Written by: Luke Sobieraj, Founder & COO, Digital Colliers
The Fed raised rates by 25 basis points on 16 September 2026, the first hike since the 2022-2023 cycle ended. Inflation prints ran hot through August, pushing the 10-year Treasury yield back above 4.6%. Most mid-market banks walked into boards the next week with the same static interest-rate sensitivity tables they used three years ago.
Those tables assume the world still works the way it did in 2022. It doesn't. Deposit behavior shifted. Loan prepayment patterns changed. Basis risk between asset and liability repricing is wider than the old models allow for.
What happened on 16 September
The Federal Reserve's September hike was smaller than the 50 and 75 basis point moves we saw in 2022, but the macro setup is different this time. Inflation re-accelerated through Q3 2026 after a brief pause, and bond markets moved faster than the Fed. The 10-year yield climbed 80 basis points between June and mid-September before the FOMC even met.
That matters because most banks model interest-rate risk as if the Fed is the only game. In reality, your loan book reprices off SOFR and your deposit costs follow the federal funds rate with a lag that varies by product and vintage. When the yield curve steepens ahead of Fed action, the old static tables miss the basis risk entirely.
Why static sensitivity tables fail now
Static tables assume parallel rate moves across the curve and linear deposit behavior. A typical mid-market bank's 2022-vintage model says a 100 basis point rate increase lifts net interest income by $X over 12 months. That number sits in a board deck for two years.
The model doesn't capture what actually happens when rates move. Deposits don't reprice in a straight line. A 25 basis point hike might move checking account costs by 5 basis points and money market costs by 20 basis points, and the mix between the two shifts as customers hunt for yield. Loan prepayment speeds change as borrowers refinance or don't. The table shows one number. Reality has twelve.
Regulatory pressure is rising too. DORA came into force in January 2025, and examiners are asking harder questions about operational resilience across risk functions. A sensitivity table from 2022 doesn't meet that bar.
What dynamic rate-risk modeling looks like
Dynamic models run scenarios, not single-point estimates. You feed in a rate path, and the model walks through how each segment of your balance sheet responds. Commercial real estate loans with rate floors behave differently than floating-rate C&I. Retail deposits in markets where you face heavy competition reprice faster than deposits in branches where you're the only game.
The shape of the solution has four pieces:
- Segment your loan book and deposits by product, vintage, and rate sensitivity. Not just "loans" and "deposits" in aggregate.
- Model deposit betas by segment. How much of a rate move passes through to each product, and how fast.
- Run multiple scenarios. Parallel shifts, bear steepeners, instantaneous shock, gradual ramp. See where you're exposed.
- Refresh the model with actual repricing data every month. Your deposit beta estimate shouldn't be static.
You're not building a PhD econometrics exercise here. You need a model that treasury can run every week and boards can read without a manual.
Refresh cadence that works
Most banks I see refresh their rate-risk view once a quarter, timed to ALCO meetings. That's not fast enough when the macro picture is moving. The operators who run this well do three things:
- Weekly sensitivity runs for treasury and the CFO. Five scenarios, updated with the latest forward curve and balance sheet position.
- Monthly full recalibration. Pull actual repricing data from the prior 30 days and update your deposit betas and prepayment assumptions.
- Quarterly deep dive for the board. Show the trend over 90 days, not just a snapshot. Boards want to see if your interest-rate position is improving or deteriorating, not just where it is today.
The infrastructure to do this isn't exotic. You need clean data feeds from your core systems, a model that runs scenarios without manual intervention, and someone in treasury who owns the refresh cadence. Most mid-market banks already have the first two. The third is the gap.