How Social Media Drove the 2016 US Presidential Election: A Longitudinal Topic and Platform Analysis

Abstract

To what extent did external events and news versus the candidates’ own actions drive the 2016 election outcome? And were candidates misled if they focused on traditional market research versus the newer probabilistic polls? Based on the dynamic political will formation framework, the authors address these questions with a national daily data set combining polling, donations and TV advertising data with social media interactions to all candidates’ posts of the two candidates on Twitter, Facebook, and Instagram. Persistence modeling reveals that donations followed rather than drove the candidates’ polls. The probabilistic polls show a different impact of candidate ads and statements, news coverage and fake news than do the traditional polls. TV ads on the economy or gun control, and on terror threats were most effective for respectively Hillary Clinton and Donald Trump. Topics matter, as fake news about a candidate hurtsher chances on one topic, but benefits her on another topic. Moreover, platforms matter: Clinton’s chances benefited from promoting women issues on Instagram, but declined from doing so on Twitter. Her moral language on Fairness Vice and social media users’ on Authority Virtue made voters less likely to vote for her, but more likely to share fake news about her and to talk positively about Trump. While news coverage had minimal impact, fake news on Clinton’s emails, shared on her Facebook page, greatly damaged her election chances. This fake news impact was most pronounced for seniors, Hispanics and high earners – demographics who moved towards Trump in the last weeks before the election. The authors draw lessons from the past election to advise where, when and how to drive the political conversation.

Keywords: politics, marketing, social media, time series, probabilistic polls, election

JEL Classification: C32, H50, M30, M31, M37, M38