Introduction
In the United States, the majority of students attend public schools based on "attendance boundaries"—catchment areas that determine which school a child attends based on their home address. Because American neighborhoods remain deeply segregated, these boundaries often institutionalize racial and ethnic separation within the school system. While redrawing the lines between different school districts is a complex political and legislative challenge, local school boards possess the authority to redraw boundaries within their own districts.
This research investigates a fundamental question: To what extent can school integration be improved using existing administrative tools while respecting practical constraints like travel time, school capacity, and geographic cohesion?
Methodology
The researchers approached school rezoning as a constrained optimization problem. Given the complexity of rezoning—an NP-hard combinatorial problem—the study utilized Constraint Programming via Google’s CP-SAT solver. This approach allowed for the handling of non-linear objectives and complex geographic constraints more effectively than traditional linear programming.
The Objective
The primary goal was to minimize racial and ethnic segregation, measured by the Variance Ratio index. This metric is particularly useful as it accounts for both the evenness of student distribution and the level of exposure between different groups.
Constraints
To ensure the proposed boundaries were realistic and acceptable to families, the authors established three primary constraints based on a survey of 250 parents:
- Travel Time: No individual family’s commute could increase by more than 50%.
- School Size: To prevent overcrowding, no school’s population could increase by more than 15%.
- Contiguity: School zones were required to be geographically "connected" (no non-contiguous islands), preserving a sense of community and neighborhood identity.
Data and Modeling
The study synthesized several high-resolution datasets to model the school assignments of over 3 million students across 98 U.S. districts:
- Attendance Boundaries: 2021/2022 school zone data provided by ATTOM.
- Demographics: 2020 US Census block-level population counts for individuals under 18 and 2019/2020 NCES school enrollment data.
- Geography: 2020 US Census block shapefiles.
- Travel Logistics: Driving times between Census block centroids and school locations were estimated using the OpenRouteService API.
Key Insights
The Efficiency Paradox
One of the study's most striking findings was that redrawing boundaries for diversity actually decreased travel times for the average student by nearly a minute. This suggests an "efficiency paradox": many current school boundaries are not only segregated but also geographically inefficient. This implies that existing zones may be inadvertently (or intentionally) designed in ways that prioritize segregation over proximity.
The Impact of Contiguity
The requirement for geographic contiguity—ensuring a school zone is a single, continuous shape—is the most significant barrier to integration. Sensitivity analysis revealed that if the contiguity constraint were removed, the potential reduction in segregation would jump from 14% to 35%.
Distribution of Impact
Achieving these diversity gains would require approximately 20% of students to switch schools. However, the researchers found that this "switching cost" was distributed relatively evenly across different racial and ethnic groups in the aggregate, rather than being borne by a single community.
Results
The optimization model demonstrated significant potential for reform:
- Segregation Reduction: A median 14% relative decrease in the Variance Ratio index across the 98 districts studied.
- Travel Stability: Despite allowing for travel increases, the average student who switched schools actually experienced a slight reduction in travel time.
- Public Accessibility: The authors launched schooldiversity.org, a public dashboard that allows stakeholders to explore alternative boundaries and outcomes for specific districts.
Limitations and Future Directions
While the results are promising, the study notes several limitations. The model relies on block-level student count estimates from Census data, which may introduce noise. Additionally, the model is a static "snapshot" and does not account for long-term behavioral shifts, such as "white flight" or residential relocation in response to new boundaries.
Furthermore, while this study focused on racial and ethnic diversity, socioeconomic status (SES) is an equally critical dimension of segregation that warrants future research. Finally, extending this work to include "feeder patterns"—ensuring diversity gains in elementary schools propagate through middle and high schools—remains an important area for future development.
Summary updated