Response to discussion on the simplification and assessment of the credit risk framework
Q1. For the purpose of reporting under CRR Article 430a, which definition of loss should be used?
We believe that the definition of “loss” referenced for Article 430a reporting requires a clearer articulation, as the terminology used in the ITS is hardly achievable, notably when it comes to SA portfolios. To avoid heterogeneity of reporting practices and to ensure comparability across banks, a more precise, universal and simpler definition is needed. We would encourage the EBA to adopt a definition that is simple, operationally feasible and aligned with existing supervisory reporting, while avoiding unnecessary pro‑cyclicality in the measurement.
- The use of the same definition as for EBA benchmarking exercise, grounded on stage 3 provisioning [“The loss ratio is computed using as numerator the sum of credit risk adjustments and write-offs for those exposures which were classified as “defaulted exposures” in the last year …”], would allow to reach this goal.
- Any move towards an annual LGD-type calculation would be inappropriate and risk introducing undue procyclicality. The proposed approach ensures comparability, avoids unnecessary complexity, and reflects the realities of data availability across institutions.
Q2. Should the loss data (CRR Article 430a) be used for the assessment of RWs of real estate exposures under CRR Article 126(4) and CRR Article 465(11)?
- Our members consider that, while the intention to introduce stronger empirical grounding in the calibration of real estate risk weights is understandable, relying mechanically on observed loss data may introduce significant pro cyclical effects. Losses in real estate markets often exhibit strong temporal clustering, and a framework directly linked to short term loss observations could reinforce credit tightening during downturns. We would instead support an approach that makes use of supervisory loss datasets in a manner that smooths volatility and avoids mechanically amplifying market cycles.
- French institutions stress that any evolution in the use of Article 430a loss data — including adjustments to its perimeter or calculation basis — should be approached with utmost caution. This metric is referenced in several CRR provisions (notably for preferential risk weights, loan splitting under the Standardised Approach, output floor transitional arrangements and LGD input floors). Any modification without prior quantitative impact assessment could therefore generate unintended and widespread prudential effects. As a result, banks strongly oppose changes to the current perimeter or methodology in the absence of a comprehensive impact study.
Q3. Which elements of the real estate framework should be further simplified?
French institutions believe that the real estate framework could be further simplified by prioritising the clarification and simplification of valuation requirements and by streamlining the loss measurement components, which today create significant operational burden. A more explicit alignment between the SA requirements and the data actually available across Member States would improve usability, and any simplification should aim at reducing interpretational uncertainty regarding collateral valuation, haircuts, and reporting thresholds. These areas represent the most significant sources of operational complexity.
Indeed, although we understand the necessity to ensure legal certainty and prudent valuation for the purpose of collateral recognition in a sound manner, the real estate framework strongly lacks of proportionality in the conception of the operational requirements to be met, entailing highly complex processes to be implemented by institutions.
- For instance, CRR Art. 229 systematically requires to liaise with an independent valuer with due expertise for the purpose of valuation and revaluation and GL EBA on LOM only offer a few derogations to be able to rely on Automated Valuation Model in specific circumstances. This does not seem appropriately proportionate when it comes to highly granular portfolio of low risk, in well-developed and mature markets, especially in the context of extensive reliance on digital solutions in the economy. We would be supportive of a simplification of the valuation framework adapted to the risk profile of the exposure and to the materiality of the transaction while ensuring a reliable, independent and prudent assessment of collateral values. This framework may include i) the use of transaction based or price data for low risk, standardised loans, where such data are of sufficient quality, granularity and frequency to ensure prudent valuation - for this risk category, prices validated by sworn officers acting in the context of a mission comparable to a public service, or used to build official statistics on real estate prices, shall be deemed as fulfilling the requirement of a reliable, independent and prudent assessment of collateral values – and ii) less constrained reliance on advanced statistical models.
- Another example is the requirement imposed to the bank of having process to monitor that the property is adequately insured against damages (CRR 208.5). According to jurisdiction, national rules on insurance schemes coverage and/or observable insured rates should be sufficient to consider that these requirements are met.
In addition to the valuation process itself, the property value to be recognized in collateral is subject to several cumulative layers of conservatism:
- haircut of 45% in SA or 40% in Foundation
- prudent valuation which must be sustainable over the life of the loan
- cap to the average value based on historic observations
At least one of these constraints should be lifted/enlightened, since they address similar concerns on valuation certainty and add extra complexity to the framework.
