Abstract

Taking Account of Estimation Uncertainty in Credit Risk Assessment

Value-at-Risk Comparison Using Bootstrapping and an Asymptotic Approach

Credit risk assessment requires both probability of default and correlation to be estimated. However, such estimation is subject to uncertainty. In order to assess the uncertainty affecting the simultaneous estimation of both parameters, the discussion in literature focuses on the use of asymptotic confidence regions. However, such regions need a very long credit history to allow such assessment to be exact. Bootstrapping represents an alternative method where the data history is short. But this method gives rise to noticeably more intense calculation work. The present article examines the minimum number of periods that must be available in order that bootstrapping and a Wald confidence region permit a comparable assessment of the credit risk. The methods applied here generate similar results where more than 100 historical periods are available.