A Remark on Infinitely Repeated Extensive Games

Let Gamma be a game in extensive form and G be its reduced normal form game. Let Gamma ^infinity (delta) and G^infinity (delta) be the infinitely repeated game version of Gamma and G respectively, with common discount factor delta. This note points out that the set of SPE payoff vectors of Gamma^infinity (delta) might be different from that of G sub infinity (delta), even when delta is arbitrarily close to 1. This difference can be substantial when G fails to satisfy the "dimensionality" condition (a-la Fundenberg and Masking (1986) or Abreu, Dutta and Smith (1992)).