Supervisory outlier tests · RiskAtoms Academy

By the end you will be able to

1State the Basel outlier test and what supervisors do with an outlier bank.
2Run the EU's EVE and NII tests on the bank, and explain why the NII test flags it at 6.99% of Tier 1 and what follows.
3Apply post-shock floors: Basel's national floor, the EU's maturity-dependent lower bound, and the UK rule with none.
4Run the UK's ICAA 9.4A test before and after 1 July 2026, and read headroom and the Tier 1 that would pass.

Why it matters

In March 2023 Silicon Valley Bank failed with large unrealised losses on its bonds (L3.6.12: Credit spread risk in the banking book and the SVB case). Outlier tests are the supervisor's screen for that kind of exposure: one number per bank, compared with a share of its capital, to decide where to look harder. RiskAtoms Bank passes every value test with room to spare. Yet if it were an EU bank, one test would flag it, and the reason is the opposite of SVB's: its earnings fall when rates fall.

Core explanation

One screen, three rulebooks

Illustrative dataRiskAtoms Bank plc's banking book, under the rules

The same made-up UK bank as Level 3, at 31 December 2025, in £ millions: the banking book of M3.6: Interest rate risk in the banking book and ALM with its cash flows, its £19.2bn macro hedge and £16bn structural hedge, and its assumptions for deposits and prepayment. Module 3.6's changes in economic value (ΔEVE) and net interest income (ΔNII) stay the bank's own figures. Here the rules of Basel, the EU and the UK are applied to them one by one; they change only the shock sizes, the floor on post-shock rates and the capital the loss is compared with.

Tier 1. Every test on these pages uses the bank's Tier 1 of £3,419.9m from M4.1: The Basel framework (CET1 plus £400m of AT1). Module 3.6 took £3,500m as a round figure, so its percentages are slightly lower. Anything a page adds to the book is labelled illustrative.

Basel: the 15% test and what follows

Principle 12 makes supervisors publish how they identify an outlier bank. They must run at least one test comparing the bank's maximum ΔEVE under the six scenarios with 15% of its Tier 1, computed as for disclosure (SRP31.82). They may add others, applied in the same form across their jurisdiction, as long as the outlier threshold is at least as strict (31.83). An outlier is presumed to have undue IRRBB and is reviewed (31.84). A bank can also be judged to have excessive risk relative to earnings, if its shocked ΔNII would leave it without enough income to maintain its normal business operations (31.87).

Where the supervisor finds the risk undue or its management inadequate, it must require the bank to reduce the exposure (by hedging, for example), raise capital, accept limits on its internal risk parameters or improve its risk management, or a mix of these (31.88), within a set time frame (31.89). Being flagged is the start of a review, not a capital charge.

The bank's worst case is parallel up: £257.23m, or 7.52% of £3,419.9m. The limit is £513.0m, so the headroom is £255.8m, almost exactly as much again. Only if Tier 1 fell below £1,715m would the same loss flag the bank.

The EU: two tests

The EU writes the screen into law. Article 98(5) of the Capital Requirements Directive has two supervisory outlier tests: (a) economic value falling by more than 15% of Tier 1 in any of six scenarios, and (b) a large decline in net interest income in either of two. Delegated Regulation (EU) 2024/856, applying since 14 May 2024, sets the scenarios, the sizes and the common assumptions, and defines the large decline as more than 5% of Tier 1 over one year (Art 5(1)). On the EU's sterling sizes the bank's worst ΔEVE is £235.42m (6.88%), a pass. Its worst NII fall is £239.21m in parallel down: 6.99% of Tier 1, against a line of £171.0m. Flagged.

Why down? In a fall of rates the bank's floating-rate loans and the SONIA legs of its pay-fixed hedges earn less at once, while much of its deposit cost cannot follow: current accounts pay 0.5% and move a tenth of the way (L3.6.09: Non-maturity deposits). Module 3.6 scored the same test with Basel's 275bp and £3,500m: a fall of £263.27m, or 7.52% (7.70% on £3,419.9m). On the EU's own 250bp the figure is 6.99%. Either way the bank is above 5%.

TimelineWhat a flag sets off

1 / 5

Measure

The bank runs the tests

On the prescribed scenarios and common assumptions (2024/856 Arts 1–4 in the EU; ICAA 9.7 and 9.11 in the UK). In the UK, SS31/15 asks for the EVE calculation at least quarterly (2.9B, from 1 January 2027).

