Automated Congressional Redistricting

Abstract

Every 10 years, when states are forced to redraw their congressional districts, the process is intensely partisan, and the outcome is rarely fair and democratic. In the past few decades, the growing capabilities of computers have offered the promise of objective, computerized redistricting. Unfortunately, the redistricting problem can be shown to be NP-Complete, but there are a number of heuristics that are effective. We specifically define the redistricting problem and analyze several variations of a new divide and conquer algorithm, comparing the compactness and population deviation of our new algorithm to existing algorithms and the actual districts. We offer a comparative component-based analysis that demonstrates the strengths and weaknesses of each algorithm component and the type of input. The comparative analysis shows that there are several ways to produce valid redistricting plans, but each approach has benefits and consequences. Our new algorithm produces valid results to the redistricting problem in almost every state that undergoes congressional redistricting, offering a new solution to this challenging real-world problem. In one version, the algorithm produces plans with the optimal population deviation in 42 out of 43 multi-district states, which is better than most algorithms in the literature. While compactness scores vary, this approach offers new opportunities to improve population deviation. Our output files comply with the accepted format at most government hearings and redistricting competitions, so the results would be compatible with most public participation efforts in 2020.

The Gerrymandering Jumble: Map Projections Permute Districts' Compactness Scores

“…[10]) and academic research alike (e.g. [4,12,17]) describe algorithmic approaches to redistricting which use geometric methods to generate districts with appealing shapes. However, these approaches ignore all of the social and political information which are critical to the analysis of whether a districting plan treats some group of people unfairly in some way.…”

Fair Minimum Representation Clustering

“…There is also a long line of research that looks at fairness in the context of gerrymandering [32,22,7,34,38,40]. While our notion of fairness shares some similarity with different notions of fairness in gerrymandering, the gerrymandering problem places different constraints on the construction of the clusters such as contiguity.…”

Identifying Communities of Interest Using Contact Graphs

“…Beyond the robustness checks identified by Duchin, additional systematic empirical evidence for the boundaries of communities should help improve these algorithms, especially in the context of large-scale computational exercises. Importantly, many of the map-drawing algorithms used today can be adapted to include communities of interest using the graph-centric method that will be described below in Chapter 3-for example, the Markov Chain Monte Carlo approach of Fifield et al [34], the minimum spanning tree method of Cannon et al [19] and Duchin [31], the sequential Monte Carlo approach of McCartan and Imai [61], and the divide and conquer approach of Levin and Friedler [55] all operate on graphs 11 .…”