Small-Balance Loan Performance by Property Type: Reading the Public SBA Tape

The public SBA loan tape does not record property type. It records an industry code, a term in months, a program, an approval amount and an outcome, and from those four things a careful reader can recover most of what a property-type question is really asking. That recovery is worth doing, because the SBA 7(a) and 504 Freedom of Information Act files are the only loan-level performance record of American small-balance commercial real estate that anyone can download without a contract, and they cover over a million loans. This article sets out how to read property type out of a tape that has no property-type field, using the borrower's North American Industry Classification System code, the loan term and the program as three independent proxies, and what that reading shows once the arithmetic is done honestly. The headline result is not the industry league table that most lenders expect. Within every property-backed industry in the file, the loans that ran 240 months or longer charged off at a small fraction of the rate of shorter loans in the same industry, and the gap between term buckets is larger than the gap between the safest and the riskiest industries. Term tells more than industry. This is the SBA chapter of the library's reference library of benchmarks built from public data, and it sits beside the structural guide to the SBA FOIA files, their fields and their suppression rules.

The field that is not there

The SBA Office of Capital Access publishes six loan-level files on its open data portal, four for the 7(a) program segmented by decade and two for 504, alongside a data dictionary. They are updated quarterly, typically about a month after the quarter closes, and the release used throughout this article is the set posted as of June 30, 2026 (U.S. Small Business Administration, 7(a) and 504 FOIA datasets, data.sba.gov, 2026). Together the three files covering approvals from fiscal 2010 forward hold 1,052,072 rows: 545,751 7(a) approvals for fiscal 2010 through fiscal 2019, 388,338 7(a) approvals from fiscal 2020 forward, and 117,983 504 approvals from fiscal 2010 forward.

What each row carries is a matter of record. Gross approval amount, SBA guaranteed approval, approval date and approval fiscal year, first disbursement date, initial interest rate and whether it is fixed or variable, term in months, the six-digit industry code and its description, franchise code and name, project county and state, SBA district office, congressional district, business type, business age, jobs supported, a collateral indicator, the loan status, and where applicable the paid-in-full date, the charge-off date and the gross charge-off amount. The 504 files add the certified development company name, the third-party lender name and the third-party dollars in the project.

What no row carries is the property. There is no property type, no square footage, no building age, no address, no appraised value, no loan-to-value ratio and no debt service coverage ratio. The collateral indicator states that collateral was taken, not what it was or what it was worth. A lender who wants to know how hotels performed, or how special-purpose assets performed, or whether a 25-year note on a car wash behaves like a mortgage or like a business loan, is asking a question the file answers only by inference. The inference is not a guess. It is a set of substitutions with stated logic, stated error and a stated count of loans behind every cell, and it produces numbers a credit committee can check, which is more than can be said for most of the figures that circulate about this asset class.

Three proxies for a missing field

The first proxy is the industry code. Some six-digit codes describe businesses that almost always occupy purpose-built real estate they own or lease exclusively: lessors of nonresidential buildings (531120), lessors of miniwarehouses and self-storage units (531130), hotels and motels (721110), car washes (811192), child care services (624410), assisted living facilities for the elderly (623312), gasoline stations with convenience stores, funeral homes and funeral services (812210), recreational vehicle parks and campgrounds (721211), general warehousing and storage (493110), offices of dentists (621210), full-service and limited-service restaurants (722511 and 722513), fitness and recreational sports centers (713940), used car dealers (441120), supermarkets and other grocery retailers (445110), and beer, wine and liquor stores. A subset of those, the ones with a single credible alternative use or none at all, is the population that the loan data on special-purpose property risk examines directly. The code is the borrower's industry, not the building's use, and the two diverge often enough that the code alone is a weak instrument.

The second proxy is the term, and it is the strongest of the three because it is regulated. Under 13 CFR 120.212 the term of a 7(a) loan shall be the shortest appropriate term given the borrower's ability to repay; ten years or less, unless it finances or refinances real estate or equipment with a useful life exceeding ten years; and a maximum of 25 years including extensions, with a portion used to acquire or improve real property allowed 25 years plus the period needed to complete construction (eCFR, 13 CFR 120.212, current 2026). Read that in reverse and it becomes a classifier. A 7(a) loan approved for 240 months or more cannot, as a matter of regulation, be a working-capital loan. It is real estate, or long-lived equipment on a real-estate-length amortization, and in the small-balance market that overwhelmingly means owner-occupied commercial property. The tape does not say "this loan is secured by a building." The maturity says it for the tape.

