Forecasting the Volatility of the Shanghai and the Shenzhen Stock Markets
预测沪深股市市场波动性

The objectives of this paper are twofold.Firstly,we evaluate the forecasting performance of GARCH((1,1),) TGARCH and EGARCH models of the Shanghai and the Shenzhen stock market volatility.Secondly,since not all investors assign equal weight to similar sized overpredictions and underpredictions,in addition to the conventional symmetric criteria considered,we also propose error statistics that designed to account for asymmetry in the loss function.Our results show that in terms of RMSE,MAE,MAPE and Theil-U criteria considered,TGARCH models perform the best in forecasting market volatility in both markets,while EGARCH models outperform regular GARCH(1,1) models.This suggests that "bad" news has a significantly stronger impact on market volatility than "good" news of the same size,and that non-linear GARCH models are able to explain the Chinese market volatility better than linear GARCH models.As for the 2 asymmetric criteria,our results show that in both markets EGARCH models outperform in terms of MME(U) criterion,which penalize underpredictions more heavily,while GARCH(1,1) models performs the best in terms of MME(O) when overpredictions are penalized more heavily.This suggests that investors should select the appropriate volatility forecasting models to tailor their individual requirements.