Moral Characters: Social Media and Elections in the U.S.

Abstract

This paper examines how social media, and the moral language used by political leaders on these platforms, affect electoral outcomes. Using a new dataset of around 13 million U.S. House congressional candidates’ tweets, we begin our analysis by describing how candidates’ social media use emerged and evolved over the last fifteen years. We use text analysis and data on social media engagement to explore the supply and demand of moral rhetoric on Twitter. We show that moral language on Twitter positively and strongly correlates with social media engagement and electoral returns. Then, using a regression discontinuity design (RDD) on primary elections, we show that social media use by politicians increases vote shares. Heterogeneity analysis provides evidence that the effect is driven by candidates who tend to use moral language in their posts