The third proxy is the program. The 504 program finances fixed assets by design: the SBA describes it as long-term, fixed-rate financing for major fixed assets, usable for the purchase, construction or renovation of existing buildings or land and for long-term machinery and equipment with a remaining useful life of at least ten years, and explicitly not usable for working capital or inventory, or for speculation or investment in rental real estate. Maturities offered are ten, twenty and twenty-five years, and the maximum loan amount is $5.5 million (U.S. Small Business Administration, 504 loans, 2026). The 7(a) program, by contrast, is deliberately mixed: the agency lists acquiring, refinancing or improving real estate and buildings alongside short- and long-term working capital, refinancing of business debt, and purchase and installation of machinery and equipment (U.S. Small Business Administration, 7(a) loans, 2026). Anything read from an undifferentiated 7(a) population is therefore a blend of two very different credits, and separating them is the analytical task.

Seasoning before comparison

Before any cohort is compared to any other, the file has to be read for what it does not show. The status codes in the 7(a) files are CANCLD for cancelled, COMMIT for approved but undisbursed, CHGOFF for charged off, PIF for paid in full, and EXEMPT for a disbursed loan that is still open. That last code is the whole problem. The Freedom of Information Act exempts from disclosure "trade secrets and commercial or financial information obtained from a person and privileged or confidential" (5 U.S.C. 552(b)(4)), and the SBA applies the exemption to loans that have not yet resolved. The public tape therefore shows outcomes for finished loans and silence for live ones. A rate computed over a cohort that is mostly still open is not a performance rate; it is a measure of how much of the cohort has had time to fail.

The open share makes the point without argument. Among 7(a) approvals, the share still carrying the EXEMPT status in the June 2026 file runs from 1.6% for fiscal 2010 to 2.2% for fiscal 2011, 4.5% for fiscal 2014, 9.4% for fiscal 2016, 22.1% for fiscal 2018, 30.8% for fiscal 2019, 53.6% for fiscal 2021, 80.1% for fiscal 2023 and 96.3% for fiscal 2025. The 504 program, whose loans are longer and amortize more slowly, runs higher still: 21.6% open for fiscal 2010 and 98.9% for fiscal 2025 (MMCG tabulation from the SBA 7(a) and 504 FOIA loan-level files as of 30 June 2026; MMCG database, 2026). A fiscal 2024 cohort with a 1.05% observed charge-off rate has not outperformed a fiscal 2012 cohort at 6.27%. It has simply not aged.

Every cohort figure in this article is therefore drawn from approvals in fiscal years 2010 through 2019, a ten-year window in which the least seasoned year is still 30.8% open and the most seasoned is 1.6% open. That window contains 478,963 disbursed 7(a) loans, where disbursed excludes the cancelled and the undisbursed, and 58,887 disbursed 504 loans, where it also excludes the 8,516 rows marked not funded. It ends at fiscal 2019 on purpose: the fiscal 2020 and fiscal 2021 cohorts were approved into a period of extraordinary federal support and deferment, and their behaviour is not yet legible. Anyone extending the window forward should say so and should publish the open share alongside the rate.

Two rates that answer different questions

There are two defensible ways to compute a charge-off rate from this file and they give different answers, both correct. The count rate divides charged-off loans by disbursed loans. The dollar rate divides gross charge-off amount by gross approval amount. Across the seasoned 7(a) window, 34,153 of 478,963 disbursed loans were charged off, a count rate of 7.13%, while $5.27 billion was charged off against $180.45 billion approved, a dollar rate of 2.92%. Alongside those, 394,721 loans, or 82.4% of the window, were paid in full, and 50,089, or 10.5%, remain open (MMCG tabulation from the SBA 7(a) FOIA files as of 30 June 2026; MMCG database, 2026).

