Attributable Risk of Race: Detecting Partisan and Racial Gerrymandering

Sidak Yntiso and Sanford Gordon (New York University)

Abstract: How can we measure racial gerrymandering? Isolating racially disparate impacts of redistricting has proven difficult as sophisticated mapmakers can often claim partisan motivations, despite evidence of racially motivated intent and effect. To address this issue, we introduce a novel measure of racial gerrymandering - the attributable risk of race (ARR) – which captures the disproportionate extent of votes wasted by racial minorities net the difference in votes wasted by each party. Our method builds on a recently developed partisan gerrymandering metric, the efficiency gap. We show that our modified metric survives extant criticisms including equivalence to a rejected `double proportionality' standard, violations of proportionality, failures to account for the spatial distribution of voters, etc. We find a large and positive ARR in North Carolina's 2012 and 2016 Congressional election results despite court ordered redistricting. Our evidence suggests that African Americans’ votes are disproportionately wasted relative to whites’ votes even after adjusting for differences in wasted votes by party. The observed ARR values exceed 95 percent of placebo values from simulated redistricting plans.

View Poster in a New Tab