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?

For the purpose of reporting under Article 430a of the CRR, the existing definition of loss should continue to apply, i.e. economic losses as defined in Article 5(2) of the CRR and reported in accordance with Annex VII of the ITS on super-visory reporting (COREP C 15.00).

At the same time, some members have expressed the view that, for certain reporting purposes, consideration could be given to a simpler and more operational definition, such as the one used by the European Banking Authority in its benchmarking exercises.

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)?

The loss data reported under Article 430a CRR should not be used as a standalone basis for the assessment or recalibration of risk weights for real estate exposures under Articles 126(4) and 465(11).

Given its aggregated, backward‑looking and statistical nature, as well as the remaining heterogeneity in loss definitions and reporting practices, Article 430a data may at most serve as one complementary monitoring indicator, but not as a prudential input for determining capital requirements.

In this context, some members have observed that the heterogeneity in definitions and reporting practices could raise questions regarding consistency with the approach outlined in response to Question 1. It is therefore important to clarify that maintaining the existing definition of loss for reporting purposes does not imply that such data are suitable for prudential calibration, given their different objectives and limitations.

Any consideration of real estate risk weights should be guided by a clear separation between statistical monitoring tools and prudential calibration mechanisms. Capital requirements under Articles 126(4) and 465(11) should remain grounded in stable, risk‑sensitive and conceptually robust prudential principles, rather than being driven by mechanically derived outcomes from aggregated historical data.

In this context, loss data reported under Article 430a may contribute to a broader understanding of market developments and loss trends, but its role should be framed within a holistic and judgement‑based supervisory perspective, recognising:

• the structural heterogeneity of real estate markets,
• the limitations inherent in backward‑looking datasets, and
• the need to preserve regulatory predictability and proportionality.

Statistical information should therefore remain contextual and complementary, and should not evolve into an implicit or de facto determinant of minimum capital requirements.

Q3. Which elements of the real estate framework should be further simplified?

The real estate framework could be further simplified by prioritising the clarification and simplification of valuation requirements. First, ESBG invites the EBA to acknowledge the practical challenges associated with the “sustainable value” requirement in Article 229(1)(b)(ii) CRR, which obliges institutions to adjust the property value “to take into account the potential for the current market value to be significantly above the value that would be sustainable over the life of the loan.” This concept stems from the Basel framework and was intended to ensure prudence in collateral valuation. However, it has proven extremely difficult to operationalise in a consistent and harmonised manner across Member States, given the absence of a clear methodology for determining long‑term “sustainable” values and the lack of alignment with national valuation standards.

We fully recognise that this requirement sits in Level 1 legislation and cannot be amended through the present Guidelines. Nonetheless, its practical limitations are directly relevant to the broader aim of simplifying the real‑estate framework and ensuring consistent supervisory outcomes. The challenges associated with this requirement have also been acknowledged outside the EU. The UK Prudential Regulation Authority decided not to implement the “sustainable value” adjustment in its Basel 3.1 package on the grounds that it is complex to apply and risks generating inconsistent approaches among institutions.[1] This demonstrates that the EU would not be alone in reconsidering the usability and proportionality of this provision.

Given these implementation constraints, we encourage the EBA to reflect these issues in its assessment and to signal the need for a review of Article 229(1)(b)(ii) CRR in future legislative revisions to support a more harmonised and operationally workable approach.

