An Experimental Study of Strategic Complementarity, Asynchronicity and Mechanism Design

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

This program of research in experimental economics incorporates bounded rationality and learning into the mechanism design framework. The project has two parts. The first consists of three experimental studies of incentive-compatible public goods mechanisms. The presence of public goods seriously challenges traditional or "natural" solutions for the allocation of private goods. Incentive-compatible mechanisms align individual self-interests with the group interest. The investigator's previous work and others' work have shown that among these incentive compatible mechanisms those which are supermodular, i.e., when players' strategies are strategic complements, converge robustly well to the Nash equilibrium prediction, and those which are not supermodular do not converge as well. This project examines whether the degree of strategic complementarity could have a significant effect on the convergence of a mechanism, i.e., whether the process is gradual. Experiments are conducted on three supermodular mechanisms for public goods provision with varying degrees of strategic complementarity to study this effect. Convergence is an important criterion for selecting a mechanism to solve a particular problem because at the end of the learning process we would like the mechanism to yield the desired outcome. The second part consists of two experimental studies of mechanism design in distributed systems, such as the Internet, which are characterized by asynchronicity and limited information. In a wide variety of real world situations a group of agents share a common production process transforming input into output. A salient example is provided by the Internet, which has becoming increasingly important in global telecommunications. In the context of several agents sharing a network link, the cost to be shared is congestion experienced. In the network context, traditional solution concepts such as Nash equilibrium or the serially undominated set might not apply. Friedman and Shenker (1998) propose new solution concepts for distributed systems. This project runs experiments to examine these new solution concepts. The project also conducts experiments to study the average cost pricing and the serial cost sharing mechanisms under complete information and synchronous moves, as well as distributed settings with limited information and varying degrees of asynchrony. This study will provide information on what types of mechanisms should be selected in distributed systems.