Interest-rate risk in the banking book (IRRBB) is the risk category that hides in the assumptions. A modest change in the deposit-decay curve moves economic value of equity (EVE) more than most trading-book positions do; a modest change in prepayment behaviour moves net interest income (NII) more than a hundred basis points of curve shift. This article treats IRRBB and its increasingly-visible sibling — credit spread risk in the banking book (CSRBB) — as a modelling problem, not a reporting problem.
For the introductory treatment see IRRBB Modelling: PCA, Behavioural Assumptions and Delta NII; this article extends the framework to CSRBB, optionality and validation.
The two lenses: EVE and NII
Regulators expect banks to measure IRRBB under two lenses. EVE captures the long-term, present-value sensitivity of the balance sheet; NII captures the shorter-term earnings sensitivity. They can move in opposite directions — a bank that is asset-sensitive on NII may be liability-sensitive on EVE — which is why running only one lens hides risk.
The BCBS IRRBB standard (BCBS 368) prescribes six standardised scenarios: parallel up, parallel down, short up, short down, steepener and flattener. National transposition (in the EU, the EBA IRRBB guidelines EBA/GL/2022/14) adds outlier tests: EVE decline above 15% of Tier 1 under any scenario, NII decline above 2.5% of Tier 1, and the supervisory outlier test (SOT) for EVE at 15% of Tier 1 across the six standardised scenarios.
EVE mechanics
EVE is computed as the present value of all repricing cash flows on assets and off-balance-sheet items, less the present value of liabilities. Under a scenario, EVE is recomputed with the shocked curve; the change in EVE is the risk metric. Two design choices dominate:
- Repricing versus contractual cash flows. Fixed-rate mortgages reprice at maturity; floating-rate loans reprice at the next reset; deposits with no maturity reprice according to a behavioural profile.
- Discount curve. The IRRBB standard specifies a risk-free discount curve for EVE; margin and credit spread components are addressed under CSRBB or excluded.
NII mechanics
NII is projected forward under each scenario over a one-to-three-year horizon. Balance-sheet composition can be held constant (constant balance sheet) or run off with new-business assumptions (dynamic balance sheet). The EBA outlier test uses a constant balance sheet for comparability; internal management uses a dynamic view for planning.
Behavioural models: where most of the risk lives
Three behavioural components dominate IRRBB.
Non-maturity deposits (NMDs)
Current accounts and savings accounts have no contractual maturity but behavioural half-lives of 3–8 years for stable segments, less for rate-sensitive ones. A robust NMD model separates:
- Core balance: the portion of the deposit book that has never been affected by rate cycles.
- Repricing beta: the fraction of a market-rate move passed through to deposit rates.
- Runoff profile: the amortisation of the core balance if new deposits stopped tomorrow.
Segmentation matters. A single blended NMD curve for retail plus SME plus corporate under-differentiates the book. The BCBS standard caps the average behavioural maturity of retail transactional deposits at 5 years, retail non-transactional at 4.5 years, and wholesale non-financial at 4 years. Internal models can use shorter — but not longer — profiles.
Prepayment on fixed-rate loans
Prepayment on mortgages and consumer loans is driven by refinancing incentive (the option value of prepaying and refinancing at lower rates) and by exogenous events (moves, refinancings, life events). A conditional prepayment rate (CPR) model links the monthly prepayment probability to the moneyness of the option and to seasoning, seasonality and burnout effects.
The model interacts with credit: prepayment competes with default for the borrower's cash flow. Modelling them independently overstates expected balances. Similar dynamics appear in IFRS 9 EAD modelling.
Early withdrawal on term deposits
Term-deposit withdrawal is smaller in aggregate but material for the NII lens because the depositor exercises the option precisely when rates rise — exactly when the bank would prefer the funding to stay locked. Model calibration uses withdrawal-fee data and observed withdrawal-versus-market-rate curves.
