统计与信息论坛

基于LASSO-QVAR模型，筛选金融机构之间有效关联，构建有向加权尾部关联网络，可以从系统视角揭示系统性风险动态特征。一方面，通过尾部关联网络构造系统性风险计分及其贡献，以度量系统性风险水平与识别系统重要性金融机构，进一步从微观层面探究系统性风险贡献的影响机制；另一方面，将尾部关联网络区分为正关联网络与负关联网络，分别提取网络整体特征以及网络节点特征，考察不同关联关系下网络特征的系统性风险预测能力。以2011—2022年32家中国A股上市金融机构作为研究对象，开展了实证研究。研究结果显示：第一，尾部关联网络具有时变特征，能够较好地吻合一些危机事件，银行业和证券业呈现出更为紧密的关联性，具有行业异质性。进一步，发现银行业风险共振(正关联网络)与风险分散(负关联网络)功能相较其他金融行业更强。第二，系统性风险计分可以作为量化系统性风险动态性的重要指标，系统性风险贡献主要集中于银行业与保险业，进一步机制分析表明公司治理水平能够显著影响到系统性风险贡献。第三，尾部关联网络特征具有很好的预测能力，网络整体特征能够预测系统性风险大小，网络节点特征能够预测金融机构系统性风险贡献。

Based on the LASSO-QVAR model, selecting effective relationships among financial institutions and constructing a directed weighted tail correlation network can capture the dynamic characteristics of systemic risk from a systemic perspective.On the one hand, the tail correlation network is used to construct the systemic risk score and its contribution, so as to measure the systemic risk level and identify systemically important financial institutions.In a further, the systemic risk contribution mechanism is investigated from a micro-level perspective.On the other hand, the tail correlation network is divided into a positive correlation network and a negative correlation network.The overall and the node characteristics of the network are extracted to examine their predictive ability for systemic risk under different linkage relationships.An empirical study is conducted on 32 Chinese A-share listed financial institutions from 2011 to 2022.The empirical results show that, firstly, the tail linkage network has time-varying characteristics, which can better match crisis events.This phenomenon is heterogeneous across different sectors, for instance the banking and securities sectors exhibit closer associations.Further, the banking sector has more vital functions of risk co-movement(positive correlation network) and risk diversification(negative correlation network) than other financial sectors.Secondly, the systemic risk score can serve as an indicator for quantifying the dynamics of systemic risk.Systemic risk contributions are mainly concentrated in the banking and insurance sectors.Mechanism analysis indicates that the corporate governance level can significantly affect systemic risk contributions.Thirdly, the characteristics of the tail correlation network have good predictive ability.More specifically, the overall characteristics of the network can predict systemic risk, while the node characteristics of the network can predict the systemic risk contributions of financial institutions.