We support the principle of proportionality, since it takes the business model of each bank into account. We do feel that the complexity of the rating “model” as a measure for applying assessment methods is not sufficient. Elements relating to the process of the rating model also need to be taken on board. Rating models can be relatively simple, but the process complex and vice versa. From our perspective, it would therefore be better to refer to the rating “system”, instead of “model” in Article 1(2,b), which would then also capture the rating process.
In principle we agree that the validation function should be independent from the credit risk control unit. The new draft RTS might however use slightly different definitions which could impact the current independence. We would like to propose to uphold the existing validation function structures, via grandfathering of the existing structure up until the moment when an internal framework receives a complete reworking (and subsequently supervisory approval).
The provisions are not completely clear. Further clarification is needed on what defines a complete economic cycle, on how representativeness should be demonstrated and on what are deemed to be acceptable reconstruction methods.
Yes, we agree with using the default weighted average LGD calculation. . As outlined under (ii) on page 80 of the draft RTS, the default weighted average should be taken for reason of consistency with respect to calibration approaches applied for other risk parameters.
We welcome the additional guidance given in Article 52 of the draft RTS. However, we would ask for more clarity on the definition of ”limited timeframe”. The RTS provides one example stating that if two defaults are observed within four months, it is expected that they will be counted as one default for PD and LGD computation, with the date of the first observed default. To achieve the goal of consistency in model outputs and comparability of RWAs, we would welcome a clear definition of ‘limited timeframe’ instead of giving an example.
The provisions introduced in Article 60 are sufficiently clear.
We support the main objective of the RTS which from our perspective appears to move towards harmonising supervisory methodologies and to provide more guidance and clarity on interpretations of various Articles of the CRR. The costs of adjusting systems and policies accompanying the changes need to be taken into account. The exact impact will depend on further clarifications on definitions and questions we have raised in our consultation response. Once we know what, for instance, “material portfolios” and “reasonable discrepancies” are, we will be able to forecast a clearer impact assessment.
In addition to the guidance and clarifications already given by the draft RTS, which is helpful, standardisation of supervisory practices is welcomed. Notably, we support the specific proposal on calculating the shortfall on the IRB (Article 73 (h)). This together our request for additional clarifications that we have put forth in this response, the draft RTS should provide the necessary harmonisation within the European context.
The implementation of some of the requirements of this RTS will likely trigger changes within rating systems. As stated in our answer to question 7 the exact impact will depend heavily on further clarifications on definitions and questions we have raised in our consultation response. Below a list of some of the main changes and the corresponding articles referred to in our response:
- Article 4: outsourcing, depending on the definition and scope of outsourcing.
- Article 10 and Article 11: could lead to changes in the validation process.
- Article 28, 1 (b): regarding retail exposures, in case of no limitation of this paragraph to non-retail exposures as is provided for by article 172 and 178 of the CRR should be factored.
- Article 28, 3 (b): in case of acquisitions.
- Article 52: the requirements outlined in the article could trigger high implementation effort changes in the source systems and underlying data.