The spread between 7.13% and 2.92% is not noise. It is the signature of a portfolio in which the loans that fail are systematically smaller than the loans that do not. A count rate answers the question a portfolio manager asks: what fraction of files will require workout attention. A dollar rate answers the question a capital planner asks: what fraction of principal is exposed. Neither is a loss rate. Gross charge-off amount is the balance written off, before recoveries and before the guaranty settles between the lender and the agency, and the SBA reports recoveries separately in its own performance tables. A credit memo that quotes a charge-off rate without saying which of these three quantities it means has not said anything.

The 504 program, computed on the same definitions and the same window, produced 887 charge-offs across 58,887 disbursed loans, a count rate of 1.51%, on an average approval of $698,498. That is less than a quarter of the 7(a) count rate on loans that are on average roughly twice as large, and the reason is structural rather than behavioural. It is the cleanest single demonstration in the public record that program design, not borrower industry, drives most of the observed difference in small-balance credit performance.

The property-backed industry table

Ranked by count charge-off rate within the seasoned window, the property-backed industries separate by a factor of twenty. Lessors of miniwarehouses and self-storage units sit at 0.61% across 979 loans averaging $1,280,641. Assisted living facilities follow at 1.91% across 1,153 loans, funeral homes at 1.93% across 1,608, recreational vehicle parks and campgrounds at 2.01% across 399, lessors of nonresidential buildings at 2.09% across 1,815, hotels and motels at 2.36% across 7,706 loans averaging $1,893,782, and offices of dentists at 2.63% across 9,072. Gasoline stations with convenience stores come in at 3.26% across 5,546, child care services at 3.73% across 5,821, and beer, wine and liquor stores at 4.19% across 5,340.

At the other end, amusement arcades charged off at 12.47% across 361 loans, used car dealers at 12.26% across 1,787, fitness and recreational sports centers at 10.08% across 7,329, supermarkets and other grocery retailers at 9.87% across 3,304, limited-service restaurants at 9.38% across 12,335, commercial and institutional building construction at 8.80% across 2,885, and full-service restaurants at 8.46% across 18,302, the largest single property-backed cell in the file. Car washes land in the middle at 5.01% across 2,195 loans averaging $1,138,793, general automotive repair at 6.42% across 6,667, and offices of physicians at 4.38% across 6,968 (MMCG tabulation from the SBA 7(a) FOIA files as of 30 June 2026; MMCG database, 2026).

Every cell quoted here holds at least 200 loans, comfortably above the ten-loan minimum below which no rate should be published or implied, and the count is quoted with the rate because a rate without its denominator is an assertion. That discipline matters most at the six-digit level, where a plausible-looking industry can turn out to hold forty loans nationally. Where a cell is thin, the honest move is to aggregate to the four-digit or three-digit level and say so, or to omit the row. The same reasoning governs geography: county-level cells thin out fast, which is why the geography of SBA lending is mapped at the state and district level before it is mapped at the county level.

Term tells more than industry

Now split each of those industries by term, and the league table stops being a table about industries. Among seasoned 7(a) approvals, hotels and motels financed on terms of 240 months or more charged off at 1.30% across 6,690 loans, while hotels and motels on shorter terms charged off at 9.35% across 1,016. Car washes: 1.63% across 1,290 long-term loans against 9.83% across 905 short-term loans. Child care services: 0.98% against 5.47%. Offices of dentists: 0.37% against 3.87%. Full-service restaurants: 2.23% against 9.60%. Limited-service restaurants: 1.37% against 10.27%. Fitness and recreational sports centers: 1.50% against 10.85%. Gasoline stations with convenience stores: 1.35% against 8.27%. Used car dealers: 1.48% against 16.12%. Supermarkets and other grocery retailers: 3.49% against 11.46%. General automotive repair: 0.73% against 8.29%. Offices of lawyers: 0.36% against 5.67% (MMCG tabulation from the SBA 7(a) FOIA files as of 30 June 2026; MMCG database, 2026).

The pattern does not have a single exception in the property-backed set. It holds at the two-digit sector level too, where cell sizes run into the tens of thousands: accommodation and food services 1.61% for loans of 240 months or more against 9.57% for shorter loans, health care and social assistance 0.62% against 5.86%, real estate and rental and leasing 0.56% against 7.17%, retail trade 1.52% against 9.28%, other services 0.82% against 8.64%, construction 0.71% against 8.43%, and professional, scientific and technical services 0.68% against 7.15%.

