Volatility Analysis of Chinese Stock Market Using High-Frequency Financial Big Data

Abstract

With the recent development of computertechnology, the high frequency financial big data have beengenerated timely and more conveniently. However, theparticularity of high-frequency big data has raised a number ofmajor challenges for data analysis. The existing mathematicalmodels that were designed for analyzing daily financial data mayno longer be suitable for studying high-frequency big data. Totackle this challenge, this work explores the appropriate modelthat is able to analyze the high-frequency financial Big Data fromthe Shanghai composite index. For analyzing market volatility,we conduct three comparison studies for different mathematicalmodels. We first compare the effect of two types GARCH(generalized autoregressive conditional heteroskedasticity)models. Numerical results suggest that the volatility proxy modelhas a better effect than the model based on the return ofShanghai composite index. This study leads to the comparisonstudy of the GARCH(1,1) model and GJR(1,1) (Glosten-Jagannathan-Runkle) model. The result show that the GJR(1,1)model is more efficient than the GARCH(1,1) model. Finally weintroduce the ARMA model based on the GJR volatility proxymodel. Analysis results indicate that the ARMA(2,1)-GJRvolatility proxy model is the most effective one to study marketvolatility. The volatility persistence parameter is 0.952, which isvery close to 1. In addition, the p-value of the Ljung-Box test is0.729, which suggests that this model can not only correct theproblem of residual but also reflect the leverage effect and longmemory character of the Chinese stock market.