中国股市波动率预测——基于已实现EGARCH模型和已实现SVL模型的实证比较研究  

Forecasting Volatility of China’s Stock Market:An Empirical Comparative Study Based on Realized EGARCH Model and Realized SVL Model

作　　者：吴鑫育[1] 王小娜 王海运 WU Xinyu;WANG Xiaona;WANG Haiyun(School of Finance,Anhui University of Finance and Economics,Bengbu 233030,China)

机构地区：[1]安徽财经大学金融学院,安徽蚌埠233030

出　　处：《重庆理工大学学报（自然科学）》2021年第7期231-241,共11页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金项目“时变崩盘风险下的期权定价及风险测度研究”(71971001);国家自然科学基金项目“时变风险厌恶下的期权定价与参数估计研究”(71501001);安徽省高校优秀拔尖人才培育项目“2017年度高校优秀青年骨干人才国内外访学研修项目”(gxfx2017031);安徽省高校自然科学研究项目“考虑期权数据信息的市场风险度量研究”(KJ2019A0659);安徽财经大学研究生科研创新基金项目“区块链技术背景下数字货币市场的风险测度研究”(ACYC2019093)。

摘　　要：通过采用上证综合指数数据,对已实现EGARCH(REGARCH)模型与已实现SVL(RSVL)模型在不同分布设定(正态分布、学生t分布与偏斜学生t分布)以及样本外阶段(预测总样本期、高波动期与低波动期)下的波动率预测表现进行实证比较研究。结果表明:在预测总样本期,REGARCH模型在3种分布下的样本外预测表现都优于RSVL模型,REGARCH模型在各分布下的样本外预测表现取决于选择的损失函数;在预测高波动期,其与预测总样本期具有相似的结论,但REGARCH模型在偏斜t分布下的预测表现最佳;在预测低波动期,其与预测总样本期的结论相反,即RSVL模型在3种分布下的样本外预测表现都优于REGARCH模型,RSVL模型在各分布下的样本外预测表现取决于选择的损失函数。Using SSE composite index data,an empirical comparison of volatility forecast performance of the realized EGARCH(REGARCH)and realized SVL(RSVL)models under different distribution specifications(normal distribution,student t distribution and skewed student t distribution)and out-of-sample periods(full sample,high volatility and low volatility periods)is investigated.The results show that the REGARCH model outperforms the RSVL model in out-of-sample forecasting in the full-sample period,regardless of the distribution specification,and the performance of the REGARCH under different distribution specifications depends on the loss function.In the high volatility period,the conclusion is similar to that of the full-sample period.However,the REGARCH model with the skewed student t distribution performs best.In the low volatility period,on the contrary,the RSVL model outperforms the REGARCH model in out-of-sample forecasting,regardless of the distribution specification,and the performance of the RSVL under different distribution specifications depends on the loss function.

关 键 词：已实现EGARCH 已实现SVL 杠杆效应 已实现核 波动率预测 偏斜学生t分布 

分 类 号：F830.9[经济管理—金融学]