The SBA Loans That Never Close: Inside the Gap Between Approval and Funding

• 

  SBA loans approved but never funded.

SBA's record fiscal 2025 was counted in approvals, and approvals include loans that are later cancelled. By one independent count, one in six fiscal 2025 7(a) approvals had been cancelled before a dollar moved, the highest share on record, with thousands more still waiting.

On September 30, 2025, the last day of its fiscal year, the U.S. Small Business Administration announced that it had guaranteed 77,600 7(a) loans for $37 billion, part of what it called the most capital ever delivered. The number was accurate as a count of approvals. It was not a count of loans that reached a business.

SBA's own definition says so. In its loan program performance tables, the agency defines gross approval as the dollar amount of loans approved in a fiscal year, including loans that are cancelled in the current and later years. Every headline SBA volume figure, in every year, is on that basis. A loan that is approved, celebrated in a year-end release and then abandoned before closing stays in the count.

How large is that gap? By an independent analysis of SBA's loan-level data, 12,924 of the fiscal 2025 7(a) approvals, worth about $4.85 billion, had been cancelled before first disbursement by March 31, 2026. That is 16.6 percent of the year's approvals, the highest rate in records going back to fiscal 2010. Another 5,491 loans were still pending, neither funded nor cancelled.

This article explains how an SBA loan is approved but never funded, how large that gap has been historically, why fiscal 2025 stands out, and what can and cannot be said about the causes. The short answer on causes is that nobody knows. SBA's public data records that a loan was cancelled but not why, not by whom and not when.

One year, three numbers

Fiscal 2025 7(a) volume appears in at least three versions, and they measure different things.

Figure	What it counts	Source
77,600 loans, $37 billion	Preliminary, rounded approvals announced on the last day of the fiscal year	SBA news release, September 30, 2025
78,078 loans, $37.29 billion	Final gross approvals as of September 30, 2025, including later cancellations	SBA fiscal 2025 year-end activity report, as compiled by independent analysts
65,154 loans, $32.43 billion	Approvals net of cancellations, as of the March 31, 2026 loan-level data	Shane Pierson, The State of SBA Lending

The first two differ because the press release was issued before the year's final approvals were recorded, and September 2025 was an unusually heavy month as lenders rushed loans through ahead of the government shutdown. The year-end figure of 78,078 is consistent with every other source we reviewed, including the 13 percent decline SBA officials cited for fiscal 2026, but we were not able to read it directly from SBA's spreadsheet.

The third figure removes the loans that were cancelled before funding. The arithmetic closes exactly: 78,078 minus 12,924 is 65,154, and $37.29 billion minus $32.43 billion is $4.86 billion, within rounding of the reported $4.85 billion. It does not remove the 5,491 loans that were still pending, so it is not a count of funded loans. It is a count of loans that had not yet been cancelled.

SBA's fiscal 2025 figures shifted again after the year closed. Its annual report, released in January 2026, rounded combined 7(a) and 504 volume to 85,000 loans and $45 billion. A November 2025 SBA release used 84,400. These are rounding and timing differences, but they show how much a single year's count can move.

The same basis applies to fiscal 2026. On November 2, 2026, SBA reported 56,130 7(a) loans for $32.4 billion and 6,075 504 loans for $7.1 billion. Those are approvals. No funded or cancelled figure for fiscal 2026 has been published. There is also a coincidence worth noting: fiscal 2026 gross 7(a) dollars, $32.4 billion, almost exactly equal fiscal 2025 net dollars, $32.43 billion. Comparisons that mix the two bases will produce a flat year that did not happen. On a like-for-like gross basis, 7(a) fell 28.1 percent by count and 13 percent by dollars.

How an approved loan gets cancelled

An SBA approval is a guaranty commitment, not a disbursement. Between the two, the borrower and lender still have to close: satisfy conditions, complete appraisals and environmental reviews, sign documents and fund. Several paths lead to cancellation instead.

The lender cancels it. According to SBA's fiscal transfer agent guidance, lenders can cancel fully undisbursed loans through SBA's E-Tran servicing system, which cancels the guaranty. Until then, lenders report the loan monthly as fully undisbursed.

SBA cancels it on a deadline. SBA policy requires 7(a) loans to be fully disbursed within 48 months of approval, after which SBA cancels any undisbursed balance. For 504 loans, a September 2025 procedural notice sets the same 48-month limit, with a 9-month limit for debt refinancing without expansion, extendable to 15 months. SBA tracks the deadlines and cancels undisbursed amounts automatically after notifying the certified development company about three months in advance.

