Herd Behavior in Artificial Stock Markets

Abstract

Herd behavior in Economics can be fruitfully represented by a generalization of the well-known Ehrenfest urn model to correlated clustering. The strategies of an agent in a stock market (planning to buy, to sell or to be inactive) are represented by three urns, and the accommodation of each agent in one of them is ruled by a random mechanism that may depend strongly on the behavior of the other agents. This mechanism is introduced in the “Genoa Artificial Stock Market” [14]. At each step, each agent chooses its strategy following the Ehrenfest-Brillouin model [10]. Given the old price, the demands of bulls and the supply of bears intersect, and generate the new price. Fat tails of price returns are obtained directly as a function of the herding parameter, without introducing any individual distinction among agents (like initial wealth, or risk-propensity). Time correlation is driven by a simple mechanism, and volatility clusters can be introduced as temporal variations of the herding parameter. The relationship between the active strategy excess (the difference “bulls minus bears”) and the price returns is studied.