Gauging Gerrymandering in Pennsylvania

A Monte Carlo Approach Using Methods from Spatial Statistics

Abstract

There is currently no widely accepted standard method to determine whether gerrymandering has occurred. To determine a cutoff for unreasonable gerrymandering, simulating collections of districting plans in the absence of partisan bias has been proposed. In simulation-based methods, real-world election outcomes are compared to results from simulated districting plans. Here, a simulation method that creates possible districts in continuous space is proposed. Existing methods use preliminary spatial discretization of the state to perform simulations. This spatial discretization can result in biased estimates, which could lead to inaccurate conclusions regarding gerrymandering. We use our continuous-space method to analyze the political districts in Pennsylvania. All of our simulated elections result in fewer than 13 Republican seats, indicating that the districting plan used in Pennsylvania prior to 2018 was likely gerrymandered. This finding agrees with and confirms the results of simulation-based discrete-space gerrymandering studies without the presence of discretization bias.