A Note on the Equivalence of Rationalizability Concepts in Generalized Nice Games

Moulin (1984) describes the class of nice games for which the solution concept of point-rationalizability coincides with iterated elimination of strictly dominated strategies. As a consequence nice games have the desirable property that all rationalizability concepts determine the same strategic solution. However, nice games are characterized by rather strong assumptions. For example, only single-valued best responses are admitted and the individual strategy sets have to be convex and compact subsets of the real line ℝ. This note shows that equivalence of all rationalizability concepts can be extended to multi-valued best response correspondences. The surprising finding is that equivalence does not hold for individual strategy sets that are compact and convex subsets of ℝn with n ≥ 2.