全球主要股市波动率的聚类特征分析

本文采用多路归一化割谱聚类方法、单变量GARCH模型和Granger因果检验相结合的模型,分阶段研究了1994-2014年间全球主要股市波动率的聚类特征。首先,利用单变量GARCH模型分别提取全球主要股市的波动率;其次,借助多路归一化割谱聚类方法的特殊性质刻画了全球主要股市波动率的聚类数目、聚类质量以及聚类结果的稳定性等特征;最后,利用Granger因果检验模型分析不同类的代表元股市间的波动溢出效应和同一类内股市间的波动溢出效应。实证结果表明,与非金融危机阶段相比,在金融危机期间全球主要股市波动率的聚类数目较多、聚类质量较高、聚类结果相对稳定、并且全球主要股市间的波动溢出效应增强。

Research on Cluster Property of Volatility of International Main Stock Markets

In this paper, multiway normalized cut spectral clustering, univariate GARCH model and Granger causality model are used to analyze cluster property of volatility of the international main stock markets. First, the GARCH model is used to depict volatility of the main markets, then partition the volatility data sets by spectral clustering method. At last, the Granger causality test model is used to analyze spillover among the main markets during different periods. The results show that during the financial crisis cluster number is larger than other periods, clustering quality is higher and clustering result is more stable. During the period of financial crisis, relations among the main markets are becoming closer.