IRRBB explained: EVE, NII, and PRA expectations

It is year-end 2021. A UK retail bank’s treasury team submits its Internal Capital Adequacy Assessment Process (ICAAP). The yield curve has barely moved in a decade. Interest rate risk in the banking book (IRRBB) is quantified and filed in three pages. Twenty months later, the Bank of England has raised Bank Rate from 0.1% to a peak of 5.25% in August 2023, the sharpest tightening cycle in a generation.¹ That ICAAP describes a balance sheet that no longer exists. Its behavioural assumptions were calibrated to a near-zero-rate world; its economic value calculations were anchored to a flat curve.

IRRBB is the risk that changes in market interest rates will affect either the economic value of a bank’s equity or its future earnings, and how well a bank answers that question rests on its behavioural assumptions as much as on the size of the shock it tests them against. It carries no Pillar 1 minimum capital requirement: firms hold capital against it through the ICAAP, which the Prudential Regulation Authority (PRA) assesses in its Supervisory Review and Evaluation Process (SREP) against two measures, economic value of equity (EVE) and net interest income (NII).

What is IRRBB, and where does the repricing gap come from?

Every bank runs a structural mismatch between when its assets reprice and when its liabilities reprice. For example: on the asset side, fixed-rate mortgages reprice only on maturity. On the liability side, the retail deposits funding those mortgages reprice whenever the bank adjusts its administered rate1, which may or may not follow Bank Rate movements, while wholesale funding reprices contractually on short tenors. The gap between asset and liability repricing is the primary source of IRRBB.

The PRA sets the IRRBB add-on under SoP 5/15,² and the risk itself decomposes the way the European Banking Authority (EBA) describes it in its guidelines.³ Three components carry most of the weight:

1. Repricing risk: how the net present value (NPV) of banking-book positions moves when rates shift across the yield curve, because assets and liabilities reprice at different times.
2. Basis risk: assets and liabilities priced off different reference rates; lending at Bank Rate funded at the Sterling Overnight Index Average (SONIA) is exposed even in a parallel move.
3. Optionality risk: customers’ behavioural flexibility, the borrower who repays early, the depositor who moves balances when it pays to.

What is the difference between EVE and NII?

EVE takes the long view of the repricing gap: the present value of every rate-sensitive cash flow over its remaining life, re-priced under a shocked curve with nothing replaced at maturity. It catches the long-dated mismatches an earnings measure will not see.

NII takes the short view: interest income and expense over the next year on a constant balance sheet, before management can reposition the book. Repricing speed matters as much as magnitude: in a rising-rate environment a bank whose asset yields reprice ahead of its deposit costs may see margin expand, while one whose deposit costs rise faster than its asset yields will see compression.

The Basel Committee on Banking Supervision’s IRRBB standard (BCBS d368) requires both, and says why: "If a bank solely minimises its economic value risk by matching the repricing of its assets with liabilities beyond the short term, it could run the risk of earnings volatility."⁴ Supervisors benchmark on EVE; banks manage the day-to-day through NII.

Where do IRRBB behavioural assumptions go wrong?

EVE and NII are only as reliable as the behavioural assumptions feeding them. The 2021 ICAAP assumed deposits that sit still, prepayments that track falling rates, and deposit pricing that barely moves. Twenty months later, none of that held.

Non-maturity deposits (NMDs), such as current accounts and instant-access savings, have no contractual repricing date, so banks assign behavioural repricing betas and stability assumptions: how quickly deposit rates respond, and how much of the balance stays put. The EBA’s guidelines put a ceiling on how long banks may assume those balances stay put: a "five years cap on weighted average repricing maturity" for "certain retail and wholesale deposits without a specified maturity", on the grounds that the item is material to the measured impact of rate changes.³

A fixed-rate mortgage hands the borrower a free option against the bank: they can repay early. They exercise it when it pays them to. Rates fall, they refinance; rates rise, they hold onto their below-market loan. Banks capture this asymmetry with conditional prepayment rates (CPRs) that vary by rate scenario. Term depositors hold a mirror option pointing the other way: rates rise, they break the deposit; rates fall, they stay put. Either way the bank gets cash back at the least useful moment and doesn’t get it when it would help most. That asymmetry is the optionality sub-risk, and the Basel framework treats it separately because the cash flows themselves depend on the shock.⁴

The failure of Silicon Valley Bank (SVB) placed NMD stability assumptions under renewed scrutiny. BCBS Working Paper 47 reviewed empirical evidence on deposit stability and documented the speed at which confidence-driven outflows can dwarf rate-driven ones. Published in February 2026, it carries its authors’ views, not official BCBS guidance.⁵ The question for SREP: how much of the assumed stable base survives a rate shock arriving with a reputational event? Firms whose NMD assumptions are calibrated only to rate-driven history should stress-test them against that combined scenario, not a rate shock alone.

What is the Basel IRRBB outlier test, and how has d578 changed it?

Behavioural assumptions determine the cash flows; the shocks decide what those cash flows are tested against. The Basel IRRBB standard (BCBS d368) requires supervisors to identify outlier banks: those whose worst EVE decline across six prescribed rate shocks exceeds 15% of Tier 1 capital.

The calibration behind the six prescribed shocks stopped at December 2015 and so reflected a near-zero-rate decade. BCBS d578, published in July 2024 and effective from 1 January 2026, recalibrates the shocks against a data window that now takes in the 2022 to 2023 tightening, and sizes them further into the tail of observed moves. The table sets out what changed.⁶

Would the new shocks have caught what the 2021 bank’s ICAAP missed? Only partly. The recalibration widens the tested range, but that ICAAP failed on behaviour: assumptions fitted to a world about to end. A bigger shock applied to the wrong cash flows still gives the wrong answer.

How does a firm defend its IRRBB Pillar 2A add-on?

The outlier test is arithmetic; the Pillar 2A conversation is judgement. Our ICAAP guide covers how that assessment is built.⁷ The PRA sets the add-on for IRRBB under SoP 5/15 and its ICAAP expectations under SS31/15.⁸ The PRA reviews the firm’s IRRBB governance and internal measurement systems directly.

What a supervisor examines goes beyond the headline EVE figure. Under the Basel principles and the EBA’s guidelines,³ an internal measurement system judged unsatisfactory can be replaced with the standardised framework,⁴ so the questions that follow are the ones worth preparing for: how NMDs are allocated to time buckets, what evidence supports the stability and repricing beta assumptions, how prepayment scalars vary across the shocks, and how EVE and NII are reconciled when they point in different directions. The ICAAP that documents these judgements, with back-tested behavioural parameters, is better placed to defend its add-on than one that produces the headline number without showing the reasoning.

What should firms do before the recalibrated shocks apply?

For the 2021 bank, running IRRBB well would not have predicted the tightening. It would have shown what the balance sheet stood to lose: deposit betas back-tested against observed behaviour, prepayment assumptions stressed both ways, shocks sized to a history containing a rate cycle.

Firms whose EVE or NII positions approach the 15% Tier 1 threshold should also review the PRA Buffer2 sizing in their ICAAP. The recalibration leaves the threshold unchanged but moves more firms close to it.

Basel set 1 January 2026 for national supervisors to implement the recalibrated shocks, so the first ICAAP cycle after that date is where firms in implementing jurisdictions should expect them to bite.⁶ The PRA’s own transposition timetable is worth confirming directly. The time to check the numbers is before they appear in a supervisory context, not after.

Frequently asked questions