A gradient boosting approach to estimating tail risk interconnectedness. Issue 8 (13th February 2022)

• Record Type: Journal Article
• Authors: Long, Yunshen; Zeng, LinQing; Wang, Jing; Long, Xingchen; Wu, Liang
• Is Part Of: Applied economics. Volume 54:Issue 8(2022)
• Page Start: 862
• Page End: 879
• Publication Date: 2022-02-13
• Subjects: Financial risk -- tail risk -- interconnectedness -- machine learning; Economics -- Periodicals; 330
• DOI: 10.1080/00036846.2021.1969002
• Languages: English
• ISSNs: 0003-6846

Abstract

ABSTRACT: Interconnectedness, an important risk indicator, should be gauged when assessing the tail risk of financial institutes. Considering potential nonlinearity and interaction in the networked tail risk contagion, we propose an interpretably non-parametric measure based on Gradient Boosting Machine (GBM) to estimate interconnectedness across financial institutes. The proposed measure can be utilized to monitor tail risk spillover channels. We apply our proposed measure to investigate the tail risk interconnectedness of the listed banks in China. We find that the tail risks tend to be spilled over within the same type of banks. The empirical result also implies that the bank heavily involved with the interbank business, such as Industrial Bank, is worthy of regulators' attention. Further analysis shows that the total connectedness of the banking system increases when the financial system is under distress. Finally, we document that state-owned banks are the main tail risk emitters during financial turbulence. Our study not only presents a flexibly data-driven way to quantify tail risk contagion but also provides useful information to policymakers for prudential supervision.