What Drives U.S. Congressional Members’ Policy Attention on Twitter?

Abstract

Understanding how Members of Congress (MCs) distribute their political attention is key to a number of areas of political science research including agenda setting, framing, and issue evolution. Tweets illuminate what lawmakers are paying attention to by aggregating information from newsletters, press releases, and floor debates to provide a birds-eye view of a lawmaker’s diverse agenda. In order to leverage this data efficiently, we trained a supervised machine learning classifier to label tweets according to the Comparative Agenda Project’s Policy Codebook and used the results to examine the differential attention that policy topics receive from MCs. The classifier achieved an F1 score of 0.79 and a Cohen’s kappa with human labelers of 0.78, suggesting good performance. Using this classifier, we labeled 1,485,834 original MC tweets (Retweets were excluded) and conducted a multinomial logistic regression to understand what influenced the policy areas MCs Tweeted about. Our model reveals differences in political attention along party, chamber, and gender lines and their interactions. Our approach allows us to study MCs’ political attention in near real-time and to uncover both intra- and inter-group differences.