In addition, institutions note that some CRR3 provisions related to the eligibility of collateral for properties under construction (notably under Article 124) have resulted in highly complex and operationally burdensome criteria, combining multiple cumulative conditions (use of the property, number of units, legal assurances, insurance mechanisms). Given the limited risk-differentiation benefits observed, banks consider that this framework could be significantly simplified and made more robust by relying on clearer and more principle‑based eligibility criteria. In addition, we would like to recall that the Draft RTS on equivalent legal mechanism state so strict and complex conditions for the recognition of the effect of a standardized by-law guarantee, both on the completion guarantee itself and the protection provider, that they simply make a level 1 text provision non applicable. We hence fully support the EU Commission’s proposal of amendments to the draft RTS.
Q4. Which other clarifications do you consider necessary to apply the new ECAI framework?
Our members appreciate the EBA’s willingness to facilitate the transition toward the new ECAI framework, yet express concern that very few rating agencies currently produce the “intrinsic” ratings required under the new rules. Overall, the current limited number of rating agencies producing an explicit rating without government support is not sufficient to cover the institution’s portfolio.
In this context, and pending the development of a sufficiently diversified supply of intrinsic ratings, French institutions consider it essential that competent authorities make full and consistent use of the discretion offered under Article 495e until its final expiry in December 2029. The EBA could play a key role by issuing an opinion encouraging authorities to apply this discretion in a harmonised manner, so as to avoid excessive reliance on a single rating provider and ensure a smooth and prudent transition.
Q5. Should the consolidation of regulatory products for credit risk be a priority or should the regulatory stability be preferable instead? Have you identified any redundancies in IRB products?
While French/EU banks support greater coherence and reduction of fragmentation in the regulatory architecture, institutions also caution against introducing further structural changes at a time when extensive CRR3 implementation work is still ongoing. A measured, stability‑oriented approach is preferable, particularly where consolidation would trigger redevelopment of existing IRB models or supervisory processes. Simplification is welcome when it reduces operational burden without undermining long‑term regulatory certainty or overwhelming existing implementation efforts.
Moreover, since the beginning of IRB repair program, publications on IRB modelling were spread over time, with sometimes several years between publications. For instance, EBA GL on PD-LGD estimation were published in 2017 while draft EBA GL on CCF estimation were shared in 2025. However, the draft GL on CCF estimation proposed changes to modelling assumptions also impacting LGD modelling (e.g. additional drawings, extrapolations) which would incur model redevelopment, when banks had mostly “IRB-repaired” their LGD models. We would like to stress that avoiding RWA volatility should also be targeted by the EBA when reflecting on simplification. Any aggregation of regulatory texts should come along with necessary adaptation of all related supervisory references to the consolidated legal basis written by the EBA, including non-binding supervisory guide and ECB expectations, in order to avoid new interpretative inconsistencies emerging from consolidation.
Q6. Do you consider that the integration of environmental and social risks into the credit risk framework could be further enhanced without undermining its simplicity? Which areas, if any, would you prioritise for further work or clarification?
We believe that ESG integration must be pursued in a gradual and proportionate manner. While the sector fully recognises the importance of reflecting ESG risks in credit processes, empirical data series remain insufficiently long and methodologically heterogeneous to support full Pillar 1 parameter integration. Enhancements should therefore prioritise areas where they can be implemented without generating double counting—most notably, improved supervisory guidance on the use of overrides and clearer articulation of the interaction between Pillar 1, Pillar 2 and stress testing. At this stage, additional complexity should be avoided to preserve simplicity and ensure data‑driven evolution.
We encourage the EBA to adopt a more collaborative working approach with the industry to ensure pragmatic, proportionate and operationally feasible outcomes.
- Overrides should be used in a proportionate and transparent manner, and the EBA’s reference to “conservative use” should be interpreted carefully so as not to impose undue conservatism or introduce a second layer of Pillar 1 add‑ons. Clarifying the expectations regarding documentation and justification of overrides would contribute to consistent supervisory treatment.
- The FBF proposes the establishment of a structured, collaborative working group between the EBA and the banking industry to address ESG risk drivers in IRB models. Given the current lack of sufficiently robust data, any integration should be gradual, with no automatic Pillar 2 add-ons, and aligned with existing regulatory timelines.
- Regarding the EBA will to ensure conservative use of overrides to incorporate ESG risks, we would like to underline that the conservative approach does not favour a best estimate approach.
Q7. Which requirements should apply in relation to the measurement of the performance of continuous models (e.g. Back-testing)? How could testing requirements be facilitated and enhanced for continuous models that are compliant with CRR, Part three, Title II, Chapter 3, Section 6 (Requirements for the IRB approach)?
/
Q8. Which requirements should apply in the application phase of continuous models (e.g. overrides)?
/
Q9. Which challenges have you encountered in implementing the new CRR definition of facility?
Please refer to Question 10.
Q10. Should a consistent and single facility definition be applied across all risk parameters?
