Characterization of Robustly Optimal Mechanisms Beyond Knowledge-Based Mechanisms

Characterize the robustly optimal mechanism when knowledge-based mechanisms are suboptimal, extending beyond the specific improvement argument provided for selected environments.

Background

The paper establishes conditions under which knowledge-based mechanisms—which condition only on the Bayesian component of an agent’s private information and ignore the ambiguous component—are robustly optimal under a maxmin criterion. It also proves a partial converse showing that, in certain environments, failure of the common-deviation condition permits profitable screening of the ambiguous component.

The authors note that their argument for improving upon knowledge-based mechanisms when those mechanisms are suboptimal applies only to specific environments and may not achieve the global optimum. They therefore leave open the general characterization of robustly optimal mechanisms outside the knowledge-based class.