Past studies show that news media delegitimizes women candidates in multiple ways, including more frequently referring women candidates by the first name (Uscinski and Goren 2011) and overly focusing on their appearance (Cummings and Terrion 2021). Meanwhile, qualitative evidence describes women politicians in the US as having a “husbands, hair, hemline problem” (Duerst-Lahti 2005), though to our knowledge there has been no large-scale quantitative study documenting whether either news media or voters overly focus on women politicians’ personal lives or appearance. This project seeks to explore how commentary on women politician’s clothing and appearance dominates online discussions of these candidates. We focus specifically on the 2020 Democratic presidential primary due to the historically large number of women running in that race. Our dataset includes over 25,000 tweets mentioning these candidates and we use novel AI With the increasing number of female politicians running for office, studies have shown that news media often delegitimizes these candidates through their rhetoric in different ways than their male counterparts. By overly focusing on appearance and demeanor (Cummings and Terrion 2021), as well as referring to females by their first name (Uscinski and Goren 2011), qualitative evidence describes women politicians in the US as having a “husbands, hair, hemline problem” (Duerst-Lahti 2005) as another battle in order to break the glass ceiling in politics. This project seeks to explore how commentary on women politician’s clothing and appearance dominates online discussions of these candidates. Focusing specifically on the 2020 Democratic presidential primary, this dataset includes over 25,000 tweets mentioning these candidates. Filtering by a novel dictionary of terms, these tweets are then compared to determine the nature and number of tweets referring to a candidate’s appearance and fashion, based on gender. Primarily focusing on mass public tweets, I aim to show how the voters are also talking about candidates and explore further direction for how this might impact women in political positions.

Previous research shows that voters are likely to engage in various forms of gender stereotyping when female candidates are on the ballot. For example, voters perceive women candidates to be less qualified than male candidates with similar qualifications (Paul and Smith 2008) while simultaneously holding women candidates much higher qualification standards than their male counterparts (Bauer 2019).  Regarding the role of the news media in gendered candidate evaluations, races with at least one woman running have been found to receive more “candidate trait” media coverage than “candidate issue” coverage (Dunaway et al. 2013). Coverage of candidates in the 2008 presidential election showed that Hillary Clinton and Sarah Palin were both referred to by gender labels more often than their male counterparts (Meeks 2013), and that Clinton was referred to more often by her first name than Barack Obama, especially by male news anchors (Uscinski and Goren 2011). 

When addressing the presumed focus on candidate clothing and appearance for women candidates, qualitative studies based on historical evidence and interviews with candidates show that journalists focus more on appearance of women candidates than male counterparts. This involves focus on clothing as well as hairstyle and physical demeanor, like whether a candidate seems tired. Some women politicians have even said in interviews that they strategically select clothing with the purpose of “blend in” (Braden 1996). More generally, coverage of women candidates may focus on whether said candidate appears “unladylike” (Braden 1996) or “too masculine” (Cummings and Terrion 2021; Messner 2007). As McGingley (2009) notes, women candidates may experience difficulties “performing gender” in a manner that is acceptable to either voters or the news media. She uses the example of Hillary Clinton in 2008 being criticized for seeming too masculine and not traditionally attractive, while Sarah Palin was seen as more traditionally feminine but criticized for seeming vain and for supposedly exorbitant clothing expenses. Notably, the challenges of “performing gender” are not solely confined to women, as Messner (2007) notes that men running for office feel pressure to appear sufficiently masculine to voters, highlighting that gendered expectations also apply to male candidates.

Overall, women candidates may have been treated differently in past election cycles due to the “novelty” aspect of women running for office (Duerst-Lahti 2006; Holt 2012). However, it is possible that as more women run for office, especially looking at the crowded 2020 Democratic presidential primary field (Hora 2019), journalists have become less preoccupied with the “husband, hair, hemline” trifecta. Notably, the extent to which discourse surrounding women candidates has changed is difficult to determine in the absence of widespread quantitative analyses on the subject. 

As Wallsten (2014) notes, Twitter users are not representative of the American public at large. As of 2019, Twitter users are on average younger, more liberal, and more educated than the average American (Wojcik and Hughes 2019). Additionally, most users on Twitter are not content creators, and many users who do post are “retweeting” rather than producing original content (Yaqub et al. 2017). However, O’Connor et al. (2010) argue that is some cases Twitter sentiment is reflective of public opinion polls, using President Obama’s polling numbers as an example. Additionally, journalists have increasingly used Twitter to find and connect with potential sources (Broersma and Graham 2012) and monitor public discussion on the platform for news stories (Wallsten 2014), demonstrating the potential for social media platforms such as Twitter to influence discourse on mainstream news outlets. Furthermore, social media platforms such as Twitter have given average users the opportunity to become “agenda-setters,” though topic analysis shows that with many political topics it is political elites that have the most influence in agenda-setting (Xu et al. 2013).

