Coordination and Decision-Making in a Market with Uncertainty

Research comparing centralized and decentralized coordination schemes is often contradictory. For example, Oskar Lange (1969) among others points out that the centralized coordination scheme can outperform decentralized schemes while Arrow and Debreu (1954) show otherwise. This paper focuses on this issue in the context of agent-based marketplaces. Specifically, we use a computational approach to study the trade off between imperfect decision-making by the central coordinating agent (the central authority) and the imperfect coordination among decentralized seller agents. Using social welfare as a metric, we study how the correlation in the quality of the seller agents, the marginal costs of product building, decision-making costs and the fraction of consumer utility transferred as compensation to the winning seller agent affect the terms of this trade-off. We find that the decentralized scheme with its parsimonious use of information and simplistic decision rule does very well in comparison to the centralized scheme, which internalizes the externalities. This is surprising since one may expect the centralized scheme to always perform better than the decentralized scheme. This paper analyzes the results and provides intuition for this apparent anomaly. Acknowledgements: This work was funded in part by NSF CISE/IIS/KDI 9873005. Authors would like to thank Amy Greenwald of Brown University for her valuable comments