在已有动态Copula模型基础上,提出可同时描述尾部相依性的非对称和长记忆特征的Copula模型.基于沪深股市数据,首次从尾部相依性的角度检验了沪深股市的长记忆效应.研究发现,沪深两市在重大利好或利空消息冲击时的相关性(即尾部相依性)都具有长记忆效应,极端事件对尾部相依性的影响比对未来收益和波动的影响更加持久.而且,样本外分析结果表明,相比已有Copula模型,具有长记忆性的Copula模型能更准确地预测未来1 周至1年的市场间相关性.

Based on existing dynamic copulas, this paper proposes a new copula model to simultaneously describe the asymmetric and long memory properties in tail dependence. Using Shanghai and Shenzhen stock exchange indices, we for the first time examine the long memory effects in stock market from the perspective of cross-market tail dependence. It is found that both upper and lower tail dependence of the two markets exhibit long memory dynamics. The impact of extremely good or bad news on the cross-market tail dependence lasts much longer than their impact on return or volatility. Furthermore, out-of-sample results show that the new model with long memory in tail dependence has higher predicative ability in forecasting the next 1-week to 1-year cross-market dependence structure than existing copula models.