On November 8, 2016, tens of million of Americans went to vote at about 110,000 polling places. Most were in and out in under 10 minutes, but many still waited in line a long while—in some cases, for hours. Until now, a comprehensive nationwide account of how much time voters spent waiting was out of reach.
In a new study led by economist Keith Chen of the University of California, Los Angeles, researchers matched anonymous location data from 10 million smartphones to 93,000 polling places to create the most extensive map to date of voter wait times across the U.S. The results, reported in a preprint paper posted on arXiv.org on August 30, showed one very clear disparity: voters in predominantly black neighborhoods waited 29 percent longer, on average, than those in white neighborhoods. They were also about 74 percent more likely to wait for more than half an hour.
This study “was a totally, totally different way to try to measure this problem than what we've seen before—and it comes to the same conclusion,” says Stephen Pettigrew of the University of Pennsylvania, an expert in voting wait time, who was not involved with the study.
Timing the Vote
Long lines at polling places have made headline after headline in recent years, and have been the subject of a flurry of research by political scientists. Many see protracted voting wait times not only as an inconvenience but as a civil rights issue.
“[It] really creates a barrier to the most vulnerable voters out there,” says Sophia Lin Lakin, an attorney at the American Civil Liberties Union (ACLU). Minority voters often have less flexible work hours, she says, so lengthy wait times can reduce their ability to vote. Long lines are estimated to have deterred between 500,000 to 700,000 people from casting their ballot in 2012. These problems led to the creation of the Presidential Commission on Election Administration, which issued a 2014 report that set forth a standard: “No citizen should have to wait more than 30 minutes to vote.”
Measuring how much time voters really spend waiting, though, is easier said than done. Previous studies have typically relied on two methods: self-reported surveys and in-person poll observers. Studies using these methods generally agreed on a basic result: in 2012 and 2016 the average nonwhite voter waited about twice as long as the average white voter. Poll observers can precisely track an individual’s wait time, and surveys are generally quite accurate. But both methods suffer from limited reach; the most comprehensive prior study covered only 528 polling places.
This data gap is where the new method shines: Chen and his colleagues collected addresses for 93,658 polling places—80 percent of the U.S. total—and converted each into latitude and longitude coordinates, effectively creating a map of voting spots across the country.
Smartphones “ping,” or send out data with their location, every few minutes on average. Using proprietary data from SafeGraph, a company that collects such information from smartphone apps, the researchers gathered pings that came from within 60 meters of a polling place during the 2016 presidential election. From this sample, they excluded people unlikely to be voters: for example, those who were in the area all day and thus likely to be poll workers, and those who were there for less than one minute and thus likely to be just passing by. After doing so, they ended up with a sample of more than 150,000 voters at around 40,000 polling locations. Using demographic data from the U.S. Census, Chen and his colleagues compared the wait times of voters by neighborhood—and found those in majority-black neighborhoods (as well as other non-white-majority neighborhoods) waited longer.
Although the wait times the researchers calculated are not as precise as past studies, which could identify individual voters, Pettigrew says the results are a trustworthy indicator of this trend. By 2016 two thirds of all Americans owned a smartphone, and voters who did not have one would presumably have been just as likely to get stuck in line. The study contended that its method was, in fact, biased toward not finding a racial disparity in wait times: false positives—nonvoters who resembled voters, based on smartphone data—would have acted like random noise and should have made it hard to see an effect. But the difference between black and white neighborhoods remains clear.
Still, the smartphone method does not answer a key question: “What it doesn’t tell us is why. What’s going on here?” says Robert Stein, a political scientist at Rice University, who did not participate in the new research. “I would want to use it as a basis for doing further observational studies. You know, you just can’t get that from the phone.”
In the new study, the researchers acknowledged the phone data were only a proxy for voting wait time, which could result from anything—ballots that take a while to fill out, a lack of parking spaces or too few voting machines. “It basically tells us who’s within 200 feet [60 meters] of a polling place on Election Day,” says Charles Stewart, a political scientist at the Massachusetts Institute of Technology, who was also not involved in the research.
Unlike explicit methods of voter suppression, such as Jim Crow–era laws, long voting wait times do not necessarily result from a direct intent to discriminate. Previous studies have found that some areas—often those that are nonwhite—lack the resources to properly accommodate their voters.
But in a 2016 study, Stein and his colleagues did find one particular culprit for long lines: voter ID laws. In many states, voters who do not have a certain form of identification must fill out a provisional ballot, a time-consuming ordeal that can clog up a polling place. Like cars on a highway, even a single stopped voter can slow down the whole line. Areas with minority voters—who are less likely to have an ID—tend to be most affected by these laws, Stein says.
The past decade has seen numerous challenges to voting, including gerrymandering (manipulating districts in a way that politically disenfranchises some voters), voter-roll purges (challenging people’s registration status to keep them from voting), and, in 2016, targeted disinformation campaigns to suppress black voters. “Significantly, the Supreme Court has neutered the core protection of the federal Voting Rights Act,” says Wendy Weiser, director of the Democracy Program at the Brennan Center for Justice at New York University. Previously, states with a history of discriminatory voting laws had to receive clearance from the federal government before making any changes to voting practices. But since the court’s 2013 decision, nearly 1,700 polling places—many in black and Latino neighborhoods—have been shuttered. In 2017 the ACLU of Georgia sued a county elections board in the state for closing several polling places without adequate notice (more than 200 voting precincts have been closed in Georgia since 2012). “From our perspective, that’s an unconstitutional burden on these individuals’ right to vote,” Lakin says.
With the 2020 elections around the corner, the authors of the new study wrote that their method is an “easily available and repeatable tool to both diagnose and monitor” voting wait times. But getting these kinds of smartphone data is difficult because they are proprietary, and getting the locations of nearly 100,000 polling places is no mean feat. “It’s an incredibly rich data set, and there would be huge opportunities for either [the study authors] or for other researchers to take what they have and build on it,” Pettigrew says. “Putting the data out there would be a huge step in the right direction.”