Which banks reprice deposits when the Fed hikes?

Three cycles, three distributions

Betas are dispersed in every cycle, which is the Drechsler-Savov-Schnabl point: deposit pricing power differs bank by bank, and it is persistent. The 2022-23 median (0.31) sits 0.07 below the 2004-06 median (0.38) and 0.07 above the 2015-19 median (0.24): the typical bank passed through less of this hiking cycle than of 2004-06, not more. Asset-weighting tells a different story because large banks reprice faster: 0.52, 0.36 and 0.49 across the three cycles.

Each panel: share of banks per 0.1-wide beta bin, same y-scale. Lighter edge bars collect betas below -0.2 and at 1.0 or above (real data, kept). 10th-90th percentile ranges: 0.23 to 0.58 (2004-06), 0.06 to 0.48 (2015-19), 0.13 to 0.56 (2022-23).

The 2022-23 pass-through built quarter by quarter

Cumulative betas start low and grind up as depositors reprice: the asset-weighted beta was 0.17 in 2022Q2, 0.34 by 2022Q4, and 0.49 by 2023Q3. Each point compares that quarter's bank-level deposit rates with the same banks' rates in 2022Q1, weighted by 2022Q1 total assets; the fed funds denominator is the quarter's average FEDFUNDS minus 2022Q1's 0.12%.

Banks in each point: those with a valid implied deposit rate in both 2022Q1 and that quarter (4,572 to 4,743 banks). The end-of-cycle point is the 0.49 headline above.

Deposit repricing and uninsured funding, shown together

The matrix crosses each bank's 2022Q1-2023Q3 cumulative beta with its uninsured share at the exact September 30, 2023 cycle endpoint. The low-beta, high-uninsured configuration contains 41 banks and 0.08% of assets in the matched sample. This is a funding configuration, not a risk score: low beta can reflect deposit mix and a valuable franchise, while uninsured concentration is not a run probability. Banks are anonymous here, and every cell carries its bank count and share of matched-sample assets.

Fixed round-number display bins, unchanged by refresh: beta below 0.25, 0.25 to below 0.50, and 0.50 or above; uninsured share below 25%, 25% to below 50%, and 50% or above. These boundaries are presentation conventions, not economic or risk thresholds; “low” and “high” identify only the outer intervals shown. Row and column bases are unconditional within the 4,572-bank matched sample. Asset shares use total assets reported on September 30, 2023, with the full matched sample as the denominator. Reported and proxy uninsured coverage are separated below the matrix.

The levels the betas summarize: FDIC national deposit rates

The betas above compress two curves into one number; this is the raw picture. The FDIC publishes national average deposit rates by product, and the gap they show is what a low beta means in dollars: in August 2023, with the average fed funds rate at 5.33%, the national average savings rate was 0.43% (it stayed between 0.33% and 0.46% across 2023) and interest checking paid 0.07%. Only term money repriced: the 12-month CD averaged 1.76%. In July 2026 the savings rate stands at 0.38% against a 3.63% fed funds rate.

Two FDIC publications, not one series: weekly national rates on non-jumbo deposits (under $100,000) from 2009-05-18 to 2021-03-29, then the revised methodology (weighted by domestic deposit share, credit unions included) monthly from 2021-04-01. No growth rate is computed across the seam. Source: FDIC National Rates and Rate Caps, all 95 series via FRED release 317.

The rate cap is the ceiling FDIC rules set for deposit rates at less-than-well-capitalized institutions (12 CFR 337.7): the higher of the national rate plus 75 basis points and, for maturity products, the Treasury-yield-based cap. Rates are national averages across insured institutions, not offer rates at any one bank; the bank-level table below uses implied rates from filings instead.

The largest banks, named

The ten largest banks by total assets at the 2022Q1 cycle start. Dispersion survives at the top: the range below runs from 0.26 to 0.75. These are implied average rates on interest-bearing deposits from public filings, not posted offer rates.

Why beta is the number to know

• Drechsler, Savov and Schnabl ("The Deposits Channel of Monetary Policy", QJE 2017) showed that banks with market power pass little of a hike through to depositors, so tightening drains deposits and contracts lending most where beta is low: doi.org/10.1093/qje/qjx019.
• Their sequel ("Banking on Deposits", JF 2021) reads a low beta as a valuable deposit franchise that behaves like a hedge: sticky cheap deposits offset the losses long-duration assets take when rates rise: doi.org/10.1111/jofi.13013.
• Drechsler, Savov, Schnabl and Wang ("Deposit Franchise Runs", JF 2025) is the 2023 caveat: the franchise only hedges if depositors stay, so a low-beta bank funded by uninsured depositors can lose the hedge exactly when it needs it, as at Silicon Valley Bank: nber.org/papers/w31298.
• DeMarzo, Krishnamurthy and Nagel ("Interest Rate Risk in Banking", NBER w33308) is the counterpoint: once the fixed operating costs of running the deposit franchise are netted out, the franchise value is smaller and the hedge weaker than the beta alone suggests: nber.org/papers/w33308.

Method, in brief

Deposit interest expense by category from Schedule RI (year-to-date, dequarterized), interest-bearing deposit balances from Schedule RC, fed funds from FRED's monthly FEDFUNDS averaged by quarter. Rates outside 0-25% are dropped as data errors and counted. The full MDRM item mapping, the dequarterization check on JPMorgan Chase, the drop counts, and the external anchor against the New York Fed's published industry betas are on the methodology page. Built by scripts/build_deposit_betas.py from the 101 quarterly FFIEC CDR bulk files; every number on this page is computed from its parquet output at build time.