Pooled across all industries, the four term buckets tell the same story at scale. Loans under 84 months charged off at 21.23% across 116,249 loans averaging $155,675. Loans of 84 to 119 months charged off at 4.54% across 152,914. Loans of 120 to 239 months charged off at 1.29% across 135,809 loans averaging $443,286. Loans of 240 months or more charged off at 1.06% across 73,991 loans averaging $1,113,340, with a dollar rate of 0.55%. The count rate on the shortest bucket is twenty times the rate on the longest, and the industry table's own spread, from 0.61% to 12.47%, is a factor of twenty as well. Term and industry appear to explain the same amount of variation, but they are not doing the same work: term is close to a clean split between mortgage credit and business credit, while industry is a mixture whose composition changes from code to code.

That is the finding, and it has a direct consequence. An industry league table computed on undifferentiated 7(a) loans is mostly measuring each industry's mix of real-estate-backed and working-capital lending, not the performance of the property. Fitness centers look like a bad asset class at 10.08%; the fitness centers that bought their buildings charged off at 1.50%, and the 6,730 shorter-term fitness loans in the window, mostly equipment and build-out credit, charged off at 10.85% and drag the blended figure with them. Used car dealers look catastrophic at 12.26%; the 472 that took 240-month money charged off at 1.48%. The property was rarely the problem. The credit structure was.

Reading the caveats honestly

A finding this clean invites three objections, and all three are partly right. The first is survivorship. Long-term loans are still open at much higher rates: 25.9% of the 240-month-and-longer bucket remains EXEMPT against 2.0% of the loans under 120 months, so a slice of the long-term cohort has not finished its life and some of it will fail later. The correction is real but small relative to the gap. Even if every currently open long-term loan in the window eventually charged off, which is absurd, the bucket would top out near 27%; and at the level of an ordinary adverse assumption the long-term rate roughly doubles while the short-term rate barely moves, leaving the ordering intact.

The second is size and selection. Loans of 240 months or more average $1,113,340 against $155,675 for loans under 84 months, and larger loans are underwritten harder, carry real property as collateral, and are made to borrowers with more equity in the deal. That is exactly the point rather than a refutation: the term proxy is doing its job of separating secured property credit from unsecured operating credit. It does mean the comparison is not a controlled experiment. It is a description of two populations that differ in more ways than their maturity, and the honest statement of the finding names them.

The third is that a count rate is not a loss rate. The dollar rates move in the same direction and by more, which strengthens rather than weakens the reading: 14.81% by dollars for loans under 84 months, 7.89% for 84 to 119 months, 0.97% for 120 to 239 months and 0.55% for 240 months or more. But none of those figures nets recoveries, and the SBA's own tables report recovery amounts and post-charge-off recovery amounts separately. A memo that treats a 1.06% count rate as a 1.06% expected loss has confused three different quantities and will be wrong in a direction that flatters the deal.

A fourth caution belongs with these, though it is a data caution rather than a statistical one. The 7(a) files carry industry codes as they stood when the loan was approved, so a single tape mixes classification vintages. Gasoline stations with convenience stores appear as 447110 on older approvals and as 457110 on newer ones; beer, wine and liquor stores appear as 445310 and as 445320. The renumbering follows the 2022 revision of the classification system, which took effect for federal statistical data covering reference years beginning on or after January 1, 2022 (Office of Management and Budget, North American Industry Classification System Revision for 2022, Federal Register, December 21, 2021), and the SBA's own size-standard table at 13 CFR 121.201 now runs on the 2022 codes. Any tabulation that groups on a raw six-digit code without a crosswalk will silently split one industry into two.

