Response to consultation on draft ITS amending Commission Implementing Regulation (EU) 2024/3117 regarding Credit Risk and IFRS 9 Benchmarking reporting
General 1: Are the instructions and templates clear to the respondents? If not, what should be clarified?
(General remarks)
While acknowledging that the proposed changes are in principle reasonable and conducive to simplification, nonetheless they imply the need for extensive operational interventions and therefore the envisaged implementation timeline appears far too challenging. In this regard application not before benchmarking 2028 (as of end-2027) would be feasible; the 2027 reporting on 2026 data could be suspended, in order to allow institutions to adapt to the new framework ahead of the 2028 reporting on 2027 data.
A significant concern is the historical depth required by the exercise, as reconstructing the time series based on the new classifications and breakdown required is inherently complex. In this regard, a staggered approach is suggested, allowing to start with the historical depth available (e.g., for the new classifications stemming from CRR3, from 2025), until the required depth is matured. An important point to be clarified concerns the “Default rate past 5 years” and the “Loss rate past 5 years” required in template C102.01. These metrics are requested according to the exposure classes aligned with the new COREP framework, and therefore with CRR3. As this classification is new, applying it retroactively would require reassigning the exposures under the new logic, since they were not reported under those definitions before 2025. As currently formulated, this requirement appears effectively not feasible on the historical side.
Since the time series that can be realistically reconstructed is currently in the order of two years, a staggered approach is suggested: starting from the two-year series and progressively building it up by adding year-end snapshots as they become available, until the five-year depth is reached. The five-year default rate and loss rate would therefore be calculated and reported only once the time series has actually reached five years, while during the transitional phase institutions would report what can be genuinely reconstructed.
The same issue would arise in the completion of the new template C104, which likewise requires historical data to be reconstructed under the new CRR3 classification.
__________
General 1
Overall, the proposed instructions and templates are understandable in their general structure. However:
- With reference to the instructions for column 050 (Rating) in template C 101.01, we seek confirmation that row 095 (PD = 100% – Total Defaulted) also includes exposures classified as “sofferenze” under the Italian framework, which do not appear to fall within the scope of row 096 or row 097. In this case, row 095 would not represent the sum of rows 096 and 097.
- With reference to the instructions for column 204 (RWA without MoC, supervisory measures at parameter level and add-ons applied at portfolio level) in template C 102.01, we would appreciate further clarification on the meaning of “add-ons at portfolio level”, including illustrative examples.
General 2: Do the respondents identify any discrepancies between these templates and instructions and the calculation of the requirements set out in the underlying regulation?
We have not identified any discrepancies.
General 3: Do the respondents agree that the amended ITS on Supervisory Reporting allows to fulfil the requirements and purpose of the Article 78 of the CRD?
We agree that the amended ITS on Supervisory Reporting allows the requirements and purpose of Article 78 of the CRD to be fulfilled, with reference to the assessment of the quality of internal approaches, the monitoring of RWEA variability, and the provision of benchmark values to the competent authorities.
General 4: what is the information (please refer to single columns) that you see as particularly costly to be provided? Please provide detailed information about the implementation and running costs of the proposed changes, also compared to the current total costs.
The derivation of historical data, considering the substantial changes in the portfolio segmentation key, is extremely complex and costly. It requires a retrospective reconstruction of data over a period of up to four years for the completion of template C 102.01, as well as additional verification efforts for template C 104.
The required implementations entail a series of activities across business and IT functions (e.g., gap analysis, drafting of requirements, testing and implementation) that banks had not planned nor included in the budget.
General 5: what is the reporting level for banking groups that you deem more appropriate?
We believe that consolidated-level reporting is more appropriate, as it provides an overall view of the application of internal models.
The EU consolidated level is considered the most appropriate reporting level for banking groups, consistent with the level at which consolidated prudential supervision is exercised and at which capital and model decisions are taken, and it ensures full consistency with the COREP data reported at the same level. This choice is also fully in line with the cost-reduction objective pursued by this exercise.
Should it be deemed necessary to preserve the ability to analyse individual significant subsidiaries, we suggest that this be achieved through an option that Competent Authorities may activate on a targeted and proportionate basis, rather than through a systematic obligation extended to all entities. Such a modular approach would preserve simplification in the ordinary case, limiting the additional burden to cases where a specific and justified supervisory need exists.
CR 1: Do you foresee any issues with supplementing the sample of counterparties listed in C 100.00 by the Large Exposures counterparties reported in the Template C 27.00 at the same reference date in order to expand the sample of low default counterparties for the reporting for template C 101.01/.02?
It is unclear why all counterparties reported in template C27.00 are included in the scope, as this template also contains counterparties (mainly from template C29) with turnover below the threshold required for classification as a "Low Default Portfolio." As a result, counterparties classified as "High Default Portfolio" would also be reported in C101.
