What Do Lenders Recover on a Charged-Off Gas Station, Car Wash or Hotel? Nobody Publishes the Answer

• 

  Exterior of a gas station

SBA has a number for how much it expects to recover on defaulted loans. It has none for the properties that make recovery hard.

Ask a loan officer what happens when a gas station loan goes bad and you will hear a confident story. The tanks have to be tested. The environmental report has to be clean. The fuel contract may not transfer. The building cannot be a pharmacy tomorrow. Ask the same officer how many cents on the dollar the lender actually recovers on a failed gas station, versus a car wash or a hotel, and the confidence disappears. No one has a number.

That is not a shortcoming of the officer. It is a gap in the public record. The U.S. Small Business Administration publishes recovery rates for its 7(a) and 504 programs as a whole. Its own budget model assumes how much it will recover on defaulted loans, and the Government Accountability Office has described how that model works. But neither SBA's published tables nor its loan-level data break recovery out by industry or by type of collateral. Anyone who quotes a recovery rate for gas stations, car washes or hotels in SBA lending is citing something that does not exist.

This article is about that gap. It lays out what SBA does publish, what its public loan file can and cannot show, what the rules say about special-purpose property, and what outside evidence from bank failures and commercial mortgage-backed securities suggests about losses on similar collateral. The goal is not to invent the missing number. It is to give lenders, borrowers and analysts an honest map of what is known, what is assumed and what is simply unmeasured.

What SBA publishes about recovery

SBA's loan program performance release carries two recovery tables. One reports total recoveries as a percentage of purchase amounts, by the year the loan was purchased. The other reports post-charge-off recoveries as a percentage of the amounts charged off, by charge-off year. Both are program-level. Neither is cut by industry, collateral type or state. The latest edition posted on SBA's site carries data through June 30, 2025 and was last updated September 15, 2025. As of late October 2026, no newer edition had appeared, so the freshest published recovery data are more than a year old.

The best primary figures on current expectations come from the federal budget. In the Federal Credit Supplement that accompanies the president's fiscal 2027 budget, the Office of Management and Budget lists the assumed default and recovery rates for each loan cohort. Default rate is gross claims divided by disbursements, and recovery rate is amounts recovered divided by defaults. For 7(a) loans, the 2027 cohort carries an assumed default rate of 6.85 percent and an assumed recovery rate of 31.16 percent. For 504 loans, the 2027 cohort carries a 5.88 percent default rate and a 44.40 percent recovery rate.

Those figures tell three stories. First, expected recovery on a defaulted 504 loan is about 13 points higher than on a defaulted 7(a) loan, which is consistent with the 504 program's real-estate-secured structure. Second, the assumed 7(a) recovery rate has been falling. It was 35.41 percent for the 2025 cohort and is about 31 percent for the 2026 and 2027 cohorts. The 504 assumption barely moved, from 43.77 percent to 44.40 percent. Third, the subsidy rate for every cohort shown is zero, meaning that on these assumptions fees are expected to cover losses.

These are budget assumptions, not observed outcomes, and the distinction matters. GAO reviewed SBA's credit subsidy model in 2020. It found that SBA assumed a lifetime recovery rate of 37.49 percent for the fiscal 2020 7(a) cohort, while no cohort since fiscal 1999 had actually recovered more than 35 percent as of March 2019. SBA officials told GAO that recoveries can extend more than 15 years past the purchase date, which is why the model's assumption could exceed what had been collected so far. The two figures are not in conflict, but they are easily confused.

GAO also described how the model works, and this is the key point for anyone hoping to find asset-class recovery data. SBA's cash-flow model applies recovery assumptions based on stratified historical averages, stratified by loan approval amount, by whether a loan is an Express or non-Express loan, and by whether it is a term loan or a revolving line of credit. Industry does enter the model, but on the default side, not the recovery side. GAO noted that the fiscal 2020 model predicted higher purchase rates for loans in accommodation, food service and entertainment. How much SBA recovers on a charged-off hotel versus a charged-off dental practice is not a parameter.

