The Design of Ambiguous Mechanisms

This paper considers the optimal mechanism design problem of an expected revenue maximizing principal who wants to sell a single unit of a good to an agent who is ambiguity averse in the sense of Gilboa and Schmeidler (1989). We show that the optimal static mechanism is an ambiguous mechanism. An ambiguous mechanism specifies a message space and a set of outcome functions. After showing that (a version of) the Revelation Principle holds in our environment, we give an exact characterization of the (smallest) optimal ambiguous mechanism. If the type set is composed of N (finite) types, then the (smallest) optimal ambiguous mechanism contains N - 1 outcome functions. We show that the share of the surplus that the designer can extract from the agent increases as the type set becomes larger and the probability of each single type decreases. In the limiting case where the agent's type is drawn from a non-atomic distribution on an interval, the optimal ambiguous mechanism extracts all the rent from the agent.