Original story by Jeremy Gelman and Steven Wilson for The Washington Post.
And while language used in tweets about Kavanaugh was less profane on average than social media in general, it was also disturbingly gendered, especially when directed at female senators.
How we conducted our research
To assess the social media response to the Kavanaugh hearing and confirmation, we collected every geocoded tweet sent in the United States from Sept. 14, the day Ronan Farrow and Jane Mayer at The New Yorker first reported an allegation of sexual assault, through Oct. 6, the day he was confirmed. Geocoded tweets are tweets that have latitude and longitude coordinates attached by a GPS-enabled device such as a smartphone. Over that three-week period, we collected 31.5 million U.S. geocoded tweets.
In addition, we collected every tweet — geocoded or not — that mentioned the word “Kavanaugh” during this period, for an additional 24.4 million tweets. These are useful for establishing whether the geocoded tweets are roughly representative of the overall social media environment. In general, we found that the non-geocoded tweets were roughly comparable, with similar proportions of individuals talking about Kavanaugh, tweeting at their senators and using profane language.
While geocoded tweets only represent about 2 percent of all tweets, they are useful for establishing comparative geographic patterns between states, because the ways in which geocoded tweets are nonrepresentative should not be different from state to state. By comparing geocoded tweets from one state to those from another state, we can draw conclusions about how relatively different the general conversation on Twitter is in those states.
The intensity of the Kavanaugh debate
Within the geocoded data, we searched each tweet for the word “Kavanaugh,” including hashtags, to measure the proportion of U.S. twitter activity related to his confirmation process. The results reveal this nomination received a staggering level of attention from Americans. While active U.S. Twitter users are on average younger, wealthier and more educated, they still make up nearly a quarter of the American population, and the bulk of trending conversation tends toward popular culture. However, over this three-week period, 1 in every 79 geocoded tweets in the United States mentioned Kavanaugh by name. On Sept. 27, the day of the Judiciary hearing, the proportion increased to 1 in 22.
This proportion of tweets overshadows every other topic, political or not, during this time. For instance, only 1 in 6,100 tweets mentioned “NAFTA” on Oct. 1, the date on which the administration announced it reached an agreement with Canada and Mexico. Over these three weeks, the National Football League was mentioned in only 1 in 183 tweets, less than half the rate of Kavanaugh.
While the Kavanaugh nomination dominated discourse, its intensity varied by state. In the figure below, we show the percentage of tweets each day that mentioned Kavanaugh from Sept. 14 through his confirmation, highlighting four critical states. Mainers’ interest in the critical role of Sen. Susan Collins (R) shows how they consistently tweeted about Kavanaugh at a higher rate than other Americans. On the date of the Judiciary Committee hearing, over 7 percent of all geocoded tweets in Maine mentioned Kavanaugh.
The map below shows the regional concentration of this intensity. The map renders the proportion of tweets discussing Kavanaugh in each state on Sept. 27. Darker green indicates higher proportions.
A spike in gendered profanity
To examine the kinds of conversations Americans were having about the confirmation, we ran all the Kavanaugh-specific tweets through a model that assesses the dominant dimension of conversation. We expected the main dimension to be partisan or ideological language.
We were wrong. What distinguished how people wrote about the Kavanaugh nomination was where they fell on the anger scale. On one end of the spectrum, people tweeted angrily, using hashtags, slurs and pejoratives. For instance, the anger in the language of those using the #stopKavanaugh hashtag was statistically almost identical to that of the #confirmKavanaugh hashtag. Those at the other end of the anger spectrum wrote in full sentences, expressing opinions or trying to reason with commenters.
To characterize tweeters’ level of anger, we searched our Twitter database for uses of any profanity in George Carlin’s once-famous list of “Seven Words You Can Never Say on Television” and cross-referenced that with our other data. By this measure, 5 percent of all U.S.-based tweets are profane; only 3.5 percent of tweets that mentioned Kavanaugh were; and only 2.1 percent of tweets directed at senators about Kavanaugh were.
However, Carlin’s list includes both gendered (terms specifically targeting women) and non-gendered profanity — and so misses just how gendered this anger was. While 17.5 percent of all profane tweets use gendered profanity, 30 percent of those directed at Collins were gendered. However, most of those were from out of state; only 16 percent of those originating within her home state of Maine were.
Key takeaways
While several prominent events have dominated headlines since those hearings, both Democrats and Republicans have claimed the Kavanaugh hearings roused their voters and motivated them to vote this Tuesday in the midterm elections. While we don’t yet know how it might affect election results, anger certainly permeated both sides of this partisan fight. Political scientists often worry that when one side of the political spectrum is not speaking to the other side, that side will only hear from like-minded partisans — while the public at large is not paying attention.
The Kavanaugh nomination defied these trends, rousing sustained public attention throughout the country. Many Americans were willing to engage in public debate and to contact their senators about this topic. While the scale of the public’s attention was staggering, and troublingly gendered anger emerged, what may be especially surprising is that the conversation about Kavanaugh was actually less profane than the average social media conversation in America.
Jeremy Gelman is assistant professor of political science at the University of Nevada at Reno.
Steven Wilson is assistant professor of political science at the University of Nevada at Reno and a project manager for the Varieties of Democracy (V-Dem) Institute.
