Response to consultation on draft ITS amending Commission Implementing Regulation (EU) 2024/3117 regarding Credit Risk and IFRS 9 Benchmarking reporting
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?
DNB raises material concerns about the introduction of template C 104.00 in its proposed form and questions whether the template adds sufficient value relative to its implementation costs. We identify two fundamental obstacles: the disproportionate burden of constructing a 15-year default history, and the structural comparability problem arising from changes in the definition of default.
The problem
Template C 104.00 requires reporting on facilities that entered default status in the last 15 years from the reference date and whose recovery process was concluded within the year before the reference date. The first reporting reference date is 31 December 2026 (Benchmark 2027), which means institutions would be required to report default vintages back to 2011.
This gives rise to two distinct and serious obstacles:
1. Disproportionate implementation cost
Constructing a reliable 15-year default history at the granularity required by C 104.00 (facility-level, by year of default, with concluded recovery outcomes) requires significant data infrastructure investment. While historical data is available, reconstructing it to the required level of granularity and consistency would require significant effort, with limited incremental analytical value given existing reporting.
EBA's benchmarking framework already captures long-run loss information through other means: column 0302 of template C 102.01 collects the model's long-run average LGD (calibrated over a minimum of 5 years per Article 181(1)(a) of Regulation (EU) No 575/2013), and column 0301 collects the realised loss rate over the past 5 years. The marginal analytical value of an additional 15-year raw dataset is therefore limited.
2. Structural lack of comparability for pre-2021 vintages
The reported series will combine structurally different default populations:
- Post-2021 defaults: classified under the harmonised definition of default (EBA/GL/2016/07), including the "unlikely to pay" criterion;
- Pre-2021 defaults: originated under earlier, divergent national definitions that did not consistently apply the harmonised criteria.
Unlike the treatment in column 0301 — where institutions are explicitly instructed to use the definition of default applicable at the time of the event (no backward simulation) — C 104.00 will necessarily mix cohorts that are not comparable across institutions. If pre-2021 data reflects different default populations across banks, cross-institution benchmarking of these vintages is inherently unreliable, regardless of how they are labelled.
Proposed solution
DNB's primary recommendation is that template C 104.00 be withdrawn. Given that the benchmarking framework already captures long-run loss information through columns 0301 and 0302 of template C 102.01, EBA should assess whether the standalone loss rate template adds sufficient marginal value to justify the implementation burden. If the information in C 104.00 is not materially additive relative to what is already reported, the template should be replaced by targeted enhancements to the existing templates.
Should EBA wish to retain template C 104.00, DNB proposes that pre-2021 vintages be made voluntary or marked as "not applicable": require mandatory reporting only for default vintages from 2021 onwards, while allowing institutions to report earlier vintages on a voluntary basis where reliable and comparable data are available under a consistent definition of default. This approach preserves the analytical design of C 104.00 while eliminating the most costly and least comparable portion of the data requirement.
Conclusion
DNB questions whether C 104.00 as currently designed delivers sufficient analytical value to justify its implementation costs. The 15-year horizon — stretching back to 2011 for the first reporting date of 31 December 2026 — creates a disproportionate data reconstruction burden, while pre-2021 vintages are structurally non-comparable across institutions due to divergent historical definitions of default. DNB urges EBA to withdraw the template, or at a minimum to make pre-2021 reporting voluntary or not applicable.