We use experiments to study the voting behavior of two groups with different material incentives in an environment with uncertain state of the world. One of the groups has information about the state of the world, whereas the other does not. Before voting individuals can communicate. We study four different communication protocols which we hypothesize to affect group identities and, hence, preferences. We find that unrestricted communication yields the highest efficiency, whereas control of the communication process by the informed group yields the lowest efficiency. Communication mainly favors the informed group. Lying and use of disrespectful language decreases efficiency.

(joint with Jordi Brandts and Lydia Mechtenberg)