Research On Herd Behavior In Stock Market Based On Multi- Agent Modeling

Finacial market is a complex adaptive system, because investors involved in the system, do not have complete information. Therefore, investors investment strategies are affected by each other, which affects the operation of the whole market. Also, investors’investing strategies are affected by the market. Thus, investors’different interactions of the information that they can get from the market, affects the formation of market price, and the price can affects the investors’strategies in turn. The aggregation, differentiation and response of the investors in the individual level emerge the complexity of the entire financial market.The traditional equation-based modeling methods are difficult to describe the variety of the types of investors, while the Agent-based modeling (ABM) methods based on complex adaptive system (CAS) can. Agent-based modeling can better reflect the investment behavior and the complexity of financial market. Also, the ABM can describe the dynamic, evolutionary and nonlinear characteristics which is different from the traditional equation based modeling methods.Till now, the researches on artificial stock market pay much attention to the micro-structure of the financial market and the learning ability of Agents. But few are focused on the study of herd behavior. In this paper, we build an artificial stock market, in which the investment strategies of agents are made on the basis of the information from the market and other agents. The innovation of this paper lie in the following aspects:Firstly, we abstract the herd behavior of the financial market, and establish an artificial stock market for the research of herd behavior using the agent-based modeling method. Secondly, the influence of the degree of herd behavior is studied through the description and adjustment of the Agents’confidence. Finally, to avoid the aggregation of all Agents’behavior, the scale that Agents can imitate is defined.The research results show that, the model constructed in this paper can significantly reflects the existence of herd behavior, which indicates that the imitation of heterogeneous Agents is the mechanism that herd behavior happens. There is a strong correlation between the volatility of the stock return and the herd behavior. The results show that Agents’herd behavior can increase the volatility and profit of the stock market. There is a positive relation between buying herd behavior and return, while negative relation between selling herd behavior and return. As Agents become more confident, the herd behavior will first increase to a maximum value then decrease, so over-confidence and self-abasement will reduce the herd behavior.