Spatial and statistical predictors of voter purge rates in Michigan

Abstract Objective Voter purges can protect election integrity by ensuring deceased or moved individuals are removed from election rolls. But they have been used to diminish voting power of marginalized groups, often by anti‐majoritarian forces seeking to undemocratically retain power. Little research has examined “who gets purged?” at the state level, especially with respect to local‐level differences. Methods We leverage Michigan's voter purge database from 2014 to 2018. Records are geocoded to their exact address, and a range of spatial correlates are identified to answer the above question. We then used generalized structural equation modeling to incorporate patterns of mobility and mortality. Results Initial results showed that more Democratic leaning areas, denser/more urban areas, and areas with more Black residents had higher purge rates. Notably, while these mediation effects were significant, racial composition and median income (i.e. more black and poorer communities) remained a significant factor in voter purge rates. These results suggest a potentially troublesome underlying element in Michigan's pattern of voter purges. We suggest this is an important first step in future research in other states and with subsequent databases, which can help strengthen the case that purges may be being used to uphold discriminatory and anti‐majoritarian goals.