The Next SBA Default Wave Is Already in the Data. It Just Doesn't Look Like One.

• 

  SBA early default watch

Charge-off tables make the newest SBA loans look like the safest ever written. Read them at the same age, and a different picture emerges.

Open any table of SBA 7(a) charge-offs by origination year and the newest loans look spectacular. According to MMCG Invest's Feasibility Index, which reflects loan status as of March 31, 2026, only 0.72 percent of fiscal 2024 loans had been charged off. For the 2015 through 2018 classes, the figures run from 6.4 to 7.6 percent. Stop reading there and you would conclude that SBA lenders spent two years writing the strongest credits in the program's history.

That conclusion would be wrong, and the reason matters to anyone who lends to, invests in or advises small businesses financed with SBA loans.

A charge-off is not a default. It is the last entry in a long ledger. A borrower falls behind, the lender works the loan, collateral is liquidated or the file is closed, SBA reviews the claim and pays on its guaranty, and only then is the balance written off. In the public record that sequence takes years. The newest loans have not had time to complete it, so comparing their charge-offs with those of loans a decade old tells you almost nothing about credit quality.

This article asks a narrower and more useful question. What do the public records say about how the 2021 through 2024 vintages are performing at comparable ages, and which industries should show strain first? Along the way we ran into something more consequential than any single rate. In April 2026, SBA changed how loan status is recorded in its public loan file, and the change has quietly broken many of the charge-off figures now circulating online.

Why charge-offs trail trouble by years

The lag is long, and independent analyses of SBA's loan-level file put numbers on it. According to SBA Loan Index data covering fiscal 2010 through March 2026, the median loan that charges off does so about 50 months after disbursement. A separate analysis of the same public file, published on GitHub, estimates that only about 23.5 percent of the charge-offs a vintage will eventually produce have happened by month 36. It puts the share at roughly 57 percent by month 60 and 97 percent by month 120.

Now apply that to a recent cohort. A research paper posted on SSRN in July 2026 examined 7(a) loans approved from fiscal 2022 through the second quarter of fiscal 2024 and counted those charged off within 18 months of approval. The answer was 361 loans out of 127,464, or 0.28 percent. The author cautioned that rates built from resolved loans are stress indicators, not measures of how much of a portfolio is troubled.

The implication is straightforward. A charge-off within 24 months of approval is a real event, and a rising rate of such events would be a serious warning. But the measure captures only the fastest and most severe failures, the loans that collapse almost immediately, often for reasons that point to underwriting or eligibility problems rather than ordinary business risk. It is a floor under early losses, not an estimate of them. Anyone presenting a low 24-month charge-off rate for the 2023 or 2024 classes as proof of health is measuring the wrong thing.

A quiet change to the data

Until this spring, the public SBA file recorded most live loans under a single status, EXEMPT, a catch-all for loans that had been disbursed but not yet resolved. SBA's disclosure regulation, 13 CFR Part 102, lists loan status other than charged-off or paid-in-full among the categories of information generally exempt from disclosure. The agency's data dictionary still reflects that approach.

The April 2026 quarterly release did not. According to LoanTape, a data provider that tracks the files, the loan-status field expanded from five codes to thirteen. Seven new codes split up the old EXEMPT group: current, delinquent, past due, deferred, in liquidation, purchased but not charged off, and sold without a charge-off. For the first time, anyone can see which open loans are behind on payments. We found no regulatory change accompanying the release.

Two other new codes matter even more for loss analysis, because they were carved out of the charged-off category. One, CLSLN, marks loans closed after a charge-off, mostly workouts and settlements, and covers 75,522 loans. The other, SOLDCO, marks charged-off loans that were sold, and covers 309. LoanTape reports that the number of 7(a) loans carrying the plain CHGOFF code fell from 216,123 in the December 2025 release to 142,898 in March 2026. That drop of 73,225 loans is about a third of everything previously shown as charged off.

The reclassification also touched the other end of the ledger. LoanTape found that roughly 10 percent of the loans moved into the two new codes had previously been labeled paid in full. Some loans that looked like clean repayments were settlements.

MMCG Invest reports that the March 31 refresh applied the full set of codes retroactively across the history of the file, back to at least fiscal 2010, so the change reaches every vintage, not just recent ones.

