Estimation Uncertainty in Credit Risk Assessment: Comparison of Credit Risk Using Bootstrapping and an Asymptotic Approach
For credit risk assessment, probability of default and correlation have to be estimated simultaneously. However, these estimates are uncertain. To assess this uncertainty the literature has discussed the use of asymptotic confidence regions. This kind of region though needs a long credit history for exact assessment. An alternative method to generate a confidence region for a short credit history is bootstrapping. Hence, it could be more appropriate to assess estimation uncertainty with bootstrapping than with asymptotic methods if only a short credit history is available. Based on a simulation study, it is analyzed how many periods should be available for assessing credit risk – taking account of estimation uncertainty – if bootstrapping and a Wald confidence region shall achieve similar results. This article shows that more than 100 cycles have to be available for similar results.