Coordination, Efficiency and Policy Discretion

Would citizens coordinate to punish a government when they observe suspicious behavior? This paper shows that under some circumstances such coordination is impossible. This fact has important implications for policy discretion. We study an environment with the following characteristics: 1) the aggregate productivity (fundamental) is stochastic, 2) only the government observes it and, 3) every agent privately receives a noisy signal about the fundamental. Item 1) implies that the best policy (tax on investment) with commitment is state contingent, while 2) and 3) make information incomplete. The main consequence of incomplete information is to make coordination among small anonymous agents harder. Since coordination is key to punishing the government when it deviates, independently of the accuracy of the signal, the set of sustainable payoffs is drastically reduced. Regardless of the size of the noise, state contingent policies cannot be an equilibrium. Moreover, the best equilibrium policy is independent of the fundamental, i.e., the optimal policy calls for strong rules rather than discretion. Finally, we show that the payoff of the best equilibrium without commitment is uniformly bounded away from the payoff of the equilibrium with commitment.