Faculty of Business and Law
School of Accounting, Finance and Economics / Finance, Economics, Markets and Accounting Research Centre
The Global Financial Crisis (GFC) highlighted the importance of measuring and understanding extreme credit risk. This paper applies Conditional Value at Risk (CVaR) techniques, traditionally used in the insurance industry to measure risk beyond a predetermined threshold, to four credit models. For each of the models we use both Historical and Monte Carlo Simulation methodology to create CVaR measurements. The four extreme models are derived from modifications to the Merton structural model (which we term Xtreme-S), the CreditMetrics Transition model (Xtreme-T), Quantile regression (Xtreme-Q), and the author’s own unique iTransition model (Xtreme-i) which incorporates industry factors into transition matrices. For all models, CVaR is found to be significantly higher than VaR, and there are also found to be significant differences between the models in terms of correlation with actual bank losses and CDS spreads. The paper also shows how extreme measures can be used by banks to determine capital buffer requirements.