Locating polling sites in highly-populated areas won’t necessarily increase voter turnout on November 5. In fact, naive approaches that rely only on population density or distance to polling sites fail to account for other important considerations, such as time constraints, which can hamper people’s ability to vote. 
 
That’s the finding of a new study by a team of UCLA mathematicians that was published today in SIAM Review, the flagship journal of Society for Industrial and Applied Mathematics (SIAM). The study suggests that filling “holes” (voting deserts where there tend to be long waiting times for voters to cast votes) in the distribution of polling sites can boost equitable voting access and allow more people to cast ballots.
 
“A lot of policies are based on more naive and arbitrary measures like determining what percentage of people are within a certain distance of a polling site, but we’ve used a mathematical approach called topological data analysis that illustrates distance alone isn’t good enough when it comes to determining where polling sites should be located,” said Mason A. Porter, a mathematics professor at UCLA. 
 
“Just as important is the issue of time – both in terms of commuting and waiting time – when determining how easy or difficult it is for people to get to the polls,” he explained.
 
The group’s research was spurred by the 2020 U.S. election, following reports of barriers to voting, especially in underserved communities.
 
“We were seeing in the news how people weren’t able to access polling sites in several cities, so we were interested in using methods familiar to us as mathematicians to quantitatively capture what we call “holes” in polling-site access,” said Abigail Hickok, a postdoctoral fellow at Yale University and a former UCLA mathematics PhD student. Hickok co-authored the study with Porter, former UCLA PhD students Jiajie (Jerry) Luo and Ben Jarman, and current UCLA PhD student Michael Johnson.
 
Using a mathematical approach called “persistent homology” (a type of topological data analysis), the study looked at current availability and coverage of polling sites in Los Angeles County, Atlanta, Chicago, Jacksonville, New York City, and Salt Lake City to identify gaps in polling-site accessibility. 
 
“With distance, you take your map and measure it, and you have a concrete number that’s approximately right, whereas with time, you have to account for waiting time at the polls and people’s ability to get themselves there, including factors such as who has a car or access to public transportation,” Porter said. “It’s a more complicated measure, but it’s a more accurate measure.”
 
The study cites the U.S. decision in March 2021 to ensure at least 90% of the adult U.S. population was within five miles of a COVID-19 vaccination site, as well as legislation in India that requires 100% of voters to live within two kilometers (1.2 miles) of a polling site. However, according to the researchers, these approaches pose several issues: they involve choosing an arbitrary geographical distance cutoff, and they fail to account for many other factors, such as population density and demographics, the availability and ease of use of public transportation, waiting times, and whether or not people can afford to take time off work.
 
“These factors can also vary by city,” Luo explained. “In a city with a high population density or a poor transportation system, for example, the time that is spent waiting at or traveling to a resource site can be a much higher barrier to access than geographical distance.” 
 
Pointing to Atlanta and New York City, which are known to have long waiting times at polling sites, and Los Angeles, which has heavy traffic, the authors note that factors like these can negatively affect the accessibility of polling sites. 
 
Using smartphone data (previously published by other researchers) of hundreds of thousands of voters to estimate waiting times and street-network data to evaluate travel times, as well as per capita car-ownership data and Google Maps, the researchers were able to suggest geographic areas where voters can be better served.
 
In Los Angeles County, for example, voter accessibility gaps were identified in northern regions such as Palmsdale and Sun Village.
 
“Our approach effectively provides a way to measure and evaluate how equitably a resource is distributed geographically,” Porter said, adding that it can be applied beyond polling sites to determine locations for other resources, such as public parks, hospitals, vaccine-distribution centers, grocery stores, Department of Motor Vehicles (DMV) locations, and Planned Parenthood clinics.