Multivariate Dynamic Analysis Of China 's Stock Returns

The return of securities has been a core topic for both researchers and practitioners for decades. People continuously pursuit the underlying mechanism for assets pricing and try to model the return by considering macro-economic environment, companies’ financial data, historical return, and many other factors.This study first focus on the applicability of Fama-French model in China stock market after the split share structure reform. Empirical study shows the good performance of FF model in Shanghai Stock Exchange(SHSE) while both size factor and value premium turn out to be significant. Furthermore, this paper considers risk effect and liquidity effect in a similar framework and proposes a modified “4-factors model”. The new model incorporate liquidity factor and possesses higher adjusted R square than Fama-French model.In addition, this paper test the robustness of this new model by modelling the average return of 25 portfolio with the consideration of stochastic volatility and time-varying coefficient. Empirical result indicates the time-varying effect of liquidity factor for those stocks with highest/lowest market size or book-to-market ratio. Furthermore, the static regression tend to underestimate the coefficient because it fails to consider volatility clustering. The time varying parameter regression with stochastic volatility(TVPR-SV) employed in this paper successfully detect high volatility period, 2007 to 2008 and end of 2014, which coincide with the period of bull market and bear market.To sum up, this paper successfully improve the 3-factors model by introduce the liquidity effect. Besides the better interpretability, the proposed dynamic 4-factors pricing model also detects the time varying effect of factors and volatility clustering phenomenon. These findings validate the employment of stochastic volatility model and shed some light for other scholars and practitioners in related area.