One more caution applies to all of these numbers. In January 2026, the independent auditors for SBA's inspector general were unable to give an opinion on the agency's fiscal 2025 balance sheet. They identified a material weakness in controls over the subsidy reestimate, finding that SBA could not support the reliability of all key data used in its 7(a) reestimate model. SBA's chief financial officer disputed parts of that finding. Whatever the merits, it means the recovery assumptions above rest on data the auditors could not fully verify.

What the public loan file can and cannot show

SBA also publishes loan-level data on every 7(a) and 504 loan through the FOIA reading room at data.sba.gov. It is the richest public source on SBA lending, and it is where an analyst would naturally turn to measure recoveries by industry. It does not work for that purpose.

The file records each loan's approval date and amount, its industry code, its lender, its status, and for loans that have been charged off, the charge-off date and the gross charge-off amount. It does not record how much SBA or the lender later recovered. No public inventory of the file's fields lists a recovery amount, a purchase amount, a sale price, a collateral description or an appraisal. One independent analysis of the file notes that, for this reason, every rate computed from it is gross.

The file's status codes have also changed. According to LoanTape, a data provider that tracks the files, SBA expanded the 7(a) status field in the March 31, 2026 release from five codes to 13, separating loans that are current, delinquent, past due or in liquidation from the old catch-all category, and splitting the charged-off category into charged off, closed after charge-off and sold after charge-off. Those new categories hint at how workouts end, but they carry no dollars. A loan marked as sold after charge-off tells you a claim was sold, probably at a discount, but not the price. Whether the gross charge-off amount represents the whole loan or only SBA's guaranteed share is not documented in anything we could retrieve.

So what can the file tell you? It can tell you how often loans in a given industry end in charge-off, how large those charge-offs are relative to what was approved, and how long the process takes. Those are measures of loss incidence. They are real and useful, but they are not recovery rates, and calling them recovery rates would be wrong.

Three asset classes, one honest table

With that limit in mind, here is what the loss-incidence data show for the three asset classes in this article. The figures come from the MMCG Invest Feasibility Index, which is affiliated with Loan Analytics. It uses SBA's loan file as of March 31, 2026 and measures seasoned 7(a) loans approved from fiscal 2010 through 2019. We were not able to recompute these figures independently from the raw file for this article, so treat them as one source's tabulation.

Asset class	Funded 7(a) loans	Charged off	Charged off or otherwise adverse	Average loan
Gas stations with convenience stores	5,545	2.78%	3.50%	$980,000
Hotels and motels	7,706	2.10%	4.44%	$1,894,000
Car washes	2,195	4.56%	6.83%	$1,139,000
All 7(a) loans	n/a	6.44%	n/a	n/a

Two features stand out. Hotels and gas stations charge off far less often than the 7(a) average, and car washes charge off more often than the other two, though still below the average. The pattern is the opposite of what the "special-purpose properties are risky" narrative would predict. The properties lenders worry most about in liquidation are among the least likely to reach it.

The second feature is the gap between Express and standard loans. In the same tabulation, car washes financed through SBA's Express program charged off at 9.89 percent on 435 loans, compared with 3.24 percent on 1,760 standard loans. Gas stations show the same split, with Express at 6.10 percent on 574 loans and standard at 2.39 percent on 4,971. Express loans are smaller and are processed with less documentation under the lender's own procedures, so the gap may reflect borrower mix, collateral and underwriting all at once. The data cannot say which. What they do show is that, for single-purpose fixed assets, the loan product matters at least as much as the asset type.

None of these figures is a recovery rate, and none says anything about what happens after a charge-off. A low charge-off rate tells you few loans fail. It tells you nothing about what a lender gets back when one does.

Why special-purpose collateral is different

SBA's rules treat these three properties as a distinct category. The agency's operating procedures have long defined special purpose property as a limited-market property with a unique physical design, special construction materials or a layout that restricts its use to the specific purpose for which it was built. When SBA issued its 2015 revision of the procedures, it listed car washes, hotels, gas stations with or without a convenience store, golf courses, medical facilities and bowling alleys as examples. Appraisal and lender summaries indicate later versions, including the revision that took effect October 1, 2026, keep gas stations, car washes and hotels on the list.