Second, we would urge the EBA to clarify the intended scope of application of Article 229 CRR, in particular by recalling that its valuation rules apply exclusively to the Standardised Approach (SA) and the Foundation IRB (F IRB) approach, and not to institutions using the Advanced IRB (A IRB) approach. We note that the EBA’s CRM Guidelines for A IRB institutions (EBA/GL/2020/05) take the position that Article 229 should apply in the A IRB context; however, this interpretation is not supported by the Level 1 legislative framework. The Basel standards clearly place the property valuation concept relevant for LTV and collateral treatment in CRE 20.75(2), which forms part of the Standardised Approach. In the CRR, Article 108(1) explicitly states that Chapter 4 (Articles 192–239 CRR), which contains Article 229, applies to SA and F IRB only. As Level 3 texts cannot extend or override Level 1 provisions, Article 229 should therefore be understood as applying solely to SA exposures and to F IRB LGD calculations. A IRB institutions estimate their own LGDs based on the requirements in Articles 161 and 164, together with the LGD input floor under Article 230. For A IRB, prudent valuation of collateral values are embedded in Article 181(1)(e), and in the EBA Guidelines 2017/16 on PD and LGD estimation. These provisions provide a dedicated, fit for purpose framework for ensuring prudence in A IRB LGD estimation, rendering the application of Article 229 to A IRB unnecessary and not aligned with the structure of the CRR. It is worth noting that the use of data that does not reflect market prices or economic reality as model input and/or for historical data could compromise the quality of the models and the output data they generate. We therefore encourage the EBA to clarify in the final Guidelines that Article 229 applies only to SA and F IRB, in line with the Level 1 text.

Fourth, the EBA should clarify that CRR only imposes the use of the valuation determined in accordance with Article 229 for the sole purpose of RWA calculations. Competent authorities should therefore not challenge efficient loan origination practices. In particular, the EBA could recall that competent authorities should not impose the use of LTV in the lending policy of European banks.

Moreover, a further simplification of the real estate framework should focus on clarifying and streamlining the role of Article 430a, ensuring it remains a statistical monitoring tool and is not used as a quasi‑prudential input for capital decisions. Other relevant actions include reducing overlaps between prudential rules, macroprudential measures and reporting requirements, which currently address similar risks through multiple instruments.

All in all, simplification should aim at improving clarity and operational usability, while preserving supervisory judgement and risk sensitivity.

[1] Quote: “The PRA agrees with respondents that the proposed requirement to adjust a valuation to reflect the value of the property that would be sustainable over the life of the loan could be complex for firms to operationalise and may lead to inconsistent approaches. Having considered the responses, the PRA has amended its draft rules to remove this requirement.”

Source: PS9/24 – Implementation of the Basel 3.1 standards near-final part 2 | Bank of England

Q4. Which other clarifications do you consider necessary to apply the new ECAI framework?

Further clarification would be useful on the treatment of new or modified ECAI rating products under the existing EU‑harmonised mapping framework, in particular to confirm when such ratings can continue to rely on already approved rating scales without requiring additional mapping or supervisory reassessment.

This would help ensure a consistent and pragmatic application of the ECAI framework across institutions, without introducing unnecessary complexity or reopening prudential calibrations.

In addition, some members have emphasised the need for timely and formal confirmation of this approach (for instance, in the form of an opinion by the EBA), given that the proposed pragmatic treatment remains subject to confirmation. This would support institutions in progressing with internal processes, including where necessary the timely initiation of licensing procedures.

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?

Overall, ESBG considers that regulatory stability should be preserved as a priority, as no significant redundancies have been identified in the existing IRB products. Maintaining a stable regulatory environment is therefore preferable to undertaking further structural changes, particularly where such changes could introduce additional complexity without a commensurate prudential benefit, and potentially new requirements.

At the same time, some ESBG members have noted that a more consolidated presentation of regulatory products related to credit risk could be beneficial, provided this does not result in substantive changes to the underlying requirements. The current framework comprises multiple documents issued by different authorities, including the ECB’s Guide to Internal Models (2025), the EBA’s Guidelines on PD and LGD Estimation (2017), the EBA’s Report on the Role of Environmental and Social Risks in the Prudential Framework (2023), the EBA’s Guidelines on ESG Risk Management (2025), and the ECB’s Guide on Climate-related and Environmental Risks (2020).

While these documents should be complementary, they partly overlap in scope and differ in emphasis, level of detail, and timing of implementation. This may give rise to interpretational challenges and, in some cases, to supervisory expectations that are not fully aligned, for example with regard to the treatment of ESG-related risk drivers in internal models.

