Inference on dynamic systemic risk measures

Abstract

Systemic risk measures (SRM) quantify the risk of a system induced by the possible distress of any of its components. Applications in economics and finance are numerous. We define a general dynamic framework for the risk factors, allowing us to obtain explicit expressions of the corresponding dynamic SRM. We deduce an easy-to-implement statistical approach which, based on semi-parametric assumptions, reduces to estimating univariate location-scale models and to computing (static) quantiles of the residuals. We derive a sound asymptotic theory (including confidence intervals, tests, validity of a residual bootstrap) for major SRM, namely the Conditional VaR (CoVaR) and Delta-CoVaR. Our theoretical results are illustrated via Monte-Carlo experiments and real financial and macroeconomic time series.