Individuals and Institutions in Strategic Network Formation

Publication Type


Publication Date



Social and economic agents are frequently presented with opportunities to form networks for mutual benefit. This paper examines the prototypical decision problem that arises in such situations, namely which other agents to link to. The problem may be difficult because the relationship between a network's structure, its value, and the share of value captured by each agent may be complex. Moreover, agents' incentives may conflict, causing network-level outcomes to be inefficient, unstable, or both. These tensions are explored in a computational model where agents of varying capabilities engage in link formation under a range of stylized conditions motivated by the example of a supply chain. Agents choose a set of links to for (i.e., upstream suppliers to purchase from) in each period. Each agent's per-period payoff depends on both its inbound and outbound links (active customers and suppliers). These payoffs are assigned randomly, in a variation on Kauffman's NK family of models, to simulate arbitrary complementarities among combinations of links. The complexity of the decision problem faced by the agents is "tuned" by a parameter K which specifies the maximum number of upstream and downstream neighbors per agent. Simulation experiments show that, as predicted by theory, agents often become locked into suboptimal network configurations or cycles of mutual frustration. However, outcomes are improved by simple localized institutions such as bilateral bargaining, which enables even severely myopic agents to rapidly achieve configurations that are stable and close to efficient. The Coase Theorem provides an interpretation of the remaining inefficiency: it is the additional surplus that could be realized through more complex and potentially more costly coordination mechanisms.


Computer Sciences | Management Information Systems

Research Areas

Information Systems and Management


5th International Conference on Complex Systems (ICCS 2004), May 16-21, Boston, MA

City or Country

Boston, MA

Additional URL


This document is currently not available here.