Why Banks Need To Question Deposit Cost Playbooks as Rates Move

# Banking

# Regional Bank Treasurers

New Empyrean report exposes the risks of constant deposit beta assumptions and makes the case for dynamic modeling.

New Empyrean report exposes the risks of constant deposit beta assumptions and makes the case for dynamic modeling.

Banks estimating deposit betas—how much of a Fed rate move they’ll pass along to depositors—risk paying higher funding costs and losing deposits by relying on models that assume beta remains constant throughout a rising rate cycle. It doesn’t.

Bank treasurers therefore need a dynamic framework for comparing modeled and actual deposit betas, recalibrating assumptions and monitoring deposit pricing as interest rates rise or fall.

Those takeaways emerge from a new report by Empyrean Solutions that shows depositor rate sensitivity during the rate-hike cycle beginning in 2022 was convex—nonlinear and increasing as rates continued to climb. “The implication for 2026 is direct,” the report says. “A deposit beta assumption should be a function of the rate level and the expected size of the move, not a single number carried across the whole curve.”

Assumptions about deposit beta are critical to a bank’s interest-rate risk management process, including how quickly non-maturity deposits reprice and how the mix of interest-bearing and non-interest-bearing (NIB) deposits shifts as rates move, the report notes.

“Those assumptions drive the net interest income forecast, the economic-value-of-equity calculation, and the hedging and funding decisions that follow,” it adds.

Empyrean’s findings are particularly relevant now, amid growing expectations the Fed’s September rate hike will be the first of several. That outlook represents a significant shift from the beginning of 2026, when markets expected rate cuts.

A potentially costly gap. Empyrean’s analysis found a large gap between what banks modeled for deposit beta and what they actually did. The median deposit beta of 16 banks during the 2022-2024 rate-hike cycle reached 68%—more than double a benchmark model’s 31% beta. On a hypothetical $50 billion in interest-bearing deposits, that gap represents roughly $850 million of unmodeled annual funding cost, the report says.

Laurent Birade , a Principal at Empyrean who authored the report, is quick to point out that those benchmark models worked reasonably well through the first 100 to 150 basis points of tightening, then diverged sharply as the increase reached about 515 basis points. “That is not a calibration error; it is convexity,” he said. “A constant beta is fine near the starting point and increasingly wrong the further rates travel.”

The size of the gap in part reflects that most of the deposit beta assumptions “were calibrated in a near-zero-rate, deposit-flooded regime,” he added. “When an ALCO sat down in early 2022 and asked, ‘What beta should we assume?’, nobody in the room had managed a real hiking cycle in over a decade. They reached for the published benchmarks.”

Addressing the potential risks requires each bank to measure its own gap. “The point is not to embarrass the team that set the 2022 assumptions; they used the best guidance available,” he said. “The point is to make the forecast error visible and measurable, so the assumption set can be recalibrated against evidence rather than against the last cycle's intuition.”

Jerry Olivo , a former bank treasurer who leads NeuGroup for Regional Bank Treasurers , observed that a static, average beta may be technically correct across a full cycle of rate increases and declines. “But it could be much less effective than dynamic betas at points in the cycle that capture convexity associated with rising and falling rates,” he said.

“By using dynamic betas at points in time through the cycle, the bank can not only better forecast the cost of deposit retention; it can also make an accurate decision whether to retain the deposit—or not.”

A monitoring framework and beta that bends. The report demonstrates that banks need to adjust deposit betas for periods like the last rate-hike cycle when rate sensitivity as gauged by actual deposit betas rose from 47% in the previous rate-hike cycle to 68%. Empyrean says the reasons include depositors who are more aware of better rates as well as technology that enables them to move quickly to higher-yielding alternatives.

The report raises two questions: Is increased sensitivity permanent, and will deposit betas move symmetrically when rates decline? (More on that in the final section of this article.) “That is precisely what a monitoring framework is built to answer,” it says.

The recommended monitoring framework examines the extent to which depositors move money out of non-interest-bearing accounts as rates rise, which is known as NIB decay. If NIB deposits decline, the analysis probes whether the bank offset the decline with growth in interest-bearing deposits or experienced an overall decline in deposits.

Offsetting can be a sign of “stickiness”—retaining deposits—and may also reflect customers bringing more money to the bank, increasing its total deposits. Empyrean calls this analysis “cannibalization and realization” or CANRE.

“Never diagnose a franchise on a single metric,” the report states. “Pair the rate-sensitivity read (NIB Decay) with the stickiness read (CANRE) before drawing any conclusion about deposit-base health.” It adds that showing only NIB may conflate rate sensitivity with franchise weakness. “That single error can send a CFO chasing a retention problem that does not exist, or worse, ignoring one that does.”

A tale of two banks. To underscore that point, the report examines two banks that both suffered significant NIB decay but had different deposit outcomes: At bank A, NIB deposits declined sharply but the bank captured $2.29 of interest-bearing deposits for every $1 of NIB deposits lost; total deposits grew 21%. Bank A’s depositors are described as rate-sensitive but sticky.

Bank B’s NIB deposits also fell, but its overall deposit base contracted about 32% and it did not replace the lost balances. Depositors were rate-sensitive and less sticky—although the report notes that a deliberate balance-sheet restructuring caused some of the deposit contraction.

One reason for the different outcomes, Mr. Birade suggested, is that bank A reacted faster to the changing rate cycle than bank B. He said this highlights a key point for bank treasurers adapting to more rate-sensitive depositors: “Ongoing model monitoring is a lot more important than it ever was.”

A few more recommendations. Bank treasurers can consider these additional suggestions made in Empyrean’s report:

Regularly compare actual deposit behavior with banks in your peer group, using updated quarterly data.

The actual-versus-modeled gap should be a formal input to each ALCO assumption review, not a report that is filed and forgotten.

Track periods of rate declines and increases against three scenarios: reversion (elevated symmetry fades); new symmetry (higher sensitivity is permanent and works both ways); new asymmetry (higher sensitivity is permanent but one-sided: depositors are fast to demand higher rates, slow to accept lower ones).

Read Empyrean’s full report by clicking here .

NeuGroup’s Regional Bank Treasurers group will benchmark deposit monitoring and model calibration frameworks in order to identify enhancement opportunities at an upcoming peer group session. Please email Jerry Olivo or Member Services if your bank would like to join this and related discussions.