|
|
The study examines the variability of risk-weighted asset (RWA) in banks that use internal models to calculate their credit risk regulatory capital requirements. It follows the Committee's 2013 report, which found considerable variation in average RWAs for credit risk in the banking book, and extends that analysis in two respects:
Data were collected from 35 major internationally active banks on their retail and small and medium-sized enterprise (SME) lending portfolios. When averaging across banks in the sample, the report finds a reasonable relationship between estimates of probabilities of default and actual default rates. A weaker relationship is observed between loss outcomes and other parameters estimated by credit risk models (such as loss-given-default). At an individual bank level, there is much greater variation between parameters estimated using credit risk models and actual outcomes. As retail business is predominantly local, these variations are partly explained by differences in local circumstances.
The report also evaluates the variability in estimates of exposure at default for all asset classes, using data from 37 banks. These reveal wide variation in bank practices, which can contribute materially to overall RWA variability. It was also found that some of the estimates used in internal credit risk models lack empirical support. Further, some data relating to one asset class are being used to support estimates for unrelated asset classes.
The report also describes sound practices relating to banks' independent model validation functions, as observed during the Committee's engagement with banks in the course of this study. These practices apply to the governance, methodology and scope of banks' validation functions, as well as to the role of validation across different phases of model development and implementation.