Research on Dynamic Contagion of Banking Risks and Identification of Systemically Important Institutions—Based on the HD-TVP-VAR-DY Model

This paper focuses on analyzing the dynamic process, strength and orientation of risk spillovers in the Chinese banking system under the exogenous shock scenario of the COVID-19 pandemic. Using the closing prices of 25 chosen banks on a daily basis, it stratifies the data into three periods: before, during, and after the pandemic. The HP-TVP-VAR-DY model is used to model risk heterogeneity and time-varying features in risk transmission processes. A dynamic topological directional graph is further used to track the core risk sources and paths in risk transmission processes. The key findings obtained from this paper are summarized below: (1) The Total Spillover Index for the banking system persisted at a high level following the outbreak of the COVID-19 pandemic, indicating its high sensitivity to abrupt and large-scale events. (2) Bank risk transmission paths are highly heterogeneous and time-varying in nature. Prior to the pandemic, CCBs were prominent in overall risk output; during the pandemic, JSCBs dominated; while in the post-pandemic period, again CCBs dominated overall risk output. In all periods, SOCBs and RCBs were identified as major risk receivables. (3) Concerning the structural change in interbank risk transmission paths, it exhibits phase-dependent features. In the pre-pandemic period, risk spillovers spread from CIB, CMB to ABC. However, in the pandemic period, interbank risk transmission paths became highly decentralized, indicating significant increases in risk outflows and inflows from RCBs and CCBs, respectively. Moreover, CMBC and SZRCB turned out to become key sources for risk radiation, while overall network mechanisms dominated risk absorption effects. However, in the post-pandemic period, interbank risk transmission paths tend to become re-centralized; BOC turned out to become a core source for risk transformation, indicating a revival in risk-output dominance in network topologies.