Ask a room of first-time buyers how often SBA acquisition loans go bad, and you will hear a frightening fraction, delivered with confidence and no source. Ask a lender, and you will hear a reassuring one. Both answers miss the point. The scary number is folklore. The reassuring number, even when it is drawn correctly from SBA data, measures the lender's experience, not yours. This lesson separates the two, shows you the number that should actually concern a buyer, and teaches you how to cite loan data so it means something.
Principle
Loan performance is not owner outcome. A default rate tells you how often lenders stopped getting paid. It does not tell you how often buyers lost their money, their savings or their house. Those are different events, and the second happens far more often than the first can show.
A lender is protected three ways on an SBA 7(a) loan: the SBA guarantees part of the loan, the business's assets are pledged as collateral, and you sign a personal guarantee. A loan can stay current for years while the owner drains savings to make payments, takes no salary, and watches the equity they put in shrink to nothing. When that loan is finally paid off or refinanced, it counts as a success in the lender's data.
So the rule is: the number that should concern you is the share of buyers who lose their equity, not the share of loans that default. And for self-funded buyers, nobody publishes it.
Case
The folklore figure
Grade: folklore. The frightening fraction you hear for SBA acquisition loans has no primary source that measured acquisition loans. As far as it can be traced, it comes from lifetime loss figures for whole decades of 7(a) loans, every purpose combined, later repeated as if they described acquisitions. For most of the program's history, the public loan data did not separate change-of-ownership loans from the rest at all, so nobody could have measured the figure people now quote.
Do not repeat it, and do not repeat any number that compares a "real" default rate against an "advertised" one unless you can name the dataset, the cohort and the year.
What the loan data can and cannot tell you
The SBA publishes loan-level data on the 7(a) and 504 programs through its open data portal, covering loans back to the 1990s (Y1). A Congressional Research Service report on the 7(a) program, first issued in 2010 and updated through 2022, explains how the program works, including that proceeds may be used to acquire an existing business (Y2). Acquisitions are a large part of the program: a 2026 Coleman Report analysis found that one in four 7(a) lending dollars finances business acquisitions (Y22).
This textbook does not print a default rate for change-of-ownership loans. Its rule, set out in lesson 0.1.3, is that a real-world figure carries a year and a source you can open, and no published analysis in the Library states that rate with all six parts listed in Step 1 below. You will still be quoted one, usually a low figure for loans approved in the last few years. When you are, read three conditions into it before you use it:
- The loans are young. Most loans in a recent cohort have had only a few years to fail. Small business loans tend to go bad over years, not months.
- A window that includes 2020 and 2021 includes pandemic-era relief, which suppressed early defaults across the program.
- It is a lender's metric. It measures lenders losing money, not buyers losing money.
The data also has gaps. A 2026 SBA Office of Inspector General evaluation found that loans from nonbank Small Business Lending Companies significantly underperformed other lenders' 7(a) loans on default rates, but it did not break out change-of-ownership loans (Y21). The official data that would answer the question precisely for acquisitions is not yet mature.
So the honest summary is: a low rate on young acquisition loans can be accurate and still be the wrong number for you. It is news for lenders. It tells you almost nothing about buyers.
The number that should concern you
The only published measures of equity loss in this field come from traditional search funds, so read them as that and nothing more. A 2020 Yale note, citing the 2018 Stanford study, reports that 29 percent of traditional search funds that closed a deal ended in partial or total loss of equity (M20). A 2025 Yale analysis of 12 serial investors found that only 51 percent of their 1,192 backings returned at least the money invested (M1), which means the other half did not.
Those are investor-backed deals, with boards, experienced investors and diversified capital behind them. They are not self-funded figures, and you should not apply them to a self-funded deal. What they show is that even in the best-studied, best-supported model, losing equity is common, while the loans in a deal can perform fine.
For self-funded SBA buyers, no comparable dataset exists. The failures are hard to count partly because they go quiet. Anecdotally, from the interview corpus of 60+ failed and struggling buyers, roughly 8 of 10 post-close failures ended in some mix of litigation, bankruptcy and a settlement carrying confidentiality terms. That is an estimate from those conversations, not a measured rate, but it explains why the failure record rarely reaches any dataset.