Against this background, a more streamlined and integrated articulation of existing requirements—without introducing new obligations—could support greater coherence, reduce interpretational complexity, and enhance regulatory certainty for institutions. Any such consolidation should, however, remain proportionate and carefully designed so as not to undermine the benefits of regulatory stability.

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?

Views within ESBG on the integration of environmental and social factors into the credit risk framework are mixed. While some members consider that this approach cannot be further enhanced without compromising simplicity, others are supportive of potential enhancements.

Within the first group, it is understood that the integration of these risks would increase complexity. Accordingly, the priority should instead be to clarify how climate-related transmission channels are to be considered, without leading to double counting, and to avoid extending modelling horizons beyond those applied under the IRB framework. In this context, the appropriate venue for addressing these risks remains ICAAP and stress testing, rather than their incorporation into Pillar 1 requirements.

Conversely, from the other standpoint, members have identified areas requiring additional clarification relate primarily to the treatment of ESG risk drivers in internal models, where current regulatory expectations across different guidance documents are not fully aligned.

For example, the ECB’s Guide to Internal Models (2025) explicitly expects institutions to incorporate all relevant and material ESG risk drivers into PD and LGD models, supported by modelling data that sufficiently captures these risks. At the same time, the guide requires the application of margins of conservatism (MoCs) where information is missing or uncertain, and the use of (conservative) overrides if ESG drivers cannot yet be directly included in the models.

The EBA’s Guidelines on PD and LGD estimation (2017) and its Report on the role of environmental and social risks in the prudential framework (2023), in turn, take a more cautious approach. They emphasise that ESG risk drivers should only be integrated into the risk differentiation model when sufficient and reliable information is available and model performance is not adversely affected. The EBA further stresses that model redevelopment or recalibration should occur only in the medium to long term, once the actual impact of ESG risks on defaults and losses becomes empirically observable. Until then, MoCs and selective, temporary overrides may be applied, but only within the boundaries of existing guidelines and only for a limited number of cases.

Additional regulatory documents highlight further operational challenges for the industry, such as the limited availability and representativeness of historical ESG‑related data, the uncertainty surrounding transmission channels, and the need to consider both micro‑ and macroeconomic effects of environmental risks. They also underline that ESG risks may affect both default likelihood and loss severity, and that institutions must ensure robust data processes, scenario-based assessments, monitoring of ESG-related losses, and sustainable collateral valuations over the life of the exposure.

Given these partly diverging expectations, it has been suggested that it would be highly beneficial for the sector if regulators provided a more harmonised and pragmatic framework for integrating ESG risks into internal credit risk models.

Additionally, some members have emphasised the importance of the following principles:

1. ESG risk factors should be treated as credit risk factors, in the same way as any other risk driver: ESG risk factors should not receive a conceptually different treatment from other credit risk drivers. Accordingly, they should only be integrated into PD, LGD or CCF models when sufficient, reliable and material information is available and when such integration effectively improves the predictive power and performance of the models.
2. Their incorporation should not be forced, nor should modelling horizons be extended beyond the IRB framework: In the absence of adequate data or robust empirical evidence, the inclusion of ESG risk factors into IRB parameters should not be forced, nor should it justify extending modelling horizons beyond one year, as this would be inconsistent with the principles of the IRB framework and could undermine model stability
3. Explicit avoidance of double counting: It is particularly important that regulatory clarification explicitly avoids any form of double counting. For instance, the effects of physical or transition risks on real estate exposures that are already reflected in collateral valuations should not be captured again through additional haircuts or parallel adjustments.
4. Very exceptional use of margins of conservatism (MoCs): In line with the treatment of other risk factors, ESG related MoCs should only be applied in genuinely exceptional cases, and should not become a generalised tool to compensate for structural data gaps or broad uncertainty.
5. A conservative use of overrides to incorporate ESG risks would not favour a best estimate approach.

