Q 40: The agencies request comment on the appropriateness of these criteria in determining whether the risk mitigation effects of a credit derivative should be recognized for risk-based capital purposes.
A risk based capital approach should make allowance for both collateral and credit derivatives.
Q 41: The agencies are interested in the views of commenters as to whether and how the agencies should address these and other similar situations in which multiple credit risk mitigants cover a single exposure.
We recommend that the agencies give the bank the flexibility to apply pragmatic rule based procedures or mathematical programming based approaches to optimally allocate the available credit risk mitigants (guarantees) to the underlying exposures. RBS approach uses Linear programming to optimise the allocation of one (or a pool of) mitigant(s). We would be happy to explain this in more detail should this be required.
Q 42: The agencies seek comment on this alternative approach’s definition of eligible retail guarantee and treatment for eligible retail guarantees, and on whether the agencies should provide similar treatment for any other forms of wholesale credit insurance or guarantees on retail exposures, such as student loans, if the agencies adopt this approach.
We support the use of common retail mitigants which are embedded in the data to be allowed for PD, EAD and LGD, not just EAD and LGD, and for a relaxed definition of eligible as it will not be possible to identify ineligible guarantees in the retail data and isolate the impact of their exclusion in the modelling.
Q 43: The agencies seek comment on the types of non-eligible retail guarantees banks obtain and the extent to which banks obtain credit risk mitigation in the form of non-eligible retail guarantees.
The impact of the guarantee / CRM is embedded in loss history (PD, EAD or LGD). The effect of the CRM compared to other more general effects is impossible to isolate, and evidencing eligibility is problematic. This leads to potential double jeopardy; the firm is unable to evidence eligibility and so has to strip out the effects, but this is not possible as the answer is embedded within historical data. The NPR assumes, incorrectly, that clean loss data exists to model PD, EAD and LGD without the effect of mitigants.