Short answer: Deposit beta is the share of a benchmark rate move that passes into what you pay for deposits. Compute it as interest expense on deposits, de-cumulated from the year-to-date filing and annualized, over average interest-bearing deposits, then divide the change between two quarters by the change in your benchmark over the same two quarters. The median FDIC-insured bank paid 1.76% on deposits in the quarter ending June 30, 2026, with the middle half between 1.37% and 2.19%. The asymmetry is the story: measured Q1 2024 to Q1 2026, the largest banks had given back about 67% of their cycle rise in funding cost, while banks between $100 million and $300 million had given back about 3%, and about 12% even when measured from their own later peak.
The funding-cost path this cycle, by bank size
Deposit beta is an argument about a path, and most beta conversations go wrong because only two points of that path are ever on the table. Here is the whole thing: the median cost of deposits for each FFIEC commercial bank peer group, every quarter from the first quarter of 2023 through the quarter ending June 30, 2026.
| Quarter | Over $100B | $10B - $100B | $3B - $10B | $1B - $3B | $300M - $1B | $100M - $300M | Under $50M |
|---|---|---|---|---|---|---|---|
| Q1 2023 | 1.38% | 1.12% | 1.23% | 1.09% | 0.92% | 0.76% | 0.50% |
| Q2 2023 | 1.55% | 1.40% | 1.41% | 1.31% | 1.09% | 0.91% | 0.60% |
| Q3 2023 | 1.73% | 1.62% | 1.60% | 1.46% | 1.23% | 1.04% | 0.74% |
| Q4 2023 | 1.82% | 1.80% | 1.75% | 1.58% | 1.36% | 1.16% | 0.86% |
| Q1 2024 | 2.48% | 2.31% | 2.29% | 2.15% | 1.92% | 1.68% | 1.26% |
| Q2 2024 | 2.55% | 2.39% | 2.34% | 2.23% | 1.96% | 1.74% | 1.34% |
| Q3 2024 | 2.55% | 2.45% | 2.37% | 2.27% | 2.01% | 1.78% | 1.43% |
| Q4 2024 | 2.55% | 2.40% | 2.39% | 2.27% | 2.01% | 1.78% | 1.39% |
| Q1 2025 | 2.04% | 2.13% | 2.16% | 2.08% | 1.95% | 1.77% | 1.44% |
| Q2 2025 | 2.00% | 2.11% | 2.16% | 2.12% | 1.95% | 1.78% | 1.41% |
| Q3 2025 | 2.01% | 2.15% | 2.15% | 2.12% | 1.95% | 1.77% | 1.42% |
| Q4 2025 | 1.99% | 2.13% | 2.15% | 2.09% | 1.93% | 1.75% | 1.35% |
| Q1 2026 | 1.74% | 1.80% | 1.94% | 1.94% | 1.77% | 1.65% | 1.29% |
| Q2 2026 | 1.74% | 1.84% | 1.96% | 1.95% | 1.78% | 1.66% | 1.27% |
Source: BankingLens, computed from FFIEC call reports for the quarter ending June 30, 2026. Quartiles are calculated across the full population of filing banks in each size band, not sampled. Non-insured non-deposit trust companies are left out of every statistic: they take no deposits and make no loans, so a margin or a funding cost computed for them has no meaning. Columns are FFIEC peer groups 1 through 8, which cover insured commercial banks; savings banks file into their own peer groups and are not in these medians. Every figure is on the year-to-date annualized basis described below, and all four first-quarter marks cover the same span of months.
How to read this table. Each figure is interest expense on deposits for the year to date, annualized, over average deposits, which is the basis the call report itself publishes on. Only the first-quarter readings look at a single quarter: Q2 averages two quarters of the year, Q3 three, Q4 the whole of it. Within-year movement is damped by construction, so Q2 minus Q1 is not a quarterly repricing move and must never be read as one. The flat runs, 2.55% three quarters together at the top, and the step up at every Q1 are that convention behaving normally, not a data error. Measure every leg first quarter to first quarter and the problem disappears.
