Sequential Elections with Limited Information

We develop theoretically and test experimentally a one dimensional model of two candidate competition with incomplete information. We consider a sequence of elections in which the same general issue predominates from election to election, but where the voters have no contemoporaneous information about the policy positions adopted by the candidates, and where the candidates have no contemporaneous information about the preferences of the voters. Instead, participants have access only to contemporaneous endorsement data of an interest group, and to historical policy positions of the previous winning candidates. We define a stationary rational expectations equilibrium to the resulting (repeated) game of incomplete information, and show that in equilibrium, all participants, voters and candidates alike, end up acting as if they had complete information: Voters end up voting for the correct candidate, and candidates end up converging to the median voter.