The 504 contrast

If term is a proxy for property collateral, the 504 program is the control group, because its entire population is fixed-asset credit by regulation. The results follow. Across seasoned approvals, the 504 count charge-off rate fell from 3.36% for fiscal 2010 to 2.06% for fiscal 2011, 1.42% for fiscal 2013, 1.04% for fiscal 2016, 0.44% for fiscal 2018 and 0.28% for fiscal 2019, while the 7(a) rate over the same approval years moved from 9.20% to 6.89%, 5.99%, 7.18%, 7.89% and 6.69%. The two series do not converge; the 504 line falls steadily while the 7(a) line stays in a band. Part of the 504 decline is seasoning, since 77.1% of fiscal 2019 504 approvals are still open against 30.8% for 7(a), and any reader who ignores that will overstate the improvement. The gap in the well-seasoned early years, 3.36% against 9.20% in fiscal 2010 and 2.06% against 6.89% in fiscal 2011, is not a seasoning artifact.

By industry within the 504 window, the ranking is compressed almost flat. Offices of dentists charged off at 0.33% across 1,796 loans, offices of lawyers at 0.46% across 1,094, general automotive repair at 0.59% across 1,184, lessors of miniwarehouses and self-storage units at 0.18% across 542, car washes at 0.89% across 787, used car dealers at 0.94% across 424, limited-service restaurants at 1.03% across 974, child care services at 1.16% across 1,374, fitness and recreational sports centers at 1.27% across 629, offices of physicians at 1.61% across 1,869, hotels and motels at 2.55% across 2,743, gasoline stations with convenience stores at 2.56% across 821, full-service restaurants at 2.68% across 2,088, and supermarkets and other grocery retailers at 4.20% across 429 (MMCG tabulation from the SBA 504 FOIA file as of 30 June 2026; MMCG database, 2026).

Compare that spread, roughly 0.2% to 4.2%, with the 7(a) spread of 0.6% to 12.5% over the same industries and the same years. Used car dealers, the worst 7(a) property-backed industry at 12.26%, are the eleventh best of twenty-six 504 industries at 0.94%. The industry did not change. The collateral and the amortization did. For anyone building the 504 side of this analysis into a workflow, the operational detail sits in the library's note on how certified development companies use data in the 504 workflow, and the shape of the small-balance market it serves in what analytics matter under $10 million.

SBA's own yardstick, and why it disagrees

The agency publishes its own performance figures, and they look nothing like the cohort rates above. The SBA's loan program performance report carries eleven tables by fiscal year for the most recent ten years, with data through June 30, 2025. Table 9, charge-off rates as a percent of unpaid principal balance, shows 7(a) Regular at 1.82% in fiscal 2016, 0.80% in fiscal 2017, 0.51% in fiscal 2018, 0.68% in fiscal 2019, 0.38% in fiscal 2020, 0.36% in fiscal 2021, 0.42% in fiscal 2022, 0.48% in fiscal 2023, 0.55% in fiscal 2024 and 0.37% in fiscal 2025, and 504 Regular at 0.84%, 0.40%, 0.40%, 0.31%, 0.29%, 0.12%, 0.31%, 0.15%, 0.12% and 0.05% over the same years. Fiscal 2025 is a partial year in this release, ending at the third fiscal quarter, and should not be read as a completed year.

These are annual flow rates on a stock, not cohort outcomes. The numerator is the amount charged off during a fiscal year; the denominator is the unpaid principal balance outstanding, which for 7(a) Regular stood at $123.40 billion in fiscal 2025 and for 504 Regular at $33.88 billion. A loan approved in 2014 and charged off in 2022 lands in the 2022 numerator and never appears as a cohort outcome at all. Because the portfolio is large and long-lived and most of it performs, an annual rate on unpaid balance is naturally an order of magnitude smaller than a lifetime cohort rate: the 0.42% posted for fiscal 2022 and the 7.13% computed for fiscal 2010 to fiscal 2019 approvals are not in conflict and are not comparable.

Table 8, purchase rates as a percent of unpaid principal balance, is the more forward-looking series, because a guaranty purchase precedes a charge-off. For 7(a) Regular it reads 0.77%, 0.78%, 0.85%, 1.18%, 0.99%, 0.57%, 0.70%, 1.00%, 1.43% and 1.37% across fiscal 2016 to fiscal 2025, and for 504 Regular 0.85%, 0.67%, 0.62%, 0.68%, 0.54%, 0.41%, 0.26%, 0.13%, 0.40% and 0.43%. The 7(a) purchase rate has risen for three of the last four posted years while the charge-off rate has stayed low, which is the pattern of stress entering the portfolio faster than it is being resolved. Table 5 puts the amounts alongside the rates: 7(a) Regular charge-offs of $1.43 billion in fiscal 2016, falling to $364.8 million in fiscal 2020, then rising through $454.4 million, $531.5 million and $643.5 million before the partial-year $457.7 million of fiscal 2025 (U.S. Small Business Administration, loan program performance tables, data through June 30, 2025). Both yardsticks belong in a credit memo. The cohort rate sizes lifetime risk for a new loan; the agency series sizes what is happening to the portfolio now.

