金融市场多尺度有向波动溢出与相依结构建模及应用研究

In recent decades, the repeated financial crisis has raised widespread concern among all levels of the society, and the risk spillovers across various markets have become a prominent feature. Studying the volatility spillovers and the dependence structure among financial markets is of great significance for preventing financial risks, while the research under multiple time scales (multiscales) such as short-, medium- and long-term has more practical guiding significance. Being aware of the obvious drawbacks of the existing research, this project takes the Chinese and internationally representative financial market as the research objects, and based on the realized volatility to study the magnitude of the directional volatility spillovers among financial markets. The specific method is combined variance decomposition and spectral representation under TVP-(S)VAR (Time-varying parameter (structural) vector autoregression) model establish a basic framework for studying the quantitative measurement of magnitude of directional volatility spillovers at multiscales; Then the research results are further used to improve the volatility forecasting at different time scales, after fully eliminating the interference of volatility information, combined with the variational mode decomposition and R-vine Copula to study the market dependence structures at different time scales; Furthermore, the results of volatility spillovers and dependence structures are used to study the systemic risk measurement, VaR (Value-at-Risk) and ES(expected shortfall) forecasting, portfolio optimization, etc., in order to provide scientific decision-making basis for government risk management departments and investors.

近几十年来，金融危机的多次爆发引起了社会各界的广泛关注，而各市场之间的风险溢出成为了金融危机的显著特征。研究金融市场之间的波动溢出和相依结构对防范金融风险意义重大，而在短期、中期和长期等多个时间尺度（多尺度）下的研究具有更强的实践指导意义。针对现有研究的明显缺陷，本项目以中国和国际代表性金融市场为研究对象，基于已实现波动率研究金融市场之间的有向波动溢出大小，具体方法是在TVP-(S)VAR模型下结合方差分解和谱表示法建立起对多尺度有向波动溢出大小进行定量测度的基本框架；然后进一步将其研究结果用于改善不同时间尺度下的波动率预测效果，在充分排除波动信息的干扰后再结合变分模态分解和R-藤Copula研究不同时间尺度下的市场相依结构；进而将波动溢出和相依结构的研究结果用于系统性风险测度、VaR和ES预测、投资组合优化等应用问题的研究，以便为政府风险管理部门和投资者提供科学的决策依据。

近几十年来，金融危机的多次爆发引起了社会各界的广泛关注，而各市场之间的风险溢出成为了金融危机的显著特征。研究金融市场之间的波动溢出和相依结构对防范金融风险意义重大，而在短期、中期和长期等多个时间尺度（多尺度）下的研究具有更强的实践指导意义。因此，我们基于TVP-VAR框架在不同时间尺度下建立起有向波动溢出测度，并用于进一步的风险预测研究之中，包括对波动率（Volatility）的预测和风险价值（VaR）的预测两个主要方面。另外，在风险预测中还考虑到了波动的波动（Volatility of Realized Volatility）的时变性、分整（Fractionally integrated）、混频数据抽样（Mixed Data Sampling, MIDAS）等多方面的影响，取得了较好的实证效果，为政府风险管理部门和投资者提供了一些科学的决策依据。