Title: Gaining Insights on U.S. Senate Speeches Using a Time Varying Text Based Ideal Point Model

Abstract: Estimating political positions of lawmakers has a long tradition in political science. We present the time varying text based ideal point model to study the political positions of lawmakers based on text data. In addition to identifying political positions, our model also provides insights into topical contents and their change over time. We use our model to analyze speeches given in the U.S. Senate between 1981 and 2017 and demonstrate how the results allow to conclude that partisanship between Republicans and Democrats increased in recent years. Further we investigate the political positions of speakers over time as well as at a specific point in time to identify speakers which are positioned at the extremes of their political party based on their speeches. The topics extracted are inspected to assess how their term compositions differ in dependence of the political position as well as how these term compositions change over time.