Can Market Participants Report Their Preferences Accurately (Enough)?

In mechanism design theory it is common to assume that agents can perfectly report their preferences, even in complex settings in which this assumption strains reality. We experimentally test whether real market participants can report their real preferences for course schedules "accurately enough" for a novel course allocation mechanism, approximate competitive equilibrium from equal incomes (A-CEEI), to realize its theoretical benefits. To use market participants' real preferences (i.e., rather than artificial "induced preferences" as is typical in market design experiments), we develop a new experimental method. Our method, the "elicited preferences" approach, generates preference data from subjects through a series of binary choices. These binary choices reveal that subjects prefer their schedules constructed under A-CEEI to their schedules constructed under the incumbent mechanism, a bidding points auction, and that A-CEEI reduces envy, suggesting subjects are able to report their preferences accurately enough to realize the efficiency and fairness benefits of A-CEEI. However, preference-reporting mistakes do meaningfully harm mechanism performance. One identifiable pattern ofmistakeswas that subjects had relativelymore difficulty reporting cardinal as opposed to ordinal preference information. The experiment helped to persuade the Wharton School to adopt the new mechanism and helped guide aspects of its practical implementation, especially around preference reporting.