Research On Herding Behavior In Securities Market Based On Computational Experiment

This paper studies the herding behavior in securities market, in which the investors making decision based not on their private information but on the influence of public opinions or other investors’ strategies. Through studying or imitating other investors, the investors make the same investment decision.Present research about herding behavior focused on the causes of the phenomenon. Empirical researches mainly apply statistical method to the testing of the investor behavior similarity. The main defect of statistical method is that it can’t distinguish genuine herding behavior from pseudo herding behavior.This paper applies the Agent based Computational Experimental method, which involves controllable and repeatable experiments based on computer technology, to the investors herding behavior in securities market from microcosmic angle. This method, which is new in this area, overcomes the defect of traditional mathematical statistical method which can’t distinguish genuine herding behavior from pseudo herding behavior.From the research, we find out that there exists obvious herding behavior in our artificial stock market and the investors’ increasing herding behavior level will promote the return and volatility level. Besides, the overall market characteristics and investors structure will affect the herding behavior level, which is just a intermediate variable affecting market return and volatility level. What really influences market return and volatility level is the market liquidity depth, the investors’ emphasis on their private information and the investors’ structure. Among our various strategies in our artificial stock market, none but the "momentum strategy " can beat the market.