NYC's Gale-Shapley High School Matching Algorithm — Scale, Audits and Unaddressed Equity Gaps

New York City's Department of Education (DOE) has used a centralised, algorithmic matching system — built on the Gale-Shapley deferred-acceptance (“stable marriage”) algorithm designed by economists Atila Abdulkadiroğlu, Parag Pathak and Alvin Roth — to assign students to high schools since 2003–04. Students rank preferred schools; schools set priority criteria such as geographic zone, academic screens or audition results; the algorithm iterates until no student–school pair would both prefer each other over their current match.

The system now processes roughly 900,000 students across some 1,600 schools, with more than 70,000 eighth-graders applying to high school programmes each year — about 75,000 for the 2026–27 cycle.

For 2026–27, the DOE reported that 92% of applicants received an offer from one of their top five ranked choices, up from 90% the previous cycle — a track record maintained since the underlying matching-market mechanism was recognised with the 2012 Nobel Memorial Prize in Economic Sciences (awarded to Alvin Roth and Lloyd Shapley).

A July 2025 audit by the New York State Comptroller found significant governance gaps behind the headline match rate. The DOE had no written policies governing a process serving hundreds of thousands of students. Roughly 7,000 students in temporary housing were not consistently given the legally required priority consideration of both current and prior addresses in the 2023–24 cycle. In a sample of 39 students, 31 were likely low-income but had not been correctly identified for priority purposes. More than 200 manual placement overrides that year were improperly documented or missing entirely, and the DOE had not met its Local Law 72 (2018) obligation to publish timely admissions data.

The case illustrates a common pattern in large-scale algorithmic public administration: the matching algorithm itself performs as designed and at genuine scale, but the surrounding data pipelines, manual-override processes and transparency obligations that determine whether vulnerable students are fairly served received far less rigour — a gap surfaced only through independent audit, not the DOE's own reporting.

Read the full analysis: https://www.osc.ny.gov/state-agencies/audits/2025/07/24/management-student-school-matching-algorithm