The residency match changed whose preferences drive the algorithm

By the mid-1990s medical students distrusted the National Resident Matching Program, which paired them with hospitals every year. Critics said the hospital-proposing algorithm served institutions, and students discussed gaming their preference lists.

In autumn 1995 the NRMP board retained economist Alvin Roth to evaluate the algorithm and design alternatives. His team built and computationally tested an applicant-proposing algorithm, completing the study in late 1996.

The board voted on 5 May 1997 to adopt the new algorithm, and the first match with it ran in 1998. Roth and Elliott Peranson later described how the design handled complementarities between applicants and positions that simple matching theory ignores.

Student groups welcomed the change as restoring faith in the system. The redesign became a landmark in market design: a matching mechanism chosen not only for efficiency, but for whose interests it visibly serves.

Let applicants propose; they get their best stable match

An applicant-proposing algorithm makes truthful preference lists a safe strategy for students.

Computational testing showed stable outcomes survive even when couples and other complications break the simple theory.

The visible choice of whose side the algorithm favors rebuilt the trust that participation depends on.

Trust is part of the mechanism. The switch to an applicant-proposing algorithm changed little visibly, but removed students' reason to distrust the match — and restored participation.

The student-proposing algorithm remains in use, and the NRMP redesign helped establish matching theory as applied economics — work for which Roth later shared the Nobel Prize.