The weeks leading up to the Wisconsin primary election on April 7 were tumultuous. On March 27, Democratic Gov. Tony Evers called for every voter in the state to be sent an absentee ballot, but the Republican-controlled legislature rejected the idea. The weekend before the election, Evers called an emergency session of the legislature, hoping to postpone the election; once again, his efforts were stymied. Evers was also blocked in the courts; the day before the election was to take place, he issued an executive order moving it to June 9, but the state supreme court struck it down.
These maneuvers occurred against the backdrop of shortages of electoral resources. On March 31, the Milwaukee Journal Sentinel described a dire poll worker shortage: Wisconsin was short some 7,000 poll workers, a situation that led to the consolidation of polling places around the state. The cuts were particularly drastic in Milwaukee City, where the number of polling places dropped from 182 in November 2016, to just five for this year’s primary. In the rest of Wisconsin, the number of polling places dropped by 11 percent. 1See https://elections.wi.gov/elections-voting/2016/fall and https://elections.wi.gov/node/6524.
As the nation prepares for the November general election, we wanted to test whether fewer polling places decreased turnout in Milwaukee, or if voters simply shifted to the vote-by-mail alternative. Our answer is no. To be clear, a surge in absentee voting may have offset at least a portion of any depressive turnout effects of consolidated polling places: while just 170,614 ballots were cast by mail in the 2016 presidential primary, 964,443mail ballots were submitted this year. However, despite this surge in absentee voting, we find that
- polling place consolidation reduced turnout by 8.6 percentage points, and
- Black turnout was especially depressed from these closures.
In order to estimate what turnout out would have been in Milwaukee City if not for the polling place closures, we used a matching model to pair voters in the city with voters outside the city. 2Potential controls came from Milwaukee, Racine, Ozaukee, Washington, and Waukesha Counties. We matched each treated voter to two control voters based on whether they voted in the 2016 and 2018 primaries; their gender, race, partisan affiliation, and ethnicity; their latitude and longitude; and estimates of their household income and education level. The data all come from L2 Political. This ensured that the “treated” voters (individuals living in Milwaukee City) and “control” voters (those living outside the city) had similar socioeconomic characteristics and voting histories. Controlling for these characteristics is important, because they are highly correlated with whether someone casts a ballot. For instance, 50 percent of all suburban voters voted in the 2016 primary, while just 27 percent of Milwaukee voters did so. After our matching procedure, 27 percent of the suburban voters who were used as controls cast a ballot in that contest. Without controlling for these characteristics, we might only be picking up on a lower propensity to vote in Milwaukee City — not the effect of the polling place closures.
In addition to the matching methodology, we only kept pairs of treated and control voters who live within a half mile of one another. Although the paired individuals live in different cities, they live so close to one another that they likely shop at the same grocery stores and eat at the same restaurants. In other words, despite living in different cities, they are probably similarly exposed to Covid-19. 3Our methodology closely follows that of this published paper. Like this analysis, that paper combined a matching model with a geographic restriction around the city border to estimate the effect of ballot initiatives on turnout in Milwaukee City.
We then ran an ordinary least squares regression to see if the voters who live just inside Milwaukee City voted at lower rates than their counterparts just outside the city. We also tested whether turnout was depressed by a different amount for Black voters than other voters in Milwaukee City.
For a fuller discussion of our data, methods, and results, please see our working paper.
Polling place closures did reduce turnout in the 2020 primary election. The table above indicates that turnout in Milwaukee City was depressed by roughly 8.6 percentage points. Considering that 26 percent of our control voters cast a ballot, this implies that polling place closures in the city reduced turnout by a third.
Of particular note, we found that this effect was larger for Black voters. Although polling place consolidation decreased turnout among non-Black voters by around 8.5 percentage points, it reduced turnout among Black voters by 10.2 percentage points.
The serious depressive effects uncovered in Milwaukee — and the racial disparities within them — are cause for concern. Clearly, not all voters who prefer voting in person will seamlessly transition to vote by mail. We cannot know whether Milwaukee residents cast fewer ballots because they were unfamiliar with the mail voting process did not trust it, or were prevented from voting in person because of the long lines. It is also possible that Black voters cast mail ballots at similar rates as other voters but had them rejected at higher rates, thereby reducing their effective turnout. However, given the magnitude of the effect, it is unlikely that this accounts for the full difference.
The case of Milwaukee is important for election administrators to keep in mind as they prepare for this fall’s election. If it can be generalized to the rest of the country, polling place closures will come at the expense of voter turnout — and particularly the turnout of Black Americans. Moreover, a recently released Brennan Center report shows that fewer polling places lead to longer lines to vote, meaning that widespread closures might make casting a ballot harder for in-person voters. If we care about the representation of nonwhite voters and voters wary of casting mail ballots, we must ensure that there are safe in-person options this fall.
2.Potential controls came from Milwaukee, Racine, Ozaukee, Washington, and Waukesha Counties. We matched each treated voter to two control voters based on whether they voted in the 2016 and 2018 primaries; their gender, race, partisan affiliation, and ethnicity; their latitude and longitude; and estimates of their household income and education level. The data all come from L2 Political.
3.Our methodology closely follows that of this published paper. Like this analysis, that paper combined a matching model with a geographic restriction around the city border to estimate the effect of ballot initiatives on turnout in Milwaukee City.