We create an environment in which congestion forces agents to match inefficiently early. We then introduce one of two centralized clearinghouse mechanisms. One of these has been successfully used to halt this kind of unraveling in a number of labor markets, while the other has failed. When it is costly for �rms and workers to be mismatched compared with the costs of matching early, the experimental observations reproduce the �eld observations. Furthermore, the experiment permits us to observe the transition between a decentralized and a centralized market, both when the centralized market fails to control unraveling and when it succeeds, at a level of detail unavailable in �eld data. Field studies of competitive entry-level professional labor markets reveal a common form of market failure involving the ‘‘unraveling’ ’ of hiring decisions, in which employment contracts from year to year become earlier and earlier in advance of employment, even when this becomes very costly {Roth and Xing 1994}. Sometimes the cause is that the market is organized in

We report on the design of the new clearinghouse adopted by the National Resident Matching Program, which annually fills approximately 20,000 jobs for new physicians. Because the market has complementarities between applicants and between positions, the theory of simple matching markets does not apply directly. However, computational experiments show the theory provides good approximations. Furthermore, the set of stable matchings, and the opportunities for strategic manipulation, are surprisingly small. A new kind of “core convergence ” result explains this; that each applicant interviews only a small fraction of available positions is important. We also describe engineering aspects of the design process.

A number of labor markets and student placement systems can be modeled as many-to-one matching markets. We analyze the scope for manipulation in many-to-one matching markets under the student-optimal stable mechanism when the number of participants is large. Under some regularity conditions, we show that the fraction of participants that have incentives to misrepresent their preferences when others are truthful approaches zero as the market becomes large. With an additional technical condition, truthful reporting by every participant is an approximate equilibrium under the student-optimal stable mechanism in large markets. The results help explain the success of the student-optimal stable mechanism in large matching markets observed in practice.