Agent-based Modeling Financial Markets, The Complexity Of Research

Lots of empirical studies show that there are many "stylized facts" in financial market, such as non-Gaussian distribution of return, excess volatility and clustered volatility etc. These stylized facts indicate that financial price fluctuations have the very complex behaviors, rather than that simple as be described by efficient market theory. In order to explore the underlying mechanism of these complexities as well as to perfect the financial theory, a great many researchers dedicating themselves to investigate rules of capital markets recently. With the development of the computer technology, complex system theory and modeling technology, agent based modeling of financial market has become one of important method to explore the financial complexity.This paper takes the bounded rationality and herding as the main assumption and constructs a microcosmic market model by agent based method, in which every agent have its own investment sentiment. Agent make decision according to the sentiment which influenced by individual information and the sentiment of their friends. We investigate the influence of herding behavior and topology structure of network on market volatility by numerical simulation experiment, and find that when agent make decision dependently, price converges to fundamental value and the market is efficient, when investors' sentiment interact with each other by word-of-Mouth communication, there is synchronization effects and the market present excess volatility, and for its "long range connections", small-world network has stronger synchronization ability then regular network, so sentiment contagion in small world network can lead to the bigger fluctuation.Furthermore, we improve our model by adding the research results of attention shift which is discovered by Sheller into the model. In extended model, agent's attention level varies with the change of price instead of keeping fixed. Experiment outcome indicate that extended model can not only produces excess volatility, but also emerges most of stylized facts, such as bubbles, crashes, fat tail distribution of return and clustered volatility etc. These volatility features were compared to the ShangHai index quantitatively, and good agreement is found between them. At the same time, we prove from the theories that these stylized facts are robust to the system size, and won't disappear while the number of agents in the system increasing.Through the above research, we illustrate how the existence of herd behavior among market participants may generically lead to excess volatility, as well as how the stylized facts such as fat tail distribution of return and clustered volatility caused by participants' attention shift, and so give the microcosmic mechanism of these macro volatility features of financial market.