杠杆效应下ARG模型的波动率预测与风险度量研究

Research on Volatility Prediction and Risk Measurement of ARG Model Based on Leverage Effect

基于EGARCH模型能够较好捕捉波动率的杠杆效应和ARFIMA-Realized GARCH模型在高频波动率预测上的优势构建了ARFIMA-Realized EGARCH模型,并在资产收益率新息序列服从偏t分布的假设条件下进行波动预测和风险度量。数值模拟结果显示,ARFIMA-Realized EGARCH模型有更好的参数估计效果。选取上证综合指数和深圳成分指数2011年1月4日至2020年12月31日的5分钟高频数据对模型进行验证,结果表明:沪深股市的资产收益率存在明显的尖峰厚尾特征,且具有高度的自相关性;在加入了杠杆函数之后,模型面对负向冲击表现出更大的波动性,且杠杆参数均在1%的水平上显著。VaR预测结果图、后验测试以及基于损失函数的MCS检验结果一致说明,相较于Realized GARCH模型,ARFIMA-Realized EGARCH模型能够更好地捕捉波动率的变化,具有更好的预测效果以及更强的稳定性。

Based on the characteristics of the EGARCH model’s ability to capture volatility leverage effects and the advantages of the ARFIMA-Realized GARCH model in high-frequency volatility forecasting,an ARFIMA-Realized EGARCH model was constructed.The model was used for volatility forecasting and risk measurement under the assumption that the asset return series follows a skewed student’s t-distribution.Numerical simulation results demonstrate that the ARFIMA-Realized EGARCH model has good parameter estimation performance.The model was validated using 5-minute high-frequency data from the Shanghai Composite Index and Shenzhen Component Index from January 4,2011,to December 31,2020.The results indicate that the asset returns in the Chinese stock market exhibit significant peak-heavy-tail characteristics and have a high degree of autocorrelation.After incorporating the leverage function,the model exhibits greater volatility in response to negative shocks,and the leverage parameters are statistically significant at the 1%level.Consistent VaR forecasting results,posterior tests,and MCS tests based on loss functions indicate that compared to the Realized GARCH model,the ARFIMA-Realized EGARCH model can better capture volatility changes,demonstrate superior predictive performance,and exhibit stronger stability.