The rules attach real costs to that classification. Under federal regulation, a 504 borrower must contribute at least 10 percent of project costs. That rises to at least 15 percent when the project involves a limited or single-purpose building, or when the business has operated for two years or less, and to at least 20 percent when both conditions apply. The regulation itself uses the phrase "limited or single purpose building" and names no property types, leaving the examples to SBA's procedures. The extra equity is, in effect, a lender's cushion against exactly the liquidation problem this article is about.

SBA also requires a different kind of appraisal. For an acquisition of a special-purpose business, the 2015 procedures call for a going-concern appraisal by a certified general appraiser who has completed multiple going-concern appraisals of similar special-use property within the prior 36 months. The appraisal must allocate separate values to the land, the building, the equipment and the intangible assets of the business.

That allocation is the heart of the collateral story. A hotel, a car wash or a gas station sells as a going concern, and part of its value is an operating history, a brand, a customer base and permits that do not survive a closure. In liquidation, the building and the land remain, and some equipment can be resold, but the intangible value largely does not. The going-concern appraisal makes that split visible before the loan is made. Nothing in the public record shows how large the loss is when it is realized.

Gas stations: the environmental tail

Gas stations carry a risk the other two do not, and it is well documented. The Environmental Protection Agency reports that more than 581,000 releases from underground storage tanks had been confirmed as of September 2025 and that nearly 528,000 sites had been cleaned up. About 54,000 sites remain to be cleaned up. EPA's year-end performance report, as summarized by an industry trainer, puts the backlog at 53,777, with about 533,000 active tanks at roughly 190,000 facilities. State cleanup funds spend about $1 billion a year.

SBA's environmental policy treats fuel sites accordingly. According to consultants who reproduce the procedures, a Phase I environmental assessment should always be obtained if a business sells, supplies or dispenses fuel, and the industry-code trigger for gas stations moved from NAICS 447 to 457 when the code was revised. Car-wash-only facilities, by contrast, may start with a lighter transaction screen, and dry cleaners with any on-site dry cleaning history need both a Phase I and a Phase II.

For a lender, the regulatory exposure is limited but not eliminated. Under EPA's lender-liability rule, a holder of a security interest is not treated as an owner or operator of an underground tank if it does not participate in management. But the rule imposes deadlines on a lender that forecloses. It must list or advertise the property within 12 months, it must act on a bona fide offer of fair consideration within 90 days once six months have passed, and if no operator is running the site it must empty the tanks within 60 days and close them permanently or temporarily. Those are real frictions on a liquidation, and they apply only to fuel.

What is missing is the dollar figure. No source we found quantifies what a lender loses to remediation, delayed sales or fuel-contract problems when an SBA gas station loan is liquidated. Environmental risk is plainly a driver of underwriting cost and liquidation friction. How much it moves recovery is unmeasured.

Car washes: boom, leverage and a bankruptcy that isn't SBA

Car washes are the newest of the three as a lending category, and the recent evidence on distress comes from a large operator, not from SBA. Zips Car Wash filed for Chapter 11 on February 5, 2025 in the Northern District of Texas with $653.9 million of funded debt, according to a summary of its first-day declaration. The plan cut about $279 million of debt and gave the lenders the equity. Zips emerged on April 30, 2025. Its restructuring included an $82.5 million debtor-in-possession loan that combined $30 million of new money with a roll-up of existing debt, and a sale of its Orlando sites to El Car Wash for $58.5 million, according to CreditSights.

The company's chief transformation officer blamed an industry that saw nearly 1,000 new car washes open over the previous half decade, according to CreditSights' account of the filing. Zips was financed with private credit, not SBA loans, so it says little about SBA recoveries. What it does illustrate is how a going-concern car wash can lose value when the market it serves fills up, and how much of a lender's outcome depends on whether the business keeps washing cars. That is the dependency the going-concern appraisal tries to measure and the real estate alone cannot cover.

We found no published recovery figure for car wash equipment or site liquidations, whether from SBA, lenders or auction data. Claims about "more foreclosures and auctions than ever" from industry vendors came with no data and should not be repeated.

Hotels: frequent defaults, uncertain severity

Hotels are the one asset class here where public data on loss severity exist, but not from SBA. The commercial mortgage-backed securities market tracks defaults and losses loan by loan. KBRA's 2025 study of conduit CMBS covered 108,096 loans and found that lodging had the highest cumulative default rate of any property type, 31.8 percent, against 16.9 percent for all loans. Health care followed at 26.1 percent and office at 22.8 percent. Office, however, had the highest loss severity. The study reported average severity of 49.7 percent on loans with losses above 2 percent and 28.9 percent across all 16,723 resolved loans. It did not single out lodging severity in its release.

