Secure implementation in Shapley-Scarf housing markets (Q639889)

This paper studies the object allocation problem [\textit{L. Shapley} and \textit{H. Scarf}, ``On cores and indivisibility'', J. Math. Econ. 1, 23--37 (1974; Zbl 0281.90014)] with strict preferences. There is a group of agents, each of whom initially owns one object. A (single-valued) solution reallocates the objects with the condition that each agent consumes one and only one object. The strict core is a central solution in the literature on the object allocation problem. The strict core is dominant strategy implementable [\textit{H. Mizukami} and \textit{T. Wakayama}, ``Dominant strategy implementation in economic environments'', Games Econ. Behav. 60, No. 2, 307--325 (2007; Zbl 1134.91539)] and Nash implementable when there are at least three agents [\textit{T. Sönmez}, ``Implementation in generalized matching problems'', J. Math. Econ. 26, No. 4, 429--439 (1996; Zbl 0876.90012)]. However, the strict core is not securely implementable [\textit{Saijo, T. Sjöström} and \textit{T. Yamato}, ``Secure implementation'', Theor. Econ. 2, 203--229 (2007)]. Secure implementation means that there is a mechanism that induces both dominant strategy implementation and Nash implementation. The paper identifies solutions that can be securely implemented in the object allocation problem. For the two-agent case the authors provide a complete characterization of securely implementable solutions; a solution is securely implementable if and only if it is either a constant solution or a ``serial dictatorship'' (Theorem 1). A serial dictatorship means that one agent chooses her best object from among the set of objects, then the second agent chooses his best object from among the set of remaining objects, then the third agent chooses, and so on; the order in which agents make their choices is fixed in advance. In the general case (with more than two agents), the authors pin down smaller classes of securely implementable solutions by adding some properties. First, the ``no-trade solution'' is the unique securely implementable one that satisfies individual rationality (Theorem 2). The no-trade solution is the one that selects the initial endowments for each preference profile. Second, a securely implementable solution satisfies neutrality (symmetric treatment of objects) if and only if it is a serial dictatorship (Theorem 3). Finally, an efficient solution is securely implementable if and only if it is a ``sequential dictatorship'' (Theorem 4). Sequential dictatorships generalize serial dictatorships. In a sequential dictatorship there is a fixed first dictator for every preference profile. The second dictator, who chooses his best object from among the set of remaining objects, is decided by the choice of the first dictator. Similarly, the third dictator is decided by the choices of the previous agents, and so on.