Network conditional tail risk estimation in the European Banking System

CoVaR is one of the most popular measures of systemic risk. It is the VaR (Value at Risk) of the system (represented as a broad market index) conditional to the fact that a certain institution is in distress (i.e. at its VaR). One of the limits of CoVaR is that it does not consider the relations among institutions in the system, failing to represent interconnectedness, that is a relevant component of systemic risk. Instead, it reflects more the systematic component of risk, that is, the one related to a common component. A popular approach to analyze interconnectedness is to consider an economic system as a network. In this work we deal with network-DeltaCoVaR, a multivariate extension of DeltaCoVaR, that measures the marginal tail dependence among institutions while controlling for the effects of the others. We discuss the properties of the model and we propose an estimation methodology based on quantile regression with Smoothly Clipped Absolute Deviation (SCAD) penalty. Finally, we use these tail risk networks to develop systemic risk indicators and to study the characteristic of the European banking system and its evolution over time.