Agent-based simulation of a financial market

This paper introduces an agent-based artificial financial market in which h eterogeneous agents trade one single asset through a realistic trading mech anism for price formation. Agents are initially endowed with a finite amoun t of cash and a given finite portfolio of assets. There is no money-creatio n process; the total available cash is conserved in time. In each period, a gents make random buy and sell decisions that are constrained by available resources, subject to clustering, and dependent on the volatility of previo us periods. The model proposed herein is able to reproduce the leptokurtic shape of the probability density of log price returns and the clustering of volatility. Implemented using extreme programming and object-oriented tech nology, the simulator is a flexible computational experimental facility tha t can find applications in both academic and industrial research projects. (C) 2001 Elsevier Science B.V. All rights reserved.