Functional Coefficients Regression Model for Security Pricing with Heterogeneous Beliefs: Application in Chinese Stock Market

A bivariate functional coefficients regression model for characterizing dynamics of security price is presented in this paper to reflect the feature of real financial market that is presence of repelling and mingling of different opinions amongst traders. The local polynomial regression approach is adopted to estimate the functional coefficients resulting from bayesian updating mechanism. After applying this model to Chinese stock market, three effects from heterogeneous beliefs are filtered , which could help investors to analyze the structure of market. Moreover, it is also tested that this model outperforms the constant coefficients regression model for out-of-smaple one-step ahead price prediction.