Abstract:
We study a setting with an uninformed decision maker who has to make a decision based on recommendations of two perfectly informed experts. The experts have opposing interests: one expert prefers the lowest feasible decision, whereas the other one prefers the highest feasible decision. We provide a tool to characterize optimal mechanisms in this environment, "constant-threat principle," which states that one can restrict attention to mechanisms in which the implemented decision after a disagreement among the experts is independent of their reports. We use this principle to characterize two classes of mechanisms: the cost-efficient mechanisms that implement the decision maker's most preferred alternative at the lowest cost and the optimal mechanisms that maximize the payoff of the decision maker net of payments to the experts.