On the Coordination and Adaptability of the Large Economy: An Application of Genetic Programming to the Cobweb Model

摘要

Modeling the cobweb model as a multiagent system, this paper analyzes the coordination and adaptability of markets composed of 500 and 100 GP (Genetic Programming)-based learning agents. By adaptability, we mean that the market can protect itself from persistent large deviations from the rational-expectations equilibrium price. The simulations conducted in this study show that the adaptability of an economy depends crucially on the size of the economy characterized by the number of firms and that a large economy is more adaptable. Since it is more difficult to coordinate beliefs and to form the representative agent in a large economy, its dynamics may not be quasi-periodic and hence it will always be able to make improvements. On the other hand, not being quasi-periodic also means that the economy will not be "stable" in the sense that the pattern of prices might be very random and volatile. The inverse relation between "stability" and "adaptability" is thus observed. According to our simulations, the large economy does not learn to be "stable" but learn instead to be more adaptable.