摘要
近年来,Realized GARCH族模型在金融市场波动率研究中展现了良好的预测效果.该文在两个Realized GARCH族模型基础上,考虑波动率存在非线性结构特征,引入基于显著跳跃方差测度的时变Markov状态转换机制以构建时变MRS-Realized GARCH族模型,推导其参数估计方法,并应用DM检验和MCS检验来评估模型的预测精度.最后,分别基于不同的评估方法,误差分布假设,滚动窗口长度,采样区间和跳跃测度对模型进行稳健性检验.以沪深300股指期货数据为例,实证研究表明:沪深300股指期货市场存在高波动和低波动状态,跳跃测度在低波动状态会对未来一期波动产生显著的正向影响,而在高波动状态会抑制未来一期波动;DM检验和MCS检验显示,时变MRS-Realized GARCH族模型在任意的损失函数指标下均具有最佳的波动预测效果;另外,稳健性检验结果证实该模型在不同情况下均具有最佳的预测表现.
In recent years,the Realized GARCH-type model has shown good prediction results in the study of volatility forecasting in various financial markets.Based on two Realized GARCHtype models,this paper considers the existence of nonlinear structures in volatility and introduces a time-varying markov regime-switching mechanism based on significant jump variance to construct the time-varying MRS-Realized GARCH-type models,derives its parameter estimation method,and evaluates the prediction accuracy of the model by applying the DM test and MCS test.Finally,the robustness tests of the model are carried out under different evaluation methods,error distribution assumptions,lengths of rolling window,sampling intervals and jump measurements.Taking the CSI 300 stock index future as an example,empirical research shows that the states of high and low volatility exist in the CSI 300 stock index futures markets,and the jump in the low volatility state will produce a significant positive impact on volatility in the future whereas the jump in the high volatility state will suppress the future volatility.The DM test and MCS test show that the time-varying MRS-Realized GARCH-type models have the best prediction effect under any loss function index.In addition,a wide range of checks confirm that the models provide good performance,which does’ t depend on the setting of various conditions.
作者
吴志敏
蔡光辉
WU Zhi-min;CAI Guang-hui(School of Statistics and Mathematics,Zhejiang Gongshang University,Hangzhou 310018,China)
出处
《高校应用数学学报（A辑）》
北大核心
2022年第4期397-414,共18页
Applied Mathematics A Journal of Chinese Universities(Ser.A)
基金
国家社会科学基金(19BTJ013)
浙江省重点建设高校优势特色学科(浙江工商大学统计学),统计数据工程技术与应用协同创新中心资助项目。