The guaranty fee is not paid. SBA's servicing pages list reinstatement of loans cancelled for nonpayment of the guaranty fee, which means some cancellations are reversible.

In SBA's public loan-level files, these outcomes appear as status codes. Cancelled loans are coded CANCLD. Approved but undisbursed 7(a) loans are coded COMMIT, and the 504 equivalent is NOT FUNDED, according to readers of SBA's data dictionary. The files are updated quarterly, about a month after each quarter ends. The latest, as of September 30, 2026, now covers the full fiscal 2026 year.

Two features of the data shape everything that follows. First, because a loan can remain undisbursed for up to four years before SBA cancels it, any recent year's cancellation rate is a floor, not a final figure. Loans pending today will eventually fund or cancel. Second, the public file contains no field for the reason a loan was cancelled, who initiated it or when it happened. Federal disclosure rules treat non-statistical information on cancelled applications as generally exempt, so the data show the outcome and almost nothing else.

How large the gap usually is

There is no official SBA cancellation series in a source we could retrieve. There are several benchmarks, and they are broadly consistent until fiscal 2025.

Source	Benchmark	Basis
Congressional Research Service, 7(a) report (2021 and 2022 editions)	Approvals typically 10% to 20% higher than loans disbursed; 10% to 15% by dollars	7(a), general
CRS, fiscal 2007 data	99,607 approved, 88,270 after full cancellations, about 11.4%	7(a), count
CRS, 504 report	5% to 15% of approved 504 loans cancelled each year, typically by the borrower	504
NAGGL, July 2025	Cancellations normally run 5% to 10% of gross approvals	7(a), trade association
Pierson, fiscal 2015 to 2024	9.7% to 13.7% cancelled	7(a), count, loan-level data
Pierson, fiscal 2010	15.1%, the prior peak	Same
Pierson, fiscal 2025	16.55% cancelled; up to 23.6% if every pending loan lapses	Same, as of March 31, 2026

Fiscal 2025 sits above every published range. On Pierson's figures, fiscal 2024 approvals of 70,242 compare with 63,096 net of cancellations, a cancellation rate of about 10.2 percent. Fiscal 2025's 16.55 percent is roughly six points higher, and it is a floor: if any of the 5,491 loans then pending were later cancelled, the rate rose.

The different benchmarks are not strictly comparable. CRS compares approvals with disbursements, which also counts loans that are not yet funded. NAGGL's range likely reflects cancellations recognized within the year, when they matter for program authority. Pierson measures loan status in data captured six months or more after the year closed. But on any of these definitions, one in six approvals cancelled is unusual.

Slower funding is part of the same picture

Loans that remain in the pipeline longer are more exposed to cancellation, and SBA loans are taking longer to fund.

According to the Coleman Report, which analyzed every Preferred Lender and general processing 7(a) loan approved in fiscal 2024 through June 2026, the median time from approval to first disbursement rose to 21 days in fiscal 2026 from 16 days two years earlier. The share of loans funded within 30 days fell from 78 percent to 65 percent, the average reached 29 days, and one loan in ten took more than 63 days. Pierson's analysis shows a similar trend on a different basis, with the median rising from 11 days in fiscal 2023 to 21 days in fiscal 2025.

Lender performance varied widely. Coleman reports that Zions Bank funded 277 loans in a median of eight days, Colony Bank 81 loans in five days, and Lendistry 360 loans in eleven days, while Newtek cut its median from 29 days to 17. Live Oak, the largest 7(a) lender, funded 737 loans in a median of 16 days in Coleman's sample, and its management has said it aims to cut average SBA closing time from about two months to two weeks.

One caution applies to these figures. Pierson notes that the share of loans with a recorded disbursement date fell from 87.4 percent of fiscal 2023 records to 76.4 percent of fiscal 2025 records. Medians calculated only on loans that have funded can understate how long the slowest loans take, and they say nothing about the loans that never fund.

Why fiscal 2025? What is documented and what is not

Several explanations are in circulation. None is backed by loan-level evidence, because the data do not record reasons. The honest approach is to separate what is documented from what is plausible.

Borrowers walking away. This is the only cause with a documented baseline. CRS describes 504 cancellations as typically borrower-driven, citing funds no longer needed or a change in ownership. Earlier CRS language attributes 7(a) cancellations to borrowers or lenders for a variety of reasons. Nothing documents how much this changed in fiscal 2025.