A recipe a lender can reproduce

The whole analysis rebuilds from public files in an afternoon, and stating the recipe is the point of publishing it. Download the 7(a) file for fiscal 2010 to fiscal 2019, the 7(a) file for fiscal 2020 forward and the 504 file for fiscal 2010 forward from the SBA open data portal, together with the data dictionary. Define the denominator first: disbursed equals every row whose status is not CANCLD, not COMMIT and, in the 504 files, not NOT FUNDED. Cancelled and undisbursed approvals never had a chance to fail and inflate the denominator if left in.

Define the outcome next. The count charge-off rate is rows with status CHGOFF divided by disbursed rows. The dollar charge-off rate is the sum of gross charge-off amount divided by the sum of gross approval amount over the same rows. Carry the paid-in-full share and the EXEMPT share alongside both, always, because those two numbers are what tell a reader whether the rate is finished. Fix the approval window at fiscal 2010 to fiscal 2019 unless there is a stated reason to move it, and report the window in the label of every figure, not in a footnote.

Then apply the three proxies in order. Group by six-digit industry code after crosswalking the 2017 and 2022 vintages so that 447110 and 457110 land in one bucket and 445310 and 445320 land in another. Bucket the term at under 84 months, 84 to 119, 120 to 239 and 240 or more, and treat 240 or more as the real-estate proxy. Keep the two programs separate, and never pool 7(a) and 504 into a single rate. Every reported cell then carries four things: the rate, the loan count, the average approval amount and the open share. Suppress any cell with fewer than ten loans rather than publishing a rate that a single outcome would move by ten percentage points, and suppress it visibly, so a reader sees that a cell exists and was withheld rather than assuming the industry was never lent to.

What goes into the file is a short table and a shorter sentence. For the subject property's industry: the seasoned count rate with its loan count, the same rate restricted to loans of 240 months or more, the average approval size of that long-term cell, and the equivalent 504 figure where the program fits. Then one sentence naming the file, the as-of date and the window. That is a sourced, checkable benchmark statement, and it is the sort of evidence the SBA file expects a lender to assemble; the wider inventory of it is set out in the library's note on the market evidence SOP 50 10 8 actually asks for. Provenance carried on every displayed value, rather than reconstructed later, is the standard the whole library works to, and the reasoning behind it is set out in the provenance standard.

What the tape cannot say

The limits are as important as the findings, and a benchmark quoted past its limits is worse than no benchmark. The file holds no property type, so every property statement in this article is an inference from industry, term and program, and the inference fails wherever a borrower in a property-backed industry leased rather than bought. It holds no collateral description and no value: the collateral indicator is binary, and the regulation confirms only that collateral exists, since SBA requires hazard insurance on all collateral for 7(a) loans greater than $500,000 and for 504 projects greater than $500,000, and generally requires guarantees from holders of at least a 20 percent ownership interest (eCFR, 13 CFR 120.160, current 2026). Recovery in these loans therefore comes from a building and from a person, and the file separates neither.

It holds no loan-to-value ratio, so the supervisory limits that govern the senior lender's side of a project, 80 percent for construction of nonresidential property and 85 percent for improved property under the interagency real estate lending guidelines (12 CFR part 365, Appendix A), cannot be tested against it. It holds no net operating income, no debt service and therefore no debt service coverage ratio, the central test of the OCC's commercial real estate lending handbook, so the coverage question has to be answered from other public evidence entirely; that is the work of stress testing coverage with market-derived inputs and of breakeven occupancy benchmarks by asset class. It holds no rent, no occupancy and no expense line, which is why property-level operating benchmarks come from the registered commercial mortgage-backed securities tape instead. And it hides open loans behind the FOIA exemption, so the most recent five years of lending, exactly the years a lender most wants to see, are the years the file says least about.

