FRM Part II · FRM Exam Part II · Governance

A bank's internal probability-of-default model for mid-market corporates has been in use for three years. The independent validation team finds that the observed default rate in the highest-quality grades has been well above model-predicted PDs for four consecutive quarters, while rank ordering remains strong. Which conclusion and action is most appropriate?

The model discriminates well but is poorly calibrated, since realised defaults persistently exceed predicted PDs. This should be escalated to model governance and recalibrated by the model owner. Rank ordering alone is insufficient because understated PDs distort capital and pricing, and validators must remain independent of model development.

1. AThe model has good discriminatory power but poor calibration; it should be escalated and recalibrated, with the issue reported to the model governance committeeCorrect
2. BThe model is fully satisfactory because strong rank ordering is the only property that matters for capital and pricing
3. CThe model should be left unchanged until a full economic cycle has elapsed, since quarterly deviations are never informative
4. DThe validation team should adjust the model parameters itself, since it is best placed to do so