Banking Systemic Vulnerabilities: A Tail-Risk Dynamic CIMDO Approach

Abstract

This study proposes a novel framework which combines marginal probabilities of default estimated from a structural credit risk model with the consistent information multivariate density optimization (CIMDO) methodology of Segoviano, and the generalized dynamic factor model (GDFM) supplemented by a dynamic t-copula. The framework models banks? default dependence explicitly and captures the time-varying non-linearities and feedback effects typical of financial markets. It measures banking systemic credit risk in three forms: (1) credit risk common to all banks; (2) credit risk in the banking system conditional on distress on a specific bank or combinations of banks and; (3) the buildup of banking system vulnerabilities over time which may unravel disorderly. In addition, the estimates of the common components of the banking sector short-term and conditional forward default measures contain early warning features, and the identification of their drivers is useful for macroprudential policy. Finally, the framework produces robust outof-sample forecasts of the banking systemic credit risk measures. This paper advances the agenda of making macroprudential policy operational.