What follows is practical. Any analysis that counts only the CHGOFF code on data from March 2026 onward now misses roughly a third of historical charge-offs. The safer definition is any loan with both a charge-off date and a charge-off amount, whatever its status code. Several charge-off rates published since April, by lenders, aggregators and trade sites, describe themselves as counting loans "marked charged off" and may carry this understatement. Readers should ask which release and which codes sit behind any rate they are handed.

Four meanings of "early default"

The phrase is used loosely, and the gaps between definitions explain much of the confusion in the public debate.

SBA's Inspector General has long treated a loan as an early default if it defaults within 18 months of disbursement. Lumos, a data vendor, uses 18 months from origination, measured against disbursed loans. The SSRN paper measures charge-off within 18 months of approval. Tax Guard, a tax-resolution firm summarizing SBA's own risk reporting, describes early defaults as those within the first 36 months. Separately, figures from SBA's internal risk analysis that reached the press put the annual guaranty purchase rate at 3.7 percent in fiscal 2024, the highest since 2012. That is a ratio of a year's purchases to the outstanding portfolio, not a measure of how new loans perform.

These are different yardsticks: default versus charge-off, disbursement versus approval versus origination, 18 months versus 36. None is wrong, but the numbers cannot be stacked side by side.

Even within a single definition, results vary sharply by loan size. Lumos reports these early-default rates, measured by loan count within 18 months:

Loan size	2024	2025
Under $150K	1.23%	1.46%
$150K–$350K	1.66%	1.88%
$350K–$500K	1.26%	1.23%
$500K–$1M	0.56%	1.41%
$1M–$2M	0.73%	0.83%
$2M–$3M	0.57%	1.13%
$3M+	0.70%	1.04%

Early default among loans of $150,000 to $350,000 reached 1.88 percent in 2025, the highest of any band, while loans between $500,000 and $1 million rose two and a half times, from 0.56 to 1.41 percent. Lumos links that jump to the period when SBA waived upfront fees on loans up to $1 million, and describes the connection as a correlation with a plausible mechanism, not a demonstrated cause. Early-default figures cited in the 2025 policy debate, roughly 1 to 1.5 percent in different accounts, sit in the same range.

Reading the vintages at the same age

The cleanest public evidence comes from comparing vintages by how much of their book has gone wrong, not by charge-offs alone. The MMCG Invest Feasibility Index, using loan status as of March 31, 2026 among funded 7(a) loans, reports both a charge-off rate and a broader adverse rate that also counts loans in liquidation and loans on which SBA has paid the guaranty without a charge-off yet.

Approval year	Charged off	Adverse (charge-off, liquidation or guaranty purchase)	Share of loans still open	Funded loans
FY2015	6.40%	7.36%	5.3%	55,421
FY2016	6.85%	8.21%	9.5%	56,789
FY2017	7.41%	9.77%	14.5%	56,079
FY2018	7.57%	10.92%	20.0%	54,195
FY2019	6.25%	10.79%	28.3%	45,670
FY2020	3.62%	7.98%	34.8%	36,463
FY2021	2.42%	7.19%	53.0%	45,147
FY2022	3.42%	10.36%	65.0%	41,737
FY2023	2.74%	10.53%	75.4%	50,137
FY2024	0.72%	6.23%	86.4%	59,893

Read the second column and recent vintages look excellent. Read the third and the picture changes. Fiscal 2022 and 2023 loans show adverse rates of 10.36 and 10.53 percent, essentially level with the 10.92 and 10.79 percent of the 2018 and 2019 classes, even though those older classes have had four to five more years to deteriorate. The 2022 and 2023 loans have also resolved far less of their books: 65 and 75 percent remain open, against 20 and 28 percent for 2018 and 2019. Because the table reflects status on a single date, shortly after the code change described above, its charge-off column may understate losses. The adverse column is the more informative one.

Other analysts see the same pattern through different lenses. Lumos reports that the share of loans defaulting in their second year was 3.0 percent for the fiscal 2016 class and 7.2 percent for fiscal 2024. In their third year, the figures were 3.1 percent for fiscal 2016 and 8.0 percent for fiscal 2023. The GitHub analysis projects that fiscal 2023 will end up with about 12 percent of loans charged off over their lives, above the 6 to 9 percent range it finds for the 2010 through 2014 classes. A projection is not an observation, and the author labels fiscal 2024 and 2025 as too young to project.

