SBA Regulations
By David Jones, MAI, CCIM
When an SBA 7(a) loan charges off, an average of 72.9% of the original approval is written off. But expected loss ranges from 0.51% to 6.73% depending on a single variable — and it is not whether the property is special purpose. Hotels, the archetypal special-purpose asset, carry the lowest expected loss of any major category in the programme.
Almost every published SBA risk figure is a charge-off rate — the share of loans that failed. It is the easier number to compute and it is the one the agency reports.
Loss severity directly shapes how lenders size equity on special-purpose collateral. Our hotel feasibility studies and car wash feasibility studies both address special-purpose designation, while multifamily feasibility studies cover a property type with deeper resale markets.
It is also only half the calculation. A lender's actual exposure is the chance of failure multiplied by how much is lost when failure happens. A 10% charge-off rate at 40% severity is a smaller problem than a 5% rate at 90%.
The FOIA release records the gross amount charged off against each failed loan, which makes severity computable directly. Across 30,839 seasoned charge-offs with a recorded amount:
That is high. It reflects what SBA 7(a) lending mostly is — working capital and business acquisition credit, frequently unsecured or secured against equipment and goodwill that has little liquidation value once the business has stopped operating. There is not much to recover.
Multiplying the two gives expected loss, the number that actually matters for pricing and for underwriting policy. The rest of this article is about what moves it.
| Loan size | Loans | Charge-off rate | Severity | Expected loss |
|---|---|---|---|---|
| Under $150,000 | 273,412 | 7.65% | 77.3% | 5.91% |
| $150,000–$350,000 | 84,502 | 6.05% | 68.4% | 4.13% |
| $350,000–$1M | 72,985 | 4.40% | 60.6% | 2.67% |
| $1M–$2.5M | 36,954 | 3.52% | 56.3% | 1.98% |
| Over $2.5M | 11,158 | 2.90% | 51.6% | 1.50% |
Small loans fail more often and lose a larger share of principal when they do. The two effects compound rather than offsetting, and the gap between the smallest and largest bands is 3.9 times.
The severity column explains why. A $100,000 working-capital loan is typically secured by a general business lien over assets that are worth very little in liquidation. A $3 million acquisition loan is usually secured by real estate that retains value regardless of whether the business survives.
Loan term is a proxy for what secures the loan. SBA maximum terms are set by use of proceeds: ten years for working capital and equipment, twenty-five for real estate. A long term is a reliable signal that there is property behind the credit.
| Term | Loans | Charge-off rate | Severity | Expected loss |
|---|---|---|---|---|
| Under 7 years | 243,799 | 9.33% | 72.1% | 6.73% |
| 7–10 years | 129,202 | 4.76% | 82.5% | 3.93% |
| 10–20 years | 42,265 | 2.67% | 52.7% | 1.41% |
| 20 years or more | 63,512 | 1.03% | 49.3% | 0.51% |
A twenty-five-year real-estate-secured SBA loan carries roughly one-thirteenth the expected loss of a short-term working-capital loan. That is the largest effect in this entire dataset, larger than industry, larger than loan size, larger than borrower experience.
Note the anomaly in the 7–10 year band: severity is highest there at 82.5%, above the short-term band. These are equipment and larger working-capital loans — big enough that the loss is substantial, but not secured by anything that holds value. It is the worst of both structures.
SOP 50 10 8 requires a third-party feasibility study when the collateral is a special-purpose property — an asset built for one use, with limited alternative use if the business fails. Hotels, car washes, gas stations, funeral homes, assisted living, bowling alleys, fitness centres.
The premise is that a special-purpose property is harder to sell and therefore riskier. That premise is correct about liquidation difficulty. It is not, in this data, correct about loss.
Expected loss by property type, seasoned FY2010–FY2019 originations:
| Property type | Loans | Charge-off rate | Severity | Expected loss |
|---|---|---|---|---|
| Specialized freight trucking | 3,051 | 9.41% | 75.2% | 7.07% |
| Food service contractors | 3,729 | 7.75% | 75.0% | 5.81% |
| Restaurants | 34,014 | 8.19% | 70.5% | 5.77% |
| Fitness and recreation centres | 11,115 | 8.29% | 69.5% | 5.76% |
| General freight trucking | 14,792 | 7.65% | 71.7% | 5.48% |
| Grocery stores | 6,418 | 7.43% | 72.9% | 5.42% |
| Personal care services | 12,755 | 7.42% | 72.3% | 5.36% |
| Home health care | 2,923 | 6.84% | 74.8% | 5.12% |
| Funeral homes | 5,500 | 6.45% | 74.7% | 4.82% |
| Beer, wine and liquor stores | — | — | — | 2.47% |
| Child care | — | — | — | 2.24% |
| Gas stations | — | — | — | 1.77% |
| Dental offices | — | — | — | 1.46% |
| Hotels and motels | — | — | — | 1.12% |
Hotels carry the lowest expected loss of any category measured — one-sixth that of specialized freight trucking, and less than a quarter of restaurants.
