INTRODUCTION

The mechanism design literature of the last thirty years has been a big success on a number of different levels. A beautiful theoretical literature has shown how a wide range of institutional design questions can be formally posed as mechanism design problems with a common structure. We can understand institutions as solutions to well-defined maximization problems subject to incentive constraints. Elegant characterizations of optimal mechanisms have been obtained. Market design has become more important in many economic arenas both because of newinsights from theory and developments in information technology. A very successful econometric literature has tested auction theory in practice.

The basic issue in mechanism design is how to truthfully elicit private and decentralized information in order to achieve some private or social objective. The task of the principal is then to design a game of incomplete information in which the agents have indeed an incentive to reveal the information. The optimal design depends on the common prior, which the principal and the agents share about the types of the agents. Unfortunately, the general theory, the applications and the empirical work have rather different natural starting points. The theoretical analysis begins with a given common prior, often over a small set of types, and then analyzes the optimal mechanism with respect to this prior. Yet, the fine details of the specified environment incorporated in the common prior are rarely available to the designer in practice.