Sample dependency during unconditional credit capital estimation

Abstract

The unconditional credit loss distribution is identified based on a long-term sample. This sample influences the capital estimate. In this study, we performed an empirical investigation of this sample dependency problem using charge-off data and by focusing on the influence of the Great Recession. The results demonstrated the significant dependency of the capital requirements on the homogeneity and cyclicality of the long-term sample. Thus, a sample containing only the Great Recession data produced lower capital requirements due to the homogeneity effect, whereas a mixed sample containing the Great Recession data produced higher capital requirements due to the cyclical effect.