This is not a paradox once term is accounted for. A hotel loan is a real-estate acquisition on a twenty-five-year term with the property as collateral. A trucking loan is a seven-year equipment credit against assets that depreciate and can be driven away. The special-purpose classification is asking whether the building is hard to repurpose. The loss data is answering a different question: whether there is a building at all.
Gas stations tell the same story at 1.77%. So do dental practices at 1.46% — a specialised fit-out, but usually real-estate backed and financed against a professional practice with durable cash flow.
None of this argues that the special-purpose feasibility requirement is wrong. It argues that it is measuring something other than expected loss.
The requirement exists because a lender holding a foreclosed car wash has a genuinely harder disposition problem than one holding a warehouse. That is true, and it is a real risk that shows up as time-to-liquidation and disposal cost rather than as write-off frequency. The severity figures here are recorded charge-off amounts, not net recoveries after sale, so the disposal drag is invisible in this data.
But the practical reading for anyone underwriting these deals is that the special-purpose label is a poor proxy for credit risk, and the variables that predict loss — term, collateral, loan size, industry cash-flow durability — are available at underwriting and are not what triggers the study requirement.
A restaurant on a ten-year working-capital loan is not classified as special purpose and carries 5.77% expected loss. A hotel acquisition on a twenty-five-year mortgage is classified as special purpose and carries 1.12%. The requirement lands on the second and not the first.
For lenders, expected loss rather than charge-off rate is the right basis for pricing and for portfolio limits. A book concentrated in short-term small-dollar credit carries roughly thirteen times the expected loss of a real-estate-secured book, and that difference is not visible in a charge-off table.
For borrowers, the structure of the loan matters more to your lender's risk assessment than your industry does. If the transaction can be structured to include real estate on a longer term, it becomes a materially different credit.
For the secondary market, guaranteed portions differ enormously in expected loss by term band, and term is disclosed at loan level.
Severity is gross, not net. The extract records the amount charged off, not amounts recovered afterwards through liquidation or guarantor pursuit. True net loss to the SBA and to lenders is lower than the severity figures shown, and the gap is likely to be widest exactly where disposition is hardest — which is to say, on special-purpose collateral. This is the single most important caveat in the article and it cuts against the headline finding.
Term is a proxy for collateral, not a measure of it. The collateral indicator field in the extract is not reliably populated in recent years and was not used. Long terms strongly imply real-estate security under SBA maximum-term rules, but the inference is structural rather than observed.
NAICS is self-declared at approval and describes the borrower's business, not necessarily the collateral. A hotel loan is coded 7211 whether or not the property secures it.
Cohort selection. FY2010–FY2019 originations, chosen for seasoning. These loans were exposed to both the post-crisis recovery and the pandemic.
Charge-off timing is not modelled. A loan that fails in year two and one that fails in year nine appear identically here.
Source. SBA 7(a) FOIA loan-level extracts, as of 31 March 2026. Records with status CANCLD excluded. Cohort: FY2010–FY2019 approvals, 479,011 originated loans.
Charge-off rate. Loans in CHGOFF status divided by all originated loans in the cohort.
Severity. GrossChargeOffAmount divided by GrossApproval, computed only on charged-off loans with a recorded amount above zero (30,839 loans), clipped at 150% to remove data errors where accrued interest and fees exceed the original principal.
Expected loss. Charge-off rate multiplied by mean severity within each group. This is a descriptive product, not a modelled estimate.
Property types. Four-digit NAICS, restricted to codes with at least 1,200 seasoned loans and 60 recorded charge-offs, then mapped to plain-language property categories.
Term bands. TermInMonths as recorded at approval.
The severity panel — charge-off rate, mean severity and expected loss by loan size band, term band and property type — is available as a CSV.
Licence. This dataset may be reproduced, analysed and republished, in whole or in part, for any purpose including commercial use, provided that sba-feasibility-study.com is credited as the source with a link to this page.
Applying this data to a live transaction? Our SBA feasibility study consultants explain scope, timing, and what lenders and CDCs expect, and the data research hub lists every study in the series.
Working a live file? Our SBA feasibility study consultants guide sets out scope, timing and what lenders and CDCs expect, the engagement record shows completed studies by asset class, and the full article index lists everything published here.
Reference material: glossary of SBA and feasibility terms, feasibility study FAQ, document checklist, educational center, calculator suite and the full resources library.
We are an independent SBA feasibility study company. Our SBA feasibility study consultants prepare lender ready studies under SOP 50 10 8 standards. See feasibility study cost and pricing, then request a fixed fee quote. Every study is benchmarked against our flagship SBA Feasibility Risk Meter research on 1,036,077 SBA loans.
Also useful: completed engagement record, our methodology, original SBA loan research and the firm and analyst profile.