Floors on post-shock rates

A 300bp or 425bp cut from today's rates can push the short end below zero. Each rulebook handles that differently (post-shock floors):

Basel National supervisors may set floors, provided they are not above zero (SRP31.93). The book uses 0%: on Basel's sizes short down would take the overnight rate to −0.33%, so the floor trims the gain from £102.99m to £99.78m.
EU A lower bound of −150bp at immediate maturity rising 3bp a year to 0% at 50 years (2024/856 Art 3(7)). On the EU's smaller sizes the lowest rate the bank's curve reaches is 0.92%, far above it, so it never binds.
UK ICAA chapter 9, as read, sets no floor, so the UK test uses the shocked rates as they are.

EU short down	Today	Shocked	Lower bound	Rate used
0 years	3.92%	0.92%	−1.50%	0.92%
1 years	4.01%	1.67%	−1.47%	1.67%
5 years	4.21%	3.35%	−1.35%	3.35%
10 years	4.29%	4.04%	−1.20%	4.04%
30 years	4.32%	4.32%	−0.60%	4.32%
50 years	4.32%	4.32%	0.00%	4.32%

The UK: one test, new sizes

Rule 9.4A of the PRA's ICAA Part asks a firm to evaluate the 9.7 scenarios regularly and to notify the PRA immediately if its EVE would fall by more than 15% of its CET1 plus AT1 capital: Tier 1. The sterling sizes in rule 9.11 were 250/300/150bp until 30 June 2026 and are 275/425/250bp from 1 July 2026. On the bank, that moves the worst loss from £235.42m (6.88%) to £257.23m (7.52%): a pass either way. There is no UK NII outlier test.

Capital for IRRBB is set in Pillar 2A. For larger or more complex firms the PRA uses its own comprehensive approach; for smaller firms it reviews their policy limits, usually set on a 200bp shift, which serve as the basis for the add-on (SoP5/15). On the bank's book a 200bp shift moves EVE by −£190.9m (up) and +£179.0m (down); a £67bn bank would more likely sit in the comprehensive approach, so this is a benchmark here, not the bank's add-on.

Sort itWhose rule is it?

Sort each rule into the rulebook it comes from.

At least one test compares the maximum ΔEVE with 15% of Tier 1; any other test must be at least as strict.
Floors on post-shock rates are up to the national supervisor, provided they are not above zero.
A fall in one-year NII of more than 5% of Tier 1 in either parallel shock is a large decline.
Post-shock rates may not go below −150bp at immediate maturity, rising 3bp a year to 0% at 50 years.
Gains in one currency count at 50% against losses in another.
Notify the regulator immediately if EVE would fall by more than 15% of CET1 plus AT1.
For smaller firms, policy limits usually based on a 200bp shift serve as the basis for the IRRBB capital add-on.

The formal version

Each test is a maximum, a minimum or a floor (M-A11: Min, max, floors and caps), with the result read as a percentage of Tier 1 (M-A01: Percentages). For scenario $i$ , $\Delta\text{EVE}_i$ and $\Delta\text{NII}_i$ signed as in L4.5.05: The IRRBB standard:

\begin{aligned} \text{EVE test:}\quad & \max_{i=1,\dots,6} \max(0, -\Delta\text{EVE}_i) > 15\% \times \text{Tier 1} \\ \text{EU NII test:}\quad & \frac{\text{NII}_{\text{shock}} - \text{NII}_{\text{base}}}{\text{Tier 1}} < -5\% \ \text{ for } i \in \{\text{parallel up}, \text{parallel down}\} \\ \text{Basel floor:}\quad & R_i(t) = \max\big(R_0(t) + \Delta R_i(t),\ f\big), \quad f \le 0 \\ \text{EU lower bound:}\quad & R_i(t) = \max\Big(R_0(t) + \Delta R_i(t),\ \min\big(\min(-1.5\% + 0.03\% \cdot t,\ 0),\ R_0(t)\big)\Big) \end{aligned}

$t$ is maturity in years; $R_0$ today's risk-free zero rate; $\Delta R_i$ the scenario shock.
EU, across currencies: losses add in full and gains count at 50% (2024/856 Art 3(8)). The bank has only sterling, so the weight does not arise.
The NII test's formula is in 2024/856 Art 5(2); NII is interest income less expense over one year, regardless of accounting treatment, on a constant balance sheet with commercial margins (Art 4).

Try it

The five tests sit on one board. Lower Tier 1 until a value test flags, and see how far it has to fall. Then remove the structural hedge: economic value stops falling when rates rise and starts falling hard when they drop, and every test now turns on parallel down.

