Blocks as Geographic Discontinuities

Do polling place movement and reassignment affect voter turnout?

August 15th, 2022

Americans have fiercely debated closing and moving polling places in recent years. Civil rights groups like the Leadership Conference Education Fund have opposed high-profile polling location closures and relocations, charging that the changes represent a “particularly pernicious way to disenfranchise voters of color.”¹ Many election administrators and other public officials have defended the changes as necessary concessions to efficiency,² to ensure compliance with regulations like the ADA,³ or as a natural response to declining numbers of in-person voters.⁴

However, a central question in the debate has remained difficult to answer: how much does changing voters’ polling places affect turnout?

To find out, we looked at hundreds of thousands of voters across ten states examining how changes in their assigned polling places affected whether they voted and how they voted—in person, early, or by mail.

Political scientists have argued—with varying levels of convincingness—that everything from the weather⁵ to shark attacks⁶ affects voter behavior. Sieving out the effects of polling place reassignment, specifically, takes some care.

To do so, we identified a natural experiment on streets where precinct boundaries run down the middle of the block, as shown below.

An illustration of our block randomization design in Milwaukee, Wisconsin. Voters on both sides of the street were assigned to the same polling place in 2012 (green, bottom panel). In 2016, voters on the east side of the block were reassigned to a new polling place 0.07 miles farther away (orange, top panel).

People rarely factor in—or even known—their new home’s assigned polling place when deciding where to live. So, in terms of polling place assignments, given that someone lives on a particular block, which side of the street they live on is “as good as random.”⁷ In particular, on the blocks split by precinct boundaries—that is, across “geographic discontinuities”—differences in polling place assignment could change voter behavior. However, the effects of age, race, gender, wealth, weather, and shark attacks should, on average, wash out.

By comparing voting rates on the side of the block closer to its polling place to turnout on the side farther away,⁸ we measured to what extent moving a polling place farther away reduced turnout. What we found, as shown in the figure below, is that assigning voters a more distant polling place reduced their probability of voting in-person on election day by about 1.5% \((\pm 0.4\%)\).

The effect of increasing distance to polling place on voting. Increased distance does not reduce probability of voting, on average, but does affect the method of voting.

However, in states where alternatives such as early voting or mail-in voting exist, assigning voters to a farther polling place increased their probability of voting using one of these substitutes by about 1.6% \((\pm 0.6\%)\). As a result, when substitutes were available, reduced in-person and increased alternative voting cancelled out to well within the uncertainty of our estimate \((0.1\% \pm 0.4\%)\).

In addition to distance, polling place reassignment itself may be a reason not to vote. It could be a source of annoyance or confusion, or it might require voters to spend additional time and effort to find out where their new polling location is and how to get there.

How much does the “shock” of polling place reassignment change voter turnout? By looking at blocks where one side of the block experienced a polling place change between the 2012 and 2016 elections while the other did not, we estimate that the “shock” of reassigning polling places reduced in-person voting by 1.3% (\(\pm 1.0\%\)) and increased voting with substitutes by 0.8% (\(\pm 0.7\%\)). Significantly, the voters living on these blocks were roughly as likely to live in counties where the shock involved adding polling places as where the shock involved reducing polling places.

The effect of shock on voting. While in-person election day voting is reduced, more than half of the reduction is substituted to other methods of voting.

In summary, in the absence of substitutes, we estimate that polling place change reduces in-person voting by around 1%.⁹ To be sure, a 1% effect on voter turnout will mostly likely result in less than a 1% change in the margin of victory in an election. Most voters’ polling places are not reassigned in a given election cycle. Moreover, partisan segregation is not total, and so increasing the distance one party’s voters have to travel to their polling place generally increases the distance for at least some voters of the other party as well.

But importantly our estimated net effect of distance on turnout when voters can substitute into other voting methods, such as early voting or vote-by-mail, is around 0.1%. While not every state affords voters easy access to substitutes, 0.1% is less than the margin of victory in every state in every presidential election since 2000.

We caution that these results are limited to the population of registered voters—and, in particular, to the registered voters who happen to live on the geographic discontinuities, who could conceivably differ from the general population in subtle ways. But our work suggests that voters may be able to make up for the challenges of traveling farther to vote and polling place assignment more generally by instead voting early or by mail when these options are available.

(Many thanks to my collaborators: Sabina Tomkins, who led this project; Keniel Yao; Tobias Konitzer; David Rothschild; Marc Meredith; and Sharad Goel.)

1. Democracy Diverted: Polling Place Closures and the Right to Vote, p. 8. ↩

2. In a Georgia County, Deep Distrust Over a Plan to Close Polling Places: “In Lincoln County, Ms. Bolton, the county elections director, argues that the change would make it easier for her to manage Election Day. Her tiny staff is stressed, she said, by the responsibility of setting up and breaking down the complicated electronic voting machines in seven locations spread around the county’s 257 square miles.” ↩

3. Dodge City polling place debacle: voter suppression or incompetence?: “Long before Cox was elected, Dodge City consolidated voting into a single polling station at the civic center in the mid-1990s after a new federal law on access for people with disabilities meant that many existing places people went to vote could no longer be used. What was intended as a temporary measure became the norm.” ↩

4. EAVS Deep Dive: Poll Workers and Polling Places: “There has been a continued decrease in physical polling places, which can likely be explained by the expansion of alternative voting options, the increased use of these options by voters, and the corresponding decrease in in-person voters on Election Day.” ↩

5. Gomez, Brad T., Thomas G. Hansford, and George A. Krause. “The Republicans should pray for rain: Weather, turnout, and voting in US presidential elections.” The Journal of Politics 69.3 (2007): 649–663. ↩

6. Achen, Christopher H., and Larry M. Bartels. 2016. Democracy for Realists: Why Elections Do Not Produce Responsive Government. Princeton, NJ: Princeton University Press. But see also, Fowler, Anthony and Andrew B. Hall. “Do Shark Attacks Influence Presidential Elections? Reassessing a Prominent Finding on Voter Competence.” The Journal of Politics 80.4 (2018): 1423–1427. ↩

7. In the jargon, voters’ voting potential outcomes—i.e., whether and how they would counterfactually choose to vote if their polling place were reassigned—should be (approximately) independent of the side of the block they live on, given that they live on a certain street. In symbols, \(V(s,d)\ \rlap{\perp}\mkern2mu\perp Z \mid B\), where \(V(s,d)\) denotes whether a voter would vote if their polling place were changed (\(s=1\)) or not changed (\(s=0\)) to distance \(d\), \(Z\) indicates which side of the street they live on, and \(B\) indicates the block they live on. ↩

8. We also required that no one’s polling place on the entire block changed between the 2012 and 2016 elections to further isolate the effect of increased distance, as distinct from the “shocks” discussed below. ↩

9. Our estimate also reflects the kind of polling place changes captured in our data. It’s possible, for instance, that if polling places were moved much farther than the typical move in our data then the reduction in in-person turnout could be correspondingly greater. ↩