Dear Sir/Madam,

Finanstilsynet and Norges Bank appreciate the opportunity to provide feedback on the proposal for reducing variation in credit risk weights for internal model approaches. We share the Basel Committee's concerns about low and diverging capital requirements among banks using internal models to calculate their capital requirements. Our experience is that approved Internal Ratings Based (IRB) models, in general, differentiate risk. While some variation in related calculations between banks can reasonably be expected to result from differing modelling approaches, we recognise the usefulness of limiting the types of exposures that can be modelled, introducing floors on parameters, and specifying estimation practices. In our view the calibrations of the measures should be consistent with the aim of risk-based capital requirements. We also recognise that these proposals will complement the introduction of overall floors based on the standardised approach.

Removal of option to use the IRB approaches for certain exposures

We agree that for certain types of exposures it is challenging for banks to obtain sufficient data to reliably estimate PD and/or LGD. We support the view that the standardised approach should be used to calculate credit risk charges related to exposures to financial institutions and the largest corporates (total assets exceeding EUR50bn), while the Foundation IRB (F-IRB) approach could be used for such calculations related to exposures to other large corporates (revenues greater than EUR200m). In our view the thresholds determining the type of exposures should be open for downwards revisions by the competent authorities. Norwegian banks' large corporates thresholds are generally considerably lower than those proposed.

Considerable difficulties are involved in modelling specialised lending exposures. These are often low-default portfolios and each exposure has specific characteristics, which makes it challenging to validate PD and LGD models. In general, we support the use of the slotting approach for specialised lending exposures. However, for certain types of specialised lending, such as income-generating real estate, our experience is that banks are able to satisfactorily validate simulation-based IRB models.

Parameter floors

In general we support floors on PDs, LGDs and CCFs at the exposure level. Such floors ensure that all exposures are measured with a minimum level of risk. These floors, together with an overall output floor, should serve to incentivise banks to develop prudent models, and to limit any actions aimed at optimising the overall impact on capital requirements.

It is challenging to validate ratings with an extremely low PD. In addition, the risk weight function is very sensitive in the region of low PDs; specifically, a small estimation error for a low-PD exposure can provide a large estimation error for the risk weight. For certain exposure types, a PD floor of 5 bps appears too low.

Validation of LGD models is challenging for some exposure types, especially since LGD estimates should reflect an economic downturn. The committee's QIS should enable a sound evaluation of the conservatism of the minimum levels proposed by the committee.

Parameter estimation practices

We support the recommendation that rating systems in general should be designed such that rating categories remain stable over time. However, this may be difficult to achieve in practice. We therefore propose that such stability be presented as a desirable feature of a rating system rather than a requirement.

For the estimation of long-run PDs, we support requiring the inclusion of downturn years in the data series. However, since various types of exposures behave differently over business cycles, national authorities should, as appropriate, have the option of setting a higher requirement than the proposed minimum requirement of 10 per cent for weighting of downturn years. For retail mortgage PDs in Norway, a minimum weight of 20 per cent for downturn years is required.

In general we support the proposals for LGD and CCF estimation. For the F-IRB approach, it is proposed that commercial real estate should have the same treatment as residential real estate. In our experience, commercial real estate is more volatile and should have a higher LGD. A higher haircut value could also be considered.

Yours sincerely

Emil Steffensen
Deputy Director General
Finanstilsynet

Torbjørn Hægeland
Executive Director
Norges Bank Financial Stability