— Clinton laughs in response to a question about whether she was alone all night on the night of the Benghazi attacks. (@WordsofSarah)
On Oct. 23, 2015, at the Benghazi hearings, Hillary Clinton laughs in response to a question about whether she was alone all night on the night of the Benghazi attacks. (@WordsofSarah)

During an interview at the Sundance Film Festival last week, actress and producer Lena Dunham described the media coverage of Hillary Clinton’s candidacy as “gendered and rabidly sexist.”

Dunham talked about her frustration with adjectives applied only to women candidates, offering a partial list of words she wished the media would stop using when describing women in general, and Clinton in particular.

In response, conservative radio talk show host Doc Thompson called on his listeners to tweet words that should not be used in reference to Clinton, with the hashtag #WordsThatDontDescribeHillary. Thompson kicked things off with some suggestions, including “cankle-less” and “honest.” Soon the hashtag was trending on Twitter, with thousands of users listing adjectives they wouldn’t apply to Hillary Clinton.

We analyzed more than 12,000 tweets containing the hashtag. While this large sample provides a good glimpse into the Twittersphere’s criticism of Hillary, it is not a representative sampling of American voters. Only about one-fifth of online Americans use Twitter at all, and anyone willing to go on Twitter to publicly discuss a political issue probably feels unusually strongly about it.

The purpose of studying Twitter conversations, then, is not to conduct representative polls but to better understand public discourse.

Here were the most adjectives people most often said did not describe Clinton:


Clinton is most commonly criticized as dishonest or untrustworthy. This parallels public opinion polls, which show that 60 percent of voters do not believe Clinton is trustworthy. To find the other most common classes of criticisms, we divided adjectives into groups with similar meanings. (You can find our adjective groupings, and all statistical results underlying this analysis, on GitHub.) We show the results in the graph below.


In addition to being criticized as untrustworthy, Clinton was frequently criticized as not likable or human; not liberal; not patriotic; and unable to relate to the middle class.

Women in politics and the double bind

The prevalence of adjectives that question Clinton’s trustworthiness and likability echoes the academic work on women in politics. Kathleen Hall Jamieson called this phenomenon the “Double Bind“: women must demonstrate ‘toughness’ and other traditionally masculine traits in order to show they are fit for office — and yet if they appear too tough they risk violating gender norms and in turn losing the trust of voters.

For male candidates, the congruence between the demands of office and their socially expected behavior makes satisfying demands of both authenticity and strength much more straightforward.

Other researchers, such as Erica Seifert, argue that authenticity has become an especially strong determinant of voting in presidential races in the past few decades, making the challenge for women in presidential races all the more salient.

Jannell Ross at The Washington Post noted that #WordsThatDontDescribeHillary contained a concerning plethora of adjectives about Clinton’s appearance and femininity, arguing “Clinton’s appearance and whether or not she meets a certain set of cultural standards of appropriate or ideal behavior and appearance for women remains top of mind for some American voters.”

We found that 3 percent of tweets referred to Clinton’s appearance in a non-sexual way (“#WordsThatDontDescribeHillary cutie.”); 2 percent referred to her appearance in a sexual way (“#WordsThatDontDescribeHillary Pretty. Would tap. 10/10”); 2 percent mentioned Bill Clinton (“#WordsThatDontDescribHillary Bill Clinton’s baby mama.”).

In total, 8 percent of tweets — nearly 1,000 — criticized Clinton for her appearance, her femininity, or her husband. On the one hand, these gendered criticisms of Clinton are far less common than criticisms of her trustworthiness. On the other hand, if one out of twelve comments you got about your job performance was sexist, you would probably find another job. These comments were also frequent enough that many tweeters noticed and complained.

What men tweeted and what women tweeted

To understand whether men and women criticized Clinton for different things, we used tweeters’ first names to identify their gender. (Using first name to infer gender is a common methodology, but it’s worth noting that it won’t work for all people — people who do not use real names, or whose gender does not fit a binary description). Thirty-nine percent of tweeters whose gender we could identify had female names. Here are the five most common tweets from tweeters with male and female names.

Most commonly retweeted by men:

  • #WordsThatDontDescribeHillary Dedicated. Honest. Trustworthy. Die Hard.
  • #WordsThatDontDescribeHillary Law abiding Important President Genuine Patriotic Married Healthy Honest Human Loving Loved Sexy
  • #WordsThatDontDescribeHillary Broke
  • #WordsThatDontDescribeHillary Principled
  • #WordsThatDontDescribeHillary: Presidential.

Most commonly retweeted by women:

  • #WordsThatDontDescribeHillary Hmm but I like Hillary…..
  • #WordsThatDontDescribeHillary Broke
  • Dear America, if you want to see misogyny in action, an avalanche of hate directed at one woman, explore #WordsThatDontDescribeHillary
  • #WordsThatDontDescribeHillary Accountable
  • Defended women Bill raped Defended Americans in Benghazi Will defend Americans as president #WordsThatDontDescribeHillary #Trump2016

The tweets most retweeted by women include one criticizing misogyny in the hashtag.

When we looked at all tweets from men and women, not just those most frequently retweeted, tweets from men were more likely to criticize Clinton’s physical appearance in a sexual way and more likely to say she was not sane or stable. Tweets from women were more likely to criticize Clinton’s likability and humility. But there were also criticisms common to both genders: both were roughly equally likely to criticize Clinton’s physical appearance in a non-sexual way, for example.

Trump supporters and Sanders supporters used the hashtag differently, predictably so

Despite its right-wing origins, the hashtag was used by both the left and right. We compared tweets from people who supported Bernie Sanders in their profile to tweets from people who supported Donald Trump, since they were the two most frequently mentioned candidates.

Aside from predictable differences (tweets from Sanders supporters were much more likely to criticize Clinton for not being liberal, while tweets from Trump supporters were much likely to criticize her for not being patriotic) tweets from Trump supporters were vastly more likely to criticize Clinton’s appearance (especially sexually), her sanity and her intelligence.

This is perhaps unsurprising from supporters of a man who has been criticized repeatedly for his sexist comments (though Sanders supporters have also been criticized for sexism). Somewhat remarkably, we were unable to find a single tweet that mentioned Bill Clinton from a tweeter who supported Bernie Sanders in their profile, though there were dozens from Trump supporters.

When we took account of a tweeter’s gender and candidate preference simultaneously, candidate preference was much more important in determining how they criticized Clinton. Candidate preference may also be especially significant in our data because of how tweets spread: people reshare tweets from their social network, and conservatives and liberals are often deeply disconnected on Twitter.

In sum, #WordsThatDontDescribeHillary demonstrates that campaign rhetoric in a forum like Twitter can be profoundly gendered, shaped by partisan affiliations but inextricably linked to the broader expectations we have for men and women.

Mary Nugent is a PhD candidate in political science at Rutgers University. Emma Pierson is a Rhodes Scholar and PhD student in computer science at Stanford who writes about statistics at Obsession with Regression