Report results from the 2025 credit risk benchmarking exercise

General Information

This report presents the results of the 2025 credit risk benchmarking exercise conducted by the European Banking Authority (EBA) in 2026. It analyzes the variability of capital requirements based on the IRB (Internal Ratings-Based) approach in European banking institutions. The scope covers European banks applying the IRB approach, with data collected as of December 31, 2024 and analyzed between April and September 2025. Topics include IRB coverage, risk parameters (PD, LGD, CCF), comparison between IRB and IFRS 9, the impact of the IRB roadmap, temporal variability and its drivers.

Executive Summary

The report addresses the variability of capital requirements of European banks using the IRB approach for credit risk, pursuant to Article 78 of the CRD. This variability affects the comparability of regulatory requirements and thus financial stability. The 2025 exercise confirms progress in implementing the EBA IRB roadmap, aiming to reduce unjustified variability by harmonizing modeling and assessment practices. A significant share of IRB models is now compliant with PD and LGD guidelines, but a non-negligible proportion remains under modification or validation, reflecting the complexity and duration of validation processes. PD variability has decreased in recent years for several asset classes, while LGD variability remains stable or slightly decreases. The conservatism margin and collateralization level partially explain this variability. The report also highlights that IFRS 9 accounting PDs are generally more variable and often lower than IRB PDs, while IFRS 9 LGDs are generally lower than IRB LGDs, notably due to the absence of crisis adjustments and regulatory floors. Supervisory authorities use benchmarking results mainly as a complement to existing measures, noting improved model alignment but also persistent risk underestimations. The report recommends continuing full implementation of the IRB roadmap to further reduce unjustified capital requirement variability (pp. 6-7).

Context and Objectives

The benchmarking exercise is mandated by Article 78 of the CRD to monitor the variability of risk-weighted assets (RWA) amounts calculated by banks using the IRB approach. Since the introduction of this approach in 2013, excessive and unjustified variability in capital requirements between institutions has been identified as a major issue. To address this, the EBA launched an IRB roadmap in 2016 aiming to harmonize modeling and assessment practices, complemented by technical standards and guidelines. Concurrently, the ECB conducted the TRIM, a targeted review of internal models. This report aims to provide an annual update on risk parameter variability and the progress of IRB roadmap implementation, based on a representative sample of European institutions. Limitations include variability in implementation timelines by institution and asset class, as well as the complexity of models and validation processes (pp. 8-9, 18-19).

Summary of Key Points by Theme

IRB Coverage: The share of exposures covered by the IRB approach has decreased to below 50% of total performing exposures in June 2025, but remains high for certain asset classes such as corporates. Large banks have higher IRB coverage (57.6%) than smaller ones (13.3%). The implementation of new Basel 3 standards led to a notable increase in FIRB exposures from 2025 (pp. 9-10).

Risk Parameters by Asset Class: Over 2020-2024, PDs weighted by EAD slightly increased for some classes (corporates, other retail, credit cards). Observed default rates (DR) have increased since 2022 for most retail portfolios, but PDs have sometimes increased less, suggesting a possible decrease in conservatism margins. LGDs are overall stable, with a slight increase for some classes (pp. 11-13).

IRB vs IFRS 9 Comparison: Since 2024, IFRS 9 data have been collected for comparison. IFRS 9 PDs are often lower than IRB PDs, with wide dispersion, reflecting notably increased sensitivity to economic conditions and absence of regulatory floors. IFRS 9 LGDs are generally lower than IRB LGDs, especially for FIRB and mortgage portfolios, due notably to absence of crisis adjustments and floors (e.g., 10% for some real estate secured loans). IFRS 9 parameter variability is higher than IRB variability (pp. 14-16).

IRB Roadmap Implementation: The roadmap includes several regulatory phases between 2016 and 2022, aiming to harmonize evaluation methodology, default definition, PD and LGD estimation, and collateral treatment. As of September 2025, a significant share of IRB models complies with guidelines, but a significant proportion remains under modification or validation, reflecting process complexity. Supervisory authorities use benchmarking results mainly as a complement to existing measures, noting overall improvement but also persistent risk underestimations (pp. 17-21).

Temporal Variability: Over 2015-2024, PD variability weighted by EAD decreased for several asset classes (corporates, institutions, secured retail), while LGD variability remained stable or slightly decreased. This PD variability reduction is attributed to regulatory and supervisory efforts. LGD variability is influenced by structural factors such as credit policies and national insolvency differences (pp. 22-25).

Drivers of Variability: PD variability is partially explained by underlying risk, measured by the 10-year average default rates, but dispersion remains wide and sometimes inconsistent with these rates, suggesting other factors such as historical data length, conservatism margins, and calibration methods. For LGD, collateralization level measured by the loan-to-value (LTV) ratio explains about 15% of variability for corporates, with a less clear relationship for real estate secured portfolios. Other factors must therefore be considered to explain LGD variability (pp. 26-32).

Main Results and Insights

- Findings: IRB coverage is below 50% of total performing exposures, higher in large banks. PD variability has decreased since 2015 for several asset classes, while LGD variability remains stable. IFRS 9 PDs are generally more variable and often lower than IRB PDs. A significant share of IRB models complies with guidelines, but some models remain under modification. Supervisory authorities observe overall improvement but also persistent risk underestimations.

- Assumptions: PD variability should be mainly explained by underlying risk (10-year average defaults). LGD variability should be related to collateralization level (LTV).

- Interpretations: PD variability reduction is linked to IRB roadmap implementation and supervisory actions. LGD variability is influenced by structural and national legal factors. IFRS 9 PD and LGD variability is higher due to sensitivity to economic conditions and differing accounting practices.

- Uncertainties: Remaining PD and LGD variability is not fully explained by studied factors, notably PD dispersion relative to default rates and low LGD-LTV correlation. Historical data length, conservatism margins, and calibration methods are areas for further investigation.

Conclusions and Recommendations

The report concludes that IRB roadmap implementation is progressing, with notable improvement in IRB model compliance with PD and LGD guidelines, contributing to reducing unjustified capital requirement variability. However, a non-negligible proportion of models remains under modification or validation, limiting short-term convergence. Supervisory authorities must continue efforts to finalize model compliance, considering portfolio and institution specificities. The report emphasizes the importance of continuing to analyze residual variability sources, notably by deepening the study of conservatism margins, calibration methods, and structural factors influencing LGD. Continued convergence of practices and reduction of risk underestimations are essential to ensure comparability and robustness of capital requirements in the EU (pp. 19-21).