Optimizing for Distributional Goals in School Choice Problems

Abstract

I investigate three goals of school choice: welfare, encouraging neighborhood schools, and diversity. I use optimization problems to find the best stable and incentive compatible match for any combination of these objectives. These problems assume there is a continuum of students and school seats, which allows me to describe the incentive compatibility conditions in a tractable form. I prove that the set of stable matchings is generically continuous in the distribution of students and the school capacities, which implies that the characterization of the possible stable matches in the continuum model approximates the set of stable matches in a matching market with a large, but finite, number of students. I then apply my framework to data from Boston Public Schools. If the mechanism conditions on demographics, the improvement (relative to the status quo) in student welfare is equivalent to moving 291 students (out of 3,479) to schools one rank higher in their preference lists. In contrast, if the mechanism does not condition on demographics, the welfare improvement is equivalent to moving only 25.1 students to schools one rank higher. Improvements in the distributional goals can be made (e.g., increasing enrollment in neighborhood schools by 50%) without reducing welfare or diversity. This paper was accepted by Gabriel Weintraub, revenue management and market analytics.