一种多资产组合风险度量解决之道:正则藤Copula

A New Approach for Measuring the Risk of Portfolios with Multiple Assets

为准确地度量包含有多项金融资产的组合的风险,本文提出使用一种新的高维Copula构建方法——正则藤Copula(Canonical Vine Copula),来对多资产间的非线性相关结构进行建模,该函数呈现以一系列成对Copula函数作为节点的"藤"的层叠结构。本文基于上海、香港和台湾三个股票市场对构建该高维Copula函数时各个节点上成对Copula函数类型的选取进行了讨论,并证实了正则藤Copula函数相比传统的多元Copula函数能够更灵活地描述各市场间尾部相关性的复杂形式。样本外风险预测绩效分析和模拟研究均表明,使用正则藤Copula函数确实能够更为稳健和准确地预测组合VaR。

In order to accurately measure the risk of portfolios with multiple assets, this paper introduces Canonical Vine Copula, which is constructed hierarchically using a cascade of pair-copulas, to model multivariate non-linear dependence structure. Based on Shanghai, Hong Kong and Taiwan stock markets, this paper discusses a proper way to choose pair-copula functions in the hierarchical construction, and verifies statistically that Canonical Vine Copula could describes the complex patterns of cross-asset dependence in tails more flexibly than traditional multivariate copula functions. Furthermore, the out-of-sample performance of risk forecasts and simulation analysis indicate that, we can obtain more robust and accurate VaR forecasts by using Canonical Vine Copula.