Determining an Expected House Majority Using Pattern Analysis

Since the early 1800s, state committees and legislatures have been drawing districts in order to win elections by splitting and grouping populations to promote their chances of victory, a process called gerrymandering. Little consensus can be found in past work as some studies claim that current partisan gridlock is partially related to this procedure, while other works have found that sorting and the spatial distribution of partisanship account for gridlock. This exercise seeks to measure the impact of gerrymandering by comparing the party makeup of the current U.S. House delegation to the natural partisan makeup of a state as determined by a randomized process. This project removes political bias in the redistricting process through a series of Monte Carlo simulations to randomly assign Census tracts to ad hoc districts that are within one percent population of each other in a given state. These aggregated districts and corresponding demographic data are then compared to historical House election results from the modern era to produce likely victors using a regression model. This approach determines a framework for the natural partisanship of the congressional delegation from each state. The delegation suggested by the regression model is then compared to the actual partisan makeup of each delegation, producing evidence that is used to evaluate the impact of current redistricting practices on the party makeup of Congress. This process can be used to predict the election results of potential redistricting plans, and potentially identify the use of gerrymandering in a way that can have serious implications for litigation and policymaking surrounding redistricting