Designing school choice for diversity in the San Francisco Unified School District

ruling that school segregation is unconstitutional, public schools across the United States are again experiencing rising segregation. In districts like San Francisco, efforts to implement district-wide choice policies have largely failed to reverse these trends. Since 2019, we have collaborated with the San Francisco Unified School District to design a student assignment policy meeting the district's goals of diversity, predictability, and proximity. We use optimization techniques to develop policies restricting choice to zones. We compare these to district-wide choice policies by estimating a choice model using historical application data and simulating assignment. Our results show multiple policies can reduce school segregation while preserving varying degrees of proximity, choice, and predictability. Zone-based policies with reserves reduce the concentration of racial minority students in high-poverty schools and maintain proximity, but reduce choice. District-wide choice with reserves and distance-based priority can further reduce segregation, but at the cost of predictability. Tradeoffs are necessary as no policy dominates another policy. In addition, we show that spatial segregation indices capture average patterns in these tradeoffs but miss important variation revealed by our simulation framework. Our work informed the design and approval of a zone-based policy in San Francisco in 2020, and supports ongoing policy implementation.