Banks observe that implementing the same level of aggregation between risk parameter raises significant operational challenges. Existing model architectures were not always designed around a single, narrow facility concept, and transitioning to this definition requires extensive remapping of portfolios, redevelopment of model segmentation, and re‑evaluation of historical datasets. These changes can also disrupt internal systems and business processes. French institutions therefore encourage the EBA to keep the current flexibility in the modelling framework.
- French banks do not support a strict and universal application of the same facility definition across all risk parameters, as such an approach would generate significant distortions and operational burden (e.g. redevelopment of a major part of modelling portfolios) without improving risk sensitivity. While greater consistency is desirable where appropriate, PD, LGD and CCF parameters each rely on different empirical foundations, and enforcing a single granularity level would risk misalignment with underlying data realities. Flexibility should therefore be preserved, and deviations should be permitted where justified. We think that a misalignment between LGD and CCF calculation granularity does exist by design (LGD granularity aligned with recovery process, CCF granularity aligned with limit granting process). If for instance, LGD calculation granularity is not in line with the recovery process, there is a strong risk that the modelling and quantification of LGD calculation would be disconnected from economic sense.
- In addition, from a PD perspective, when the definition of default is at obligor level, it is not possible to ensure the same level of aggregation with both LGD and CCF. Besides, the restriction of IRB-CCF models to revolving commitments only makes it all the more difficult to align with LGD calculation granularity (as LGD models would cover both revolving and non-revolving products) with the CCF calculation granularity.
Q11. Are adjustments proposed in the representativeness requirement for the CCF parameter also suited for PD and LGD risk parameters? Which amendments would be needed to accommodate PD and LGD specificities?
The EBA stance is not sufficiently detailed at this stage to understand the concrete proposals of the EBA and would welcome more clarity on these points. If this is the case, extension of any flexibility to PD and LGD are welcome. If the proposed simplification is limited to the dimensions of representativeness, then this seems interesting to us but we believe that the CCF framework in its entirety should not apply
We would like to highlight the general difficulty to meet representativeness requirements especially for portfolios characterised by low default incidence or structural scarcity of observations. PD and LGD model specificities therefore require tailored expectations rather than a simple transposition of CCF concepts. For instance, possible use of external data should not be the only solution.
Q12: Do you consider further simplification of the representativeness requirement, as proposed for the CCF parameter, as necessary for PD and LGD and if so, what kind of simplification?
Yes. Our members consider simplified representativeness criteria as necessary for PD and LGD, particularly for Low‑Default Portfolios, specialised portfolios, and exposures for which empirical data is inherently limited. Simplifications could include broader tolerance ranges, alternative qualitative measures, or explicit allowances for expert‑based assessments where statistical representativeness cannot be achieved without compromising model stability.
- In general, more flexibility in the representativeness requirements is welcome under the conditions that MoC should not be systematically applied.
- In addition, we understand that EBA is possibly planning to write policy regarding likely range of variability of one-year default rates, we would like more flexibility in meeting the related EBA requirements.
Q13. Should these simplifications be pursued? Do you have any preferred approaches with respect to these simplifications?
As a general comment, we think that the EBA should clarify that an "optional fallback approach" is a choice of the bank. We also think that the conditions of application of such approaches should be flexible and feasible enough in order to provide simplification for the bank. For instance, it should not be expected from banks to apply all the usual EBA detailed requirements before applying the "optional fallback approaches". On top of that, we understand that there is optionality in the application of such proposed "optional fallback approaches", in order for banks to choose the most appropriate timing to implement (and avoiding mandatory model changes).
In addition, although we see some hint at low default portfolios in the paragraphs dedicated to Margins of Conservatism, we would welcome further EBA intention to provide more adapted requirements to low default portfolios.
We support pursuing simplifications that reduce modelling burden without eroding risk sensitivity. EU institutions welcome simplifications that provide clarity and reduce unnecessary divergence but insist that the EBA must avoid introducing simplifications that are in fact punitive, overly conservative or operationally disproportionate. The objective should remain to strengthen usability of the IRB framework, not to discourage its adoption.
- Regarding LGD: we think that it is positive to account for direct and indirect costs by using a relative increase of realised LGD, as an optional fallback approach.
- Regarding LGDD: we welcome a SA-like approach which brings an acceptable cost-effectiveness regarding the framework for defaulted exposures.
Regarding CCF: we welcome the possibility to introduce elements of the cohort approach having understood that it allows the possibility to use year-end modelling snapshots and believe this flexibility could be extended to other alternative approaches. Moreover We think that some elements will deserve further clarify from EBA.
See our answer to question 14.As a side comment also expressed during the consultation on GL on CCF estimation, we would like to remind to the EBA that level 1 text does impose a specific CCF-in-default approach.
