Improving American Democracy through Optimization: Evaluating Fairness in Congressional and State Districting

Abstract

Gerrymandering has become a central issue that has dominated the headlines of contemporary American politics, contributing to declining trust in the fairness of our democracy. Although the United States maintains a robust democratic system, recent years have intensified perceptions that representation is unfair and, at times, systematically distorted. The 2016 election of President Donald J. Trump was particularly controversial, as President Trump won the electoral vote over Hillary Clinton despite losing by almost 3 million votes. At the legislative level, several states such as Texas, California, Illinois and North Carolina, have implemented state house and congressional district maps where the distribution of votes does not align with proportional representation. This thesis will explore how issues surrounding gerrymandering and fairness ambiguity metrics can be improved using computational methods. The exploration will demonstrate that an optimization-based framework grounded in the concept of a state-specific feasibility frontier is the best tool to evaluate fairness in our democracy. The analysis below demonstrated that fairness should not be treated as a single uniform scalar score that can be used to compare all types of districts in different states. Instead, the idea of fairness is state-dependent and is bounded by a state's inherent characteristics. While early stages of my work looked to incorporate partisan data, the results of my analysis suggest that geometric fairness is a far better starting ground. This distinction motivates a more nuanced approach to evaluating redistricting outcomes and lays the groundwork for eventually incorporating partisan data as tools become more sophisticated.