#22 2003 · NYC Department of Education · School choice
New York ran its high schools like the residency match and the unmatched fell ninety percent
the problem
Thirty thousand students matched to nothing
background
Every eighth-grader applying to a New York City public high school listed five preferred schools and hoped for the best; schools made offers directly, with no coordination between them, and a student could hold multiple offers while others held none. The congestion this created was the visible failure: nearly half of students ended up assigned to a school that wasn't even on their five-school list, sent through an unpopular administrative placement instead.
In 2003-04, economists Atila Abdulkadiroğlu, Parag Pathak and Alvin Roth — extending the same deferred-acceptance matching theory behind the medical residency Match — redesigned the process for New York City students entering high school in September 2004: one round, ranked preferences submitted once, computed centrally instead of negotiated school by school.
what everyone would do
Try to fix the process incrementally — let students submit more preferences, give schools more time to make offers, add administrative staff to manually resolve conflicts. None of that touches the actual structural flaw: schools making uncoordinated individual offers with no visibility into what other schools are offering the same student will always double-book some students and leave others unmatched, no matter how the details around that structure are tweaked.
what they saw
Abdulkadiroğlu, Pathak and Roth saw that New York's high-school admissions had the exact same structural shape as the medical residency Match — many applicants ranking many providers, providers making offers with no coordination — a problem matching theory had already solved decades earlier for doctors. The fix wasn't inventing a new algorithm for schools, it was recognizing school choice was the same problem wearing different clothes.
the move
Replaced the strategic high-school application scramble with a deferred-acceptance match designed by the same economists.
why it works
Deferred acceptance replaces every school's uncoordinated, independent offer with one centralized computation that considers all students' rankings and all schools' capacities simultaneously, so no student can hold multiple offers while another holds none purely because of timing or which schools happened to respond first. Because the algorithm was already proven stable and strategy-proof in the residency match, importing it meant New York didn't need to invent or validate a new coordination mechanism from scratch — it only needed to adapt existing, already-trusted math to a new population, which is why the number of unassigned students fell from roughly thirty thousand to three thousand in the very first year it ran.
the payoff
Students left unassigned by the process fell from roughly thirty thousand to three thousand in the first year.
where it breaks
The transplant only works when the target problem genuinely shares the source problem's structure — many-to-one matching with ranked preferences on both sides — rather than merely resembling it superficially; a selection process with different constraints (a single provider choosing among many applicants with no applicant preference, for instance) wouldn't benefit from the same algorithm. It also depends on participants submitting genuine preferences rather than gaming the system, which deferred acceptance is specifically designed to make unnecessary — but only if participants understand and trust that honesty is their best strategy, which requires the mechanism to be clearly explained rather than assumed self-evident.
what came after
That 31,000-to-3,000 drop is why the NYC redesign stands alongside the residency Match and kidney exchange as one of matching theory's three flagship applications. Boston adopted a comparable redesign for its own school-choice system in 2006, using the same team's research on the trade-offs between strategy-proofness and efficiency.
references
- [1]The Design of School Choice Systems in NYC and Boston: Game-Theoretic IssuesAlvin Roth, Stanford (course slides), 2008web.stanford.edu
- [2]The Deferred Acceptance Algorithm: The Economics of Matchmaking in the High School Application ProcessCornell University (Networks course blog), 2017blogs.cornell.edu