Complexity and Organizational Architecture

This paper revisits the literature on modelling organizations by means of networks of agents. Individual agents are engaged in screening projects, and architectural features of organizations, that is how each agent’s decision combines with those of others, a®ect the organization’s screening performance. It emphasizes how an organization of several agents may be improve upon individual performance by a suitable arrangement of the flow of decisions. The paper is motivated, in part, by a theorem due to Von Neumann, Moore and Shannon on how to build reliable networks using unreliable components and extends previous contributions by Sah and Stiglitz by recasting their original model in standard firm-theoretic terms and endogenizing its features. For an organization’s screening performance to improve over those of an individual’s, it must be sigmoid in individual performance, as measured by the probability that a good (bad) project be accepted (rejected). This is indeed the case for organizations with mixed Sah-stiglitz architectures, such as hierarchies made up of components that are polyarchies, and polyarchies made up of components that are hierarchies, give rise to such functions. This property is in turn critical for determining of the optimal number of levels of a hierarchy, and for endogenizing individual screening performance. The models are extended to allow for individuals’ own screening to be influenced from the opinions of superiors and subordinates. The paper examines the implications of such interactions for the limits to organizational performance.government to influence the real value of assets using fiscal and monetary policy.