主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2020, Vol. 28 ›› Issue (7): 68-76.doi: 10.16381/j.cnki.issn1003-207x.2018.1833

• 论文 • 上一篇    下一篇

基于局部相关系数和截尾扭曲混合Copula的杠杆效应识别和度量

沈根祥1, 邹欣悦2   

  1. 1. 上海财经大学经济学院, 上海 200433;
    2. 上海对外经贸大学统计与信息学院, 上海 201620
  • 收稿日期:2018-12-25 修回日期:2019-03-20 出版日期:2020-07-20 发布日期:2020-08-04
  • 通讯作者: 沈根祥(1964-),男(汉族),河南许昌人,上海财经大学经济学院,教授,博士生导师,研究方向:金融计量经济学、金融市场数量分析,E-mail:sgxman@shufe.edu.cn. E-mail:sgxman@shufe.edu.cn
  • 基金资助:
    国家社科基金重大项目(16ZDA031)

Identification and Measurement of Leverage Effects Using Local Correlation and Truncated Distorted Mix Copula Constructing

SHEN Gen-xiang1, ZOU Xin-yue2   

  1. 1. School of Economics, Shanghai University of Finance and Economics, Shanghai 200433, China;
    2. School of Statistics and Information, Shanghai University of International Business and Economcs, Shanghai 201620, China
  • Received:2018-12-25 Revised:2019-03-20 Online:2020-07-20 Published:2020-08-04

摘要: 作为风险资产收益和波动之间关系的度量,杠杆效应是金融市场数据三大分布特征之一,在波动预测、资产定价和风险管理中起着重要作用。日内高频数据计算的已实现波动作为波动的代理变量,解决了波动不能观测的问题,实现了用波动和收益直接建模捕捉杠杆效应。深入了解收益和已实现波动的相关模式并以此构建二者的联合分布是正确度量杠杆效应的关键。本文以局部相关系数为工具研究收益和波动在不同取值范围内的相关性变化,实证研究结果表明,与负收益冲击引起波动增加一样,正收益冲击也会引起波动增加,这与传统杠杆效应理论并不一致,与Chen和Ghysels(2011)对美国股票市场的实证结果一致。为正确捕捉和度量实证结果反映出的杠杆效应,在扭曲混合Copula构造方法基础上,本文用截尾扭曲函数构造扭曲混合Copula,以此作为收益和已实现波动的联合分布,再现收益和已实现波动的局部相关性特征。以上证综指2013.1.29日至2017.4.30区间内日内1分钟高频数据为样本进行实证分析表明,本文构造的Copula函数具有和实际数据一致的局部相关特征,能够正确刻画市场表现出的杠杆效应。Copula拟合优度的非参数检验表明,实际数据不拒绝本文构造的Copula函数,而现有文献采用的单成分Copula函数和两成分混合Copula函数均被拒绝。本文为收益和已实现波动的联合建模提供参考,具有基础重要性。

关键词: 杠杆效应, 已实现波动, Copula, 局部相关系数

Abstract: It is found the empirical evidence in China stock market that the dependence structure between the asset's return and its volatility measured by realized volatility, so-called the leverage effect, have a special correlation pattern in terms of local correlation, which is not consistent with that implied by typical leverage effect assumption,but consistent with the findings about American equity market in Chen and Ghysels (2011). The distortion mixture method of Li et al. (2014) is employed to construct Copulas to capture the tail dependence in real data, and tailor the quadratic distortion functions by truncation to mitigate the confounding of the components Copula in the mixture. The closeness of the local correlation pattern of the estimated local correlations using simulated data from the constructed Copulas to that of real data shows that the Copulas proposed in this paper capture the correlation features in real data well, and the nonparametric goodness-of-fit test confirms the validity of the Copula further.

Key words: leverage effect, realized volatility, copula, local correlationship

中图分类号: