A dynamic model of equilibrium selection in signaling markets

In his work on market signaling, Spence proposed a dynamic model of a signaling market in which a buyer revises prices in light of experience and sellers choose utility-maximizing signals given these prices. Spence also suggested that subjecting the dynamic process to rare perturbations might allow one to choose between multiple equilibria. This paper examines the effect of introducing such perturbations into Spence's dynamic model. We find that refinement results arise naturally from the dynamic analysis. In a broad class of markets, our model selects a separating equilibrium outcome if and only if the equilibrium outcome satisfies a version of the undefeated equilibrium concept, whereas a pooling equilibrium outcome is selected if and only if the equilibrium outcome is both undefeated and satisfies D1.