The pandemic years need their own footnote. The 2020 and 2021 vintages look unusually clean. The GitHub analysis shows about 1.4 to 1.5 percent of those loans charged off over 45 months, compared with 2.5 to 2.8 percent for the 2018 and 2019 classes. Relief measures during the pandemic, including SBA payments on borrowers' loans, likely flattered them. Lumos calls fiscal 2022 the first genuinely post-stimulus vintage, and it is also the first to show clear deterioration.

Where early distress is likely to show up first

No public source yet publishes charge-off rates for the 2021 through 2024 vintages by industry at a fixed age. The best available guide is the seasoned record: how loans from fiscal 2010 through 2019 performed over their lives, as measured by the MMCG Invest Feasibility Index with status as of March 31, 2026.

Industry	Seasoned charge-off rate	Adverse rate	Loans
Residential remodelers	11.05%	13.03%	6,426
Fitness centers	9.29%	11.71%	7,329
Dry cleaning	8.82%	10.37%	2,391
Limited-service restaurants	8.19%	10.43%	17,057
Long-distance truckload	7.71%	9.01%	7,961
Full-service restaurants	7.58%	9.88%	25,182
Local freight trucking	7.50%	9.28%	6,403
General auto repair	5.68%	6.72%	6,667
Child day care	3.25%	4.57%	5,821
Gas stations with convenience stores	2.78%	3.50%	5,545
Dentists	2.28%	3.14%	9,072
Hotels and motels	2.10%	4.44%	7,706

Several patterns stand out. Food service, trucking, fitness and personal services carry seasoned charge-off rates near or above 7.5 percent, against 6.44 percent for 7(a) overall in the same data. Health care, hotels and gas stations sit far below it. At the sector level, MMCG reports arts and entertainment at 8.36 percent and transportation at 8.20 percent, against 4.09 percent for health care.

Current data point the same way for trucking. Lumos reports an annualized default rate of 7.6 percent for transportation and warehousing across 15,057 active loans in the first half of fiscal 2026, against a portfolio average of 5.4 percent and 3.6 percent for health care.

These are lifetime rates, not early ones, and they describe the past. They are a map of where trouble has historically concentrated, not a forecast. Still, the industries with the worst seasoned records are the ones where thin margins, heavy fixed costs and variable-rate debt collide. About 85 percent of 7(a) loans carry variable rates, according to an analysis of the FOIA file by sba-feasibility-study, and prime climbed from 3.25 percent to 8.50 percent between 2022 and 2023. Borrowers in these industries who took loans near the bottom of the rate cycle had the most room for payment shock. Price also tracks loss in the seasoned data: MMCG finds that loans priced above 11 percent charged off at 14.93 percent, against 2.94 percent for those at 4 percent or below. Riskier borrowers pay more, so that is no evidence that rates cause losses.

Industry comparisons across vintages also require care. The federal NAICS classification was revised in 2022, which moved some codes, including gas stations from 447110 to 457110, so a naive comparison can split one industry into two. And with event rates near 1 percent, an industry cell with a few hundred loans may contain only a handful of events.

How the 2021 through 2024 vintages were made

The environment these loans were born into was unusual in several respects.

Fees. According to Lumos' summary of SBA notices, upfront guaranty fees were waived entirely from late December 2020 through September 2021, waived on loans up to $350,000 in fiscal 2022 and up to $500,000 in fiscal 2023, and waived on loans up to $1 million from October 2023 until March 2025, when fees returned. Tax Guard reports that SBA estimated it failed to collect more than $460 million in fees during 2022 through 2024. Whether the waivers changed who borrowed is unproven.

Standards. SBA's Inspector General was blunt in its fiscal 2026 report on management challenges. Changes adopted in 2023 sharply lowered underwriting standards in the 7(a) program, the agency reversed many of them in 2025, and the loans made in between remain exposed to greater risk and loss. SBA's operating procedures changed twice in 2023, with SOP 50 10 7 taking effect August 1 and version 7.1 in mid-November, and again in 2025, when SOP 50 10 8 took effect May 31 and, according to NAGGL, returned eligibility determinations to delegated lenders. The guaranteed 7(a) portfolio stood at about $163 billion in June 2025, per the Inspector General.

