Faculty, Arches

A Close Look at Redistricting

Students use math, geography, and other disciplines to examine the way electoral maps are drawn.

In 2019, as the federal government began preparations for the 2020 U.S. Census, Associate Professor of Mathematics and Computer Science Courtney Thatcher was doing a lot of thinking about the problem of redistricting.

Every 10 years, the states use population data from the census to redraw election maps. In theory, this redistricting process ensures that there are the same number of people in each district, giving them equal representation in Congress. In practice, though, the process is often mired in partisan attempts to gerrymander districts to benefit one political party over another. Thatcher figured there had to be better ways to think about drawing district maps, so she decided to look for them.

Prof. Courtney Thatcher stands behind students working on their laptops during class

It was important to Prof. Courtney Thatcher to involve students in her redistricting research project, providing them with hands-on research experience and real-world applications of their classroom studies.

"There are lots of ways to think about redistricting," Thatcher says. "In recent years, policy-makers have turned to statistics, machine learning, and data science to make districts using algorithms. That’s one way to do it, but without taking in the geographical, social, and political context, you’re really missing the point."

Working with geographer Jim Thatcher at University of Washington Tacoma and several other senior personnel, including computer scientist Adam Smith at Puget Sound, Thatcher developed a proposal that resulted in a $343,000 grant from the National Science Foundation’s Research Experiences for Undergraduates program. The grant funded a four-year investigation into whether machine learning algorithms that take social and geographic factors into account could shed light on how to create more equitable districts.

"Something we had to keep in mind is that there is no ‘correct’ district. We’re not trying to create new districts; we’re trying to evaluate existing districts to see if there are ways to make them work better for the people who live there," Thatcher says.

The research project ran for four summers, from 2019 through 2022. For eight weeks each summer (with the exception of 2020, when COVID-19 forced Thatcher to switch to a remote program), 12 students from Puget Sound, University of Washington, and across the country came to live and work on the University of Puget Sound campus.

Looking over the shoulder of a student to see the screen of the student's laptop

Running for four summers from 2019 to 2022, 12 students from Puget Sound, University of Washington, and across the country came to live, learn, and work on the Puget Sound campus.

Each student was selected in order to bring a unique skill set to the team: distill the key points of a political science paper, use geographic information software, or hard-code computer programs. Together, these students from different backgrounds and academic disciplines gained hands-on experience applying social theory to empirical data, building spatial models to evaluate district maps, and teaching—and learning from—each other.

“Teaching at a liberal arts college, providing undergraduate research opportunities is a core part of what I do,” Thatcher says. “So, another goal of the project was to get students involved and help them build critical thinking and problem-solving skills to help them be successful in the things they want to do after college.”

Drawing on the skills of an interdisciplinary research team helped Thatcher and her students understand the real-world implications of redistricting, particularly how the lived experience of people in a given district often differs from the way policy experts think about the region.

“As a mathematician, it’s fascinating to work with geographers to think through how people actually live in the space and how redistricting affects them,” Thatcher says. “Two points that look close together on a map may be separated by a river or a mountain range or a prison that you have to travel around. In those districts, someone may have to drive an incredibly long distance to vote, for instance.”

Thatcher and the research team used census data to train their machine learning models and to predict what factors cause a district plan to be challenged in court or thrown out as unconstitutional.

When Thatcher and her students started the project, they had to rely on data from the 2010 census to train their machine learning models. Now that population data from the 2020 census has been released and states are actively engaged in the process of reapportioning their congressional districts, the researchers can apply their models to the process in real time and see the implications of the newly drawn maps. Over the past two summers, the team also worked on ways of predicting what factors will cause a district plan to be challenged in court or thrown out as unconstitutional.

So far, the team has produced one paper and a peer-reviewed abstract, with more publications to come. While Thatcher is hopeful that the research may help reduce gerrymandering and produce fairer election districts, a primary benefit has been to build up the confidence and experience of the students who have passed through the program.

“We’ve sent students off to funded Ph.D. programs; one of them got a National Science Foundation graduate research fellowship; several of them have gone to highly competitive data science programs and research groups; and others have presented our findings at their home institutions,” Thatcher says. “So, there’s the research component, but we also try really hard to give students a leg up in understanding where they want to go and how to get there.”

EXPLORE MORE STORIES