中国金融业风险溢出预警研究——基于藤Copula CoVaR模型
Research on Risk Spillover Early Warning of China's Financial Industry Based on Vine Copula CoVaR
闫海波 1沙龙1
作者信息
摘要
在经济全球化的复杂环境下,当前各个行业之间的联系日益密切,因此如何有效地测度行业间上下行风险的相关性,对于中国金融业风险防范与化解有重要意义.本文选取A股房地产业、货币金融服务业、资本市场服务业、保险业和其他金融服务业市值最高的50家上市公司日度收盘价计算对数收益率作为样本.藤Copula函数是一个概率密度函数,用于描述多维随机变量的联合分布.通过藤Copula函数,可以更准确地评估多个随机变量之间的联合风险.相较于传统相关性模型其更能反映不同市场条件下真实资产之间的相依结构.此外,为了更好地描述尾部风险溢出强度,利用不同置信水平下CoVaR模型量化分析行业间在市场高涨和市场下跌时的风险溢出强度.研究结果表明行业间风险溢出强度有明显差异且模型能够识别重大事件的发生.最后利用LSTM神经网络模型对CoVaR值进行拟合训练,训练集与测试集RMSE均小于0.38.综上所述,说明该模型可以很好地建立行业间尾部关系及测度风险溢出强度.在此基础上结合LSTM神经网络模型可充分挖掘金融时间序列的非线性特征优势,对行业间风险溢出进行预警,对于风险溢出监测有重要的现实意义.
Abstract
In the complex environment of economic globalization,the current links between various industries are increasingly close.Therefore,it is of great significance for China's financial industry to scientifically and accurately measure the intensity of financial risk spillover and carry out early warning work.This paper selects the daily closing price of 50 listed companies with the highest market value of A-share real estate industry,monetary and financial services industry,capital market services industry,insurance industry and other financial services industry to calculate the logarithmic rate of return as a sample,and uses Vine-Copula to establish the dependence of risk spillover between industries.The Co VaR model at different confidence levels is used to quantitatively analyze the risk spillover intensity between industries when the market is rising and the market is falling.The results show that there are significant differences in risk spillover intensity between industries.Finally,the LSTM neural network model is used to fit the CoVaR value,and the RMSE of the training set and the test set is less than 0.38.In summary,the Vine-Copula-CoVaR model can well establish the tail relationship between industries and measure the risk spillover intensity.On this basis,combined with the LSTM neural network model,the nonlinear characteristics of financial time series can be fully explored,and the risk spillover between industries can be warned,which has important practical significance for risk spillover monitoring.
关键词
风险溢出/藤联系函数(Vine/Copula)/条件在险价值(CoVaR)/神经网络模型(LSTM)/预警Key words
Risk Spillover/Vine Copula/CoVaR/LSTM/Early Warning引用本文复制引用
出版年
2024
NETL
NSTL
万方数据