Learning From Likes: The Effect of Social Engagement Feedback on Politicians' Social Media Communications

Ryden Butler (Washington University in St. Louis)

Abstract: Considerable research indicates that politicians adapt their rhetoric in order to appeal to the preferences of their constituents. However scant research has examined this phenomenon in the context of politicians' social media communications. In this study I use time-series causal inference methods to estimate the degree to which politicians are attentive to the engagement activity of their online followers when crafting messages on social media. I use a novel dataset containing the complete history of Facebook posts authored by legislators in the 115th U.S. Congress. The data includes more than 800,000 Facebooks posts spanning the decade before July 2017. I estimate and validate a topic model to represent the set of possible issues on which legislators can comment via social media. With these topics I use multiple time series models to examine how the prevalence of issues addressed by politicians in a given period is affected by the amount of social engagment with posts on that topic in prior periods. Initial results suggest that social engagement feedback affects the stylistic evolution of politicians' social media rhetoric. These results contribute to a growing body of evidence that politicians actively cultivate a virtual constituency and speak to the increasing role of social media in mediating the relationship between politicians and the electorate.

View Poster in a New Tab