Overall, the objective would be for the EBA to clarify this framework without over‑emphasising ESG risks relative to other credit risk drivers, while preserving proportionality, the simplicity of the IRB framework, and avoiding developments that could generate inconsistencies or unnecessary duplication.

In this regard, it is suggested that harmonisation would reduce interpretational differences, avoid unnecessary supervisory discussions during IMIs and OSIs, and support a more consistent, efficient, and proportionate integration of ESG risks into the credit risk framework.

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)?

A clarification would be welcome regarding the conditions under which continuous PD and LGD models are considered acceptable within the IRB framework.

While certain provisions of the CRR (e.g. Article 169(3)) could be interpreted as allowing continuous estimation approaches, current supervisory practices and the interpretation of the EBA Guidelines on PD and LGD estimation appear to favour discrete rating grades or pools. In practice, continuous models that are not embedded within a sufficiently granular segmentation framework have often not been accepted, creating uncertainty for institutions.

To ensure consistency with existing IRB requirements and supervisory expectations, it would be helpful to clarify that continuous models may be used as estimation tools, provided that their outputs are appropriately discretised into rating grades or pools that meet established minimum standards. In particular, such discretisation should ensure an adequate level of homogeneity within segments and differentiation of risk across segments, comparable to traditional grade structures.

Furthermore, the discretised framework should enable performance measurement and validation at a level of granularity equivalent to rating grades, including back-testing, calibration, and discriminatory power assessments in line with established IRB practices.

Clarifying this point would help align regulatory interpretation with supervisory expectations and provide greater certainty to institutions regarding the acceptable use of continuous modelling approaches.

Q8. Which requirements should apply in the application phase of continuous models (e.g. overrides)?

Clarification would be welcome on how override requirements should be applied in the context of continuous models.

Existing regulatory provisions, in particular Article 172(3) CRR and the EBA Guidelines on PD estimation, LGD estimation and the treatment of defaulted exposures, already set out general principles for overrides, including requirements on justification, governance, and documentation. These principles remain relevant and applicable irrespective of whether models are based on discrete grades or continuous risk measures. As such, additional prescriptive requirements may not be necessary.

At the same time, the application of overrides in continuous models raises specific operational considerations. In particular, where continuous risk estimates are used, overrides should be designed in a way that preserves the integrity of the model’s structure and risk differentiation. One possible approach is to implement overrides as continuous adjustments (e.g. through shifts in log-odds or equivalent transformations), ensuring consistency with the functional form of the model and maintaining the rank ordering of obligors.

Many institutions already complement continuous models with a discretised framework (e.g. mapping to grades or pools) for validation and governance purposes. Such discretisation can also provide a practical basis for applying overrides, for example by assigning exposures to an adjusted grade or pool and applying the corresponding representative risk parameter.

Overall, it would be beneficial to clarify that institutions may apply overrides either at the level of the continuous estimate or via an associated discretised framework, provided that:

• the approach is consistent with the underlying model structure,
• risk differentiation and rank ordering are not unduly distorted,
• overrides are properly justified, documented, and subject to governance in line with existing regulatory requirements.

Such clarification would ensure consistency with current regulation while providing flexibility for institutions to implement overrides in a manner appropriate to continuous modelling approaches.

Q9. Which challenges have you encountered in implementing the new CRR definition of facility?

None.

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 proposed adjustments to the representativeness requirements for CCF are broadly supported and appear suitable for extension to PD and LGD risk parameters. A more harmonised framework across all three parameters would enhance consistency, reduce ambiguity, and help avoid situations where differing interpretations of regulatory requirements could lead to overlapping or conflicting supervisory expectations.

