Asymmetry in Political Geography and Compactness in Districting: a Computational Analysis of Bias

TL;DR

This study analyzes how districting biases and asymmetries in political geography affect election outcomes, revealing that partisan advantages can persist despite efforts to create more compact and community-cohesive districts.

Contribution

It provides a computational analysis of partisan asymmetries, showing how districting constraints influence bias and proposing methods to reduce gerrymandering effects.

Findings

Democrats win fewer precincts but with larger majorities.

Republican control persists even with statewide Democratic majorities.

Compactness and community constraints can influence partisan bias.

Abstract

We investigate the distribution of partisanship in a cross-section of ten diverse States to elucidate how votes translate into seats won and other metrics. Markov chain simulations taking into account partisanship distribution agree surprisingly well with a simple model covering only equal voting population-weighted distributions of precinct results containing no spatial information. We find asymmetries where Democrats win fewer precincts than Republicans but do so with large marjorities. This skew accounts for persistent Republican control of State Legislatures and Congressional seats even in some states with statewide vote majorities for Democrats. Despite overall results showing Republican advantages in many states based on mean results from simulations covering many random scenarios, the simulations yield a wide range in metrics, suggesting bias can be minimized better by selecting districting plans with low values for efficiency gap than by selecting plans with values near the means for the ensemble of random simulations. We examine constraints on county splits to achieve higher compactness and investigate policies requiring cohesiveness for communities of interest as to screen out the most obvious gerrymanders. Minimizing county splits does not necessarily reduce partisan bias, except for Pennsylvania, where limiting county splits appears to reduce bias.