A Framework for Studying Economic Interactions

Most economic models implicitly or explicitly assume that interactions between economic agents are ‘global’ in other words, each agent interacts in a uniform manner with every other agent. However, localized interactions between microeconomic agents are a pervasive feature of reality. What are the implications of more limited interaction? One set of mathematical tools which appears useful in exploring the economic implications of local interactions is the theory of interacting particle systems. Unfortunately, the extant theory mainly addresses the long-time behavior of infinite systems, and focuses on the issue of ergodicity; many economic applications involve a finite number of agents and are concerned with other issues, such as the extent of shock amplification. In this paper, I introduce a framework for studying local interactions that is applicable to a wide class of games. In this framework, agents receive shocks which are stochastically independent; payoffs depend both upon the shocks and the strategies of other agents. In finite games, ergodicity is straightforward to determine. In finite games which evolve in continuous time, the stationary distribution (if it exists) may be computed easily; furthermore, in this class of games, I prove that any stationary distribution may be attained by suitable choice of payoff functions using shocks which are distributed uniform on (0, 1). In systems in which all interactions are global, I prove that nonlinear behavior can arise even in the infinite limit (thus demonstrating that laws of large numbers can fail in systems characterized by interaction), despite the fact that the only driving forces are agent-level iid disturbances. Using numerical methods, I investigate the properties of the processes as one passes from discrete to continuous time, as one alters the pattern of interaction, and as one increases the number of interacting agents. In so doing, I provide further evidence that the existence of local interactions can change the aggregate behavior of an economic system in fundamental ways, and that the form of that interaction has important implications for its dynamic properties. ________________________ A previous version of this paper circulated under the title, "Interaction in finite economies." Thanks to the Santa Fe Institute for its gracious hospitality during the 1992 Complex Systems Summer School, where I initially began thinking about these models; also, thanks to Russell Cooper, Boyan Jovanovic, Michael T.K. Horvath, seminar participants at Virginia Tech, the 1996 SEDC conference in Mexico City, and the 1998 NBER macroeconomic complementarities workshop. I am entirely responsible for all errors. A Framework for Studying Economic Interactions 2