Econophysics and Agent-Based Modeling of Financial Market

Abstract

The availability of large amount of financial data has led to the discovery of a number of universal features found in financial time series, including fat-tail distributions and long memory. Such universality has attracted the attention of physicists to unravel the underlying mechanisms in market dynamics. Starting from certain empirical behaviors of agents in financial market, this work presents an agent-based model capable of quantitatively reproducing those two universal features. Explicit behavioral origin of such features is demonstrated ? fat tail arises as technical traders converge in opinions; long memory is due to the heterogeneous investment horizons of agents. The apparent criticality similar to phase transition in physics is also explained through a one-to-one correspondence. The stochastic volatility model extended from the agent-based model provides a behavioral interpretation to the general ARCH formulation, and creates a new way of calibrating stochastic volatility models through empirical behaviors instead of statistical estimation.