In particular, the discourse on education and health care seems to be driven by women during the debates.

We can now answer RQ1. We find there is some alignment of feminine topics with the top women candidates. However, of more significant interest is co-occurrence with Trump, which is significantly overtaken by men and associated with the masculine issues around national security and immigration.

Candidates Framed in Opposition

To further delve into why certain topics emerge as more salient, we now turn our attention to RQ2: What topics (gendered issue topics, electability, and incumbent comparisons) have the highest association with individual contenders? We further separate these comparisons into the top 3 men and women candidates, showing precisely who the actors are in Figure 5. Here, the absolute number of articles is shown, with the top 3 men (Biden, Sanders, Buttigieg) and top 3 women (Warren, Harris, Klobuchar). The story becomes clear here. The co-occurrence with Trump is tied heavily with Joe Biden, including the discourse regarding National Security. Education was initially driven by Elizabeth Warren, then overtaken by Sanders. Interestingly, Kamala Harris is one of the first to focus on health care and preserves her top spot against traditional candidates. This may be attributed to her shifting position on health care (Laszewski, 2020). Naturally, we also observe a decrease in coverage after Harris drops out.

OPEN IN VIEWER

Figure 5.

Discourse topics decomposed into top individual candidates.

Of course, this data does not exist in a vacuum. To address what appears to be the most drastic shifts in discourse, we provide some event-based relationships for consideration. Looking to the “Trump” segment of Figure 2, a visible, disproportionate spike of male discourse surrounds the September to October 2019 timeframe—This is almost entirely due to a series of articles discussing Trump’s references to Joe Biden via Hunter Biden’s Ukrainian connections, such as Washington Post’s September piece titled “Trump was fixated on attacking Biden long before his call with Ukraine.” Figure 3 is similarly explicable. “National Security” topic discussion rose noticeably for candidate Biden during the impeachment hearings, likely due to addresses by President Trump alleging some misconduct and resulting national insecurity.

The disconnect between male and female candidates in this graphic reduces significantly as the impeachment hearings return to center stage and candidates’ response to those proceedings is made public. In addition, the two primary topic sectors where major spikes were driven by female candidates, “Health Care” and “Education,” centered on candidates Warren and Harris’s policy addresses on those topics. Note as well that when isolated by the candidate the role of candidates Warren and Sanders in driving the “Health Care” and “Education” topics becomes apparent, especially around the middle and then end of 2019. Both occurrences seem to be tied to instances where one or both candidates explicitly targeted Biden (Warren) or Warren and Biden (Berman, 2019) on those topics, especially the shortcomings of Biden’s proposed Affordable Care Act amendments and student lending record.

Part of the limited success of female candidates may be attributed to these discourse topics. We observe how the main discourse topics of female candidates—education and health care—are not associated with topics (national security) more salient to Trump. To make this precise, Figure 6 shows the cluster correlation maps of the discourse time series, with the colors representing the time-series correlation between discourse topics. Correlation maps are standard visualizations for understanding how closely related topics are to each other. Red indicates two topics were mentioned together; blue indicates less frequently. Clustering adds an additional layer of hierarchical information, by ensuring multiple issues that have high correlations are plotted together.

OPEN IN VIEWER

Figure 6.

Cluster map of discourse topics, by pairwise correlation.

What we observe is that health care and education have a much lower correlation with electability and Trump, indicated in blue, possibly due to their complexity and political sensitivity encouraging more dedicated coverage. Issues such as immigration, national security, and the economy have a much higher correlation. Previous research has demonstrated the former topics tend to be stereotyped as “feminine” issues and the latter as “masculine” issues (Heldman et al., 2005; Huddy & Terkildsen, 1993; Lavery, 2013). While previous literature has argued that such stereotyping can be advantageous for female candidates, who are perceived as more adept at handling these issues (Bystrom & Dimitrova, 2014; Kahn, 1992), our research suggests that these issues were marginalized and therefore negatively impacted the perceived electability of the candidates who associated with them. Furthermore, we contend that policy issues that correlate with electability and Trump will perform better. The corollary here is that women candidates disproportionally mention issues around social policy, particularly Warren and Harris.