Try itFive tests, one bank

Tier 1 capital£3,420m

Structural hedge (£16bn receive fixed)

Basel SRP31.82, ΔEVE

7.52% / 15% · parallel up

EU SOT, ΔEVE

6.88% / 15% · parallel up

EU SOT, ΔNII one year

6.99% / 5% · parallel down

UK ICAA 9.4A, to 30 Jun 2026

6.88% / 15% · parallel up

UK ICAA 9.4A, from 1 Jul 2026

7.52% / 15% · parallel up

each test's limit, as a share of Tier 1

Tests that flag the bank1 of 5

Tier 1 that clears every test£4,784m

The binding test is the one with the largest loss ÷ limit: EU SOT, ΔNII one year (£239.2m ÷ 5%). With the hedge in place the value tests are driven by parallel up and the earnings test by parallel down.

Worked example

The EU NII test on RiskAtoms Bank

£ millions, sterling, EU sizes (±250bp parallel), one year, constant balance sheet, no floor binding. The bank is a UK bank, so the EU test is applied to show how it works.

From the two parallel shocks to a flag

Step 1 of 5

1
Run the two shocks
$\begin{aligned} \Delta\text{NII}_{\uparrow} &= \text{£}231.40\text{m} \\ \Delta\text{NII}_{\downarrow} &= -\text{£}239.21\text{m} \end{aligned}$

Regulatory lens

Basel: SRP31.82–31.89 and 31.93, with the July 2024 sizes to be used from 1 January 2026. Statements on jurisdictions are as at 3 November 2026 as at 3 Nov 2026.

In the UK

ICAA 9.4A (in the version of 31 December 2021; the version from 1 January 2027 keeps the same test) with the 9.11 sizes that changed on 1 July 2026 (LIAF02/25). From 1 January 2027 SS31/15 adds that a firm breaching 9.4A is an outlier firm, that the PRA will review it, and the actions it may expect (2.7B–2.7C). Capital for IRRBB is a Pillar 2A matter (SoP5/15); the PRA also notes that it assesses credit spread risk in the banking book as part of IRRBB (PS15/26, May 2026).

In the EU

Article 98(5) of the CRD and Delegated Regulation (EU) 2024/856, in force since 14 May 2024. The EBA first proposed a large decline of 2.5% of Tier 1. The Commission wanted to amend that, and in its opinion of April 2023 the EBA proposed 5% instead, the level the Regulation adopted. The EU still uses its 2016 sterling sizes; we found no amendment adopting Basel's 2024 sizes, and in January 2026 the EBA said it was still considering their impact on EU banks.

In the US

We found no numeric IRRBB outlier test in the US rules. Supervisors expect banks to measure earnings and economic value under shocks beyond ±200bp, such as up and down 300bp and 400bp (SR 10-1, 2010), and judge the results in examination.

Spark Atoms

Spark AtomWhy 2.5% was the first answer

The EBA calibrated its NII threshold so that the earnings test would flag no more banks than the 15% value test. On its December 2020 data that gave −2.5% of Tier 1, without fair value changes (final draft RTS, October 2022). At 2.5% the bank's line would be £85.5m, and its 6.99% would be nearly three times over it.

Spark AtomThe hedge that passes one test fails the other

Without its £16bn receive-fixed structural hedge the bank would lose £771.6m of economic value in parallel down, 22.6% of Tier 1, and its NII would fall by £549.2m on the EU sizes. The hedge cuts both, but leaves economic value exposed to rising rates. Every hedge trades one measure against the other.

Key takeaways

Basel requires at least one outlier test: the maximum ΔEVE over the six scenarios against 15% of Tier 1 (SRP31.82). An outlier is reviewed; measures follow only if the risk is undue.
The EU adds a large decline in one-year NII of more than 5% of Tier 1 under the parallel shocks (2024/856). The bank is flagged at 6.99%, on parallel down.
Floors: Basel's are national and not above zero; the EU's lower bound is −150bp rising 3bp a year to 0% at 50 years; the UK rule has none.
The UK's ICAA 9.4A test uses 15% of CET1 plus AT1, on 275/425/250bp from 1 July 2026: the bank is at 7.52% (6.88% on the old sizes).
Read a test as headroom (limit less loss) or as the Tier 1 that would pass (loss ÷ limit). No test sets a capital charge by formula.