Further clarification of how to assess and address potential lack of representativeness across samples would be beneficial. Aligning the treatment of this aspect across PD, LGD, and CCF would contribute to greater transparency and consistency in both model development and supervisory assessment. At the same time, certain adaptations may be necessary to reflect the specific characteristics of PD and LGD models. For instance, the identification and treatment of key risk characteristics may require adjustment to ensure appropriate application to PD models, while LGD models may involve additional considerations related to recovery processes and collateral. Additionally, it should be recognised that the framework for assessing representativeness as described in the CCF guidelines may not be fully applicable to PD and LGD parameters. While a simplification focused on representativeness is welcome, certain elements, such as the distinction between the development sample and the testing sample within the risk differentiation sample, may be difficult to implement in practice, notably due to constraints in data availability.

Finally, a more risk-based and performance-oriented approach to representativeness could be considered, particularly in the context of development samples. Allowing a degree of flexibility where data limitations exist (such as in low-default portfolios) while maintaining robust requirements for calibration and validation datasets, would support model reliability without introducing unnecessary complexity.

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?

ESBG considers that further simplification of the representativeness requirements for PD and LGD would be welcome. In particular, aligning these requirements with the approach proposed for the CCF parameter would enhance consistency across risk parameters, reduce ambiguity, and help avoid situations where differing interpretations of regulatory frameworks could lead to overlapping or conflicting supervisory expectations.

The proposed clarifications in the CCF framework, including the explicit treatment of potential lack of representativeness across different samples, are viewed positively. Extending a similar approach to PD and LGD would improve transparency and reduce the scope for divergent interpretation in both model development and supervisory assessment.

It should be noted that certain targeted adjustments may be necessary to reflect the specific characteristics of PD and LGD models. For instance, the formulation in paragraph 52 of EBA/CP/2025/10 regarding key risk characteristics may require refinement to ensure appropriate application to PD models, where the identification and role of risk drivers differ from those typically considered in the CCF context.

Q13. Should these simplifications be pursued? Do you have any preferred approaches with respect to these simplifications?

The proposed simplifications are generally welcomed and should be pursued, as they have the potential to reduce unnecessary complexity and improve the usability of the regulatory framework. However, simplifications should be designed in a way that ensures they constitute a genuine reduction in burden for institutions, rather than creating additional interpretational challenges or becoming a de facto supervisory benchmark.

In this respect, simplified approaches should be applied only in clearly defined situations and should be understood as practical alternatives for cases where the calculation of a given component is particularly challenging or where the necessary data are not available. They should not be regarded as a floor or default minimum standard across all cases. Should their use be subject to extensive justification, strict back-testing requirements, or repeated supervisory discussion, the intended benefit of simplification would be significantly reduced.

More generally, simplifications should preserve sufficient flexibility and risk sensitivity. In particular, fallback approaches should not become effectively mandatory where institutions are able to support more risk-sensitive estimates through robust data and analysis. For example, making fallback approaches for downturn LGD or CCF effectively compulsory could lead to overly conservative outcomes, particularly in jurisdictions where no downturn has been identified over extended periods. This could result in own funds requirements that do not adequately reflect the underlying risk and introduce inconsistencies across jurisdictions depending on supervisory interpretations.

At the same time, simplifications are particularly valuable in areas where the current framework is overly complex relative to the benefits in terms of risk sensitivity. For instance, the proposed simplifications regarding direct and indirect costs are welcomed, as they enhance clarity and comparability across institutions. Similarly, simplifying the treatment of LGD for defaulted exposures would help reduce modelling burden without materially weakening the quality of risk estimates.

Caution nevertheless is warranted in areas where standardisation may not be appropriate. For example, standardising certain components such as MoC C may not be suitable in view of the heterogeneity of the portfolios to which they are applied (C10). 

Likewise, some members are of the view that introducing additional alternative approaches, especially for the calibration of long-run averages (LRA) for credit conversion factor (CCF), may increase modelling burden if institutions are required to justify differences between methods, particularly where existing regulatory provisions already prescribe a specific approach (C15).This proposal comes at an unfortunate point when it is clear to most banks that the subsample of observations 12-months prior to default is not a random sub-set of all facilities defaulting within the next 12 months. In this vein, the CRR clearly prescribes the fixed-horizon approach.

