Hiroshi Takahashi and Takao Terano
GARCH (Generalized Autoregressive Conditional Hetero- scedasticity) is a macro level model to estimate the volatility of financial markets. Although the model is very fundamental in the financial and economic domain, however, there have been no clear explanation about the model from the micro-level financial behaviors. This paper develops agent-based simulation models, which consists of simple agents with rational and/or non-rational decision making functionalities for investment. Using the simulation model with both rational and non-rational agents, the paper has shown that the behaviors of the non-rational agents with the characteristics of prospect theory coincide with the estimation by GARCH model.