The EBA NII methodology is constant balance sheet, volume constant and behaviourally capped. Banks that run stress NII like an internal IRRBB exercise get floored by the top-down: the EBA challenger typically reads 150-300 bps worse than a bank's internal model, and the gap is almost always explained by one of three bank-side errors.
Audience: Head of ALM, Group Treasurer, Head of IRRBB Modelling, Head of Behavioural.
The EBA EU-wide stress test runs NII on a constant balance sheet — no new lending, no new deposits, no asset rotation through the horizon. Volumes are held constant. Behavioural assumptions are constrained by caps the EBA methodological note specifies explicitly. A bank that treats the EBA exercise like its internal IRRBB run — with dynamic rebalancing, full behavioural freedom and asymmetric deposit-beta — gets a submission that the top-down methodological challenger flags as “materially below peer.”
The challenge cycle is brutal. The EBA reads the bank's NII projection, compares it against the top-down benchmark (built from aggregated COREP, FINREP and macro inputs), and writes a methodology challenge when the gap is unexplained. In 2023 and 2025, roughly 40% of the major NII methodology critiques concerned one of three bank-side errors we document below.
Every EBA cycle runs NII on three constraints that interlock. Understanding which constraint binds first — and on which book — is the fastest diagnostic for a projection that looks weak.
Every asset and liability is projected by contractual re-pricing date under the scenario interest-rate path. No rebalancing. A 5-year fixed-rate mortgage re-prices once at maturity, regardless of the bank's internal expectation of refinance. Floating-rate positions re-price on their schedule. Basis spreads (OIS vs IBOR vs risk-free) are held.
Methodology ref: EBA methodological note of the cycle — re-pricing gap section; sets the contractual treatment.
Behavioural pass-through on non-maturity deposits is governed by the deposit-beta cap. The EBA sets a floor on beta under rising rates (to prevent banks under-shocking) and a ceiling on beta under falling rates (to prevent banks over-claiming NII benefit). Prepayment is also capped — a bank cannot assume retail mortgage prepayment responds as aggressively to rate shocks as the internal model might suggest.
Methodology ref: EBA methodological note — behavioural floor/ceiling caps on deposit beta and retail prepayment.
Total assets held constant at T-0 in volume terms, across the horizon. No growth, no deleveraging, no strategic shift. When a loan amortises, it is replaced at the contractual re-pricing rate — not the market rate, not the bank's strategic pricing. When a deposit runs off, it is replaced at the EBA assumed rate. This is the constraint most banks under-respect.
Methodology ref: EBA methodological note — constant balance-sheet assumption; static rails.
Drawn from EBA public transparency disclosures for both cycles. The figures below are bank-level NII cumulative deviation vs the 2022 baseline (2023 cycle) and vs the 2024 baseline (2025 cycle), across the 3-year adverse horizon. Full dataset downloadable at our public stress-test dataset.
| Quartile | 2023 cycle (3-yr cumulative) | 2025 cycle (3-yr cumulative) | Pattern |
|---|---|---|---|
| Best quartile | −5% to −6% | −5% to −6% | Well-hedged floater books, modest deposit beta |
| Sector median | −8% to −9% | −8% to −9% | Reasonable hedging, central methodology |
| Worst quartile | −10% to −12% | −10% to −12% | Concentration in fixed-rate assets, asymmetric deposit beta |
Source: EBA 2023 and 2025 EU-wide stress test bank-level transparency disclosures, NII projection under adverse. Cross-checked in the Ezelman EU G-SIB stress-test dataset.
The sector median has been remarkably stable across the two cycles — roughly 300 bps of NII cumulative loss over 3 years under adverse. A bank whose 2025 projection was materially outside this range (either much worse, or much better) should have a specific diagnostic to run before 2027 templates land.
Every NII submission we review for a client shows at least one of these three errors. In 2025, across the 20+ bank-level transparency templates we benchmarked, all three appeared in the worst-quartile cohort.
