局部多元分位数模型及尾部风险溢出研究

As the integration of the global economics and business, financial risk tends to transmit among different countries and various markets, which intensifies the global risk contagion and even may lead to an international economic crisis. The traditional risk models perform poorly in depicting the dynamic evolution of financial extreme events, which makes how to model financial extreme risk contagion a challenging issue. In this project, we propose a new localized multivariate autoregressive quantile model to analyze the dynamic structure change and time-varying features of tail risk contagion among financial markets. We will try to extend the multivariate quantile regression by exploiting the local adaptive approach and discussing the statistical properties under finite sample estimation. Further, after deriving the related theoretical properties of local change point detection, the estimation method and the corresponding theorems of the localized multivariate autoregressive quantile regression model will be smoothly constructed. Besides, we will utilize the proposed model in two important financial applications: (1) the analysis of financial risk contagion among international asset markets, especially comparing the dynamic risk structure change before and after the financial crisis periods; (2) the analysis of tail risk spillover among Chinese important financial institutions.

随着金融经济一体化的发展，金融风险跨境和跨市场传递趋于频繁，加剧了金融风险的国际漫延，严重时可能引发全球性经济危机；同时，金融市场极端事件的动态演变，使得传统风险测度模型面临严峻考验，从而使研究金融市场极端风险溢出成为富有挑战性的课题。本项目试图提出一个新的局部多元分位数模型，研究金融市场极端尾部风险传染的动态结构和时变特征。本项目拟基于局部适应性方法扩展分位数回归并推导其有限样本性质，论证局部断点检验在多元分位数回归中的相关理论，从而建立局部多元分位数模型的估计方法和理论依据。在此基础上，本项目将提出的局部多元分位数模型应用到两个实证研究：（1）考察美国、德国和中国等全球主要股指之间的尾部风险溢出，特别是金融危机前后的动态变化；（2）考察中国金融体系内银行、证券和保险等主要金融机构间的时变尾部风险传染。本项目的研究有助于揭示金融尾部风险的传导机制，为金融监管和金融稳定政策的制定提供依据。

随着经济金融一体化的发展和金融业务复杂程度加深，金融体系之间的联系愈加紧密，金融风险跨境和跨市场传递趋于频繁，加剧了金融风险的国际漫延。特别是2019年新冠疫情的爆发，各国股市大幅动荡，甚至危及金融体系的整体稳定性，使得研究极端事件金融风险的内在传染机制富有重要意义。本项目提出了一个新的局部多元分位数模型，构建了其估计方法和理论框架，研究金融市场极端尾部风险传染的动态结构和时变特征。本项目基于局部适应性方法扩展分位数回归并推导其有限样本性质，论证局部断点检验在多元分位数回归中的相关理论，从而建立局部多元分位数模型的估计方法和理论依据。在此基础上，本项目将提出的局部多元分位数模型应用于实证研究，考察了美国和德国等主要发达国家之间的尾部风险的动态溢出效应。本项目的研究有助于揭示金融尾部风险的传导机制，为金融监管和金融稳定政策的制定提供依据。