School Matching, Mechanism Design, Networks and Advice: An Experimental/Field Study

New York University, New York NY

Investigators

Abstract

In recent years a number of school systems have used various schemes to match students with schools. In these mechanisms students rank the schools they would like to attend and then are placed into schools by a process that takes these rankings into account along with the capacities of the schools and any special priorities granted to some students. The resulting matches may be good or bad depending on the behavior of the students (or their parents) in the matching process (i.e., the strategic dissembling they employ.) Theoretical work done to design matching mechanisms that mitigate (or eliminate) such strategic dissembling assumes the matching process occurs once and is not repeated again. As a practical matter, this isn't how school matching markets work. Instead, the matching process is repeated prior to the beginning of each school year. Parents who have participated in prior years may pass along advice to newcomers they know regarding how to behave in these markets. Over time groups of people will develop "traditional wisdom" as to how to behave. If this traditional wisdom leads people to behave optimally then the outcome of the match is likely to be consistent with the intentions of the mechanism designers. If, however, the traditional wisdom reinforces bad behavior, the resulting allocations may be sub-optimal. In the funded research the PI will conduct a series of experiments to examine the impact of intergenerational advice and network structure (who talks with who) on the performance of matching mechanisms. Matching markets occur is a variety of situations from matching students to schools to interns to hospitals. Many share the feature that the market reoccurs at specific intervals with information on "how to play" the game being passed down over time. Understanding how the resulting evolution of traditional wisdom and how the structure of connections among market participants (e.g., which parents talk to who) influences the outcome of matching market is crucial to designing matching mechanisms with desirable properties.