Measuring Partisanship and Representation in Online Congressional Communications

Abstract

Social media and the internet have created new ways for representatives to communicate. How have members of Congress responded to these opportunities? We introduce a multi-platform dataset of congressional communications extending back to the onset of the social media era. Using computational language processing, we classify approximately 4.7 million tweets, 2.4 million Facebook posts, and 184,000 email newsletters authored by members of Congress between 2009 and 2022 based on intended purpose, and scale the partisanship of each message along a continuous left–right dimension. After validation, we demonstrate how our data can be used to study partisanship and representation in the contemporary Congress. Importantly, our data show congressional rhetoric has become more partisan and negative as social media usage has increased. We identify one potential mechanism contributing to this trend: partisanship and negativity receive inflated levels of positive engagement on social media relative to other forms of messaging like credit claiming or constituency service.