This is collaborative research with Stefan Reichelstein at Stanford University (Grant 8704065). The research concerns the design of efficient organizational schemes. In particular, it considers the comparative advantages of hierarchical arrangements. The authors posit a group of agents who are to take actions in a changing environment. The joint action has to meet a performance standard, i.e., it has to lie in a set that depends on the environment. Each member observes some component of the actual environment and chooses her action after receiving messages from others. A network mechanism prescribes messages sent and actions taken and may be hierarchical. A network mechanism is efficient if it satisfies the performance standard and no other mechanism does so at a smaller informational cost. Cost is measured by the size of the set of messages that agents can send or receive. General techniques for constructing efficient mechanisms have been applied to a multi-plant firm example and an exchange-economy example. The broader significance of this work is that it seeks to generalize these results to find those basic properties of performance standards that make hierarchies efficient. In the same context the researchers study: (1) dynamic issues, i.e., the number of steps until actions are obtained; (2) reward structures, i.e., contracts among the group's agents, so that hierarchies can be examined in light of the incentive problems; and (3) promising links to the "communication complexity" literature in computer science.
