504 defaults less than 7(a) — but almost entirely because its loans are bigger

Dollar-weighting — charged-off dollars over approved dollars — flatters whichever program writes bigger loans, because big loans default less often. 504 loans are big (a median $708,000 second, on top of an $880,000 bank first mortgage). So by dollars, 504’s losses look tiny: 0.84% against 2.11%. Count each loan once instead, and the picture is ordinary: 5.0% against 5.7%. Same federal file, same seven-year window — a completely different headline.

504 is genuinely safer only at the small end (under $350k). From $350k to $2M — where most 504 projects sit — 504 defaults more than 7(a), peaking at 8.0% versus 2.4% in the $1M–$2M band. 504’s overall edge comes almost entirely from 7(a) carrying an enormous tail of small, riskier loans, not from 504 borrowers being safer at a given size.

Two things stack up. First, the widely-quoted figures are dollar-weighted, which rewards 504 for lending big. Second, even the loan-counted overall number (5.0% vs 5.7%) is a composition effect — a textbook case of Simpson’s paradox — because 58% of all 7(a) loans are under $150,000, the band that fails most. Neither means a $600,000 504 borrower is safer than a $600,000 7(a) borrower. The data says the opposite.

None of this makes 504 a worse choice — it just isn’t a lower-default one. Its genuine advantages are structural: a fixed rate for up to 25 years, which removes the payment shock that pushes marginal variable-rate 7(a) borrowers over; a lower blended cost of capital on owner-occupied real estate; and a smaller down payment on big property deals. If you are buying a building, those are the reasons to run a 504 quote — not a default rate that, loan for loan, is no better. Compare the two on your numbers →

Source. SBA’s 7(a) and 504 FOIA releases of 2026-06-30, public domain. 146,851 seasoned 504 loans and 433,292 seasoned 7(a) loans — approved through fiscal 2018 and old enough to judge. “Default” is a charge-off within seven years of approval, counted by loan (one loan, one vote), which is why every rate equals a plain count you can check. Loans are grouped by fiscal year of approval so a fast-growing program cannot hide behind young loans.

One honest caveat. The size bands use the SBA-guaranteed amount — for 504 that is the CDC second only, so a 504 loan in a band sits on a larger total project than a 7(a) loan in the same band. We compare by SBA exposure because that is the number both programs report; read the bands as “SBA dollars at risk,” not total project size.

Choose 504 for owner-occupied property and heavy equipment, where its fixed long-term rate and lower down payment shine — not because it defaults less. Choose 7(a) for flexibility, speed and everything that isn’t real estate.

Compare on your numbers → How 504 works → Find a CDC →