Consistent measures of systemic risk

This paper presents a methodology to infer multivariate densities that characterize the asset values for a system of financial institutions, and applies it to quantify systemic risk. These densities, which are inferred from partial information but are consistent with the observed probabilities of distress of financial institutions, outperform parametric distributions typically employed in risk measurement. The multivariate density approach allows us to propose complementary and statistically consistent metrics of systemic risk, which we estimate using market-based data to analyze the evolution of systemic risk in Europe and the U.S., throughout the financial crisis.