Probabilistic Voting and Accountability in Elections with Uncertain Policy Constraints

We consider accountability in repeated elections with two long‐lived parties that have distinct policy preferences and different levels of valence. In each period the government faces a privately observed feasibility constraint and selects a publicly observed policy vector. While pure strategy equilibria do not exhibit tight control on government policy making, complete control is possible in mixed strategies. In optimal equilibria voters use reelection functions which depend on policy in a manner that causes the governing party to internalize voter preferences. In these optimal equilibria the voters use different reelection functions for different parties.