In recent years, increasing political polarization has been the subject of much discussion. There are several ways to measure aspects of partisanship, including surveys of voter attitudes, the voting behavior of politicians, and measures of political speech based on new text-analytic, machine-learning techniques. This study examines the trends in these distinct measures and assesses the likely causes of changes in partisanship in the US Congress using econometric techniques such as interrupted time-series designs and structural-break analyses.