基于交错取样门限多幂次变差的中国股市波动细分及非对称性建模

Chinese Stock Markets' Volatility Distinction and Asymmetric Modeling Using Staggered Threshold Multi-power Variation

以多幂次变差的测量为理论基础,考虑到有限样本规模的局限以及市场微观结构噪声的影响,提出交错取样门限多幂次变差方法并将其用于中国股市高频已实现波动的细分,区分出连续波动与跳跃波动。根据已实现波动、连续波动与跳跃波动的不同统计特征,分别为已实现波动与连续波动建立LHAR-V-CJ模型,为跳跃波动强度建立LHAR-SJ-C模型,为跳跃波动间隔时间建立LACH-DJ-C模型,引入异质非对称性。使用沪深300指数的实证表明,已实现波动及连续波动与跳跃强度、跳跃间隔时间呈现出不同的非对称性特征,且本文提出的各非对称性模型较现有模型均有较明显的拟合能力改进。

Based on the theory of multi-power variation measures and considering the limitation of finite sample size as well as market microstructure noises, we propose a staggered threshold multi-power variation, which is then used in Chinese stock markets＇ realized volatility distinction, generating continuous-time volatility and jump volatility estimators. According to different statistical characteristics of realized volatility, continuous-time volatility and jump volatility, we set up LHAR-V-CJ model for realized volatility and continuous-time volatility, LHAR-SJ-C model for jump sizes, LACH-DJ-C model for jump intervals, all of which introduce characterization of heterogeneous asymmetry. Empirical results using CSI300 index indicate that, realized volatility and continuous-time volatility, jump sizes, jump intervals show different asymmetric effects. In addition, our proposed asymmetric models all lead to significant fitting capability improvements.