Furthermore, the inclusion of C27 counterparties would entail a significant implementation effort from an IT perspective.
Clarification is also sought on whether High Default Portfolio counterparties should be excluded from the integration of the C100.00 list. In addition, template COREP C27.00 also reports groups of connected clients: it should be clarified whether this type of counterparty shall likewise be excluded, considering the difficulty of providing data at the granularity required by template C101.01. Lastly, the inclusion of C27.00 counterparties would entail a significant implementation effort from an IT perspective, as well as changes to the current (as-is) processes.
CR 2: Do you identify any errors in the list of counterparties in C 100.00? If so, please report it to us.
Four counterparties have been identified in template C100.00 for which the “Portfolio Name” is missing. The affected counterparties are identified by the following Counterparty Codes:
- LC_2000089
- LC_200115
- LC_200152
- LC_200177
CR 3: Do you foresee any issues with the transfer of the portfolio/model split from the former template C 105.02 to the template C 101.01 and the new template C 102.01? Do you agree with the EBA’s assessment that the portfolio/model split in the new template C 102.01 is the preferred option compared to designing a separate template increasing the reporting details of the former templates C 105.01 and C 105.02?
As regards the proposed transfer of the portfolio/model split from the former template C105.02 to templates C101.01 and C102.01, clarification would be needed about how to represent portfolios that are associated with multiple model IDs (e.g., PD model, LGD model, ELbe model) in templates C101.01 and C102.01.
CR 4: Do you think it is possible to provide the details of the impacts of the RWEA add-ons and MoCs (columns 0203, 0204, 0205 of Template C 102.01) at the required segmentation level?
The proposed reporting requirement is considered feasible in principle. However, the detailed allocation of the impacts attributable to RWEA add-ons, Margins of Conservatism (MoCs) and supervisory measures at the granularity requested for columns 0203–0205 of template C102.01 is not currently available and would require specific system enhancements. Its implementation would require changes to the existing reporting infrastructure and data architecture, resulting in increased operational complexity, additional implementation effort and associated costs.
CR 5: Do you agree with the proposed simplification in the reporting for column 0202 (RWEA Standardised) of the template C 102.01? Do you deem the segmentation level for which the information is requested sufficient to highlight the key differences between the IRB approach and the Standardised Approach?
No comment
CR 6: The split by collateral type (column 0061 of new template C 102.01) differs from the split used in the current Benchmarking portfolios. Do you foresee any issues or disadvantages with the new definition and instructions as compared to the ones currently in use?
The new definitions may require re-mapping of collateral types, leading to implementation challenges and increased operational burden.
With a view to further simplifying the compilation of column 0061, we propose not to consider the coverage percentage, and to clarify how to report exposures secured by different types of collateral at the same time. For instance, an additional classification (e.g., "mixed collaterals") could be introduced to flag exposures associated with more than one type of collateral. This option would remove the need to perform comparisons on coverage percentages and on the prevalence of the different types of collateral.
CR 7: Do you see material obstacles in providing the information required by the column 0302 of the template C 102.01 (Long-run average realized loss rate of assigned facility grades)?
The requested field "Long-run average realised loss rate of assigned facility grades" appears to coincide with column 0153, "LGD without MoC, supervisory measures and downturn component". In both cases the reference is to the estimated LGD LRA which — consistently with the treatment described for column 0153 — coincides with the observed (realised) long-run average.
In this regard, it has to be noted that reporting at the individual grade level is not already required as mandatory and its inclusion in this template would entail a significant operational burden for banks not applying such granularity. This requirement thus appears only partially consistent with the existing regulatory requirements and might give rise to considerable implementation and reporting challenges:
- regulatory and modeling misalignment: in line with paragraph 161(b) of the EBA Guidelines on PD and LGD estimation (EBA/GL/2017/16), many institutions calculate and apply the LRA realized loss rate strictly at the calibration segment level, rather than at the individual grade or pool level.
- operational burden: as this metric is not inherently built or stored at a granular grade/pool level within the approved risk models, extracting it would require significant ad-hoc computational effort and structural data adjustments, without providing additional risk-management value.
- data quality and volatility risk: providing the LRA at a grade or pool level, where it is not directly used for modeling purposes, could lead to highly volatile, unrepresentative, and counter-intuitive results. This would undermine the comparability of the benchmarking exercise itself.
We would therefore suggest making this column optional or, alternatively, allowing institutions to report the LRA realized loss rate at the calibration segment level, consistently with their internal modeling approach approved under the EBA GL.
CR 8: Do you see material obstacles in providing the information required by the new loss rate template (C 104.00)? Would alternative measures of loss rates be more appropriate?
