Capturing systemic risk by robust and sparse network estimation

Network analysis is becoming a fundamental tool in the study of systemic risk and financial contagion. Still, the network structure has to be typically estimated from noisy data, as the true network structure is usually unobservable, and standard statistical methods return dense network structures, which are hard to be interpreted. We introduce an approach that allows to estimate sparse networks, capturing only the relevant links, and better deal with estimation error due to outliers. Empirical analysis on CDS spreads and equity returns highlights the ability of our approach to capture/infer the most relevant European bank system interconnectedness and contagion dynamics.