How Did Your ALCO Model Hold Up?

My firm did a sample data run for a client that included all commercial banks in NY, NJ, PA, and MD between $500 million and $1.5 billion in total assets to see how various banks did in balance sheet and income statement ratios during the course of the Fed tightening run from year end 2021 until the third quarter 2023. Some interesting insights relating to their 1-year cumulative repricing gap that the banks reported on their call reports:

• At 12/31/21, of the 68 banks that met the criteria, only 10, or 15% had a 1-year cumulative negative gap. This is defined as rate sensitive assets (assets that are expected to mature or reprice within 1 year) less rate sensitive liabilities (liabilities that are expected to mature or reprice within 1 year). If rates went up, so the theory goes, the 85% of banks with a positive 1-yr cumulative gap, should see net interest margin go up as assets reprice faster than liabilities. This made sense because the Fed Funds Rate at this time was 0-25 bps and bankers positioned their balance sheets accordingly.

 

• At 12/31/22, after 450 bps of Fed rate hikes, 48 of the 68 banks, or 71%, had a better net interest margin for the quarter ended 12/31/22 than the quarter ending 12/31/21. Since 85% of them had positive one-year cumulative gaps, their ALCO assumptions mostly worked.

 

• At 9/30/22, only 13 banks, or 19% showed a negative one-year cumulative gap. Meaning 81% thought their net interest margin would increase in a rising rate environment. Between 9/30/22 and 9/30/23, 53 banks, or 78%, had a lower net interest margin. How could their ALCO assumptions be so wrong?

• By 9/30/23, the 1-year negative cumulative gap had nearly tripled to 30, or 44%. Interesting because the Fed Funds Rate was zero-25bps at 12/31/21 and almost everyone knew rates would inevitably go up. It makes sense that so few considered themselves liability sensitive at 12/31/21. I'm actually surprised so few (44%) consider themselves negatively gapped right now. Declining rates are far more likely than rising rates. The most recent Fed dot plot predicts Fed Funds declining in 2024.

When I asked my colleagues what they thought, here is what a couple of them had to say:

If I recall from my ALCO committee days… ALCO models largely did not rate shock 450-500 points and if they did, that type of move seemed quite far out of the realm of possibilities.  In a 200-300 model, spreads would have mostly held up. My guess is that the duration of money market accounts in most ALCO models were in the 3-6 year time frame but when rates went up 500 points in the real world, these longer duration "core deposits" actually left the bank or repriced much faster than anticipated as banks worked to retain these accounts.  Also, many banks were using CDs to retain these accounts and shifting deposits out of these longer duration products into 6-12 month CDs shortened the liability duration averages (in models) and increased the liability sensitive nature of most banks.

Deposit duration assumptions in ALCO models built for 'normal' markets simply did not hold up in recent quarters.

~ Ben Crowley, Managing Director, The Kafafian Group, Inc.

I think bankers overestimated the loyalty of their depositors following the pandemic & PPP, coupled with a sustained low rate/high liquidity environment. These factors led to a false sense of security that low-cost deposits were there to stay. When the national and super regional banks began raising rates they were reluctant to follow – until it was too late. They quickly learned that customers were not loyal and deposit attrition happened so fast that they had to raise rates more aggressively than anticipated to retain remaining deposits and attract funds to replace what they had lost.  

Service is important. But you still have to price competitively.

~ Chris Jacobsen, Managing Director, The Kafafian Group, Inc.

~ Jeff