Test on the Accuracy of Basel II by Simone Varotto, ICMA Centre, University of Reading
The problem has interested me for some times. The Basel II IRBA approach of calculating risk capital does take into account the individual risk level and portfolio effect. The paper shows that by prescribing single-factor correlation, the accuracy of calculated capital is not much better than the standardized approach. They demonstrate this by comparing capital requirements calculated by more sophisticated model (CreditMetrics).
In addition to capital over-estimation bias, it also means that the risk sensitivity might not be good enough for transfer pricing or effective portfolio optimization.
Interestingly, the paper also shows that SA approach could yield lower capital requirement than IRBA in portfolios with high averaged PD and long duration. This raises the possibility of regulatory arbitrage by SA banks building portfolios with long duration, exposing them to greater economic risk without having to hold proportionate regulatory capital.