This paper proposes a class of games called revision games . In a revision game, players start with initially prepared actions, followed by a sequence of random revision opportunities until a predetermined deadline. In the course of revisions, players monitor each other's behavior. It is shown that players can cooperate and that their behavior under the optimal equilibrium is described by a simple differential equation. We present the necessary and sufficient conditions for cooperation to be sustained in revision games. We also present applications to the preopening activities in the stock exchange and to an electoral campaign.

We construct a dynamic model of election campaigns. In the model, opportunities for candidates to refine/clarify their policy positions are limited and arrive stochastically along the course of the campaign until the predetermined election date. We show that this simple friction leads to rich and subtle campaign dynamics. We first demonstrate these effects in a series of canonical static models of elections that we extend to dynamic settings, including models with valence and a multidimensional policy space. We then present general principles that underlie the results from those models. In particular, we establish that candidates spend a long time using ambiguous language during the election campaign in equilibrium.