Skew GARCH Model and Its Application in Chinese Stock Market

Volatility models such as GARCH with normal, Student-t or GED conditional error distributions are proved to be not good enough to Capture the skewness, leptokurtosis and volatility clustering of financial asset return time series of emerging equity markets. Based on the research of Branco et al. (2001)([1]), Sahu et al. (2003)([2]), and Azzalini et al, (2003)([3]), GARCH models with skew normal and skew t error distributions are put forward in this paper. Using daily A share aggregated market returns of Shanghai and Shenzhen stock markets in China over the period from 2 January 1996 to 31 December 2005, empirical results prove that GARCH-St model fits data best for these two markets and return series of Shanghai stock market is more non-normal comparatively. The GARCH-St appears to be a promising specification to accommodate high peakedness and thick tails in data series characterized by skewness and volatility clustering.