​The EBA publishes its annual assessment of banks’ internal approaches for capital requirements calculation

The European Banking Authority (EBA) today published its 2025 Reports on the annual market and credit risk benchmarking exercises, highlighting continued progress in the consistency and reliability of banks' internal models across the EU, while identifying areas for supervisory attention as key regulatory reforms approach full implementation.

For market risk, results confirm stable and relatively low dispersion across key risk metrics, reflecting improved data quality and convergence in modelling practices. For credit risk, the variability of the estimated PD shows a gradual reduction over the longer term.

Market risk: progress ahead of FRTB

The 2025 EU market risk benchmarking exercise indicates tangible improvements in data quality, consistency, and comparability across banks.

As introduced last year, the assessment is presented in two dedicated reports covering the Internal Model Approach (IMA) and the Alternative Standardised Approach (ASA). This improves transparency and analytical clarity, particularly given that the ASA is expected to play an increasingly central role under the forthcoming FRTB framework.

The IMA exercise, covering 43 EU banks across 13 jurisdictions, shows a marked improvement in data quality for Initial Market Valuation (IMV), with reduced dispersion across asset classes. Variability in Value-at-Risk (VaR) remains at historically low levels. However, higher dispersion persists in stressed VaR (sVaR) and Incremental Risk Charge (IRC).

Results for the ASA confirm its role as a more stable and comparable framework, with continued improvements in consistency. In particular, dispersion in the Sensitivities-Based Method (SBM) has declined further, reaching an average of 8%.

Key findings

IMA

• Greater consistency in FX instruments following supervisory guidance.
• Persistent dispersion in commodities due to small samples and varied practices.
• Higher variability remains in sVaR and IRC, despite stable VaR.
• Limited supervisory concerns; most deviations were considered justified.

ASA

• Further convergence strengthens ASA as a regulatory benchmark.
• Methodological issues remain in FX translation and equity sector mapping.
• SBM consistency improved, with dispersion falling to 8%.
• Seven banks were flagged for further supervisory review.

Credit risk: stable outcomes with gradual structural improvements

The credit risk benchmarking shows a stable overall picture alongside structural improvements linked to ongoing regulatory reforms. The share of Exposure at Default (EAD) under the Internal Ratings Based (IRB) approach has continued its gradual decline in recent years. Furthermore, the increase in approved material model changes across all asset classes suggests that the implementation of the IRB roadmap is advancing.

Over the 2015-2024 period:

• PD variability declined across several asset classes.
• LGD variability remained broadly stable, with a slight downward trend.

The underlying portfolio risk, represented by the long-run default rate computed over ten years, can partly explain PD variability. As expected, there is some correlation between PDs and average default rates, although some variability remains. For LGD, differences in collateralisation levels-captured by loan-to-value (LtV) ratios-explain only part of the observed variability.

Note to the editors

The EBA's annual benchmarking exercises are a key supervisory tool to assess the consistency and comparability of internal models across EU banks. They support supervisory work and contribute to enhancing confidence in the regulatory framework.

For credit risk, these exercises complement the EBA's broader roadmap to repair IRB models, a central pillar of the ongoing review of the IRB framework, alongside the finalisation of Basel III reforms.

By providing regular benchmarks and peer comparisons, the exercises enable competent authorities to identify undue variability, promote best practices, and ensure a level playing field across institutions.