Good and pissed: gendered emotional appeals on Twitter

ABSTRACT

Persistent gender stereotypes portray women as pleasant and polite, but in the wake of the #MeToo movement and polarized politics, female candidates are turning to Twitter and they aren’t hiding their frustration. Congressional candidates use Twitter to connect with voters, but political stalemates over health care, reproductive rights, and pay equity are the fodder for female candidates’ emotionally charged rhetoric on Twitter. Women are running and winning at rates comparable to men, but female candidates are relying on emotional appeals in distinct ways from their male counterparts. We use a dataset of tweets by candidates for the U.S. House from 2016–2020 to evaluate gender-based differences in the emotional appeals candidates make on Twitter. We find that women running for office adopt a unique style of angry emotional appeals on Twitter, as female candidates defy stereotypes by incorporating more angry rhetoric in their tweets. These differences persist after accounting for differences in party, electoral success, district competitiveness, and other potential confounds. Our research demonstrates that women seeking congressional office act differently than men in their self-presentation online, and offers insight into how anger has become central to online messaging.

KEYWORDS:

• Twitter
• emotion
• congress
• communication

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. We would like to thank Derek Willis for his help collecting the tweets from 2020.

2. The reason we treat number of tweets as the denominator and not the number of words in tweets (the default setting in LIWC), is that using the latter can overestimate the degree to which a legislator uses angry rhetoric when they issue short tweets, relative to more garrulous tweeters. For instance, suppose we have two legislators, who each post four tweets. Legislator A’s four tweets consisted of two words each, and two of those tweets feature one angry word. Legislator B’s tweets consisted of five words each, and every one of those tweets featured one angry word. If we used words in tweets as the denominator, Legislator A’s rate would be 0.25 (2 angry words divided 8 total words) and Legislator B’s rate would be 0.20 (5 angry words divided 20 words total). Thus, even though Legislator B is using angry rhetoric more often, her rate is smaller than that of Legislator A. Using total tweets as the denominator allows us to avoid this issue; with this formula, Legislator A’s rate is 0.5 (2 angry words divided by 4 tweets), while Legislator B’s rate is 1.0 (5 angry words divided by 5 tweets). We think these scores better capture frequency of angry word use.

3. The distributions of our dependent variables are heavily skewed toward zero, However, a zero-inflated negative binomial (ZINB) specification is not appropriate. This specification assumes that a zero outcome is due to two different processes, and that the researcher is ignorant as to which process is behind any or all zeroes. In our case, the two different processes for scoring “0” on our DVs could be that a legislator candidate tweeted during the collection period but did not incorporate any angry words, or it could be that she failed to tweet altogether. We, however, are aware of which process is behind each zero because we know the total amount of tweets issued by each member and exclude any candidate who did not tweet at all from the analysis and include the total amount of tweets issued by each account as a control variable.

4. Race ratings collected on September 9, 2016, September 7, 2018 and September 11, 2020.

Additional information

Notes on contributors

Annelise Russell

Annelise Russell is assistant professor of public policy at the University of Kentucky. She can be reached at [email protected].

Heather Evans

Heather K. Evans is the John Morton Beaty Professor of Politics at the University of Virginia’s College at Wise. She can be reached at [email protected].

Bryan Gervais

Bryan T. Gervais is associate professor in political science and geography at the University of Texas at San Antonio. He can be reached at [email protected].