Prior to 2016 there was little opportunity to analyze discourse of candidates based on gender on social media. Though analysis of traditional media shows that news anchors referred to Hillary Clinton as a presidential candidate more by her first name (Uscinski and Goren 2011) and were more likely to use gender labels to describe her (Meeks 2013), we do not know whether this was also true of social media content creators. In one of the few large-scale social media studies that involve sentiment analysis on social media of a woman candidate, Yaqub et al. (2017) find that tweets mentioning Clinton were more negative than tweets mentioning Trump. However, this may have reflected the candidates’ own tweets, as Clinton’s tweets were overall more negative in tone than Trump’s. Therefore, it is difficult to determine if tweets referencing Clinton were more negative in tone due to her gender or other factors.

Tweets for this analysis were collected from the “#Election2020” dataset, which includes over 600 million “U.S. politics -and election-related tweets” from December 2019 to June 2021 (Chen et al. 2020). For purposes of this study, I have only included tweets from dates of Democratic presidential primary debates: December 19, 2019, January 14, 2020, February 7, 2020, February 19, 2020, February 25, 2020, and March 15, 2020. Retweets and duplicate tweets were removed from the dataset. 

To analyze the occurrences and likelihood of my dictionary terms along gender dimensions, I run a logistic regression model for the different terms in the two dictionaries. Because I want to determine how being a female affects the likelihood of occurrence of terms, my dependent variable is “female” (coded as a dichotomous variable) and my independent variables are the various dictionary terms.

Potential discrepancies in clothing/appearance dialogue may impact voters’ perceptions of candidates and whether voters believe these candidates for whom coverage is disproportionately focused on appearance to be well-suited for office.  Additionally, such dialogue may potentially affect support for women candidates via framing affects. Framing refers to subjects forming an opinion about a subject based on how it is presented. Framing does not necessarily provide subjects with new information, but may instead “guide” consumers of media content to think about what they already know in a way that emphasizes certain considerations over others (Chong and Druckman 2007). Therefore, frequent explicit mentions of clothing and appearance may lead to voters fixating on these factors as well.

Eventually, my goal is to conduct a similar analysis using cable news transcripts rather than just election-related tweets. As Suhay, Grofman, and Treschel (2020) note, news coverage affects voter support of candidates, even if unintentionally, when journalists covertly compliment or criticize a candidate. Therefore, if it is found that cable news commentators are priming viewers to consider women candidates’ fashion choices or other factors related to their appearance, this may delegitimize these candidates in the minds of these viewers. Additionally, focusing on what women candidates are wearing may prime viewers to consider candidate gender while focusing on what male candidates wear might have little to no effect. There is evidence that priming voters to think of certain considerations, like “electability,” results in voters being less supportive of women candidates (Bateson 2020; Masket 2020). Therefore, it is possible that drawing attention to clothing and appearance could make voters feel that women candidates are less competent or qualified for office. 

I have conducted preliminary data gathering using Proquest with the following search:

noft("Elizabeth Warren" OR "Warren" OR "Amy Klobuchar" OR "Klobuchar" OR "Kamala Harris" OR "Harris" OR "Kamala" OR "Kristen Gillibrand" OR "Gillibrand" OR "Marianne Williamson" OR "Williamson" OR "Tulsi Gabbard" OR "Tulsi" OR "Gabbard" OR "Andrew Yang" OR "Yang" OR "Bernie Sanders" OR "Bernie" OR "Sanders" OR "Beto O’Rourke" OR "Beto" OR "O’Rourke" OR "Cory Booker" OR "Booker" OR "Joe Biden" OR "Biden" OR "Joe" OR "Michael Bloomberg" OR "Bloomberg" OR "Pete Buttigieg" OR "Pete" OR "Buttigieg" OR "Bill de Blasio" OR "Deval Patrick" OR "Eric Swalwell" OR "Jay Inslee" OR "Joe Sestak" OR "John Delaney" OR "John Hickenlooper" OR "Julian Castro" OR "Michael Bennet" OR "Mike Gravel" OR "Seth Moulton" OR "Steve Bullock" OR "Tim Ryan" OR "Tom Steyer" OR "Wayne Messam") AND at.exact("Transcript") AND publication.exact("CNN Newsroom" OR "New Day" OR "Early Start" OR "Inside Politics (CNN)" OR "Mornings with Maria" OR "The Lead With Jake Tapper" OR "@This Hour" OR "The Situation Room (CNN)" OR "CNN Tonight" OR "NBC Nightly News" OR "Hannity" OR "Fox News@Night" OR "Hardball w / Chris Matthews" OR "CNN Special / Live Event" OR "The 11th Hour with Brian Williams" OR "Special Report (FOX)" OR "Cuomo Prime Time" OR "Tucker Carlson Tonight" OR "Anderson Cooper 360" OR "All In with Chris Hayes" OR "The Rachel Maddow Show" OR "Lou Dobbs Tonight" OR "Fox News Sunday" OR "State of the Union with Jake Tapper")

This has resulted in 4,000+ transcripts using the same date range as the tweets.

Additionally, I would like to conduct interviews with female candidates and politicians to discuss their feelings on how they are portrayed in the media and online. I would like to see if that has any determinance on how they approach their campaigns and if it is somehthing they consider when running for office.