A Fed Study Puts a Lending Cost on CECL

A Federal Reserve staff working paper posted September 25, 2026 estimates that the current expected credit losses standard, or CECL, reduced annual loan growth by about 77 basis points in the banks it studied. The finding matters beyond bank accounting: it suggests that credit can tighten through reserve and capital mechanics even when policy rates and borrower demand do not change.

The judgment

• Implication: Loan-loss accounting is part of credit supply. A lender’s expected-loss model can affect how much balance-sheet capacity remains for new loans, especially when capital is already constrained.
• What would change the conclusion: The conclusion would weaken if later research fails to replicate the estimated effect outside this sample, or if banks respond over time by raising and retaining enough capital to prevent allowances from constraining lending.
• Management action: Borrower CFOs should add lender-capacity and renewal risk to liquidity forecasts; bank CFOs should separate expected loss, capital usage and liquidity cost when explaining loan pricing and growth limits.

The reported finding

CECL replaced the incurred-loss model for U.S. generally accepted accounting principles. Most public banks adopted it on January 1, 2020; remaining banks generally followed on January 1, 2023. Instead of waiting until a loss is probable, CECL requires lenders to recognize expected credit losses over the remaining life of a loan.

That timing difference changes more than reported earnings. Loan-loss provisions reduce earnings and therefore capital. Because equity capital is generally more expensive than deposits or other bank funding, a larger allowance can make lending capacity more costly in much the same way as a higher capital requirement.

The paper analyzes regulatory data from 2016’s first quarter through 2024’s second quarter. Its final sample contains 364 bank holding companies and standalone banks with $1 billion to $100 billion of assets as of late 2019, after excluding very large and very small institutions and applying other comparability filters. The authors also remove Paycheck Protection Program balances and control for local employment and regional economic conditions.

Using differences in adoption timing, loan products and the day-one effect of CECL, the authors’ preferred instrumental-variable estimate finds that a one-percentage-point increase in allowance coverage reduces lending by about 0.85% per quarter. Applied to CECL’s estimated 24-basis-point increase in allowance coverage in the sample, that becomes an estimated 77-basis-point reduction in annual loan growth. Average annual loan growth in the sample was about 6.4%.

Those are estimates, not observed mechanical rules. The paper is preliminary staff research and explicitly does not represent a conclusion of the Federal Reserve Board.

Why the capital channel matters

The most useful result may not be the 77-basis-point headline. It is the evidence about how banks absorbed the accounting change.

The paper reports that the negative relationship between allowances and loan growth was stronger at more capital-constrained banks. For each standard-deviation increase in excess capital, the estimated effect fell by roughly one-third of the average effect. The authors also found no statistically significant effect on common-stock distributions or share repurchases.

The authors interpret those results as evidence that the allowance shock was accommodated primarily through reduced lending rather than lower shareholder distributions. Numbers & Judgment’s analysis is that this creates an important forecasting distinction: a bank can remain solvent, meet regulatory ratios and continue paying shareholders while still becoming a less willing source of incremental credit.

For borrowers, “the bank has capacity” is therefore not a binary statement. Capacity depends on the lender’s capital buffer, portfolio mix, expected-loss assumptions and the type and maturity of the proposed loan. A borrower may look unchanged while the lender’s economics have moved.

Do not turn 77 basis points into a universal forecast

The study’s design is unusually careful, but its estimate has boundaries.

• The sample deliberately excludes banks below $1 billion and above $100 billion in assets, so the result should not be applied mechanically to community banks or the largest global institutions.
• CECL’s first large-bank adoption occurred immediately before the pandemic. Emergency capital-transition rules muted part of the regulatory effect, while extraordinary fiscal and monetary programs changed credit conditions.
• The magnitude varies across methods. The paper says portfolio weighting produces stronger results, while synthetic-control estimates are weaker, although the overall conclusion remains.
• The analysis ends in 2024’s second quarter. Banks may continue to adapt their capital plans, pricing and product mix.

The right use of the estimate is not “subtract 0.77 percentage point from every lending forecast.” It is to treat expected-loss accounting as one of the mechanisms that can shift credit supply at the margin.

What borrower finance teams should change

A financing forecast should distinguish borrower risk from lender-capacity risk. That means tracking more than the interest rate.

• Map renewal timing. Identify which facilities mature, reprice or require approval during the forecast horizon.
• Watch concentration on both sides. A company may rely heavily on one lender, while that lender may be reducing exposure to the company’s industry, collateral type or loan maturity.
• Separate availability from usability. An undrawn line with a springing covenant, borrowing-base limit or discretionary renewal is not the same as unrestricted cash.
• Build a slower-credit case. Model the cash effect of a delayed renewal, smaller commitment or higher equity contribution without pretending to know a universal probability.

This extends the argument in CFO Confidence Masks a Cash-Access Divide: aggregate optimism can coexist with sharply different access to funding. It also complements The Fed’s Faster Discount Window Raises the Bar for Liquidity Drills, because nominal funding access is not the same as a tested path to cash.

The free 13-Week Cash Flow Forecast Template can provide the timing layer for a delayed or reduced borrowing scenario. The judgment remains management’s: deciding which commitments can proceed before credit is certain.

What bank boards should ask

CECL model governance is not only an accounting-control issue. Forecast horizons, reversion methods, loss histories and portfolio segmentation ultimately feed capital allocation and product economics. A board reviewing credit growth should be able to reconcile four things: expected losses, capital consumption, loan pricing and risk appetite.

The paper does not show that CECL is a mistake. Earlier recognition can make reserves more forward-looking and reduce the “too little, too late” problem exposed by the financial crisis. It does show that the benefit is not free. When expected losses enter the accounts earlier, some of the adjustment may appear as credit that is not originated.

Bottom line

Accounting standards can change real financing decisions. The Fed paper’s estimate is bounded, preliminary and sensitive to method, but its direction is operationally important: when allowances absorb scarce capital, lending can slow before dividends do.

For CFOs, that means credit availability belongs in the forecast as a variable, not a promise. For bank boards, it means reserve methodology belongs in the capital-allocation discussion, not only the audit committee binder.

Sources

• Federal Reserve Board, “How Did CECL Affect Bank Lending?”, FEDS 2026-066, posted September 25, 2026.
• Ranish and Vojtech, full working paper, August 2026.
• Financial Accounting Standards Board, Accounting Standards Update 2016-13, June 2016.

Reported facts and estimates are attributed above. Interpretations, operating implications and management recommendations are Numbers & Judgment analysis.