Set against those limits, the scale is the compensation. The SBA approved 78,078 7(a) loans for $37.29 billion and 6,762 504 loans for $7.80 billion in fiscal 2025, on data as of September 30, 2025, and the loan-level files reach back to fiscal 1991 (U.S. Small Business Administration, 7(a) and 504 Monthly and Yearly Activity Report, 2025). No other public record of small-balance commercial property credit is remotely this deep. MMCG Analytics is a map-first commercial real estate analytics platform built on federal, state and public data with source and vintage carried on every displayed value, and its SBA layer is derived from these same public FOIA files, covering over one million loan records under a minimum-cohort standard that shows no performance rate for a cohort under ten loans. A separate published series, the MMCG SBA 7(a) Performance Series, reports an all-industry adverse-resolution rate of 7.6% for loans approved in fiscal 2014 to fiscal 2021, counting charge-off, liquidation and guaranty purchase together (MMCG Research, 2026). That is a different definition over a different window, and it should never be plotted on the same axis as the 7.13% charge-off count rate computed here. The analytics supply the benchmark and its provenance; the credit judgment stays with the lender.

Frequently asked questions

Does the SBA loan data show property type?

No. The 7(a) and 504 FOIA files carry the borrower's industry code, loan term, program, approval amount, county and outcome, but no property type, address, square footage, appraised value, loan-to-value ratio or debt service coverage ratio. Property type is inferred from three proxies: the industry code, a term of 240 months or more as a real-estate marker under 13 CFR 120.212, and the 504 program's fixed-asset mandate.

What is the SBA 7(a) charge-off rate by industry?

Across seasoned approvals from fiscal 2010 to fiscal 2019, the property-backed industries run from 0.61% for self-storage lessors (979 loans) and 1.91% for assisted living (1,153) up to 10.08% for fitness centers (7,329) and 12.26% for used car dealers (1,787), against an all-industry 7(a) count rate of 7.13% on 478,963 disbursed loans. Figures are an MMCG tabulation from the SBA FOIA files as of 30 June 2026.

Why do longer SBA loans default less?

Because term is close to a clean split between property credit and operating credit. Under 13 CFR 120.212 a 7(a) term above ten years requires real estate or long-lived equipment, so 240-month loans are effectively mortgages. In the seasoned window they charged off at 1.06% across 73,991 loans averaging $1,113,340, against 21.23% for loans under 84 months averaging $155,675. The gap holds inside every property-backed industry tested.

How does 504 performance compare with 7(a)?

On the same definitions and the fiscal 2010 to fiscal 2019 window, 504 charged off at 1.51% by count across 58,887 disbursed loans against 7.13% for 7(a) across 478,963. By approval year the 504 rate fell from 3.36% in fiscal 2010 to 0.28% in fiscal 2019 while 7(a) moved from 9.20% to 6.69%. Part of the 504 decline is seasoning: 77.1% of its fiscal 2019 approvals are still open.

What does the EXEMPT status mean in the SBA files?

It marks a disbursed loan that has not resolved. The SBA applies FOIA Exemption 4 at 5 U.S.C. 552(b)(4), covering confidential commercial or financial information, so the public tape reports outcomes only for finished loans. Open shares run from 1.6% of fiscal 2010 7(a) approvals to 96.3% of fiscal 2025, which is why only seasoned cohorts can be compared.

Is a count charge-off rate the same as a loss rate?

No. The count rate was 7.13% and the dollar rate 2.92% over the same seasoned 7(a) window, because failed loans are systematically smaller. Neither nets recoveries, which the SBA reports in separate tables. The agency's own annual charge-off rate on unpaid principal balance, 0.37% for 7(a) Regular in fiscal 2025, is a third quantity again and is not comparable to a cohort rate.

How many loans does a cohort need before a rate can be published?

At least ten. The SBA analytical standard applied here shows no performance rate for a cohort under ten loans, and thin cells are aggregated to a broader industry level or withheld. Every figure in this article names its loan count for the same reason: a rate without a denominator cannot be checked.

