Testing the rational expectations hypothesis with agent-based models of stock markets

Using agent-based models of stock markets, this paper examines the rational expectations hypothesis from a bottom-up perspective. We apply standard linear and nonlinear econometric tests to artificial time series generated from two artificial stock markets composed of bounded-rational traders. The two artificial stock markets differs in their architectures: one has a business school, and one does not. While the linear test shows that the market with the business school fails to reject the rational expectations hypothesis quite often, the nonlinear one does not. Therefore, strictly speaking, these two agent-based markets of bounded-rational traders do not collectively behave as what rational expectations hypothesis predicts, and hence do not lend support to rational expectations hypothesis.