The Empirical Study Of Conditional Beta Pricing Model On Shanghai Stock Market

Since CAPM was proposed half century ago, scholars from various countries correct and develop the model, from nonconditional(static) models to conditional models. In the last decade, Chinese researchers have done some empirical study on our stock or fund market. Overall, their researches could be classified in two ways, which are modifying variables and changing estimation methods, and the conclusions also have some discrepancies. They have some argument on the choice of risk factors, while most of their studies on the beta coefficients are the significant test of the long-term average value.This paper, which made the deep study on Fama-French three-factor model, is based on a new two-stage nonparametric estimation method recently proposed by some foreign researchers, whose advantage is to relax the limitation of the model assumptions and to approach the real value of risk factors by nonparametric iteration methods. Not only the test was running, but also much detail analysis for market price of risk and their beta coefficients was mentioned in this paper.First of all, conditional betas are estimated nonparametrically for each asset and period using the time-series of previous data. In this step, we pick up four commonly used state variables based on other scholars' study, and use only two of them at a time. The pair of instrument variables with the best fit degree was choosed for the further study.Secondly, we estimated the time-varying risk factors from the cross section data of the returns and the obtained beta coefficients. Meanwhile, the optimal bandwidth was calculated by minimizing the penalized sum of squared residuals, which will be returned back to the iterative approach for beta coefficients to acquire more accurate results.And then, this paper provided some detailed analysis on the properties of market price of risk and beta coefficients estimated from the empirical study. In our conclusion, FF three-factor model is supported by Chinese stock market, where the existence of three factors is statistical significant, but not obvious at all time. Moreover, all the coefficients get smaller when the size factor of portfolio become bigger, but the relationship is different on the book-to-market factor, at least not the simple linear function proposed in some articles. However, the volatility of three betas change in the same way----they all become lower when the size and book-to-market factor get bigger. Some of their density curves show twin peaks with closed values. This could be explained by the inadequate data sample of our market.At last, this paper also studied the feature of betas in about45industries. The empirical results show that the market factor is statistical significant and exists in Chinese market with no doubt. Specially, values of betas are bigger in emergent industries, so do the volatilities. However, size factor and book-to-market factor are existed indeed, but they sometimes couldn't be observed or tested under the effect of industries or time zones the researcher choosed.