A Novel Method for Showing Racially Polarized Voting: Bayesian Improved Surname Geocoding

Abstract

In Thornburg v. Gingles, the Supreme Court provided the elemental test for vote dilution claims under § 2 of the Voting Rights Act. In part, § 2 requires Plaintiffs to prove that voting patterns within the challenged jurisdiction are polarized by race. Because most states do not track the race of voters, social scientists developed statistical methods to make the evidentiary showing required in Gingles. These methods are decades old and are often the subject of intense scrutiny in vote dilution trials. In some cases, the size of the jurisdiction and the quality of the voter file and voting records prevent plaintiffs from meeting their burden of proof. Analyzing the presence of racially polarized voting will be one of the most important issues during and after the 2021−22 redistricting round. Within the last year, an innovative method adapted from other fields of study has been applied to the racially polarized voting analysis in vote dilution cases and has been approved by a federal district court and the Second Circuit: Bayesian Improved Surname Geocoding (BISG). BISG has received little scholarly attention in legal scholarship addressing voting rights yet promises to be the most critical advancement in detecting vote dilution in decades. This Article seeks to showcase this method, equipping voting rights advocates and governments alike in their effort to secure equal voting rights nationwide. This Article argues that BISG should be used by voting rights advocates as an additional method to bolster racially polarized voting analysis conclusions when the necessary data is available and of sufficient quality. Further, BISG should be utilized by governments in jurisdictions with limited access to American Community Survey or decennial census block data to redistrict in compliance with § 2.

Keywords: Voting rights, civil rights, racially polarized voting, vote dilution, gerrymandering, Voting Rights Act, Gingles, expert witness, quantitative methods, Bayesian