Guide

Bank deposit beta explained

Harbor Credit Union's $4.8B balance sheet ran an asset-liability management review in March 2022 when the Federal Reserve began hiking from near zero. The ALM model assumed deposit costs would rise roughly one-for-one with the policy rate — a deposit beta of 1.0. Reality was slower: savings and money-market yields lagged fed funds by months, while wholesale funding and reserve-linked spreads moved immediately. Net interest margin (NIM) expanded in early 2022, then compressed through 2023 as lagged deposit repricing caught up. The desk's quarterly NIM forecast missed by 19 bp at the worst point because it treated deposits like a floating-rate liability instead of a sticky, segmented book.

The team rebuilt the liability stack by product tier — non-interest-bearing checking, interest-bearing checking, savings, money-market, and brokered CDs — and estimated cumulative and marginal deposit beta against the effective fed funds rate (EFFR). Forecast error on forward NIM fell from 19 bp to 5 bp over six quarters. This guide explains what deposit beta measures, how to estimate it without fooling yourself, why high-beta online banks differ from sticky retail franchises, how deposit beta drives monetary policy transmission and NIM cycles, the Harbor Credit Union refactor, a technique decision table versus mortgage pass-through and yield-curve models, pitfalls, and a production checklist for bank treasury and macro desks.

What deposit beta measures

Deposit beta is the elasticity of a bank's deposit funding cost to changes in short-term market rates, usually benchmarked against the effective federal funds rate or SOFR. In plain terms: when the Fed hikes 100 bp, how many basis points does your blended deposit cost actually rise?

Two definitions matter:

Cumulative beta — total change in deposit cost divided by total change in the policy rate over a full hiking or cutting cycle. Answers: “Over the last 400 bp of hikes, how much did we pass through?”
Marginal beta — the pass-through on the next 25 bp move, often estimated with rolling regressions or bucketed FOMC windows. Answers: “If the Fed hikes again next meeting, what happens to our cost of funds?”

Marginal beta typically rises late in a hiking cycle as competition for deposits intensifies and customers chase yield in money-market funds or Treasury bills. Cumulative beta can look low early (sticky checking and savings) then jump as CD specials and promotional rates reset. Treating beta as a single constant across the cycle is the most common modeling mistake.

How to estimate deposit beta in practice

A workable production pipeline:

Segment liabilities by product and behavioral stickiness (NIB demand, IB demand, savings, retail MM, jumbo MM, brokered CDs, public funds).
Compute blended cost per segment: interest expense divided by average balance, annualized.
Align rate benchmarks — EFFR or SOFR for policy; compare segment yields to competitive tables (e.g. national savings averages) to spot lag.
Regress changes in segment cost against cumulative EFFR changes over FOMC windows; use robust errors because outliers cluster around quarter-end and liquidity stress episodes.
Separate cumulative vs marginal — cumulative for full-cycle NIM bridges; marginal for next-quarter scenario shocks.

Public banks disclose deposit costs in 10-Q filings; aggregating peer groups gives industry beta bands. Regional banks with heavy commercial operating accounts often show lower beta on NIB balances (operational stickiness) but higher beta on public and brokered funding. Online-only banks cluster at the high-beta end because customers comparison-shop rates monthly.

Deposit beta is not the same as mortgage pass-through: mortgages reprice through MBS convexity and primary-secondary spreads over quarters; deposits reprice through franchise loyalty, regulation, and competitive pressure over weeks to quarters. Mixing the two in one “policy beta” breaks scenario analysis.

High-beta vs sticky deposit franchises

Deposit beta is a function of product mix and customer behavior, not just management willingness to pay up.

Segment	Typical beta range (hiking cycle)	Why
Non-interest-bearing checking	0 – 0.1	Operational balances; rate-insensitive until stress migration
Retail savings	0.2 – 0.5	Sticky households; banks lag on headline savings APY
Money-market / HISA	0.5 – 0.9	Rate-sensitive; competes with T-bills and MMFs
Brokered / listing-service CDs	0.8 – 1.1+	Price-takers; reprices with market in days

During the 2022–2023 hiking cycle, industry cumulative deposit beta estimates clustered roughly 30–50% for large regionals with sticky retail books, versus 70%+ for deposit-gathering fintechs. When ON RRP balances drained and money-market fund yields rose, marginal beta on retail savings accelerated — the lag was not infinite, just nonlinear.

