A herding perspective on global games and multiplicity

Recently, it has been claimed that full-information multiple equilibria in games with strategic complementarities are not robust, because generalizing to allow slightly heterogeneous information implies uniqueness. This paper argues that this "global games" uniqueness result is itself not robust. If we generalize by allowing most agents to observe a few previous actions before choosing, instead of forcing players to move exactly simultaneously, then multiplicity of outcomes is restored. Only a small sample of observations is needed to make our herding equilibrium behave like a full-information sunspot equilibrium instead of a global games equilibrium