New gastroenterologists participated in a labor market clearinghouse (a "match") from 1986 through the late 1990's, after which the match was abandoned. This provides an opportunity to study the effects of a match, by observing the differences in the outcomes and organization of the market when a match was operating, and when it was not. After the GI match ended, programs hired fellows earlier each year, eventually almost a year earlier than when the match was operating. It became customary for GI program directors to make very short offers, rarely exceeding two weeks and often much shorter. Consequently many potential fellows had to accept positions before they finished their planned interviews, and most programs experienced cancellations of interviews they had scheduled. Furthermore, without a match, many programs hired more local fellows, and fewer from other hospitals and cities than they did during the match. Wages, however, seem not to have been affected. To restart the match, we proposed a policy, subsequently adopted by the gastroenterology professional organizations, that even if applicants had accepted offers prior to the match, they could subsequently decline those offers and participate in the match. This made it safe for programs to delay hiring until the match, confident that programs that did not participate would not be able to "capture" the most desirable candidates beforehand. Consequently it appears that most programs waited for the match in an orderly way in 2006, when the GI match was reinstated. The market for gastroenterologists provides a case study of market failures, the way a centralized clearinghouse can fix them, and the effects on market outcomes. In the conclusion we discuss aspects of the experience of the gastroenterology labor market that seem to generalize fairly widely.