KBRA's study of single-borrower CMBS, released in December 2025, found a lower lodging default rate. Of 244 lodging loans, 31 had defaulted, 12.7 percent, behind office at 19.7 percent, against 12.9 percent overall. Average severity across all resolved loans in that population was 11.2 percent, rising to 40.3 percent where the loss exceeded 1 percent. A KBRA review of eight single-borrower loss loans in March 2026 concluded that property type did not drive loss severity. Losses instead reflected macroeconomic disruptions that hit lodging and retail first and office later.

The takeaway is double-edged. Hotels default more often than most commercial property, but the best available evidence does not show that they lose more when they do. These are large CMBS loans, not owner-operated limited-service hotels financed with SBA guarantees, and the two populations should not be treated as the same. They are context, not estimates.

Outside benchmarks: what a failed bank's commercial loans lost

The most useful outside yardstick for small-business real estate losses comes from bank failures. Researchers at the Federal Deposit Insurance Corporation examined 14,225 defaulted commercial real estate loans at 295 failed banks between 2008 and 2013 and found a mean loss given default of 44 percent and a median of 41 percent. About 30 percent of defaulted loans were cured with no loss, and about 10 percent lost more than the loan's value, once costs were counted. The paper found that foreclosure added about 20 percentage points to expected loss, and that each additional quarter spent in workout added roughly three points. It did not break results down by property type.

Benchmark	Population	Loss measure
FDIC failed-bank study	14,225 defaulted commercial real estate loans, 2008–2013	Mean 44%, median 41%
KBRA conduit CMBS	16,723 resolved loans	28.9% average severity
KBRA conduit CMBS	10,591 loans with losses above 2%	49.7% average severity
SBA budget assumption, 7(a)	Defaulted loans, 2027 cohort	31.16% recovered, so about 69% not recovered
SBA budget assumption, 504	Defaulted debentures, 2027 cohort	44.40% recovered, so about 56% not recovered

Read together, these figures bracket the question without answering it. Losses on defaulted commercial real estate in a crisis ran roughly 30 to 50 percent of the balance. SBA's budget assumes that it will lose more than half of what it pays out on defaulted 504 loans and about two-thirds on defaulted 7(a) loans. The two are not directly comparable. SBA's recovery rate is measured against the guaranteed amount it paid, after lenders have already liquidated collateral, so the "unrecovered" share is net of what lenders captured first. But the table shows the scale of what is at stake.

A current example of collateral repricing

A live example of special-purpose collateral losing value came this month, though it involves none of the three asset classes. Live Oak Bancshares, one of the nation's largest SBA lenders, reported third-quarter 2026 results on October 21. Its provision for credit losses was $54.7 million, up from $25.8 million in the second quarter, driven largely by a $17.0 million charge related to its distillery portfolio, which it has said it is exiting. Net charge-offs were $32.1 million, 1.00 percent of loans on an annualized basis. On its earnings call, management said it records a charge-off for the excess of what it does not expect to collect.

Distilleries, like hotels and car washes, are operating businesses with purpose-built real estate and specialized equipment. When conditions in a niche turn, collateral values can reprice faster than appraisals update.

What this means for lenders, borrowers and analysts

The honest conclusion is narrower than a recovery table would be, but it is still actionable.

For lenders, the going-concern appraisal split is the most important tool, and it is worth treating as more than a compliance step. The gap between the going-concern value and the sum of land, building and equipment is the cushion at risk in a liquidation. Underwriting a hotel, car wash or gas station should include a scenario in which the business is closed and only the hard assets are sold, with time and costs for that sale built in.

For borrowers, the higher equity requirement on a single-purpose building reflects SBA's judgment about downside, and a borrower who understands the logic can plan for it.

For analysts and anyone publishing on this topic, the most important discipline is labeling. Charge-off rates from the public file are loss incidence, not recovery. Program-level recovery rates are budget assumptions, not observed outcomes, and are not industry-specific. Benchmarks from CMBS and failed banks are context from different populations. Mixing these into a single "recovery rate for hotels" would make the number look more precise than any source can support.