Eligibility. A separate Inspector General review, Report 26-07, examined loans approved and disbursed between August 1, 2023 and December 31, 2024. It found that SBA's screening framework did not fully check three of six eligibility requirements: for-profit status, size and ineligible business type. The Inspector General could offer only limited assurance that borrowers on 73,302 loans totaling about $32 billion were eligible. In a sample of 188 flagged loans, 71, or 38 percent, lacked sufficient documentation, about $60.7 million in all. SBA agreed to flag all 73,302 loans for review when a guaranty purchase is requested.

That last commitment has a consequence for anyone watching charge-offs. A loan flagged for extra review at purchase may take longer to reach purchase and, eventually, charge-off. If so, the loans most exposed to the eligibility problem, those approved from late fiscal 2023 through early fiscal 2025, could show their losses later than their underlying distress would imply. It is one more reason to rely on status-based measures alongside charge-offs.

Purchases. The Inspector General's fiscal 2027 challenges report, issued October 16, 2026, states that SBA purchased more than $5 billion in defaulted guaranteed loans in fiscal 2025, more than it had resources set aside to cover, according to the Coleman Report's account of the document. The same account notes that the report does not establish that those payments were improper.

A lender-type warning

In September 2026 the Inspector General published a review of Small Business Lending Companies, the nonbank 7(a) lenders SBA licenses and oversees. The report's summary says SBLC loans approved and disbursed from fiscal 2016 through 2023 significantly underperformed on default rates and that early defaults rose substantially. It also says SBA's Office of Credit Risk Management has not conducted a safety and soundness examination of any SBLC since April 2023, because of contracting problems.

The detailed figures come from press accounts of the report. According to Inc., SBLC loans defaulted at 14.87 percent against 9.79 percent for other lenders, and early defaults, those within 18 months of disbursement, ran 5.38 percent against 2.62 percent. Two unnamed SBLCs originated 69 percent of SBLC loans but accounted for 84 percent of the defaults. Coleman Report adds that SBA's own response table shows a cumulative default rate of 23.4 percent for the fiscal 2023 SBLC cohort, 20.2 percent without the largest lender, against 11.9 percent for 7(a) overall.

The report and SBA disagree on how fast early defaults have grown. The Inspector General counted early defaults rising from 20 in fiscal 2016 to 180 in fiscal 2023, an increase of 800 percent, while SBLC originations grew from 719 to 1,945 loans. SBA called the comparison misleading and put the increase at 112.5 percent after adjusting for disbursements, while acknowledging that is still elevated. Dividing early defaults by originations in the reported figures gives roughly 2.8 percent for fiscal 2016 and 9.3 percent for fiscal 2023. Funder Intel reports that the Inspector General also warned that SBLC loans approved in fiscal 2024 and 2025 are already defaulting faster than other lenders' loans, with the caveat that those loans are young.

What lenders and the market are saying

Bank earnings offer one read. Live Oak Bancshares, one of the program's major lenders, reported third-quarter 2026 net charge-offs of $32.1 million, or 1.00 percent of loans annualized, up from 0.61 percent a year earlier, along with a provision of $54.7 million. Of that provision, $17.0 million related to a stressed distillery portfolio, according to the company's filing with the Securities and Exchange Commission. On its earnings call, management described past dues of 9 basis points and said small business charge-off and nonaccrual levels were consistent with recent quarters. The distillery exposure is a single-industry problem, not evidence of a broad early-default wave. It is a reminder, though, that concentrated sector stress can surface in a lender's results long before it shows up in vintage tables.

Origination data show a market that is contracting. American Banker reported 56,130 7(a) loans totaling $32.4 billion in fiscal 2026. SBA's fiscal 2025 year-end figures, as relayed by CT Acquisitions, were 78,078 loans and $37.3 billion, so volume fell about 28 percent by count and 13 percent by dollars. Part of the gap reflects the autumn 2025 government shutdown, which Lumos and American Banker say pulled loans into late fiscal 2025. Lumos counts 1,141 participating 7(a) lenders in 2026, a 30-year low.

Rates have turned against variable-rate borrowers again. The Federal Open Market Committee voted 12 to 0 on September 16, 2026 to raise its target range a quarter point, to 3.75 to 4 percent, according to its statement. That put the prime rate at 7.00 percent from September 17, after cuts through 2025 had taken it to 6.75 percent.

What to watch in the next release

SBA posts its loan-level file quarterly, about a month after each quarter closes. The June 30, 2026 extract was the latest available in late October, and the September 30 file had not yet posted. When it does, approvals through September 2024 will have had a full 24 months to charge off, and fiscal 2024 can be read as a complete vintage for the first time.

