Debating debate: Measuring discursive overlap on the congressional floor.

摘要

The study of how elites communicate to each other is an understudied topic largely because we lack a viable, large-scale, measure of discursive overlap. Discursive overlap is the extent to which parties and partisans talk to and past each other. In this paper, I introduce a repurposed measure - cosine similarity scores - and a method of measurement that concisely quantifies discursive overlap. I compare this measure to two others - overlap coefficients and Wordfish scores Slapin and Proksch (2008). To compare the scores, I first examine the distribution of the scores and then compare how well each does in a series of tests, including how well each reflects reality and how well each responds to different aspects of communication that increase or decrease discursive overlap. Throughout the paper, I use the 2008 Farm Bill as an ongoing case. I conclude that cosine similarity scores do indeed capture discursive overlap and show that it is the best measure among the three considered.