Association for Public Policy Analysis & Management

As urban school choice programs have proliferated, city leaders have turned to unified enrollment systems to manage student applications and placements. These systems ask families to submit ranked lists of school requests so that the system can place students in schools based on families’ requests, seat availability in schools, and various types of priority classifications (e.g., sibling and neighborhood priority). Compared to decentralized choice processes administered by individual schools, a unified enrollment system can ease the burden on families and school leaders while more efficiently matching students to schools and limiting the opportunities for schools to admit students unfairly.

At the heart of unified enrollment systems is a placement algorithm that determines how seats are allocated when the number of requests for a seat exceeds the number of seats available. A rich and growing body of research assesses the efficiency of these algorithms, with roots in the Nobel Prize-winning work of Alvin Roth and Lloyd Shapley. Relatively little research, however, examines the decisions the algorithms leave to policymakers and their implications for equity. In deciding which groups of students should have priority and how the priorities are operationalized, policymakers can shape which students attend which schools. In many cases, policymakers must make these decisions without clear information about their consequences—and without the public able to scrutinize their decisions.

This study uses several years of data from the country’s most comprehensive unified enrollment system—the New Orleans OneApp—to examine how the use of priority groups affects which students are assigned to which schools. We have reproduced the algorithm that places students in schools, which enables us to simulate how placements would differ under alternate formulations of the algorithm. For example, we simulate placements if the system gave less (or no) preference to neighborhood students, newly gave preference to low-income students, or used no priority groups at all.

Our outcomes of interest include common segregation measures (e.g., a dissimilarity index), the rates at which disadvantaged students are placed in high-demand schools, and match rates. The key assumption underlying these simulations is that families’ ranked school requests would not change under different formulations of priority groups. This assumption seems plausible, since the OneApp’s algorithm was designed to be strategy-proof and induce parents to rank schools in their true order of preference, but we can explore it further by examining whether parents’ requests have changed in response to year-to-year changes in the algorithm. This will provide additional insights into the degree to which families’ ranked requests reflect their true preferences.

Policymakers and administrators cannot avoid making decisions about whom, if anyone, to prioritize in unified enrollment systems. Any algorithm reflects normative decisions about how to allocate scarce seats, whether those decisions are made carefully and explicitly or not. By exploring the implications of these decisions, this study will both inform a growing academic literature and provide tangible guidance to education policymakers.