Professeur invité Robert Hammond (North Carolina State University)
In collaboration with Umut Dur (North Carolina State University) and Onur Kesten (Tepper School of Business).
We analyze sequential preference submission in centralized matching problems such as school choice. Our motivation is school districts and colleges that use an application website where students submit their preferences over schools sequentially, after learning information about previous submissions. Comparing the widely used Boston Mechanism (BM) to the celebrated student-proposing Deferred Acceptance (DA) mechanism, we show that a sequential implementation of BM is more efficient than a sequential implementation of DA under a natural equilibrium refinement. For any problem, any equilibrium outcome under BM (weakly) Pareto dominates the student optimal stable matching (SOSM). These gains occur because sequentiality serves as a coordination device. We present two sets of empirical tests. first, we study a field setting in which sequential BM was used in practice. The field data provide suggestive evidence that is consistent with our theory. Second, we conduct a laboratory experiment to compare the sequential mechanisms in a controlled environment. We find that BM Pareto improves upon DA when students submit sequentially but not when students submit simultaneously. We conclude that sequential preference submission allows students to overcome the coordination problem in school choice.
Keywords: School choice, student assignment, matching theory, sequential-move games
JEL classiﬁcation: C78, D61, D78, I20