Author(s)

Jialin Xu

Corresponding Author

Jialin Xu

ABSTRACT

In this paper, by using closing prices of CSI All Share Health Care Index on a daily basis from January 4, 2005 to November 29, 2019, constituting 3624 daily in-sample data points, the ARMA-GARCH model under different error distributions are employed to analyze and predict the index returns and prices. The empirical results demonstrate that the return rate of CSI All Share Health Care Index is volatility clustering, with non-normal characteristics of high kurtosis and fat tail, while no strong evidence of the existence of leverage effect has been found. Contrary to the expectation, although the ARMA(1,1)-GARCH(1,1) model under student’s t distribution shows the best fitting effect, it presents a poor short-term forecasting performance for the index. Based on the result, a recommendation can be made that it is necessary to further consider relevant factors and explore the combination of nonlinear models so as to improve the prediction accuracy.

KEYWORDS

ARMA-GARCH; forecasting performance; CSI All Share Health Care Index