In matching problems with minimum and maximum type-specific quotas, there may not exist a stable (i.e., fair and non-wasteful) assignment (Ehlers et al., 2014). This paper investigates the structure of schools' priority rankings which guarantees stability. First, we show that there always exists a fair and non-wasteful assignment if for each type of students, schools have common priority rankings over a certain number of bottom students. Next, we show that the pairwise version of this condition characterizes the maximal domain of two schools' priority rankings over same type students to guarantee the existence of stable assignments. To prove the existence theorem, we propose a new mechanism Deferred Acceptance with Precedence Lists (DAPL), which is feasible, non-wasteful, strictly PL-fair and group strategy-proof for any priority rankings. Strict PL-fairness is weaker than fairness, but DAPL satisfies fairness under our sufficient condition. We also show that there is no strategy-proof mechanism that Pareto dominates DAPL whenever the outcome of DAPL is Pareto dominated by a stable assignment. * This is originally Chapter 3 of my Ph.D. thesis at Harvard University. I am very grateful to my advisors Eric Maskin, Scott Kominers and Edward Glaeser for their constant guidance. I am grateful to Lars

No abstract

We propose a new solution concept in the roommate problem, based on the "robustness" of deviations (i.e., blocking coalitions). We call a deviation from a matching robust up to depth k, if none of the deviators gets worse off than at the original matching after any sequence of at most k subsequent deviations. We say that a matching is stable against robust deviations (for short, SaRD) up to depth k, if there is no robust deviation up to depth k. As a smaller k imposes a stronger requirement for a matching to be SaRD, we investigate the existence of a matching that is SaRD with a minimal depth k. We constructively demonstrate that a SaRD matching always exists for k = 3, and establish sufficient conditions for k = 1 and 2.

This paper identifies a condition for an efficient social choice rule to be fully implementable when we take into account investment efficiency. To do so, we extend the standard implementation problem to include endogenous ex ante and ex post investments. In our problem, the social planner aims to achieve efficiency in every equilibrium of a dynamic game in which agents strategically make investments before and after playing the mechanism. Our main theorem shows that a novel condition commitment-proofness is sufficient and necessary for an efficient social choice rule to be implementable in subgame-perfect equilibria. The availability of ex post investments is crucial in our model: there is no social choice rule that is efficient and implementable in subgame-perfect equilibria without ex post investments. We also show that our positive result continues to hold in the incomplete information setting.

No abstract