Stressing of migration matrixes for International Financial Reporting Standard 9 and Internal Capital Adequacy Assessment Process calculations

Rating transition matrixes are commonly used for computing expected credit losses under International Financial Reporting Standard 9 (IFRS 9) and for stress testing under the Internal Capital Adequacy Assessment Process (ICAAP). The standard method for stressing a through-the-cycle transition matrix requires estimation of the default correlation parameters. In this paper we show that the correlation estimates are sensitive to both the model assumptions and the estimation methodology. We compare Gaussian and logistic distributions, flat and rating-dependent correlations, and asymptotic and default-count-based estimation approaches. For example, we demonstrate that the estimated unexpected losses under the ICAAP may be twice as high when using rating-dependent correlations compared with a flat correlation approach. We conclude that there are significant model and estimation risks associated with the use of rating transition matrixes for IFRS 9 and the ICAAP.