基于BEMD-Copula-GARCH模型的股票投资组合VaR风险度量研究

鉴于股票波动具有显著的多尺度特征,本文引入二元经验模态分解(EMD)与二元CopulaGARCH算法,提出一种新的VaR风险度量模型,即BEMD-Copula-GARCH模型.具体地,新BEMD-Copula-GARCH模型可分为三个主要步骤:数据分析,分风险估计和总风险集成.首先,基于二元EMD模型,将复杂且相互作用的股票对分解为若干组较为简单且相互独立的分量,以降低建模难度.其次,引入二元Copula-GARCH模型,刻画各组分量间的相互关系,以度量股票投资组合在不同尺度上的分VaR值.最后,集成各分VaR值以得出最终VaR风险度量结果.实证研究以恒生指数与上证综指为数据样本构造投资组合,结果表明:本文所构建的新模型能有效度量投资组合风险,其估计精度显著优于DCC-GARCH和Copula-GARCH等现有模型.

VaR measurement for stock portfolio based on BEMD-Copula-GARCH model

Given that stock fluctuation has significant multi-scale features, a novel Value-at-Risk (VaR) model is proposed by combining the binary empirical mode decomposition (BEMD) and Copula-GARCH algorithm, i.e., BEMD-Copula-GARCH model. In the proposed model, three main steps are included, i.e., data decomposition, individual risk measurement, and total risk integration. First, the binary EMD technique is employed to decomposed the pair of complex and interactive stock series into pairs of relatively simple and independent components, to reduce the modeling difficulty. Second, Copula-GARCH model is introduced to individually capture the dynamic dependence between the decomposed components in pairs, in terms of VaR at different time-scale. Finally, the individual results are integrated into the final VaR measure. In the empirical study, the portfolio with equal-weighted Hang Seng Index and Shanghai Composite index is analyzed, and the results indicate that the proposed model outperform the benchmark DCC-GARCH model and Copula-GARCH model in terms of the risk measurement accuracy.