Unveiling Early Warning Signals of Systemic Risks in Banks: A Recurrence Network-Based Approach

TL;DR

This paper introduces a recurrence network-based framework using high-frequency bank stock data to identify early warning signals of systemic risks and potential bank crises.

Contribution

It presents a novel approach combining nonlinear time series analysis and recurrence networks to monitor banking system stability with high-frequency data.

Findings

Average mutual information indicates periods of extreme volatility

Recurrence network indicators reveal nonlinear dynamics of bank systems

High-frequency data improves early warning signal detection

Abstract

Bank crisis is challenging to define but can be manifested through bank contagion. This study presents a comprehensive framework grounded in nonlinear time series analysis to identify potential early warning signals (EWS) for impending phase transitions in bank systems, with the goal of anticipating severe bank crisis. In contrast to traditional analyses of exposure networks using low-frequency data, we argue that studying the dynamic relationships among bank stocks using high-frequency data offers a more insightful perspective on changes in the banking system. We construct multiple recurrence networks (MRNs) based on multidimensional returns of listed banks' stocks in China, aiming to monitor the nonlinear dynamics of the system through the corresponding indicators and topological structures. Empirical findings indicate that key indicators of MRNs, specifically the average mutual information, provide valuable insights into periods of extreme volatility of bank system. This paper contributes to the ongoing discourse on early warning signals for bank instability, highlighting the applicability of predicting systemic risks in the context of banking networks.