By Kareem Haggag and Devin Pope
Kareem Haggag is a professor of behavioral economics at Carnegie Mellon University. Devin Pope is a professor of behavioral science at University of Chicago’s Booth School of Business.
Until now, research into how long American voters spend waiting to cast their ballots has been limited by the means of data collection, largely self-reported surveys and the accounts of observers at a few hundred polling places. Studies suggested African Americans were consistently forced to wait longer — and thus discouraged from voting — than the population at large.
Now a much more extensive form of data collection — from millions of GPS-equipped smartphones, at tens of thousands of polling stations, in the 2016 presidential election — confirms that voters in majority-black neighborhoods are likelier to wait longer than those in majority-white neighborhoods, often considerably longer.
That is just one of the findings we report, with co-authors Keith Chen and Ryne Rohla, in a National Bureau of Economic Research working paper.
The study began by collecting the addresses of more than 90,000 polling stations and using that information to establish their longitude and latitude coordinates. Working with proprietary data from a company that collects smartphone “pings” indicating the smartphone’s location at regular intervals (typically every five minutes), we started off with information from more than 10 million smartphones. Eventually we winnowed the sample down to 150,000 smartphone users whom we identified as likely voters at about 40,000 polling stations. (People were only included who spent a reasonable amount of time at the polling place on Election Day, but did not visit the polling place on other days leading up to or after Election Day.)
We estimate that the average time it took Americans to vote was 19 minutes, with 18 percent of individuals spending more than 30 minutes at their polling place. These averages mask significant geographic differences. For example, voting times are more than twice as long in some states (South Carolina, Indiana and Utah) than in others (Connecticut, Delaware and Massachusetts).
But the starkest finding is that areas with a high percentage of black residents face longer average wait times. For example, relative to entirely white neighborhoods (identified using census data), residents of entirely black neighborhoods waited 29 percent longer to vote and were 74 percent more likely to spend more than 30 minutes at their polling place. A disparity exists for entirely Hispanic neighborhoods, too, but it is not as pronounced.
That wait of more than half an hour is significant. As President Barack Obama’s Commission on Election Administration declared in its 2014 report, “No citizen should have to wait more than 30 minutes to vote.”
Who or what is behind these racial disparities in voting wait times? Our sample focuses on just one election, so our ability to establish causes was limited, but we looked for clues in the correlations.
Our first instinct was to look at partisan bias. Republicans have an incentive to suppress the black — predominantly Democratic — vote, and in several states, including Alabama, Georgia and Ohio, they have pursued voting restrictions perceived as attempts to limit minority turnout. But we found that the racial disparities in voting wait times are similar, if not worse, within Democratic-majority states and counties. For instance, the wait-time gap is larger in solidly blue California than in solidly red Alabama.
We looked for other patterns in the data but found that racial disparities are uncorrelated with a host of other plausible drivers, including early-voting laws, strict ID laws, income inequality and racial segregation. In case other researchers would like to examine the data for so-far-undetected patterns, we provided state and congressional district-level estimates of the racial disparities in wait times in the paper (including estimates adjusted for measurement error).
The apparent lack of partisan bias in the racial disparity may seem surprising, but it echoes results of a prior study showing that politicians across both parties were similarly less responsive to voter registration help requests made by black constituents. That discrimination may extend to an imbalanced allocation of resources: A 2014 Brennan Center for Justice study documented that polling places with higher proportions of black residents in Florida, Maryland and South Carolina had fewer polling machines and workers in the 2012 election. It seems likely that such unequal resourcing remained in the 2016 election.
The first step toward addressing the racial disparity in voting wait times would be to ensure the equal distribution of polling machines and workers. But that might not be sufficient. As the Pew Charitable Trusts reported last year, “In the five years since the U.S. Supreme Court struck down key parts of the Voting Rights Act, nearly a thousand polling places have been shuttered across the country, many of them in southern black communities.” Fewer places to vote can of course mean longer waits at the surviving sites.
With the 2020 election 11 months away, it may still be possible to correct the racial disparity in voting wait times. Ultimately, we hope smartphone data from the election will be closely monitored and analyzed, leading to greater accountability and eventually to parity in voting wait times across the country.