Assessment of systemic risk through multilayer financial networks and the dynamics of shock propagation using percolation methods and stochastic modeling on real data

Abstract

This paper develops a comprehensive framework for quantifying systemic risk based on a multilayer representation of the financial system that combines three key channels of contagion: interbank lending, over the-counter derivatives exposures and common asset holdings. The empirical analysis relies on an aggregated and anonymized dataset for 520 European financial institutions over 2019–2023, including global systemically important banks (G-SIBs), other banks, insurance companies and investment funds. For each layer we compute core network statistics (density, degree distribution, clustering coefficient, average path length) and then assess system resilience using percolation-based simulations of random and targeted shocks, complemented by stochastic agent-based modelling under several macroeconomic stress scenarios. The results show that the derivatives layer exhibits the highest density, clustering and core–periphery structure and thus acts as the fastest and most fragile channel of shock transmission; the default of only 1% of the most central institutions can trigger cascades involving more than one third of the system. The strongest increases in systemic-risk indices are generated by equity-market crashes and credit-spread spikes, which activate the common-asset channel. A decomposition by institution type reveals a disproportionately large contribution of G-SIBs to aggregate systemic risk, while also highlighting the relevance of herd risk stemming from numerous medium-sized banks and non bank intermediaries. Overall, the findings underscore the need for a multilayer network perspective in macroprudential policy design.