ABSTRACT
Women running for Congress make different choices from men about how to connect with constituents on social media, and the increasing number of women running for Congress from both parties suggests that further assessment of the gendered patterns of emotional appeals is needed. We use this opportunity to assess the joint influence of gender and partisanship on patterns of emotional appeals, showing how party moderates the distinct appeals women candidates make on social media. We use a dictionary-based computational approach to catalog congressional candidates’ emotional rhetoric on Twitter during the 2020 election year, finding Republican women use more joyful appeals and fewer angry appeals compared to both Republican men and Democratic women, suggesting a gap in emotive appeals and differing expectations for how women communicate that varies with party. Our results underscore the importance of accounting for relative partisanship in developing a more nuanced explanation of how and when women adopt stereotypical styles of campaign communication as the number of Republican women running for Congress continues to increase.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Supplemental data
Supplemental data for this article can be accessed on the publisher’s website at https://doi.org/10.1080/1554477X.2023.2194232
Notes
1. An exception being Gervais and Morris (2018).
2. The NRC Word-Emotion Association Lexicon (Mohammad and Turney 2013) is an open-access resource available at: https://saifmohammad.com/WebPages/NRC-Emotion-Lexicon.htm. It contains 13,901 unigrams across 10 categories: two broad sentiment categories (negative and positive) with four discrete emotive categories contained in each.
3. Specifically drawn from 1) the Macquarie Thesaurus (which includes phrases); 2) the Ekman subset of the WordNet Affect Lexicon; and 3) all terms in the General Inquirer. See Mohammad and Turney (2013) for further detail.
4. Hua and Macdonald’s (2020) revised negative sentiment dictionary (which contains about 20% less unigrams) is found to accurately classify 76.5% of messages with negative language in comparison to hand-coding, improving upon the original dictionary’s performance accuracy metric by 5%.
5. We estimate alternative models aggregated at the Twitter account level and find substantively similar results (see Appendix ).
6. There are no significant gender differences found in candidates’ use of trust appeals, however, which we aim to investigate further in the next iteration of this research.
Additional information
Notes on contributors
Annelise Russell
Annelise Russell, PhD is an Assistant Professor of Public Policy at the University of Kentucky. She is also a faculty associate of the US Policy Agendas Project and a member of the Comparative Agendas Project. Dr. Russell’s research interests include questions about how policymakers communicate their agendas and the role of the media, particularly social media, in the political process. Much of her research is on congressional decision-making and communication, including an active research agenda in the intersection of social media and political institutions.
Maggie Macdonald
Maggie Macdonald an incoming Assistant Professor at the University of Kentucky Fall 2023. She received her PhD in political science from Emory University in 2020. Her research agenda centers around how American political elites, such as candidates for office or interest groups, publicly communicate, with a focus on their use of new technologies like social media.
Whitney Hua
Whitney Hua is currently the director of applied data and science at the Center for Election Science. She holds a PhD in political science from USC and a bachelor’s degree in Political Science from the University of California, Irvine. Her research focuses on political communication and behavior with an emphasis on Congress, social media, race and gender politics, and computational methods.