Two things jump off that table once you read it on first-quarter marks. The first is that the peak did not arrive at the same time for everybody. Every cohort above $1 billion set its highest first-quarter reading in Q1 2024; every cohort below kept climbing and did not set its own until Q1 2025, a full year later. The two ends crossed in between. The over-$100 billion median came into Q1 2025 fifty-one basis points below its 2024 full-year average, while the under-$50 million median came in five above: half the industry was still repricing upward while the other half had already turned.
The second is what has happened to the community bank funding advantage. The gap between the largest banks and the smallest was 88 basis points in Q1 2023 and 122 at the Q1 2024 mark. By Q1 2026 it was 45, the narrowest first-quarter reading of the cycle. Nothing improved at the top; the bottom simply stopped moving.
What deposit beta actually is
Deposit beta is the fraction of a benchmark rate move that reaches your funding cost. If the benchmark moves 100 basis points and your cost of deposits moves 40, your beta is 0.40. That is the whole definition, and everything contentious about it lives in three choices that are usually left unstated: which benchmark, which window, and which deposits.
Use whatever driver rate the ALCO model already runs on, because a beta computed against one index and reported against another cannot be reconciled by anybody. State the window in the sentence: a beta without a start and end quarter attached is a rumor. And say whether the rate sits on total deposits or interest-bearing deposits, because both get called "cost of deposits" and they differ by the size of your noninterest-bearing book. That is not cosmetic. The median bank funded 20.98% of deposits with noninterest-bearing balances in Q2 2026, 23.11% under $100 million against 19.52% between $1 billion and $3 billion. A beta on total deposits blends in that free money and always reads lower. Interest-bearing is the cleaner denominator, because it is the book that can reprice.
Exactly which call report lines to use
The numerator is interest expense on deposits, Schedule RI, item 2.a. The denominator is average interest-bearing deposits, from Schedule RC-K, the quarterly averages page. Use RC-K rather than averaging two period-end balances off Schedule RC-E: it is already an average taken inside the quarter, which is what the numerator's expense was incurred against, and one large balance sitting on the books on the last day cannot throw it off.
The trap is that Schedule RI is year to date. A June 30 filing reports six months of interest expense, a September 30 filing nine. Divide a September number straight into RC-K and you have a nine-month average dressed up as a quarterly rate, which is exactly what makes a beta look stable while it is in fact moving. De-cumulate first: subtract the prior quarter's year-to-date figure, then multiply by four. Our guide to reading a call report covers the same mechanic for every other Schedule RI line. A de-cumulated series will not reproduce the table above except in first quarters, and that is expected: the table is on the published year-to-date basis, and Q1 is the one quarter where the two definitions agree.
So the rate is:
Cost of deposits (quarterly, annualized) = (RI 2.a year to date − RI 2.a prior quarter year to date) × 4 ÷ average interest-bearing deposits from RC-K
Deposit beta = (cost of deposits in the end quarter − cost of deposits in the start quarter) ÷ (benchmark in the end quarter − benchmark in the start quarter)
Note what the call report does and does not give you. It gives you the numerator, cleanly, for every bank in the country. It does not give you the denominator: that comes from your own rate deck. Board presentations on beta most often fall apart because the numerator was pulled from the filings and the denominator was remembered from a headline.
Working it with real numbers
Take the $1 billion to $3 billion peer group, the band most community bank ALCOs benchmark into. Its median cost of deposits was 1.09% in Q1 2023, 2.15% in Q1 2024 and 1.94% in Q1 2026. The up-leg numerator is 106 basis points over four quarters; the down-leg numerator, so far, is 21 over eight. Divide each by the benchmark move across the matching window and the arithmetic is finished.
Every mark is a first quarter for the reason above: on a year-to-date basis, Q1 against Q1 is the only pairing where both observations cover the same span of months. Pair a Q1 against a Q3 and the beta is part repricing and part accounting, in a proportion you cannot recover.
If you would rather not argue about the benchmark at all, there is a version that removes it. Divide your move by a peer cohort's move over the identical window and the rate denominator cancels, leaving a relative beta: how much of the industry's repricing you took. Over Q1 2024 to Q1 2026 the over-$100 billion median fell 74 basis points and the $1 billion to $3 billion median fell 21, a relative down beta of 0.28. Over the four quarters ending Q1 2024 the same two moved 110 and 106, a relative up beta of 0.96. Same peer group, same source, same method: 0.96 up, 0.28 down, and nobody has to agree on the benchmark.
