Non-Performing Loans and Systemic Risk in Financial Networks

In this paper we study the implications of non-performing loans (NPLs) for financial stability using a network-based approach. We start by combining loan-level data from DealScan and firm-level data from Orbis to reconstruct the empirical global financial network in the period 1991-2016 and identify a series of stylized facts. Based on these findings, we develop a model in which two types of agents, banks and firms, are linked in a network by their reciprocal claims and analyze how an increase in NPLs affects the stability of the system. We study the model analytically and with numerical simulations, deriving a synthetic measure of systemic risk and quantifying the threshold level of NPLs that triggers a systemic crisis. Our model shows that there exist a level of connectivity that maximizes the fragility of the financial system and that small changes in the initial NPLs shock can have very different consequences at the aggregate level.