Best-Reply Matching in an Experience Good Model

Abstract

The paper studies a class of experience good models in a new way. We focus on signaling games close to Akerlof’s market for lemons, in which a seller sells a good to a buyer, who ignores the quality of the good during the transaction. In this context, we first establish some properties of the mixed Perfect Bayesian Equilibria. Then we turn to the concept of bestreply matching (BRM) developed by Droste, Kosfeld & Voorneveld (2002, 2003) for games in normal form. BRM equilibria respect a consistency which is different from the Nash equilibrium one: in a BRM equilibrium, the probability assigned by a player to a pure strategy is linked to the number of times the opponents play the strategies to which this pure strategy is a best reply. We extend this logic to signaling games in extensive form and apply the new obtained concept to our experience good models. This new concept leads to a very simple rule of behavior, which is consistent, different from the Perfect Bayesian Equilibrium behavior, different from Akerlof’s result, and can be socially efficient.