Regarding Downturn LGD and CCF: we understand that the spirit of the EBA Guidelines on Downturn LGD and CCF estimation is to allow to use an observed impact when data is available, thus to provide more accurate downturn estimation when banks invest in retrieving the data and investing the downturn analysis. Thus, going for an approach similar to reference value or the fallback approach of the EBA Guidelines would not favour the best estimate approach, leading to conservative results which could be detrimental to most portfolios if applied (in particular with a paradigm shift for LDP).
We would welcome EBA views on replacing the fallback approach (LRA + 15%) of the EBA Guidelines on Downturn LGD and CCF estimation by an approach similar to reference value.
Regarding MoC: Our members welcome the proposed simplified MoC framework, particularly as a means of supporting a more level playing field and which would, in some cases, allow for a reduction in the often costly efforts undertaken by banks.
- For MoC A and MoC B, particularly regarding the requirement to quantify each deficiency separately, we suggest allowing a single aggregated quantification for groups of deficiencies, where appropriate. This would be particularly relevant where multiple deficiencies have a similar impact on risk parameters, for example when they affect risk differentiation rather than risk quantification. This could mitigate undue complexity and avoid potential double counting.
- For MoC C, we consider that the framework should continue to provide institutions with sufficient flexibility in its technical calibration, including the choice of appropriate methodologies and percentiles, in a manner proportionate to the nature, materiality, and complexity of their activities, as well as to the deficiencies identified. We strongly suggest clarifying MoC C through a principles‑based approach. Any detailed formula, if introduced, should be available only as a fallback option and should adequately capture the differences between high‑default and low‑default portfolios.
However EBA’s proposal is not sufficiently detailed, therefore we can only opine on the principles without having evidence that the outcomes will practically simplify the calibration of the different MoCs.
Q14. Do you have any comments and suggestions with reference to the calibration of the fall back approaches?
Yes. French banks consider that several fallback calibrations—including notably the LGD downturn add‑on and the fixed CCF fallback values—are excessively conservative, such that they may not constitute workable alternatives in practice. Overly punitive fallback parameters risk rendering optionality ineffective and could disincentivise the maintenance of IRB models. We encourage the EBA to reconsider these calibrations to ensure they remain proportionate and risk‑sensitive.
Regarding the incorporation of elements of the cohort approach in CCF and making a comparison with long-run average based on 12-month fixed horizon approach, we would welcome confirmation that the long-run average is performed the same way that under 12-month fixed horizon approach under draft EBA Guidelines on CCF estimation, thus using an arithmetic average of realised CCFs weighted by the number of default. We are however concerned that the condition to apply this fallback approach requires to explain any significant deviation from the long-run average under the 12-month fixed horizon approach. Such comparison between the two calculations is operationally burdensome especially for models with high volumetry of defaults as this will imply for instance to construct two different Reference Datasets (one for the calculation under 12-month fixed horizon approach, one for the calculation under cohort approach). Performing both exercises would therefore comes with significant struggles. We would welcome EBA further possibilities to integrate elements of the cohort approach under more simplified conditions (e.g. developing the model under cohort approach and LRA comparison with 12-month fixed horizon on a sample).
Q15. Do you see other potential simplification areas where the modelling burden is not commensurate to the gain in risk sensitivity?
EU institutions highlight portfolio areas such as Low‑Default Portfolios, such as specialised lending as example, where high modelling burden produces limited marginal gains in risk sensitivity.
Simplification efforts should prioritise these areas and look for ways to reduce complexity while maintaining meaningful supervisory assurance.
Regarding CCF: we think that the approach to propose a fixed CCF which is at least 100% is deterrent to use such optional fallback approach.
Q16. What do you perceive as challenges in your capacity to collect appropriate data, in particular in relation to indirect costs?
Indirect cost data remains particularly challenging to collect and validate. Historical tracking of restructuring costs, legal fees and other indirect components has not always been undertaken in a manner which is granular enough. Banks therefore support simplified, standardised approaches for indirect costs, provided they remain empirically grounded (e.g. no collection line by line, allocation rules can be used to redistribute the global costs at facility level).
Q17. Do you agree with the approach proposed by EBA? Do you see further measures as necessary?
The FBF appreciates the EBA’s commitment to strengthening and simplifying the credit risk framework and supports efforts to improve coherence, reduce fragmentation and promote harmonised supervisory practices. Our members nevertheless believe that many aspects of the Discussion Paper require careful refinement to ensure that simplification does not create unintended conservatism or operational disruption. We trust that the detailed feedback provided through this consultation—particularly with regard to fallback calibrations, facility definition in modelling, ESG integration, ECAI treatment and real estate exposures—will support the next steps of the EBA’s work and contribute to a more efficient, proportionate and risk‑sensitive prudential framework.