This paper proposes a new autoregressive conditional volatility-skewness-kurtosis model: GJRSK-M model.Setting techniques about GJRSK-M model are discussed,such as model identification,confirming model' step,model estimation.We use this model to study the higher-moments volatility characteristics of China stock market and use out-sample-forecast technique to compare the forecast ability between GJRSK-M model and the other existing models.Research results show that just like conditional volatilit...