Sources

1. U.S. Small Business Administration, Office of Capital Access, 7(a) and 504 FOIA loan-level datasets and data dictionary, files as of June 30, 2026 (dataset last modified April 28, 2026). https://data.sba.gov/dataset/7a-504-foia
2. U.S. Small Business Administration, Small Business Administration loan program performance, eleven tables by fiscal year with data through June 30, 2025 (file updated September 15, 2025): Table 1 unpaid principal balance, Table 3 number of approved loans, Table 5 charge off amount, Table 8 purchase rates, Table 9 charge off rates. https://www.sba.gov/document/report-small-business-administration-loan-program-performance
3. U.S. Small Business Administration, 7(a) and 504 Monthly and Yearly Activity Report, fiscal 1991 to fiscal 2025, data as of September 30, 2025. https://data.sba.gov/dataset/7a-504-activity-reports-fy2025-year-end
4. 13 CFR 120.212, What limits are there on loan maturities, Electronic Code of Federal Regulations, current 2026. https://www.ecfr.gov/current/title-13/section-120.212
5. 13 CFR 120.160, Loan conditions (personal guarantees, appraisals, hazard insurance), Electronic Code of Federal Regulations, current 2026. https://www.ecfr.gov/current/title-13/section-120.160
6. 13 CFR 121.201, Small Business Size Standards by NAICS Industry, Electronic Code of Federal Regulations, current 2026 (table published on the 2022 classification codes). https://www.ecfr.gov/current/title-13/section-121.201
7. U.S. Small Business Administration, 504 loans: program description, eligible uses, maturities and the $5.5 million maximum, 2026. https://www.sba.gov/funding-programs/loans/504-loans
8. U.S. Small Business Administration, 7(a) loans: program description and eligible uses, 2026. https://www.sba.gov/funding-programs/loans/7a-loans
9. U.S. Small Business Administration, CDC/504 loan program, lender resources page, 2026. https://www.sba.gov/partners/lenders/cdc504-loan-program
10. U.S. Small Business Administration, 7(a) loan program, lender resources page, 2026. https://www.sba.gov/partners/lenders/7a-loan-program
11. Freedom of Information Act, 5 U.S.C. 552(b)(4), text as published by the U.S. Department of Justice, Office of Information Policy, 2026. https://www.justice.gov/oip/freedom-information-act-5-usc-552
12. Office of Management and Budget, North American Industry Classification System Revision for 2022; Update of Statistical Policy Directive No. 8, Federal Register notice of December 21, 2021. https://www.federalregister.gov/documents/2021/12/21/2021-27536/north-american-industry-classification-system-revision-for-2022-update-of-statistical-policy
13. U.S. Small Business Administration, SOP 50 10 8, Lender and Development Company Loan Programs, effective June 1, 2025. https://www.sba.gov/document/sop-50-10-lender-development-company-loan-programs
14. U.S. Small Business Administration, SOP 50 57 4, 7(a) Loan Servicing and Liquidation, effective November 1, 2025. https://www.sba.gov/document/sop-50-57-7a-loan-servicing-liquidation
15. Federal Deposit Insurance Corporation, Interagency Guidelines for Real Estate Lending Policies, 12 CFR part 365, subpart A, Appendix A (supervisory loan-to-value limits), Electronic Code of Federal Regulations, current 2026. https://www.ecfr.gov/current/title-12/part-365
16. Office of the Comptroller of the Currency, Comptroller's Handbook, Commercial Real Estate Lending, version 2.0, March 2022. https://www.occ.gov/publications-and-resources/publications/comptrollers-handbook/files/commercial-real-estate-lending/pub-ch-commercial-real-estate.pdf
17. Board of Governors of the Federal Reserve System, Senior Loan Officer Opinion Survey on Bank Lending Practices, July 2026. https://www.federalreserve.gov/data/sloos/sloos-202607.htm
18. Board of Governors of the Federal Reserve System, Financial Stability Report of May 2026 (asset valuations section). https://www.federalreserve.gov/publications/2026-may-financial-stability-report-asset-valuations.htm
19. MMCG Research, SBA 7(a) Performance Series: adverse-resolution rate by industry for loans approved fiscal 2014 to fiscal 2021, loan-count basis; MMCG database, 2026. https://mmcganalytics.com/sba-default-rates/