Considering the frequency domain and nonlinear characteristics of financial risks, we measure the multiscale financial risk contagion by constructing EMD-Copula-CoVaR models. Using a sample composed of nine international stock markets from January 4, 1999, to May 13, 2021, the empirical study reveals that: (1) EMD-Copula-CoVaR models can effectively measure the multiscale financial risk contagion, and the financial risk contagion is significant at all time scales; (2) The high-frequency component is the major contributor of financial risk contagion; meanwhile, the low-frequency component is the smallest among all time scale components; (3) The risk export of the US financial market to other markets, except the UK under the original and medium-frequency component, is higher than that it receives; and (4) Even though the magnitude of overall financial risk contagion is similar for the COVID-19 pandemic, Subprime Crises, 9/11 terrorist attack and other crises, the relative importance of different frequency components is heterogeneous. Therefore, the countermeasures of risk contagion should be designed according to its multiscale characteristics.

考虑到金融风险的频域和非线性特征，我们通过构建EMD-Copula-CoVaR模型来衡量多尺度金融风险传染。以1999年1月4日至2021年5月13日9个国际股票市场为样本，实证研究表明：（1）EMD-Copula-CoVaR模型可以有效衡量多尺度金融风险传染，金融风险传染在所有时间尺度上都是重要的；(2)高频成分是金融风险传染的主要贡献者；同时，低频分量在所有时间尺度分量中最小；(3)美国金融市场在原始和中频分量下向除英国以外的其他市场输出的风险高于其接收的风险；(4) 尽管 COVID-19 大流行、次贷危机、9/11 恐怖袭击和其他危机的整体金融风险传染程度相似，但不同频率分量的相对重要性是异质的。因此，应根据其多尺度特征设计风险传染的对策。