Reaction

Check what stuck

Question 1 · objective 2

Under the EU's parallel down shock of 250bp, RiskAtoms Bank's one-year NII falls by £239.21m. Its Tier 1 is £3,419.94m. What is the fall as a percentage of Tier 1?

%

Question 2 · objective 4

How much Tier 1 would the bank need for that fall to be no more than 5% of it, in £m?

£m

Question 3 · objective 3

What is the EU's lower bound on a post-shock rate at a maturity of 10 years, in %?

%

Question 4 · objective 2

An EU bank's NII test shows a large decline. What follows?

AIts supervisor assesses its IRRBB and its management, and may use its powers, such as a capital requirementBAn automatic Pillar 2 add-on equal to the excess over 5%CNothing: the NII test is for disclosure onlyDThe bank must close its structural hedge within a year

Question 5 · objective 3

From 1 July 2026, which test does a UK bank run under ICAA 9.4A?

AA fall in EVE of more than 15% of CET1 plus AT1, on sizes of 275/425/250bp for sterlingBA fall in EVE of more than 15% of total capital, on sizes of 250/300/150bpCA fall in NII of more than 5% of Tier 1DA fall in EVE of more than 20% of own funds under a 200bp shift

Glossary

Outlier bank: Outlier bank: a bank whose worst fall in economic value under the prescribed shocks is more than 15% of Tier 1, or that fails another published test; it is presumed to have undue IRRBB and is reviewed (SRP31.82–31.84).
Supervisory outlier test (SOT): Supervisory outlier test (SOT): the EU's name for the two tests of Article 98(5) of the CRD, one on economic value and one on net interest income.
Large decline: Large decline: a fall in one-year net interest income of more than 5% of Tier 1 under either parallel shock (Delegated Regulation (EU) 2024/856, Art 5).
Post-shock floor: Post-shock floor: a lowest value for shocked rates. Basel lets supervisors set one no higher than zero (SRP31.93); the EU sets a lower bound that rises with maturity (2024/856, Art 3(7)).
EU lower bound: EU lower bound: −150bp at immediate maturity, rising 3bp a year to 0% at 50 years and beyond; where today's rate is already lower, today's rate applies.
Headroom: Headroom: how much more loss a bank could take before a test flags it: the limit less the loss.
Pillar 2A: Pillar 2A: the PRA's firm-specific capital add-on for risks Pillar 1 misses or understates, including IRRBB.

Deep Dives & Origins

Deep Dives

L4.9.01: Pillar 2 and the SREP: Pillar 2 and the SREP, where an outlier review leads. Intermediate.
L3.6.12: Credit spread risk in the banking book and the SVB case: SVB, the case these tests are measured against. Intermediate.

Origins

Basel Committee on Banking Supervision, SRP31: Interest rate risk in the banking book, paras 31.82–31.93, version in force from 1 January 2026 and version effective 15 December 2019, bis.org.
Basel Committee on Banking Supervision, Recalibration of shocks for interest rate risk in the banking book, July 2024 (SRP98.63), bis.org.
Commission Delegated Regulation (EU) 2024/856 of 1 December 2023, Arts 1, 3, 4 and 5 (English text via lexaris.eu; Italian Official Journal text, OJ L, 24 April 2024).
European Banking Authority, Final draft RTS on supervisory outlier tests (October 2022) and press release EBA updates definition of large decline in net interest income, 27 April 2023, with the Opinion on regulatory technical standards on supervisory outlier tests(EBA-Op-2023-03); Report on IRRBB heatmap implementation, January 2026.
Directive 2013/36/EU (CRD), Articles 98(5) and 104(1), consolidated text (lawplayer.com); Commission Implementing Regulation (EU) 2024/855 of 15 March 2024 (IRRBB supervisory reporting).
Prudential Regulation Authority, PRA Rulebook: Internal Capital Adequacy Assessment Part, rules 9.4A, 9.7, 9.11, as at 3 November 2026; LIAF02/25 (October 2025); SS31/15, October 2025 near-final version, paras 2.7B–2.7C and 2.9B.
Prudential Regulation Authority, SoP5/15: The PRA's methodologies for setting Pillar 2 capital, May 2026 update, chapter on IRRBB; PS15/26 Pillar 2A review: Phase 1, 28 May 2026, para 5.10.
Board of Governors of the Federal Reserve System and others, Interagency Advisory on Interest Rate Risk Management (SR 10-1), January 2010.

Watch this lesson

Video version

Prefer to watch? The same lesson as a 12-minute video will sit here, made from this page's script.