Here is the same calculation for every size band, with the share of the cycle rise given back through Q1 2026.
| Bank size (total assets) | Q1 2023 | Q1 2024 | Q1 2026 | Up leg | Down leg | Given back |
|---|---|---|---|---|---|---|
| Over $100B | 1.38% | 2.48% | 1.74% | +110 bp | −74 bp | 67% |
| $10B - $100B | 1.12% | 2.31% | 1.80% | +119 bp | −51 bp | 43% |
| $3B - $10B | 1.23% | 2.29% | 1.94% | +106 bp | −35 bp | 33% |
| $1B - $3B | 1.09% | 2.15% | 1.94% | +106 bp | −21 bp | 20% |
| $300M - $1B | 0.92% | 1.92% | 1.77% | +100 bp | −15 bp | 15% |
| $100M - $300M | 0.76% | 1.68% | 1.65% | +92 bp | −3 bp | 3% |
| $50M - $100M | 0.65% | 1.51% | 1.48% | +86 bp | −3 bp | 3% |
| Under $50M | 0.50% | 1.26% | 1.29% | +76 bp | +3 bp | −4% |
Source: BankingLens, computed from FFIEC call reports for the quarter ending June 30, 2026. Quartiles are calculated across the full population of filing banks in each size band, not sampled. Non-insured non-deposit trust companies are left out of every statistic: they take no deposits and make no loans, so a margin or a funding cost computed for them has no meaning. Up leg is the change in median cost of deposits from Q1 2023 to Q1 2024; down leg is Q1 2024 to Q1 2026; given back is the down leg as a share of the up leg. All readings are first-quarter marks, covering the same span of months.
The gradient is almost perfectly monotonic with size, and it runs the opposite way to the up leg. The smallest banks took the least of the rise, 76 basis points against 110 at the top, and on this common window have given back none of it. The largest took the most and have returned two thirds.
One qualification, and it is this article's own advice turned back on its own table. Q1 2024 was the peak first-quarter mark only above $1 billion; the cohorts below kept repricing into Q1 2025. Measured from their own peak they have given back 12 to 18 basis points, roughly 12% to 18% of their rise, rather than the 3% and minus 4% a common window reports. That is still well under a third of what the largest banks have returned, and both readings earn their place: a common window says where you stand against the industry, an own-peak window what you have done since you turned.
Why the down beta lags
The lag is not inertia or bad management. It is four mechanisms, only one of which is discretionary.
Time deposits reprice on their own calendar. A thirteen-month CD booked at the top of the cycle pays the top of the cycle until it matures, whatever the benchmark does in month three. A bank that funded heavily with term money at the peak bought a dated schedule it cannot accelerate. This is the largest component of the lag and the most forecastable: the maturity ladder says when relief arrives, quarter by quarter, before it does.
Exception pricing unwinds one relationship at a time. Rates conceded to large depositors during a funding scare are not administered rates. Each is a conversation with a customer who remembers what was promised, so the bank moves at the speed of its relationship managers rather than the curve.
Administered rates get cut slowly on purpose. Savings and money-market rates are the bank's own decision, and the asymmetry is behavioral: depositors notice a cut and do not notice a raise they never received. Cut too fast and you find out what your balances were really worth. Most ALCOs would rather find out gradually.
Mix does not come back. Balances that migrated from checking into money market and CDs on the way up stay there. Every product rate can fall all the way back and the blended cost still will not, because the weights changed. This is the part that never reverses and the part most often left out of a forecast.
Size sits on top of all four. Banks over $100 billion carried a 4.75% median brokered share of deposits in Q2 2026 and banks between $10 billion and $100 billion 3.69%, while the median bank under $1 billion carried none. Brokered and wholesale money is the fastest-repricing liability on the sheet in both directions, which is why the largest cohort led up and leads down. A community bank's beta is low because its funding is genuinely stickier, and stickiness does not choose a direction.