Six habits will keep readers out of trouble when the new numbers arrive.

Count charge-offs by date and amount. A loan with a charge-off date and a charge-off amount is a charge-off, whatever code it carries.

Compare vintages at the same age. Comparing them at the same calendar date rewards the youngest loans for being young.

Pair charge-offs with a status-based measure. Liquidation and guaranty purchase are earlier signals than the write-off itself.

Hold the industry code constant. Use one NAICS vintage, and report counts and confidence intervals alongside rates.

Treat 2020 and 2021 as flattered. Pandemic relief makes them poor benchmarks.

Ask who measured, and when. A rate computed before April 2026, or on only the CHGOFF code, describes a different population than one computed after.

For lenders and investors, the practical lesson is to underwrite and monitor the 2022 and 2023 classes as seasoned problem vintages in waiting, not as loans too young to judge. For feasibility work in the highest-loss industries, it is to test debt service coverage at today's prime rate and a stressed revenue case, not at the rate and demand conditions of origination. For borrowers, it is that lenders are tightening, SBA is reviewing eligibility more closely at purchase, and documentation matters more than it did three years ago.

Where the published record goes wrong

This series documents not only what the data say but where commonly repeated claims fail against them. For this topic:

• Charge-off rates that count only the CHGOFF status code after April 2026 miss about a third of historical charge-offs, and some loans once shown as paid in full were settlements.

• SBA Loan Index states that SBA's 504 loan file never carries charge-off status. Counts of the same files from two other data vendors, along with MMCG's 504 rate, point to roughly 12,100 charged-off 504 loans.

• The oft-quoted 4 percent figure for 7(a) is a cumulative share of all loans since fiscal 2010 marked charged off, not an annual rate. NADCO's chief executive cited SBA data showing a one-year 504 charge-off rate of 0.07 percent as of August 31, 2025. The two are different measures and should not be compared.

• "Early default" is defined four different ways across the Inspector General, Lumos, academic research and trade commentary.

• Fiscal 2025 7(a) volume appears as 77,600 loans in SBA's preliminary release, 78,078 in final approvals and 65,154 funded loans in one analysis of the FOIA file. They measure different things.

• The more-than-$5 billion fiscal 2025 purchase figure comes from the Inspector General's fiscal 2027 challenges report, not the fiscal 2026 edition.

• The Inspector General's 800 percent and SBA's 112.5 percent figures for SBLC early-default growth use different bases and are both described as correct by their authors.

• A median time to charge-off of about 50 months measured across all loans since 2010 understates the true lag, because recent loans can only have charged off quickly so far.

• SBA's data dictionary still describes the EXEMPT status as protected from disclosure, though the files now carry thirteen status codes.

• The 7.00 percent prime rate reflects a September 2026 increase, not continued easing.

Sources:

• U.S. Small Business Administration, 7(a) and 504 FOIA loan-level data and data dictionary, data.sba.gov, extracts through June 30, 2026

• SBA Office of Inspector General, Report 26-01, Top Management and Performance Challenges Facing the SBA in Fiscal Year 2026

• SBA Office of Inspector General, Report 26-07, SBA's Screening of 7(a) Loan Applications Under its Risk Mitigation Framework

• SBA Office of Inspector General, Report 26-14, Small Business Lending Companies' Performance in the 7(a) Loan Program and SBA's Oversight

• SBA Office of Inspector General, Report 27-01, Top Management and Performance Challenges Facing the SBA in Fiscal Year 2027

• 13 CFR Part 102, Subpart A, Appendix A, Records Maintained by SBA

• Federal Open Market Committee, statement of September 16, 2026

• Live Oak Bancshares, Form 8-K and third-quarter 2026 earnings materials, Securities and Exchange Commission

• National Association of Development Companies, letter to the House Small Business Committee, November 14, 2025

• National Association of Government Guaranteed Lenders (NAGGL)

• American Banker

• Inc.

• Coleman Report

• Funder Intel

• MMCG Invest, Feasibility Index, Q2 2026

• Lumos, SBA 7(a) default rate and program performance analyses, 2026

• LoanTape, SBA loan-level delinquency analysis, 2026

• SBA Loan Index

• SSRN, Predicting Early Default in SBA 7(a) Loans: Public-Data Risk Ranking Across a Regime Shift