Comparing covariance measures: a systemic risk perspective

Abstract

In dieser Arbeit wird der bedingte Value at Risk (CoVaR) mittels eines multivariaten HEAVY- und GARCH- Modells auf Basis von verschiedenen realisieten Kovarianzmaßen geschätzt. Die systemischen Risikoschätzer werden mithilfe der gefilterten historischen Simulation und für bedingt normalverteilte Renditen gebildet. Die Modellvarianten bilden das systemische Risiko während der Finanzkrise nur unzureichend ab.

In this paper I address the issue of predicting Conditional Value at Risk (CoVaR) by using a multivariate HEAVY and GARCH model and different realized covariance measures. Systemic risk forecasts are produced by Filtered Historical Simulation and conditional normality in the return distribution. The performance of the CoVaR is evaluated for all models, indicating the failure to describe systemic risk adequately during the financial crisis.