Feng Yongfu a, Hua Xia b and Gao Jinkang c

　　a and b School of Finance, Southwestern University of Finance and Economics

　　c China Research Center of Financial Law, Southwestern University of Finance and Economics

　　Abstract：This study investigates the basic numeric characteristics of Chinese A-share market index volatility (i.e., the clustering, heteroscedasticity, and jumps) from the perspective of data mining. It presents a theoretical-empirical model based on these three major characteristics and conducts a maximum likelihood estimation to study Chinese A-share market return data empirically. Results show that, in full-sample or special periods, this model calibrates the A-share index volatility well and simulates in-sample volatility better than the four major empirical models adopted to study volatility. In out-of-sample forecasts, this model performs better than the other four models on the value-at-risk dates, which are the volatile days. This model can also decompose and explain the volatility of the Chinese A-share index. On the basis of GARCH, this study revises the volatility model proposed by Maheu and extends Engle’s research framework. Thus, this model is of theoretical significance. This model’s simulation and forecast functioning can contribute to regulatory expectation management and investment portfolio construction.

　　Keywords: index, volatility model, calibration forecast