Defending the number to a board
Bring three numbers, not one: the cycle-to-date up beta, the down beta since your own funding-cost peak, and the peer-relative version of both. When a director asks why the bank's beta is not the number in a trade press headline, the answer is almost always that the headline is a cycle-to-date average over a window starting three years ago, and a long window dampens the current quarter mechanically. Name your own peak quarter and start the down beta there: peaks did not arrive on a common date, and one run from an industry peak that is not yours misstates the position in whichever direction flatters or damns you.
Then say what a low down beta is costing, because it is not a virtue. On the way up it was evidence of a franchise; on the way down it is margin sitting in someone else's account. A bank in the $1 billion to $3 billion band that has moved 21 basis points while the largest moved 74 should be able to price that gap as a deliberate retention decision, or it is not a decision at all.
Finally, show cost of funds beside cost of deposits. The median bank's cost of funds was 2.34% in Q2 2026 against a 1.76% cost of deposits, and that 58 basis point wedge is borrowings. If yours is wider than your peers', your beta reads better than it is, because the fastest-repricing part of your funding is not in the number you are quoting. Our note on what a good cost of funds looks like has the full distribution.
Frequently asked questions
What is deposit beta?
Deposit beta is the share of a benchmark rate move that passes into what a bank pays for deposits. If the benchmark moves 100 basis points and a bank's cost of deposits moves 40, its beta is 0.40. It is computed over a window, not a point, and the window has to be stated for the number to mean anything.
How do you calculate deposit beta from the call report?
Take interest expense on deposits from Schedule RI, item 2.a. Because Schedule RI is year to date, subtract the prior quarter's figure to isolate the quarter, then multiply by four to annualize. Divide by average interest-bearing deposits from the quarterly averages on Schedule RC-K. Beta is the change in that rate between two quarters over the change in your benchmark across the identical window. Skipping the de-cumulation step is the easiest way to get it wrong: a year-to-date figure divided straight into average deposits gives a smoothed average of the year so far, not a quarterly rate.
What is the difference between cycle-to-date beta and down beta?
Cycle-to-date beta measures the whole rising-rate cycle from its starting point, so it is a long-window average that gets harder to move as the window lengthens. Down beta measures only the easing leg, from the peak in funding cost forward. Different windows, different numbers, and quoting the first when the board asked about the second is the most common error in a beta discussion.
Why is the down beta lower than the up beta?
Time deposits reprice on their own maturity schedule rather than the benchmark's, exception-priced relationships are renegotiated one at a time, administered savings and money-market rates get cut slowly because customers notice a cut more than a missed raise, and balances that migrated out of checking on the way up do not migrate back. The lag is visible in the filings: on first-quarter readings every cohort above $1 billion peaked in Q1 2024, while every cohort below kept repricing and did not peak until Q1 2025.
What has the industry deposit beta looked like this cycle?
Measured Q1 2024 to Q1 2026, banks over $100 billion gave back 74 of the 110 basis points their median cost of deposits rose over the four quarters ending Q1 2024, about 67%. Banks between $1 billion and $3 billion gave back 21 of 106, about 20%. Banks under $1 billion barely moved on that window because it was not their window: they kept repricing into Q1 2025, and from their own peak have given back roughly 12% to 18% of their rise.
Is a low deposit beta always good?
On the way up, yes: it means funding held while rates rose. On the way down it is the opposite, a bank still paying for a rate environment that has already passed. The same number is a franchise strength in one direction and unharvested margin in the other, which is why the two have to be reported separately.
Where these numbers come from
Every figure on this page comes from the FFIEC call reports that 4,296 banks file each quarter, current through the quarter ending June 30, 2026. That 4,296 is what BankingLens covers; the medians and quartiles here are computed over 4,238 of them, because the 58 non-insured non-deposit trust companies in FFIEC peer group 401 are left out of every statistic on the site. Nothing here is modeled, estimated, or sampled. You can pull the quarterly cost of deposits, cost of funds and deposit mix for any individual bank behind these medians on the dashboard, or benchmark a single institution against its own FFIEC peer group on Bank Peer Intel. Plans start at $29 a month (pricing).