Modelling exposure at default without using conversion factors

Banks accredited by their regulator to use the Advanced Internal Ratings Based (A-IRB) approach are required to provide their own estimates for calculating their minimum credit capital; these estimates rely on statistical and analytical models to predict Probability of Default (PD), Loss Given Default (LGD) and Exposure at Default (EAD). This thesis focusses on estimating EAD for banks granting revolving loans to large corporates and leverages the Global Credit Data (GCD) database. This thesis briefly discusses why risk management, particularly credit risk management, is important for banks and we survey the existing EAD modelling literature which to date has had less focus than PD and LGD modelling. Our prosed methodology models both loan balance at default (EAD) and changes in loan limit at default as random variables, modelling their joint dynamics via a two stage model– the first stage estimates the probability that limits decrease while the second stage estimates EAD conditional on changing limits. To the best of our knowledge, our approach is the first to estimate EAD and changes in loan limit directly for large corporate revolving facilities using the GCD database. Our model suggests that the key drivers of EAD include: limit; balance; utilisation; risk rating; and time to maturity. We also find evidence that banks actively manage limits in the lead up to default, and that these changes in limits have substantial effects on the outcomes of realised EAD.