Jonathan Mattingly: Using Computational Sampling to Quantify Gerrymandering

Submitted by Tony I Garcia on

The Department of Applied Mathematics is pleased to host this series of colloquium lectures, funded in part by a generous gift from the Boeing Company. This series will bring to campus prominent applied mathematicians from around the world.

Name: Jonathan Mattingly (Duke University)

Date: March 10th, 2022

Location: (MLR 301) 

Using Computational Sampling to Quantify Gerrymandering

Abstract: The US political system chooses representatives to localized geographic districts. Every 10 years, the census counts the population, requiring new districts. Gerrymandering is the harnessing of this administrative process for partisan political gain. With the  release of the new Census numbers in the Fall of 2021, redistricting has again moved  into the national conspicuousness. Society is confronted with the need to create/evaluate distracting maps. Can we recognize gerrymandering when we see it? Is proportionality relevant? What is fair? How does the geopolitical geometry inform these answers?  What is the effect of incumbency protection or the Voting Rights Act or more generally the preservation of communities of interest?

Since 2013, my group has developed methods using Monte Carlo sampling to reveal the structure of the map between votes and political outcomes under typical distracting. I have testified in Common Cause v. Rucho and Common Cause v. Lewis which resulted in the redrawing of the NC State Legislative and Congressional map for the 2020 elections. More recently, I testified in the NC State case (Harper and Common Cause v Hall) which led to all of the new maps being declared unconstitutional.

This is a story of the interaction between lawyers, mathematicians, and policy advocates; each group pushing the other. The problem of understanding gerrymandering has also prompted the development of several new computational algorithms which come with new mathematical questions. Our most recent work mixes multiscale graph algorithms with ideas from parallel tempering to produce robust samples from high-dimensional distribution dictated largely by non-partisan policy concerns.

Youtube: Watch the talk online here