The bank builds a replication portfolio for non-maturity deposits in the internal IRRBB framework — typically a ladder of 1-to-5-year bonds with dynamic rebalancing. When the adverse scenario hits, the internal model continues to rebalance the ladder, locking in higher forward-rates as they arise. The EBA constant-balance-sheet constraint does not permit this. The replication must be frozen at T-0 and allowed to run off naturally. The projection gap is often 50-120 bps of NII across the 3-year horizon.
Diagnostic: if the bank's internal ALM committee runs quarterly replication-portfolio rebalancing, the EBA submission is at risk unless the T-0 freeze is explicit.
Historical deposit-beta is measured over cycles where retail rates were generally falling or stable. The resulting beta is low (typically 20-40% for retail current accounts in the euro area). The bank applies the same low beta under the adverse rising-rate scenario — but retail competition is asymmetric: deposit betas accelerate as rates rise (depositors notice competitor offers) and are sticky as rates fall. Under-shocking the rising-rate beta is the single largest contributor to the EBA NII challenge.
Diagnostic: if the bank uses a single symmetric beta number rather than asymmetric rising/falling betas, and the rising-beta is < 50% for retail current accounts, the EBA will challenge it.
Under the internal IRRBB framework, the bank typically nets macro hedges (interest-rate swaps) against the underlying book — the hedges are treated as permanently effective. Under the adverse scenario, basis moves (OIS vs IBOR, euro-area vs US$, vanilla vs inflation-linked) can break the effectiveness. The bank that assumes full hedge effectiveness across the 3-year adverse horizon is over-stating the hedged-NII offset. Typical projection gap: 30-80 bps.
Diagnostic: if the bank does not run an explicit hedge-effectiveness test under the stress basis paths (and document it in the submission), the EBA will ask, and the bank will be unable to defend.
The EBA requires a non-maturity deposit run-off profile under stress. Most banks use their internal operational assumption (typically 5-10% annual run-off for stable retail). Under adverse macro — rising unemployment, falling HPI — run-off accelerates. The EBA expects the bank to apply a stressed run-off (often 15-25% for non-core retail). Under-modelling the run-off means under-modelling the funding cost hit.
Diagnostic: if the non-maturity deposit run-off under adverse is the same as under baseline, it will be flagged.
In the mandates we run, each of the three core errors is addressed through a named workstream with explicit deliverables. These are not theoretical remediations — they are what we ship.
We produce a T-0 replication-portfolio snapshot, signed by the Head of ALM, that is frozen for the full stress horizon. Every quarterly internal rebalancing is excluded. A separate “EBA-view” replication portfolio runs in parallel to the internal model, documented and reconciled monthly. Written into the ANCHOR-Harmonise pillar of our framework.
We rebuild the behavioural engine on asymmetric betas: rising-rate beta is calibrated to the 2022-2024 rate-rise experience (where empirical retail betas in the euro area were 45-65% on current accounts, 60-80% on savings); falling-rate beta calibrated separately. This produces two numbers, not one, and the difference is documented in the methodology pack. Also runs through a peer-benchmark gate: if the rising-beta is materially below peer median, a challenger note is triggered.
We run the swap book through the full EBA adverse macro path — including basis shocks (OIS vs IBOR vs inflation-linked) — and measure effectiveness quarter by quarter. Where effectiveness breaks, the hedged-NII offset is reduced accordingly. The resulting adjustment is typically 20-60 bps on NII, and it is documented in the submission rather than discovered in the EBA challenge.
The 2025 EBA methodology carried the 2023 treatment forward with only technical refinements — tighter behavioural caps on deposit beta, revised prepayment tables, a more explicit treatment of cross-currency funding. The 2027 cycle is expected in the same direction: the EBA is unlikely to loosen the constant-balance-sheet rails or the behavioural caps. Banks that had a 2025 submission materially below peer median should run a specific diagnostic before the 2027 templates land.
The three diagnostics we recommend: