Monte Carlo Simulations of Contagion in the US Banking System A Network Theory Approach to Systemic Risk

Abstract

Over the past decade, economists have developed an alternative approach to systemic risk that draws extensively from network theory. This network-based approach sees the systemic risk as heavily dependent on the contingent arrangement of financial linkages within the system. The past research in this network approach has relied upon three assumptions: exogenous shocks are idiosyncratic, loss-given-default is constant and the financial network resembles an Erdös-Rényi random network. However, there has been recent criticism that these assumptions do not reflect the actual financial network. This thesis uses a Monte Carlo simulation methodology to analyze contagion in the US financial markets with a set of assumptions that helps to address these criticisms. Specifically, exogenous shocks are aggregate, loss-given-default is increasing for higher order defaults and the financial network is a small-world network. We find results that differ significantly from previous literature. First, we find that contagion is increasing in the homogeneity of bank balance sheet composition. Second, we find the distribution of contagion is highly bimodal, not normal. Lastly, we find that the increase in contagion is primarily in the tail of the distribution. These results have large implications for regulators, as it implies that current regulatory incentives for banks to converge on similar balance sheet exposures may in fact be increasing systemic risk.