No one ever likes long lines. Waiting in line may be inconvenient at the coffee shop or the bank, but it's a serious matter at voting centers, where a long wait time can discourage voters—and can be seen as an impediment to democracy.
However, with millions of Americans showing up at the polls, can long lines really be avoided on Election Day? By developing a tool to help better prepare polling places, Caltech sophomore Sean McKenna is using his Summer Undergraduate Research Fellowship (SURF) project as an opportunity to address that problem.
Over the summer, McKenna, an applied and computational mathematics major who works with Professor of Political Science Michael Alvarez, has been building a mathematics-informed tool that will predict busy times in precincts on Election Day and allocate voting machines in response to those predictions. This information could help election administrators minimize wait times for millions of voters.
"My project is based on a report from the Presidential Commission on Election Administration, which asserted that no American should ever have to wait more than 30 minutes to vote," McKenna says. "And so we're trying to see if we can help reach that goal by allocating voting machines in a new way."
McKenna's work is part of the Caltech/MIT Voting Technology Project (VTP), which has been working on voting technology and election administration since the 2000 election. At a June workshop for the collaborative VTP project, which aims to improve the voting process through research, McKenna met with academics and election administrators who suggested how he might apply his background in mathematics to create a tool for voting administrators to use on the VTP's website.
The tool he is developing uses a branch of applied mathematics called queueing theory to quantify the formation of lines on Election Day. "Queueing theory assumes that arrivals to a system like a polling place have a random, memoryless pattern. Under this assumption, the fact that one person just showed up to the precinct doesn't tell us whether the next person will show up two seconds from now or two minutes from now," he says. "Furthermore, queueing theory predicts line lengths and wait times as long-term averages, which scientists might call a steady-state approximation."
Although queueing theory provided a good jumping off point, there were a few real-world problems that an analytical model on its own couldn't address, McKenna says. For example, voter arrival behavior is not completely random on Election Day; early morning and late afternoon spikes in arrivals are the norm. In addition, polls are usually only open for 12 or 13 hours, which is not considered to be enough time for steady-state queueing approximations to be applicable.
"These challenges led us to review the literature and determine that running a simulation with actual data from administrators, as opposed to attempting to adjust strictly analytical models, was the best way to represent what actually happens in an election," McKenna explained.
The goal of the research is to create a simulation of an entire jurisdiction, such as a county with multiple polling places. The simulation would estimate wait times on Election Day based on information election administrators enter about their jurisdiction into the web-based tool. Administrators would then receive a customized output prior to Election Day, suggesting how to allocate voting machines across the jurisdiction and detailing the anticipated crowds—information that could both predict the severity of long lines and prompt new strategies for allocating voting machines to preempt long waits.
Several other Caltech undergraduates in Alvarez's group also have been working on alternative ways to improve the voting process. Senior physics major Jacob Shenker has been developing a system for more secure and user-friendly postal voting, and recent graduates Eugene Vinitsky (BS '14, physics) and Jonathan Schor (BS '14, biology and chemistry) produced a prototype of a mobile phone app that could help voters determine if there is a long line at their polling place.
While these projects were completed separately, McKenna says there may be room for collaboration in the future. "One thing that we're hoping my tool will be able to do is to predict for administrators what times are going to be busiest, and we could also export this information to the app for voters," he says. "For example, the app could alert someone that their polling place is very likely to have long lines in the morning so they should try to go in the afternoon."
The technologies that McKenna and his student colleagues are developing could change the way that millions of Americans participate in democracy in the future—which would be an impressive accomplishment for a young student who has yet to experience the physical aspect of lining up to vote.
"So that's one kind of sticky situation about my working on this project: I've never actually been in to vote in person. I've only been able to vote once, and since I'm from Minnesota, it had to be absentee by mail," he says.