The Stable Long-run Capm And Nonparametric Method Estimation

This thesis is mainly concerned with statistical inferences and modifying of classical CAPM based on varying-coefficient partially linear models and Counterfactual semi-parametric model.Varying-coefficient partially linear models are useful combination and extension of classical linear models. In practice, the difficult point is to estimate the varying function coefficients and the inferences on them. In this thesis, we study the estimations of varying-coefficients and parametric component using profile least-squares technique. The main focus is to possess the generalized likelihood statistics developed by Fan et al. (2001) for the inferences on the varying-coefficient and parametric component. Finally, a result of a Monte Carlo experiment is presented to examine the finite sample performances of the proposed procedures and a power compare of proposed factor bootstrap and wild-bootstrap is discussed.This thesis proposed that price relationship between an asset of interest and the bench market portfolio into account, and proposed Long-run function coefficient CAPM. In the positive analysis, we get the conclusions as follows:we proposed a new method to perform the information of asset price using entropy, and we found that the information of price has a negative impact on asset return. The return can capture the characteristic of multi-peak after considering the information of price. The market beta is a time-varying and influence by information of price. Volatility has a nonlinear impact on return. The conventional CAPM is based on excess returns and contains no information of stock prices. Note that when two stock prices have a common stochastic trend and build a long-run equilibrium, we can use this information to improve the performance of the model. The long-run beta has an negative correction on return and so on.This paper provides a non-parametric method for solving the information of price indicator from a new perspective, and proposes long-run function coefficient CAPM, then tests predictive power of liquidity to returns from the Long-run CAPM and the return distribution perspective. They are applied to China's stock market research, the empirical test verifies the points that raised by the assumption that the parameters of market risk is time-varying, volatility has a non-linear impact on the income, and the price's common trend information exists on the adjustment of return and so on.