Deposit beta, NIM and monetary transmission

Net interest margin equals asset yield minus funding cost, scaled by the balance sheet. In a hiking cycle with fixed-rate loans and floating-rate assets:

Early phase — asset yields reprice faster than deposit costs (low marginal beta). NIM expands; bank stocks often rally on “beneficiary of higher rates” narratives.
Mid phase — deposit competition wakes up; marginal beta rises. Funding costs catch assets; NIM plateaus.
Late phase — cumulative beta materializes; inverted yield curves pressure asset yields while deposits remain elevated. NIM compresses; credit provisions may rise in parallel.

For macro desks, deposit beta is the bridge between forward guidance and real-economy lending conditions. Slow pass-through dampens the tightening bite of hikes on household cash flows (sticky savings APY) but can prolong inflation pressure if demand deposits stay abundant. Fast pass-through (high-beta franchises) transmits policy more quickly but can trigger deposit flight to money funds — visible in H.8 deposit aggregates and MMF inflows.

Harbor Credit Union refactor

Harbor segmented its $3.1B deposit base into six behavioral buckets and assigned time-varying marginal beta curves per bucket, keyed off EFFR level and local competitor rate scans. Key changes:

Replaced a single 0.85 deposit beta with bucket curves (NIB 0.05, savings 0.35 rising to 0.55 above 4% EFFR, brokered 0.95).
Linked MMF migration triggers to ON RRP rate and 3-month T-bill spread monitors — when spread exceeded 25 bp, savings marginal beta stepped up one notch.
Fed NIM bridge reports weekly instead of quarterly; cumulative beta tracked against peer 10-Q disclosures.

Over six FOMC meetings in 2023, NIM forecast error fell from 19 bp to 5 bp. The largest fix was recognizing that Harbor's jumbo MM promo rates were driving marginal beta above 0.9 even while headline “blended beta” looked moderate. Without segmentation, the model smoothed away the very product that was moving.

Technique decision table

Approach	Use when	Skip when
Segmented deposit beta	Bank ALM, NIM forecasting, rate-risk KPIs	Pure asset-side duration books with minimal deposits
Single blended beta	Quick peer sanity check; stable product mix	Hiking/cutting cycles; mix shift toward brokered funding
Mortgage pass-through model	MBS convexity, housing affordability, primary-secondary spread	Deposit cost forecasting; different lag structure
Yield-curve shift scenarios	Duration risk, bond portfolio marks, pension ALM	Short-end deposit repricing without curve parallel shift
Fed funds futures strip	Market-implied policy path for marginal beta timing	Long-run franchise stickiness absent rate competition

Common pitfalls

One beta for all products — masks jumbo MM and brokered spikes that drive marginal funding cost.
Using only headline savings APY — ignores NIB checking that dominates balance but not interest expense.
Confusing cumulative and marginal — board slides show low cumulative beta while next-quarter shock needs marginal.
Ignoring MMF migration — deposit outflows raise wholesale funding beta separately from in-place deposit repricing.
Peer beta without mix adjustment — a sticky retail bank is not comparable to a brokered-funded model.
Static beta through cutting cycles — banks often lag on the way down too, but competitive dynamics differ from hikes.
Treating IORB as deposit floor — IORB sets a corridor for banks, not a retail savings floor; see discount window context for institutional rates only.

Production checklist

Segment deposits by behavioral stickiness and funding stability (ASF tiers if reporting NSFR).
Estimate cumulative beta for full-cycle NIM bridges; marginal beta for next-quarter scenarios.
Benchmark EFFR/SOFR changes against segment-level cost, not blended APY alone.
Monitor MMF and T-bill spreads as leading indicators of rising marginal beta.
Reconcile internal beta to peer 10-Q deposit costs quarterly.
Stress-test deposit outflows separately from repricing (liquidity vs margin).
Document beta assumptions in ALM committee minutes; update after mix shifts.
Pair deposit beta with asset-side repricing for true NIM scenario grids.
Track forecast error on NIM and funding cost; retire static beta assumptions that drift.
For macro views, connect deposit stickiness to policy transmission lag versus Taylor-rule implied rates.

Key takeaways

Deposit beta measures how much deposit costs move when policy rates move.
Cumulative and marginal beta answer different questions — use both.
Product mix drives beta more than a single bank “franchise score.”
Harbor Credit Union cut NIM forecast error from 19 to 5 bp with segmented beta curves.
Deposit beta is the liability-side counterpart to mortgage and bond pass-through on the asset side.