Deposit Beta in Bank ALM Models: The 2022-2025 Calibration Problem
Most US bank ALM models use deposit beta from the 2014-2019 cycle. Actual betas in 2022-2025 ran 2-3x higher. The recalibration framework, by tier, with EVE/NII shock implications.
Most US bank ALM models still use deposit beta assumptions calibrated against the 2014–2019 rate cycle. The 2022–2025 cycle showed those calibrations were materially wrong. Banks running stale betas understated their NII risk by ~3x heading into the SVB crisis.
What is deposit beta?
Deposit beta is the percentage of a base rate (typically Fed Funds for US, Bank Rate for UK) that gets passed through to the rate paid on a deposit product.
If Fed Funds rises 100bps and a savings account rate rises 30bps, the deposit beta is 30%.
Deposit beta varies dramatically by:
- Product type: instant access < notice < fixed term
- Customer type: retail < small business < commercial < institutional
- Insurance status: insured (FDIC-covered) < uninsured
- Channel: branch < online < broker
What 2022-2025 changed
The previous decade (2010–2021) was characterised by ZIRP and modest rate moves. Deposit betas calibrated against this period showed:
- Retail savings: 20–30%
- Commercial deposits: 30–50%
- Institutional / brokered: 60–80%
When rates rose 525bps over 2022–2023, actual betas were materially higher:
- Retail savings: 35–50% (up from 20–30%)
- Commercial deposits: 70–90% (up from 30–50%)
- Uninsured commercial: 80–95% (the SVB cohort)
The reason: depositors had access to better rate alternatives (T-bills, money market funds yielding 5%+) and digital channels made it easy to move money.
Why this matters for ALM models
ALM models project NII under interest rate scenarios. Two key levers:
- Asset side: how fast do loan yields reprice
- Liability side: how fast do deposit rates reprice (driven by deposit beta)
If your model assumes 30% beta when actual beta is 80%, in a +200bps shock your modelled NII is materially overstated. Your EVE / NII shock numbers are similarly distorted.
For the 2022–23 cycle, banks running 2019-vintage betas were modelling NII benefits of ~10–15%. Actual NII benefits were closer to 3–5% — and for some banks (those with high uninsured commercial mix), NII actually compressed.
How to recalibrate
1. Segment deposits by beta tier.
Don't average across the whole book. Split into:
- Insured retail (low beta)
- Insured commercial (medium beta)
- Uninsured commercial (high beta)
- Brokered / wholesale (highest beta)
Each gets its own beta assumption.
2. Use 2022-2025 actuals as base case.
Calibrate against your bank's actual experience over the rising rate cycle. Don't blend with pre-2022 data — the rate environment was structurally different (digital access, money market yields, etc.).
3. Flex beta in scenarios.
Base case = 2022–2025 actuals. Adverse scenario = 1.2x base (faster pass-through under stress). Severely adverse = 1.5x base (run-off scenario).
4. Re-calculate EVE / NII shock.
Re-run your standard shock scenarios (+200bps parallel, etc.) with new betas. Expect your numbers to look materially worse than the previous cycle's output. That's the point — the previous cycle's output was wrong.
What about the deposit beta floor?
Some banks model a "floor" beta below which the deposit rate doesn't fall in a falling rate environment. For example, retail savings rates might have a 0.10% floor regardless of how low Fed Funds goes.
The floor matters in falling-rate scenarios. If you model 30% beta on the way up but ignore the floor on the way down, your NII benefit in a rate cut scenario is overstated.
For a complete bank long-term plan model that handles deposit beta tiering, EVE / NII shock, and floor logic, see our Bank Long-Term Plan Model.
What to do if you can't recalibrate immediately
If your ALM model is stuck on old betas and a recalibration is months away, the interim hack: apply a "beta haircut" to your modelled NII benefit.
For US regional banks, halving the modelled NII shock benefit gets you roughly to where the recalibrated number would land. Not perfect, but better than presenting a number you know is wrong.
Get the Bank Long-Term Plan Model
The model that puts the principles in this post into practice. Bank-grade build, open formulas, no VBA. Same-day delivery.
SFS Models builds institutional-grade Excel financial models for banking and finance professionals. Browse the catalogue or get the free sample.