“An Experimental Study of Matching Markets with Incomplete Information,” Leeat Yariv, Marina Aganov, and Ahrash Dianat, California Institute of Technology; Larry Samuelson, Yale.
Exchange in some important markets occurs through two-sided matching processes: matching hospitals with interns, workers with employers, organ donors with patients, children available for adoption with adoptive parents, and students with schools. Market design for these applications apply algorithmic procedures for efficient matching—say schools with students—based on one or each side’s rank order preference for the other side. The theoretical literature has dealt with the problem under the assumption that the participants have complete information on the preferences of others. But recent theory enables the prediction of final outcomes in match-rematch processes that are expected to emerge as preferences are revealed by decisions in decentralized repeat interactions. The theory, for example, might predict convergence to efficient stable outcomes in matching temporary workers with firms through an agency. But, how would you test that prediction? In the field, precise preferences cannot be readily and independently elicited or deduced for comparison with final allocations. In the lab, one can use payoffs to induce preferences, vary payoff structures and initial allocations (which define stable efficient outcomes). In the baseline experiments, subjects know all others’ value types. In the incomplete information setting, all know firms’ value types, but workers’ value types are strictly private. Hence, a target prediction is well defined; then, information and the sophistication requirement of firms’ decisions about workers is varied across treatments. By inaugurating the study of matching markets under incomplete information, this research holds promise for improving and expanding the many applications.