The loss rate template (C 104.00) has been introduced as a new reporting requirement and therefore does not constitute a simplification, as there are new fields not already processed for prudential reporting. While the requested information is available, there are initial challenges in reconciling such information with the data required for benchmarking purposes.
The requested information is deeply rooted in the LGD model estimation phase, whereas the existing benchmarking architecture primarily relies on application data. Merging these two dimensions (e.g., mapping model-specific estimation parameters directly into regulatory Asset Classes) creates a severe operational mismatch. To populate these fields, institutions would likely have to rely heavily on proxies, reducing the accuracy and consistency of the data provided to the EBA.
It has also to be considered that information previously used for model estimation purposes will now no longer be managed for this purpose due to mandatory switch from AIRB to FIRB for certain portfolios (e.g. large corporates). More generally speaking, based on the current instructions, most fields (excluding columns 0031 and 0032 related to LGD models) seem to apply across all asset classes, including those under the Standardized Approach (STD) or FIRB. This creates a disproportionate burden, given that, under STD and FIRB frameworks, institutions do not maintain or continuously monitor updated LGD/loss datasets, as these exposures do not rely on internal estimates. Forcing institutions to build ad-hoc tracking and monitoring systems solely for benchmarking purposes is highly inefficient.
The main concern is the historical depth required by the exercise, as reconstructing the oldest observations in the time series is inherently complex.
Clarification is also sought about whether, where repossession information is not available, may the corresponding columns be left missing.
Lastly, fulfilling the exercise requires a whole-life Reference Data Set which, for modelling purposes, is currently used solely as a calibration exercise for the validated component-based models.
In light of the above, we recommend narrowing the scope of template C104.00 to portfolios under the Advanced IRB (AIRB) approach and clarifying the exact granularity required for the asset class mapping, to avoid the systemic use of proxies.
CR 9.1: The split by facility type (column 0062 of template C 102.01) is designed to collect information on the functioning of the CCF models. The classification into buckets (facility types) of off-balance-sheet items is based on Annex I of Regulation (EU) No 575/2013. Do you think there exist other classifications (for example product types) of off-balance-sheet items that would be more useful for representing the CCF models?
In our view, the classification based on Annex I represents an appropriate and pragmatic choice, as it facilitates comparability across institutions and ensures coherence with the existing regulatory framework.
CR 9.2: Template C 101.01 and C 102.01 propose two alternatives to collect the information on the CCF, either by one column (column 0110 in C 101.01 and column 0120 in C 102.01) or by two separate columns (columns 0111/0112 in C 101.01 and columns 0121/0122 in C 102.01). Which of these reporting alternatives do you prefer?
Retaining a single column, in continuity with the current configuration, is deemed preferable in order to avoid an increase in complexity and implementation costs that would not be justified by significant informational benefits.
CR 10: Do you support the use and the proposed granularity of the common master scale to limit the number of reported rating grades? What alternative would you propose.
We support the use of a common master scale, as it improves comparability across institutions and the stability of benchmarking analyses over time. Its adoption should, however, be weighed against the implementation effort required to adapt to the new scale. In this respect, we would propose considering a master scale aligned with the one currently used for regulatory purposes in COREP (ref. template C08.03) and in Pillar III.
IFRS9 1: Do you agree with the proposed definition of SA institutions as relevant for the benchmarking reporting?
We concur with the definition of ‘SA institution’ as set out in the document: “It is also required to extend such reporting to relevant institutions that apply the Standardised approach under Part Three, Title II, Chapter 2, of that Regulation (SA institutions) that adopted the IFRS 9. It is proposed to define as relevant SA, those SA institutions falling in the definition of Large institutions according to Article 4 (1) (146) of the Regulation (EU) 575/13 (CRR) and provided that they are subject to IFRS 9 requirements.”
IFRS9 2: Do you agree with the proposal to refer to SA exposure classes in defining the portfolios for SA institutions?
We support the proposal to refer to SA exposure classes when defining portfolios for SA institutions. However, this change will require a material revision of the current setup of preparatory activities for the reporting templates.
Despite the expected simplification benefits in the medium to long term, significant constraints affect the feasibility of the proposed implementation timeline. In particular, the introduction of the new rules for the identification of counterparties and HDP clusters presents substantial operational challenges.
These challenges arise both from misaligned data production timelines and from structural differences in counterparty clustering methodologies between the IFRS 9 framework and the COREP/FINREP reporting frameworks.
As a result, the implementation will require a substantial and resource-intensive data reconciliation process. Under the current timeline, this is not considered operationally feasible without material execution risks.
For these reasons, we strongly recommend postponing the entry into force of the new framework to September 2028 and not conducting the 2027 exercise.