Assessing the suitability of regulatory asset correlations applied to South African loan losses
Stoffberg, Hestia Jacomina
MetadataShow full item record
The Basel Committee on Banking Supervision (BCBS) designed the Internal Ratings Based (IRB) approach, which is based on a single risk factor model. This IRB approach was de-signed to determine banks’ regulatory capital for credit risk. The asymptotic single risk factor (ASRF) model they used makes use of prescribed asset correlations, which banks must use for their credit risk regulatory capital, in order to abide by the BCBS’s rules. Banks need to abide by these rules to reach an international standard of banking that promotes the health of the specific bank. To evaluate whether these correlations are as conservative as the BCBS intended, i.e. not too onerous or too lenient, empirical asset correlations embedded in gross loss data, spanning different economic milieus, were backed out of the regulatory credit risk model. A technique to extract these asset correlations from a Vasicek distribution of empirical loan losses was proposed and tested in international markets. This technique was used to extract the empirical asset correlation, and then compare the prescribed correlations for developed (US) and developing (South Africa) economies over the total time period, as well as a rolling time period. For the first analysis, the BCBS’s asset correlation was conservative when com-pared to South Africa and the US for all loan types. Comparing the empirical asset correlation over a seven-year rolling time period for South Africa and the BCBS, the specified asset cor-relation was found to be as conservative as the BCBS intended. Comparing the US empirical asset correlation for the same rolling period to that of the BCBS, it was found that for all loans, the BCBS was conservative, up until 2012. In 2012 the empirical asset correlation sur-passed that of the BCBS, and thus the BCBS was not as conservative as they had originally intended.