Geographic Policy Evaluation in US State and Local Politics

Abstract

This dissertation investigates sources and solutions of geographic inequality in the United States through the creation and analysis of novel datasets on school assignment, legislative redistricting, and local government meetings. The study of state and local politics in the United States has long been limited in scope by a lack of readily accessible and centralized data. Here, I presents three studies that advance our understanding of subnational politics and public policy through their use of geographic, text, audio, and video data. The first study focuses on school rezoning. Policy-makers are resurfacing longstanding debates about school integration, driven in part by a series of state lawsuits and the approach of the 70th anniversary of the Brown v. Board of Education decision. Focusing on New Jersey, one of the most segregated school systems in the United States, I adapt algorithms designed for use in political redistricting to simulate school assignment boundaries under a set of commonly proposed policy solutions, including attendance zone integration, redrawing school districts, and district consolidation. I find that school segregation in New Jersey could be substantially reduced while keeping students close to home and without requiring any new school infrastructure. Further, I demonstrate that these gains are only possible if policy-makers are willing to change school district lines. In the second study, co-authored with members of the ALARM Project, we examine partisan gerrymandering in the 2020 redistricting cycle. The study applies a complete workflow for simulation-based redistricting analysis to the congressional districts of all 50 states in 2020. The simulations combine geographic and election data to draw valid plans that meet federal and state-specific redistricting requirements. We find that partisan gerrymandering is widespread in the 2020 redistricting cycle, as many states enacted more consistently partisan plans than would be expected under our non-partisan simulations. However, we also find that most of the electoral bias created by this gerrymandering cancels at the national level, resulting in two expected additional seats for Republicans. Finally, we find that this partisan gerrymandering makes elections less responsive to voter preferences, in part by reducing the number of competitive districts. The third study, co-authored with Soubhik Barari, details the collection and analysis of LocalView, a dataset of more than 130,000 videos of local government meetings and their corresponding textual and audio transcripts of local government meetings publicly uploaded to YouTube – the world’s largest public video-sharing website – from 1,012 places and 2,861 distinct governments across the United States between 2006-2022. This approach aims to address two longstanding issues in the empirical study of US local politics. First, large-scale data sources are difficult to collect. Second, data across disparate local governments are difficult to compare, due to local idiosyncracies in reporting practices. LocalView contributes to both issues by presenting a centralized and standardized method for collecting data on local policy-making from local governments across the country.