The Impact of Constituent Unemployment on the Ideology of Congressional Representatives

Abstract

Unemployment is perhaps one of the most tangible aspects of an economy for the average citizen. It upends their lives and is often almost completely out of their control, leading to frustration and increased demands for protection from one’s government. This thesis contends that this increased demand on government manifests itself in changes in political ideology, using evidence from both cross-sectional analysis for the years 2020, 2022, and 2024, along with time series analysis for the years 1997-2014. To measure ideology, this project debuts a novel ideology index created from the speeches of the members of the US House of Representatives using Natural Language Processing (NLP) techniques. When using the more traditional political mapping of the US House – the DW-NOMINATE score system (Poole & Rosenthal, 1985) – results showed that districts with higher rates of unemployment were associated more with Democrats than Republicans, although the directionality of this effect is unclear. These findings motivate the creation of a new, more dynamic measure of political ideology in congress using means unavailable to Poole & Rosenthal. Using this measure in a time series analysis with in-district fixed effects yields findings that suggest members of congress change the composition and content of their speech as unemployment increases, all the while getting more polarized along party lines. This thesis, with its innovative methods and question, fill unexplored gaps in the literatures of computational political science and economics, while also introducing exciting prospects for future research using similar NLP methods.