However, on this point, other members have highlighted that the implications of this approach could benefit from further clarification. In particular, a cohort-based approach incorporating a variable time horizon may better capture client behaviour between 12 months prior to default and the date of default, and could therefore be more appropriate for modelling CCF in low-default portfolios.

Q14. Do you have any comments and suggestions with reference to the calibration of the fall back approaches?

ESBG considers that fall-back approaches can play a useful role within the framework, but their calibration requires careful consideration to ensure that they remain meaningful and proportionate.

These approaches should not be designed or applied as backstops. Their calibration should adequately reflect the underlying credit risk, rather than embedding excessive conservatism that may distort risk sensitivity. While it may be appropriate to acknowledge that simplified approaches are conservative by nature, this should be clearly articulated to avoid them being interpreted or used as benchmarks in supervisory assessments.

In this context, ESBG believes that overly conservative calibration may undermine the practical usability of fallback approaches. If such approaches are systematically calibrated at levels that are significantly above risk-sensitive estimates, they are unlikely to be used effectively, thereby limiting their effectiveness as simplification tools and potentially increasing modelling burden rather than reducing it.

Finally, fallback approaches should remain consistent with the CRR framework. In particular, the introduction of fixed approaches for entire types of exposures may raise concerns where it appears to deviate from the allocation of IRB and standardised approaches as prescribed in the CRR and related regulatory technical standards. Such elements could introduce additional complexity and uncertainty regarding method selection.

Q15. Do you see other potential simplification areas where the modelling burden is not commensurate to the gain in risk sensitivity?

Additional areas for simplification can be identified where the current modelling burden appears disproportionate to the resulting gain in risk sensitivity.

Additional simplification of the estimation of downturn LGD would prove beneficial, especially in situations where more than one downturn period is identified. Under the current framework, institutions may be required to perform full calibrations for multiple candidate downturn periods, including the incorporation of margins of conservatism, in order to identify the most adverse outcome in terms of RWA. This approach is operationally burdensome and often provides limited additional insight, as the downturn period that is most adverse from a calibration-sample perspective typically also leads to the most conservative RWA outcome. Where outcomes are similar across periods, the selection may also become unstable over time and sensitive to minor portfolio or data changes, potentially resulting in frequent re-identification of the “worst” period without a commensurate increase in risk sensitivity.

More generally, introducing a concept of “non-compliance with negligible effect” could significantly reduce unnecessary modelling burden. According to paragraph 50 of the EBA Guidelines on PD and LGD estimation, institutions are expected to correct models to ensure full compliance with the CRR and to rectify all data and methodological deficiencies within a reasonable timeframe. In practice, however, addressing certain deficiencies may require substantial redevelopment efforts despite having an immaterial impact on model outputs. For example, it is a standard practice in many institutions to use the outstanding amount at default to isolate outliers in LGD models. However, this stands in contradiction with paragraphs 122 and 162 of the EBA Guidelines on PD and LGD estimation, which require the use of appropriate risk driver values within the year prior to default for this purpose. Identifying such alternative drivers may not only be operationally burdensome but could also result in relationships that are unstable over time. Similarly, supervisory challenges related to discriminatory power or homogeneity in immaterial sub-segments of the portfolio may trigger remediation efforts that are disproportionate to their impact on the estimates.

Further simplification may also be warranted in the rating process for exposures with partial information. Current requirements may lead to overly conservative ratings where certain data are not available, even when other relevant information exists. For instance, for newly established companies where financial information is not yet available, ratings may still be derived from behavioural information. In such cases, the application of overly conservative assumptions may not accurately reflect the underlying risk.

Q16. What do you perceive as challenges in your capacity to collect appropriate data, in particular in relation to indirect costs?

