By Kevin Morris (Brennan Center NYU) and Peter Miller (Brennan Center NYU)
The year of 2020 was one marked by disruption and upheaval as a consequence of the COVID-19 pandemic. The United States proceeded with scheduled elections during the pandemic, forcing voters to reconsider whether and how they would be involved in the contests. We take advantage of a natural experiment to assess how COVID and the substantial reduction in polling places affected turnout in Milwaukee during the April presidential primary election relative to a set of voters largely unaffected by closed polling places. Unlike previous cases of polling place consolidation in the literature, the episode in Milwaukee was brought about by a natural disaster at the last minute before an election rather than an administrative decision made well in advance of election day.
The initial phase of the pandemic was marked by a palpable sense of uncertainty and confusion among state officials. The governor declared a state of emergency, banned gatherings of more than 10 people, and called for absentee voting for all Wisconsin voters in the weeks leading up to the election. The legislature refused these initiatives, even challenging the governor’s actions in state and federal courts. The speaker of the state Assembly, by contrast, worked at a polling place on election day, dressed in protective equipment and remarked “You are incredibly safe to go out. People have to make their own best judgment.” The city of Milwaukee closed 97% of the planned polling places for the primary election because of a shortage of poll workers and higher prevalence of COVID cases.
Our research is designed to assess two hypotheses. First, we expect that the consolidation of polling places in Milwaukee will reduce turnout relative to voters in the surrounding area of the state, but outside the city itself. Second, we expect greater exposure to COVID in the city will reduce turnout above and beyond the effects of polling place consolidation.
We deploy a regression discontinuity in space model to assess these hypotheses. In other words, we match voters in Milwaukee city to voters outside the city but in an adjacent county in a series of increasingly wide buffers around the municipal boundary. This model allows us to compare, as nearly as possible, voters who share a social milieu and presumably have similar levels of exposure to COVID. The voters within the city, however, are also exposed to the stark reduction in the number of polling places; voters in the surrounding counties had a relatively light reduction in the number of polling places.
At our most restrictive buffer—a quarter mile on either side of the Milwaukee city limits—we find voters in Milwaukee city were almost 9 points less likely to vote—regardless of method—than voters on the other side of the municipal boundary. As we increase the width of that buffer, and in so doing start to compare voters in areas with greater risk of exposure to COVID, the demobilizing joint effects of polling place consolidation and COVID exposure further reduce the likelihood of casting a ballot among the voters in Milwaukee city. At our most permissive, 10-mile buffer the likelihood of voting by a voter in the city is reduced by 12 percentage points. We also find these effects are stronger when we focus our analyses on Black voters living in Milwaukee. At our most restrictive specification, the likelihood a Black voter in Milwaukee cast a ballot in the primary election was reduced by more than 10 points. This effect grows to 15 points as we increase the width of the buffer. While polling place consolidation in Milwaukee reduced turnout among voters in the city, this effect was heaviest among minority voters. We summarize our findings in the graph below.
We see a few areas where this research can inform future work at the intersection of voting behavior and election administration under adverse and inclement circumstances. These findings suggest a natural disaster occurring before an election is difficult for election officials to counteract using their discretionary authority. We are also aware that this study is located in the most segregated city of its kind in the United States. Research in other cities would establish if the effect with regard to Black voters in Milwaukee we report here is a function of the social geography of residential patterns.
Kevin Morris is a quantitative researcher in the Voting Rights and Elections Program at the Brennan Center for Justice at NYU School of Law and a PhD student in the sociology program at the CUNY Graduate Center. His work focuses on felony disenfranchisement, voter purges, and access to the ballot box.
Peter Miller is a researcher at the Brennan Center for Justice at NYU School of Law, where he focuses on redistricting, voting, and elections. His research interests include U.S. and comparative politics, voting behavior, political institutions, and public opinion. He was a Fulbright Scholar at Tampere University, in Finland, during the 2016 to 2017 academic year. He holds a doctorate in political science from the University of California, Irvine and a bachelor’s degree from Reed College.