An experimental study of strategic information transmission

We examine strategic information transmission in an experiment. Senders are privately informed about a state. They send messages to Receivers, who choose actions resulting in payoffs to Senders and Receivers. The payoffs depend on the action and the state. We vary the degree to which the Receivers' and the Senders' preferences diverge. We examine the relationship between the Senders' messages and the true state as well as that between actions and the true state and contrast the ability of different equilibrium message sets to explain the data. When preferences are closely aligned Senders disclose more. We assess two comparative statics: (i) as preferences diverge, state and action are less frequently matched, and (ii) messages tend to become less informative as preferences diverge. The first result is weakly confirmed for adjacent treatments but is considerably stronger when non-adjacent treatments are compared. We find that as preferences diverge messages become less informative. While the ex-ante Pareto-optimal Bayesian Nash Equilibrium does not explain our conditions, the equilibrium message sets supported by the data are similar to the ex-ante Pareto Optimal message sets.