According to participating credit institutions, there does not appear to be significant challenges related to the capacity to collect appropriate data for indirect costs. Banks generally consider that the necessary data can be obtained without undue difficulty, and this aspect of the framework does not seem to constitute a major operational constraint.

However, other data-related requirements may give rise to more substantial challenges. In particular, the requirement to fully replicate (i.e., back-simulation) the current Definition of Default over extended historical data series, potentially spanning more than 20 years, can be operationally complex. In many cases, the necessary data are either not available or only available to a limited extent, making full implementation difficult.

This may result in significant additional effort without a commensurate improvement in risk measurement. A more proportionate approach to historical data requirements in this area could therefore be considered, factoring in data availability constraints and the practical limitations of reconstructing long historical time series.

Q17. Do you agree with the approach proposed by EBA? Do you see further measures as necessary?

While we acknowledge the objective of enhancing simplicity and supervisory convergence, we consider that the approach proposed by the EBA may result in unjustified increases in risk‑weighted assets, as it could be interpreted and applied in practice as introducing an implicit floor for IRB models. In particular, the increased reliance on aggregated, standardised or simplified metrics—when used as reference points for assessing the appropriateness of existing prudential calibrations—may effectively constrain the outcomes of internal models, even where those models are compliant, risk‑sensitive and well‑performing. This creates a risk that simplification tools or monitoring indicators gradually evolve into de facto benchmarks, limiting the ability of IRB models to reflect portfolio‑specific risk characteristics.

In addition, we consider that further clarification would be particularly important in banking consolidation scenarios, where historical default and recovery data from integrated entities may not be representative of the acquiring institution’s current underwriting standards, recovery processes or risk profile. In several consolidation cases, the acquired entity’s historical lending and recovery practices have materially differed from those of the acquiring bank, often translating into structurally higher observed defaults in legacy datasets. In such circumstances, mechanically incorporating the acquired entity’s historical data into the acquiring institution’s IRB calibration may distort risk estimates and generate capital outcomes that are not reflective of the current and forward-looking risk characteristics of the portfolios to which the parameters are applied.

The IRB framework is grounded in the principles of representativeness, relevance and risk sensitivity, as reflected in CRR Articles 180(1)(e) (PD estimates reflecting current underwriting standards) and 180(1)(h) / 180(2)(e) (use of sufficiently long historical periods, subject to relevance), and further elaborated in the EBA Guidelines on PD and LGD estimation (EBA/GL/2017/16), which explicitly identify lending standards and recovery policies as key dimensions of representativeness. These provisions imply that, following a consolidation, institutions should be allowed to assess whether historical data from integrated entities remains predictive of the current portfolios and underwriting standards of the acquiring institution.

In this context, the option to exclude non‑representative historical data should be explicitly recognised as a fully acceptable and compliant outcome, alongside the possibility of applying appropriate adjustments and margins of conservatism. This is particularly important to ensure that requirements in CRR Articles 181(1)(a) and 182(1)(a)—to use “all observed defaults within the data sources” for LGD and CCF estimation—are not misinterpreted as a requirement to include all available historical data irrespective of representativeness. Rather, these provisions should be understood as preventing cherry‑picking within the selected relevant data sources, after a prior representativeness assessment consistent with Article 180 and the EBA Guidelines.

From an implementation perspective, this could be operationalised through a pragmatic framework that: (i) allows specific, portfolio‑level treatment for the back‑book of the acquired entity (e.g. segmentation or transitional arrangements where its legacy standards remain relevant), while (ii) avoiding mechanically importing and needing to adjust non‑representative legacy data into calibrations for the acquiring institution’s back‑book and new originations post‑merger when empirical analysis shows it is not predictive of the current risk profile.

Without such clarification, there is a risk that simplification tools, aggregated indicators or standardised benchmarks could effectively override institution‑specific representativeness assessments, leading to undue conservatism, operational burden and volatility, without commensurate gains in risk sensitivity, especially following mergers or acquisitions.

Upload files

Name of the organization