We investigate the pairwise sentiment between candidates. To do this, we consider the networks of candidate co-occurrences. Figure 7 shows the networks of portrayals across candidates during different time frames. We first separated the articles into three time periods—before September 2019, between September 1, 2019 and February 8, 2020 (the eighth debate), and February 8, 2020 until March 12, 2020. We observe a few trends. First, the strengths of the connections shift from the minor, nonconventional candidates such as Harris, Gabbard, and Yang, to the center and newcomers, such as Bloomberg. More importantly, the Washington Post—the outlet with the most articles—disproportionately published articles that link Biden and Trump within this central period. Two likely explanations may be traced here. First, Biden has significant institutional connections with the Washington Post, a newspaper located in the national capital and deeply interested in national establishment politics, and as such comparison between the establishment frontrunner and incumbent may be anticipated. Second, looking beyond the Washington Post, the discussion between the two is significant across several of the outlets during different time phases. Some of this is likely due to Biden’s role as Obama’s heir apparent, but it is also likely that Biden and those supporting his nomination understand that comparison with Trump is the strongest possible argument for the former vice president’s nomination. Focusing on defeating Trump allows Biden to more easily deflect from talking about policy.

OPEN IN VIEWER

Figure 7.

Network evolution of candidate co-occurrence in the Washington Post.

To summarize, with the media highlighting the impeachment process, Biden not only received the most amount of media coverage, he was also at the center of the most highlighted issues during the Democratic presidential primaries, that being which candidate had the best odds of defeating Trump. Given that the media devoted significant coverage to the impeachment hearings rather than focusing on the specific policy proposals put forward by the candidates, the female candidates were in effect marginalized in the media discourse. A Washington Post opinion piece titled The No.1 reason Biden is likely to beat Trump is a good example of the dominant media framing that helped Joe Biden become the nominee. Boot (2020) explained Joe Biden as the preferred choice because “he doesn’t have a radical agenda. He isn’t a new face; he has been involved in national politics longer than the median American has been alive,” and argues that “ultimately, I suspect, the outcome can be ascribed to the simple fact that most people like “Uncle Joe.” In other words, the most mainstream and centrist candidate Joe Biden could subvert the notorious novelty of Trump’s administration, whereas women candidates associated with new novelty could not.

Sensitivity to Voter Sentiment

We have now shown that which topics—across policy, electability, and candidate comparisons—intersect most closely by gender, how these topics relate to one another, and which candidates seem to benefit the most in framing. Finally, we consider electability directly in relation to polling results, as part of RQ3. Figure 8A shows the percentage of articles that cover women and men in regard to electability, overtime. We see quite clearly that at no point were the number of articles that discussed the electability of women greater than that of men. Note, these percentages do not sum to one as candidates can be mentioned in tandem. Figure 8B shows the polling results over time and shows men were polled to be much more likely over the entire primaries.

OPEN IN VIEWER

Figure 8.

(A) Shows the proportion of press coverage for men and women, as relating to electability; (B) shows the polling results aggregated across men and women; (C) shows the standardized difference between press and polling for women with an absolute difference of 9.992; and (D) for men with an absolute difference of 3.597.

What is more useful in our context is the discrepancy between the press and polls. Figure 8C and D shows the standardized difference for women and men, respectively. A standardized difference of 0 indicates the press accurately represents polling numbers in amount of coverage; a value greater than 0 means overrepresentation and less than 0 means less representation.

We find that the electability coverage fluctuates with polling three times as much for women than men, shown in yellow. Causality is impossible to pin, but as in most social systems, these are likely feedback loops. As such, we evaluate both directions. The first possibility is that polling results impact electability discourse, specifically by reinscribing “pragmatic biases” toward women candidates. Pragmatic bias occurs when “when people withhold support for a member of a group because they see success to be difficult or impossible for members of that group to achieve” (Corbett et al., 2022). This bias qualitatively differs from the personal biases, toward women as a group, that affect more abstract notions of ability to address issues rather than their sheer ability to win in political contests. In sum, the quantity of electability discourse for women candidates may fluctuate significantly in either direction based on their electability as indicated by poll numbers. This can arise from both journalistic coverage or the candidate’s campaign themselves. Inversely, the amount of electability discourse in the media may shape poll numbers. In this regard, there is wide literature showing how news coverage of campaigns influences voters one way or another, broadly and specific to gender (Druckman, 2005).

In either case, we show that either news coverage for women appear to be more reactive than men to polling numbers, or voters respond stronger to how women appear in the press. Whether the change in news coverage is driven by the candidates themselves or the press remains an open question. What we can infer is that, because of the masculine centric topics set by Trump, women candidates were constantly on a backfoot adjusting their communication and strategy.