Excess Volatility and Herding in an Artificial Financial Market: Analytical Approach and Estimation

Several agent-based models have been proposed in the economic literature to explain the key stylized facts of financial data: heteroscedasticity, fat tails of returns and long-range dependence of volatility. Agentbased models view these empirical regularities as emerging properties of interacting groups of boundedly rational agents in financial markets. The complexity of these interacting agent models has largely constrained their analytical treatment, limiting their analysis mainly to Monte Carlo simulations. In order to overcome this limitation, we introduce a ‘minimalist’ model of an artificial financial market, along the lines of our previous contributions, based on herding behavior among two types of traders. The simplicity of the model allows for an almost complete analytical characterization of both conditional and unconditional statistical properties of prices and returns. Moreover, the underlying parameters of the model can be estimated directly, which permits an assessment of its goodness-of-fit for empirical data. While the performance of the model for domestic stock markets has been the focus of a previous contribution, in this paper we report results for selected exchange rates against the US dollar.