Market Analysis

Vintage Watch: The FY2023 SBA Cohort Is Failing Faster Than Any Book Since the Financial Crisis Era

By David Jones, MAI, CCIM

Loans approved in fiscal 2023 have charged off at 0.917% within their first 24 months — the highest early-failure rate of any origination year measured. They borrowed at an average initial rate of 10.12%, 85% of them on variable terms, and 11.57% of the book is currently in some form of distress.

Dataset: the cohort panel behind this analysis is available for download at the end of this article under an attribution licence.

Why the usual numbers hide this

The SBA’s published charge-off statistics are a portfolio figure. They mix every origination year together, and they are dominated by loans that have already run most of their term. In a period of rising origination volume, that mix flatters the recent book: FY2024 and FY2025 loans have had almost no time to fail, so they pull the aggregate rate down while telling you nothing.

The FOIA release makes a better measure possible. Every loan carries its approval date, its current status, and — where applicable — the date it was charged off. That allows a like-for-like comparison: what share of each origination year had charged off by the time it was 24 months old?

That single change reorders the picture entirely.

FY2019 through FY2021 are artificially clean — pandemic-era deferment, forbearance and stimulus suppressed failures that would otherwise have surfaced. The meaningful comparison is FY2023 against the last normal pre-pandemic cohorts, FY2016 through FY2018. Against those, FY2023 is worse than all three.

What the FY2023 borrower actually signed

The cost of capital moved further and faster between FY2021 and FY2023 than in any comparable period in the programme’s recent history.

A borrower who closed in FY2021 on a ten-year term at 5.49% and a borrower who closed in FY2023 at 10.12% are servicing very different obligations against the same kind of business. On a $500,000 loan over ten years, that difference is roughly $1,300 a month.

And 85% of the FY2023 book is variable. Those borrowers did not fix the higher rate — they accepted it as a floating spread over prime, at a moment when prime was at its highest level in two decades. A file underwritten to a debt-service coverage ratio of 1.25 at closing has very little room in it.

The distress that hasn’t resolved yet

Charge-offs are the end of a process, not the beginning. Before a loan is written off it typically passes through delinquency, deferment, transfer to liquidation, or purchase of the guaranty without charge-off. The FOIA extract records all of these as live statuses, which means the deterioration is visible well before it becomes a loss.

Current distress by origination cohort:

Read these with care: they are not age-matched. Older cohorts show lower distress partly because their troubled loans have already resolved one way or the other. FY2024 and FY2025 show low distress because they are young.

What the table does show is the size of the unresolved problem sitting on the FY2022 and FY2023 books right now. Roughly one loan in nine from FY2023 is currently in a non-performing status of some kind. Some of those will cure. Historically, most will not.

Neither cohort’s losses are close to fully expressed.

Which businesses are failing

Within the FY2022–FY2023 cohorts, distress is not evenly spread.

Transportation and warehousing leads, and it is the most legible result in the table. Freight rates peaked in 2022, small carriers and owner-operators expanded on the strength of those rates, and the subsequent freight recession left them servicing equipment debt against materially lower revenue. Nearly 6,000 SBA loans were written into that sector across the two peak years.

Accommodation and food service carries the largest absolute exposure at 12,289 loans and is the sector most people would have named first — but it is fifth by rate, not first.

What this means

For lenders holding FY2022–FY2023 paper, the unresolved distress on those books is the number worth stress-testing. A portfolio-level charge-off rate that blends in a large, clean FY2024–FY2025 book will understate what is coming from the peak-rate vintages.

For the secondary market, guaranteed portions from FY2022 and FY2023 originations carry measurably different loss experience from adjacent vintages, and the cohort data is public. Pricing that treats vintages as interchangeable is pricing on incomplete information.

For borrowers who closed in FY2023 on variable terms, the arithmetic cuts the other way for once. Rate relief flows straight through to a floating-rate loan. Where a fixed-rate borrower would need to refinance and pay for the privilege, a variable-rate borrower gets the benefit automatically.

For anyone underwriting a new file, the FY2023 cohort is a useful reminder that a debt-service coverage ratio calculated at closing is a snapshot, not a forecast, and that a 1.25 on an 85%-variable book is thinner than it looks.

Limitations

The 24-month measure requires a complete window. FY2024 originations have not all reached 24 months, so their figure is partial and is flagged as such. FY2025 is excluded from the comparison entirely.

Distress rates are not age-matched. The cohort distress table compares books at different stages of life and should be read as a snapshot of current condition, not as a like-for-like failure comparison. The 24-month charge-off table is the age-matched measure and is where the central claim rests.

Pandemic cohorts are not comparable. FY2019 through FY2021 benefited from deferment, forbearance and stimulus that suppressed failures. They are shown for completeness and excluded from the comparison.

Charge-off is not loss. The extract records written-off amounts but not recoveries.

Rates are as approved. Initial interest rate is recorded at approval. For the 85% of loans on variable terms, the rate actually paid has moved since.

This is a cohort analysis, not a forecast. It measures what has already happened to loans approved in specific years. Whether the FY2023 book continues to deteriorate depends on rate policy, sector conditions and the economy, none of which this dataset speaks to.

Frequently asked questions