Abstract: In order to get the more accurate estimation of volatility of daily return series of Shanghai and Shenzhen Stock market with regime switching, volatility of these stock index return series are divided into three regime states: rising, falling and consoliclation in the paper. Return series of Shanghai Composite Index and Shenzhen Component Index are chosen as study sample and January 4, 2000 to 2011 December 30 is set as the sample period and January 4, 2012 to January 17, 2012 is set as out of sample period. Then GARCH model, RS-GARCH model, APGARCH model and RS-APGARCH model are applied to estimation and forecasting of volatility of these two return series. Finally MSE₁, MSE₂ and QLIKE are used to evaluate the performance of these models. The results show that APGARCH model is more accurate in estimation and prediction of the volatilities of the series than the GARCH model, models with Markov regime switching are more accurate in estimation and prediction of the volatilities of the series, and the models with normal error distribution are more accurate in estimation and prediction of the volatilities of the series than the models with the error distribution following t-distribution.

Key words: Markov regime switching, APGARCH model, volatility