A House’s Speech Divided: Novel Applications Of Text-As-Data For The Study Of Elite Polarization In The U.s. House Of Representatives (1983-2016)

Current models of elite polarization imply that the behaviors and ideologies of Democrats and Republicans have become increasingly distinct. The congressional roll-call voting record is the most relied-on indicator of congressional polarization, however, voting behavior is limited in its scope, ability to provide deeper insights into the nature of elite polarization, and can be affected by external non-ideological factors. This dissertation leverages the richness of the congressional record and introduces a flexible computational method, the dynamic topic model, to study three unique but related indicators of political polarization across three decades of debate from the floor of the House of Representatives (1983-2016). Using the output of the dynamic topic mode – and through the lens of political communication – this dissertation reveals patterns of increasing polarization in not only what Democrats and Republicans talk about, but also how political issues are discussed. Furthermore, this dissertation interrogates elite ideologies through belief network analysis and finds that the networks of political beliefs held by Democrats and Republicans have not significantly diverged since 1983. This dissertation introduces a novel approach to the study of political polarization in Congress and provides three applied use-cases for studying political polarization through text-as-data and relevant quantities to political communication.