Performance of Selected Models with Heterogeneous Expectation Formation

The Efficient Market Hypothesis (EMH) asserts that the prices of securities correctly and fully reflect all available information. But there are some empirical facts in capital markets thatEMHis not able to explain. Recently, a lot of new models with heterogeneous agents in expectation formation of future asset prices have arisen. The main aim of this paper is to compare three of them - model introduced by C. Chiarella in 1992, where the equilibrium price is determined by the excess demands of different groups of agents, model developed by W. Brock and C. Hommes with additional variable of proportion of agents with different strategies and further extended D. Goldbaum's model. Statistical properties of the time series of S&P 500 index are compared with the price time series generated by these models. This can be done due to chaotic price behaviour, which occurs for partially estimated, partially selected model values of parameters in these nonlinear dynamical systems. The results show that the third model achieves the best performance. This model is the most complex one and thus the most closely to reality but at the expense of mathematical tractability.