Abstract
Previous research has shown the importance of Donald Trump’s Twitter activity, and that of his Twitter following, in spreading his message during the primary and general election campaigns of 2015–2016. However, we know little about how the publics who followed Trump and amplified his messages took shape. We take this case as an opportunity to theorize and test questions about the assembly of what we call “attentive publics” in social media. We situate our study in the context of current discussions of audience formation, attention flow, and hybridity in the United States’ political media system. From this we derive propositions concerning how attentive publics aggregate around a particular object, in this case Trump himself, which we test using time series modeling. We also present an exploration of the possible role of automated accounts in these processes. Our results reiterate the media hybridity described by others, while emphasizing the importance of news media coverage in building social media attentive publics.
Acknowledgments
The authors wish to thank Jennifer Stromer-Galley, Ralph Schroeder, participants in the workshop on “Order and disruption in the attention economy,” at the Weizenbaum Institute, Berlin, and anonymous reviewers for their constructive comments on earlier versions of this research.
Notes
1 Supplementary Materials can be found at: https://osf.io/d8ytr/?action=download%26mode=render.
3 Botometer is described by its creators as “a machine-learning framework that extracts and analyzes … over one thousand features, spanning content and network structure, temporal activity, user profile data, and sentiment analysis” (Bessi & Ferrara, Citation2016). See details of the Botometer API at: https://botometer.iuni.iu.edu/#!/api.
4 Note that we first-difference the Trump follower count variable. This was done to induce stationarity in the series. Both the Dickey-Fuller test and the KPSS test showed strong evidence of a unit root in the series. Even the differenced follower count, however, still possessed residual auto-regression, thus we utilized Prais-Winsten regression.
5 We also tested the models with only accounts for which we obtained CAP_Universal scores, i.e., excluding suspended and protected accounts; the results are nearly identical.
Additional information
Notes on contributors
Chris Wells
Chris Wells (Ph.D., University of Washington, 2011) is an assistant professor in the Division of Emerging Media Studies and Department of Journalism at Boston University. His research interests include the formation of public opinion and civic engagement in the hybrid media system.
Yini Zhang
Yini Zhang (M.A., University of Wisconsin-Madison, 2016) is a Ph.D. candidate at the School of Journalism & Mass Communication, University of Wisconsin-Madison. Her research interests include attention economy and public opinion in the hybrid media system.
Josephine Lukito
Josephine Lukito (M.A., Syracuse University, 2015) is a Mass Communication & Journalism Ph.D. Candidate at the University of Wisconsin, Madison. She holds Ph.D. minors in English Linguistics and Political Science. Her research interests include natural language processing and sociolinguistic theory to study international political communication.
Jon C. W. Pevehouse
Jon C. W. Pevehouse (Ph.D., Ohio State University, 2000) is Vilas Distinguished Achievement Professor of Political Science at the University of Wisconsin-Madison. His research interests include the area of international relations and time series methodology.