For feasibility work, the missing data are an argument for conservatism. When recovery cannot be benchmarked for the asset class, a study should state that plainly and test debt service coverage and collateral coverage under stress, not rely on an assumed recovery.

How to get the real answer

SBA holds the underlying data, and a request under the Freedom of Information Act is the route to it. A useful request would ask for aggregate, not borrower-level, figures by six-digit industry code and approval year: purchase dates and amounts, recoveries before and after charge-off, collateral liquidation proceeds, and whether claims were sold. Aggregating by industry and year reduces the privacy objections that often block loan-level requests. Expect processing to take months, and expect SBA to say that some of what is requested does not exist in the form asked.

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:

• No public source reports what SBA or lenders recover on defaulted loans by asset class. Statements that SBA recovers a particular percentage on gas stations, hotels or car washes have no public basis, because SBA's recovery model is not stratified by industry and the public loan file contains no recovery field.

• SBA's assumed lifetime 7(a) recovery of 37.49 percent for the fiscal 2020 cohort, the observed cumulative recoveries below 35 percent through 2019, and the budget's 31.16 percent for the 2027 cohort measure different things. The first and third are assumptions and the second is an observation.

• Gas station default rates in circulation range from about 3 percent to about 15 percent. The differences come from denominators: funded loans versus resolved loans, count versus dollars, and seasoned versus all vintages.

• The claim that 504 loans never report charge-offs is wrong. Data vendors count more than 12,000 charged-off 504 loans in SBA's file.

• Lists of special-purpose property vary by source. Gas stations, car washes and hotels appear in SBA's own lists. Self-storage and RV parks appear only on some vendors' lists.

• Rankings of which commercial property type loses the most in default are not stable. FDIC researchers noted that one study put retail among the lowest-severity categories while another put it highest. Rankings depend on the period.

• Lodging delinquency figures differ by source because the populations differ: 6.67 percent from S&P in April 2025, 5.81 percent from Trepp in September 2025 and 3.63 percent from KBRA, which covers only the deals it rates, in October 2025.

• The size of EPA's tank-cleanup backlog appears as about 53,000, about 54,000 and 53,777. They are the same quantity at different levels of rounding.

• A reported figure of more than $5 billion in defaulted SBA loans purchased in fiscal 2025 comes from a trade-press account of the inspector general's fiscal 2027 challenges report. Whether it covers only 7(a) loans is unclear.

• A figure of under 1 percent recovered, sometimes quoted about SBA, refers to emergency pandemic disaster loans, not to 7(a) or 504 loans.

Sources:

• U.S. Small Business Administration, Loan Program Performance tables, data through June 30, 2025

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

• U.S. Small Business Administration, Information Notice 5000-1340, SOP 50 10 5(H)

• U.S. Small Business Administration, SOP 50 10 8 and SOP 50 10 8.1, Information Notice 5000-880695

• Office of Management and Budget, Federal Credit Supplement, Budget of the U.S. Government, Fiscal Year 2027

• U.S. Government Accountability Office, GAO-20-618, SBA Generally Incorporated Key Elements for Its Credit Subsidy Estimates

• SBA Office of Inspector General, Report 26-03, Independent Auditors' Report on SBA's Fiscal Year 2025 Financial Statements

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

• 13 CFR 120.910, Borrower contributions

• 40 CFR Part 280, Subpart I, Lender Liability for Underground Storage Tanks

• U.S. Environmental Protection Agency, Releases from Underground Storage Tanks

• KBRA, Conduit CMBS Default and Loss Study, July 2025

• KBRA, Single-Borrower CMBS Default and Loss Study, December 2025

• KBRA, Anatomy of Loss in Single-Borrower CMBS, March 2026

• Federal Deposit Insurance Corporation, Center for Financial Research Working Paper 2015-03, Loss Given Default for Commercial Loans at Failed Banks

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

• Coleman Report

• CreditSights

• ElevenFlo

• PASS Testing

• LoanTape

• Apify

• MMCG Invest, Feasibility Index, Q2 2026

• Reliant Business Valuation

• Partner ESI