Basis risk and optionality
Two additional risk families are increasingly emphasised by supervisors.
Basis risk arises when assets and liabilities reprice to different indices — for example, a loan book referencing an interbank benchmark and a deposit book referencing an administered rate. Even under a parallel curve shock, basis risk generates residual sensitivity. Post-benchmark reform (LIBOR to risk-free rates), basis risk between compounded RFRs and legacy indices remains material in some portfolios.
Automatic and embedded options — caps, floors, prepayment options embedded in loans, early-termination rights — have to be revalued under each scenario. Closed-form Black or Hull–White models are common; Monte Carlo is used where path dependency matters.
CSRBB: the newer sibling
The EBA CSRBB guidelines make explicit what supervisors have implied for years: credit spread risk on the banking book is a distinct risk category. CSRBB captures the sensitivity of instruments — typically the liquidity buffer, securitisation positions and non-liquid assets — to changes in credit spreads, separately from underlying interest rates and separately from name-specific credit risk.
The measurement approach mirrors EVE and NII under CSRBB scenarios. The scope debate — which instruments belong in CSRBB and which in default credit risk or market risk — is where most implementation friction lives. A defensible scoping paper is a validation prerequisite.
Numerical intuition
For a mid-sized retail bank with 40% NMDs on the liability side and a mortgage book with 20% fixed-rate loans:
- A +200 bp parallel shock on a constant balance sheet typically moves NII by 4–8% of baseline in year one, depending on repricing betas.
- The same shock on EVE typically moves 5–12% of Tier 1 through the combination of repricing gap and NMD-duration assumptions.
- Halving the assumed core-NMD fraction moves EVE by more than doubling the shock — a reminder that behavioural assumptions dominate the numbers.
These illustrative ranges vary widely with product mix; the point is the order of magnitude of behavioural sensitivity relative to curve sensitivity.
Validation and independent challenge
A robust validation programme covers:
- Behavioural back-testing: were the last 12–24 months of NMD balances, prepayment rates and withdrawal rates consistent with model expectations?
- Repricing-beta stability: did the model's repricing betas move with the empirical betas observed in the 2022–24 hiking cycle?
- Basis-risk decomposition: does the model produce sensible sensitivities to individual index-basis moves?
- Scenario coverage: do internal scenarios include the six BCBS shocks plus non-parallel, historically-observed and idiosyncratic scenarios?
See The Model Validation Lifecycle for the broader framework.
Governance implications
Boards should see EVE and NII under the six standardised scenarios plus internal scenarios, in absolute terms and against limits. Limits should be set at the metric level, at the sub-metric level (per scenario, per behavioural component) and at the earnings level (rolling 12-month NII at risk versus budget). ALCO minutes should record the assumptions that most moved the numbers each quarter — usually one or two behavioural parameters.
Limitations
Every IRRBB number is conditional on behavioural stability. The 2022–24 hiking cycle showed that repricing betas move — sometimes doubling within a year — as competitive dynamics change. Models that hold betas constant to their pre-2022 calibrations understated risk materially. Continuous back-testing and expert-judgment adjustments are not a compliance overhead; they are the model.
Conclusion
IRRBB and CSRBB reward institutions that treat behavioural modelling as their primary risk-management skill and treat the scenario prescription as the reporting overlay. The reverse — treating the six scenarios as the model — mislocates the effort and the risk.
Related reading
- IRRBB Modelling: PCA and Delta NII
- Liquidity Risk: LCR, NSFR and ILAAP
- The Model Validation Lifecycle
References and further reading
- Basel Committee on Banking Supervision, Interest rate risk in the banking book, BCBS 368.
- European Banking Authority, Guidelines on the management of interest rate risk arising from non-trading book activities (EBA/GL/2022/14).
- European Banking Authority, Guidelines on the management of credit spread risk arising from non-trading book activities (EBA CSRBB guidelines).
About the author
Part of an ongoing series on ALM and financial risk modelling — more about the author.