Annual is not cumulative
The other common error is mixing an annual rate with a cumulative one. Illustrative arithmetic, with an invented rate: suppose a group of loans defaults at 2 percent a year. After one year, 2 percent have defaulted. After ten years, the cumulative share is about 18 percent (the share surviving is 0.98 multiplied by itself ten times, roughly 82 percent). Quote the 2 percent next to someone else's 18 percent and you have "proven" a ninefold difference that does not exist.
Coastal Comfort: the lender's view and yours
In the fictional Coastal Comfort HVAC deal, the buyer borrows $3,030,000 under SBA 7(a), puts in $210,000 of their own cash beside $210,000 from outside investors, and owes the seller $750,000 across two notes: $500,000 that waits on full standby until the SBA loan is paid off, and $250,000 that pays from closing. Under SBA SOP 50 10 8.1, effective October 1, 2026 (Y25), every individual who owns 20 percent or more of the borrower must provide an unlimited full personal guarantee.
Picture the bad-but-not-catastrophic year. In the bear case this textbook models on the closed deal, Coastal Comfort's earnings fall from the quality-of-earnings figure of about $920,000 to about $583,000. That still covers the roughly $524,000 of annual payments on the SBA loan and the paying seller note about 1.11 times, at an illustrative 10 percent rate on the SBA loan, so every payment is made. After capital spending and tax, though, the business is about $77,000 short of cash in the first year, and the buyer covers the gap from savings. The lender's data records a performing loan. The buyer's records show a year of writing checks and a shrinking bank account. If it keeps going, the buyer's $210,000 is gone long before the loan ever shows up as a default. Your outcome and the loan's outcome separate early, and only one of them is measured.
Steps
1. When you hear a default number, ask what it measures
Every loan-performance figure has six parts. If any is missing, you do not yet know what the number means:
- The metric. Default, guaranty purchase (the SBA paying the lender on its guarantee) or charge-off (the loss written off). These are different.
- The basis. Share of loans by count, or share of dollars.
- The purpose. All 7(a) loans, or change-of-ownership loans only.
- The cohort. Which approval years.
- The window. Annual, cumulative to a date, or lifetime.
- The source and year. Which dataset, pulled when, analyzed by whom.
2. Rewrite before you repeat
A correct citation names all six parts. Use this as the template: "Change-of-ownership 7(a) loans approved in fiscal years [cohort] show [rate] [defaulted, purchased or charged off], by [count or dollars], [annual, cumulative to a date, or lifetime], in [dataset], pulled [date] and analyzed by [whom]." Then add the caveats that apply, such as a young cohort or a relief period. Anything shorter is missing something.
3. Never compare across windows
Do not compare a young cohort's rate with a lifetime rate, an annual rate with a cumulative one, or a dollar-basis rate with a count-basis rate. If two figures differ on any of the six parts, they are not comparable.
4. Model your own equity loss, not the loan's default
For your target, or for Coastal Comfort, write down the year in which you would stop paying yourself, the year your savings would run out covering payments, and what the personal guarantee would reach after that. The SBA Stress-Tester at /sba-stress-test lets you run the debt side of that test on your own numbers.
5. Ask your lender the right question
Lenders can tell you about their loan performance. Ask instead: "Of the acquisition borrowers you have financed, how many are still running the business, and how many sold or closed it at a loss?" Most will not know. That tells you something too.
Deliverable
Two short documents for your deal file:
- A Loan Data Citation Card: the six-part template above, filled in for every default or loss figure you have heard during your search, each graded confirmed, misapplied or folklore.
- A Personal Downside Sheet: for your target or Coastal Comfort, the earnings decline at which you stop drawing a salary, the months of payments your savings cover after that, and a list of every asset your personal guarantee reaches. Lesson 1.2.2 shows how to trace each guarantee in full.
This lesson is education, not lending, legal or investment advice. SBA rules and loan data change; confirm current terms with an SBA lender and current data at its source.