Researchers Used AI to Rezone School Districts. Here’s What They Found

Efforts to desegregate schools through rezoning often spark a backlash among parents citing concerns over school crowding and commute times.

But ill-fitting attendance boundaries (and the formulas used to make them) may make districts less efficient as well as less equitable, and new research suggests education leaders can better integrate schools without lengthening the amount of time students spend on the bus.

In a new study in the journal Educational Researcher, education scientists from Northeastern University and the Massachusetts Institute of Technology simulated new attendance zones for nearly 100 of the nation’s largest districts.

Updating attendance boundaries using algorithms based on district goals and parent preferences, they found, could reduce the level of segregation between white students and students of color across district schools by 14 percent on average, while slightly reducing travel times and requiring about a fifth of students to change schools.

“[Attendance] boundaries are such a sticky default setting in districts,” said Nabeel Gillani, an assistant professor of design and data analysis at Northeastern and the lead author of the study, noting that on average, only 15 percent of districts update their attendance zones each year.

“I wonder if these boundaries have been set, and the world changed, and we haven’t updated in ways that could actually lead to wins across the board” in efficiency and diversity, he said.

More diverse schools, shorter bus rides

The analysis is part of a massive zoning simulator that uses elementary school data from more than 4,000 districts. The tool allows districts to realign attendance zones to see how they might increase racial diversity, while capping increases in travel times and the student populations at individual schools. A 2007 U.S. Supreme Court ruling limits schools’ ability to use race as a deciding factor in assigning students to schools, but the researchers are expanding the tool to simulate zoning plans that would reduce income disparities in schools as well.

The Civil Rights Project at the University of California, Los Angeles, which tracks school segregation over time, finds that both racial and economic isolation have risen significantly in the last 30 years, particularly in the same 100 largest districts whose zoning plans were simulated in the current study. For example, before the pandemic, Black students attend schools where they account for, on average, 47 percent of the population, though they make up only 15 percent of all public school students nationwide.

Differences between school districts account for an estimated two-thirds of total racial segregation in schools, but superintendents have much more control of within-district segregation, Gillani said.

For example, nearly 70 years after the U.S. Supreme Court’s landmark school desegregation ruling, students of color in Topeka, Kansas, still attend schools largely separated from their white peers. The data show more than 65 percent of Topeka elementary students are non-white, but they attend schools where 78.5 percent are students of color. Redrawing attendance zones based on the study’s algorithm would cut average travel time there by about a minute and increase racial integration by about 3 percentage points.

In the analysis of the largest school districts, researchers structured new zoning plans to prevent commute times from increasing by more than half or student populations from growing by more than 15 percent at any school.

Gillani and his colleagues are working with school districts to use algorithm-based tools to plan both how to overhaul attendance zones and to engage their communities in the discussion.

“To get board members to pay attention to this and consider policy changes,” based on new attendance zones, he recommended “focus on the travel times, focus on school utilization, and let desegregation be the Trojan horse, in some ways, in the discussion.”