Research · published 2026-07-28
504 borrowers default at less than half the 7(a) rate
We measured every SBA loan approved since 2010 — 890,648 of them — and grouped them by the year they were approved so that fast-growing programmes cannot hide behind young loans. Within seven years of approval, 2.11% of 7(a) dollars had been charged off against 0.84% of 504 dollars. The gap survives controls for industry and for loan size. It does not survive going up-market: above $2 million it disappears entirely.
The headline, and why it is not the whole story
Of loans old enough to judge, 24,754 of 432,960 7(a) loans have charged off. For 504 the figure is 608 of 53,621. On a dollar-weighted basis that is 2.11% against 0.84%.
The obvious objection is that the two programmes lend to different people. They do, dramatically. 71% of 7(a) loans are under $350,000; only 9% of 504 projects are. If small loans fail more often — and they do — then some of the gap is just size.
Share of each programme's loans by total financed amount. The programmes barely overlap.
Control 1: reweight 7(a) to 504's loan sizes
does the gap survive?If you take the 7(a) default rate within each size band and reweight it to match 504's mix of loan sizes, 7(a) falls from 2.11% to 1.87%. So size explains part of the gap — but only part.
A 2.2× gap remains after adjusting for the single most obvious confounder. Whatever is going on, it is not simply that 504 does bigger deals.
Control 2: compare the same industries
7-year charge-off rate, both programmes| Industry | 7(a) | 504 | Ratio |
|---|---|---|---|
| Full-service restaurants | 3.17% | 1.41% | 2.2× |
| Fitness and recreation centres | 2.86% | 1.05% | 2.7× |
| Beer, wine and liquor stores | 2.58% | 1.63% | 1.6× |
| Offices of dentists | 1.66% | 0.70% | 2.4× |
| General automotive repair | 1.61% | 0.62% | 2.6× |
| Real estate offices | 0.78% | 0.31% | 2.5× |
Six industries with meaningful volume in both programmes. The ratio clusters between 1.6× and 2.7× — close to the size-adjusted figure, and consistent enough that it is unlikely to be an artefact of any single sector.
Where it breaks down
The gap narrows steadily as loans get larger, and at the top it reverses. Under $350,000 7(a) defaults at four times the 504 rate. Between $2M and $5M, 504 is marginally worse — 1.03% against 0.96%. Whatever advantage 504 has is a small-loan phenomenon, and it is gone by the time you are financing a $3 million building.
The gap by loan size
Ratio of 7(a) to 504 charge-off rate within each band. Above $2M the programmes perform the same.
They also fail at different times
share of all failures, by year after approval7(a) failures cluster early — by year three, 28% of them have already happened, against 12% for 504. More than half of all 504 failures arrive after year six. That is the signature of a longer, mostly fixed-rate loan secured on property: when these deals go wrong, they go wrong slowly.
What we think is happening
We can measure the gap. We cannot prove its cause from this data, and it would be dishonest to pretend otherwise. Three explanations fit what we see, and they are not mutually exclusive.
What we did not find, and the numbers we distrust
Two industries move against the pattern, and in both cases the 504 sample is thin enough that we would not lean on them. Long-distance trucking shows a 24.7× gap — but on only 625 seasoned 504 loans with 5 failures, where a single additional default would move the figure by a fifth. Assisted living reverses, with 504 worse than 7(a), on 676 loans and 8 failures. Both are reported here because leaving them out would be picking the data that agrees with us.
We also cannot compare interest rates between the programmes. SBA's 504 release contains no rate field at all, so the fixed-versus-floating explanation above is inference from loan structure and failure timing, not something we measured directly.
Method
so you can check itSource. SBA's 7(a) and 504 FOIA releases, as of 2026-06-30, US government public domain. 795,652 funded 7(a) loans and 94,996 funded 504 loans approved FY2010 onward. Cancelled and never-funded approvals are excluded — 137,000 of them in 7(a) and 23,000 in 504. Leaving them in the denominator deflates every rate by roughly 15% and is the most common way these figures are published wrong.
Vintages, not pools. Charge-offs surface three to five years after approval, so an all-time average makes any recently grown portfolio look pristine. Loans are grouped by fiscal year of approval and compared only at equal ages. "Seasoned" means every loan in the cohort has had at least seven years to fail.
Dollar-weighted. Charged-off dollars over approved dollars. For 504, "approved" is the CDC second only; the bank first mortgage is not SBA debt and is not counted, though it is included when classifying project size.
Reproducing this. The pipeline is straightforward and the source files are public. Write to corrections@sbadecoded.com if you want the working, or if you think we have this wrong.
If you are choosing between them
None of this says 504 is right for you. It says that among borrowers who took each, 504 borrowers failed less often — and that the difference shrinks as deals get larger. If you are buying property and have been quoted only a 7(a), it is worth getting a 504 quote before deciding.
Compare the two on your numbers → How 504 works → Find a CDC →