Lenders clearing their books. Cancellation is partly an administrative act. In July 2025, NAGGL's president, Tony Wilkinson, wrote to members that program availability is measured on net approvals, so cancellations are subtracted from gross volume. He urged lenders to cancel any loans that will not be closed, for whatever reason. In a year when program authority was a concern, an active cleanup of dormant approvals would raise the measured cancellation rate without any change in borrower behavior.

The shutdown. The 43-day federal shutdown that began October 1, 2025 closed E-Tran to new applications, loan increases and reinstatements of cancelled loans. Servicing continued, and loans that already had SBA loan numbers could close. SBA estimated the shutdown blocked $5.3 billion of lending to about 10,000 businesses, an extrapolation from fiscal 2025 run rates. Before it began, lenders rushed approvals into September 2025, and Lumos Data estimates about $1.7 billion was pulled forward into that month. Two points follow. Loans that were never approved because of the shutdown cannot appear as cancellations, because they never received a loan number. And it is plausible, though undocumented, that loans approved in a rush were more likely to fall apart afterward. SBA's Dan Pische, director of the Office of Financial Assistance, told American Banker the pre-shutdown surge helped explain fiscal 2026's lower volume.

Rule changes. SBA's standard operating procedure 50 10 8 took effect June 1, 2025, tightening equity, credit score and documentation requirements. From March 1, 2026, a citizenship rule required all owners to be U.S. citizens or nationals living in the United States, making businesses with any lawful permanent resident ownership ineligible. The rule applies to loans receiving an SBA loan number on or after that date, and an SBA notice bars increases to existing loans for borrowers who no longer qualify. The Government Accountability Office concluded in July 2026 that the notices constituted a rule under the Congressional Review Act, which SBA had not submitted. Lenders have described lost deals: Gulf Coast Bank's SBA lending head told American Banker his team walked away from a handful of deals over the new eligibility rules, and a Bank of America regional executive said a narrower borrower pool means a smaller pipeline. Because these rules key on loan-number issuance, their main effect should be fewer approvals, not cancellations of loans already approved. Any link to fiscal 2025's cancellation rate is unproven.

Fees. SBA raised guaranty fees in March 2025. There is historical precedent for fee changes creating an incentive to cancel and resubmit loans: in 2009, SBA explicitly barred borrowers from cancelling pre-Recovery Act approvals to reapply under lower fees. Whether anything similar happened around the 2025 fee change or the fiscal 2026 fee relief for manufacturers is unknown.

Interest rates. Rates do not explain fiscal 2025. The prime rate fell during the year, which, as Pierson notes, rules out rate shock as a blanket explanation. The Federal Reserve's September 16, 2026 increase came at the very end of fiscal 2026 and cannot explain anything before it.

Pierson's own caveat is the right one: the data cannot say why. The most defensible reading is that fiscal 2025's record rate likely reflects several forces at once, including a record surge in approvals, a pre-shutdown rush, tighter rules and administrative cleanup, in proportions no public source can establish.

The 504 program needs separate treatment

Cancellation in the 504 program cannot be measured the same way. The 504 debenture, the SBA-backed second loan, typically funds only after a project is completed, which can take many months, sometimes years. A 504 loan recorded as NOT FUNDED is usually waiting on construction, not abandoned. Comparing 504 NOT FUNDED rates with 7(a) COMMIT rates would overstate 504 attrition.

That caution is not theoretical. SBA reported 6,750 fiscal 2025 504 loans in its year-end release, while a vendor count of loans net of cancellations and unfunded approvals, from the March 2026 data, put the figure at 3,799. A gap of 44 percent is far larger than any 7(a) cancellation rate and almost certainly reflects funding timing, data filtering or both. Until someone replicates it, there is no reliable 504 never-funded rate for fiscal 2025. The only published benchmark remains CRS's historical range of 5 to 15 percent of approvals cancelled each year.

Who cancels: what the data can and cannot answer

No primary source, including SBA, its inspector general, GAO or CRS, publishes cancellation rates by lender, loan size, industry, state, subprogram or franchise status. We found no inspector general or GAO study of approval-to-funding outcomes in the 7(a) or 504 programs at all. Most of those breakdowns can nonetheless be computed from the public files.

Question	Computable from public data?
Cancellation rate by lender	Yes, using the lender name and FDIC or credit union identifier, with minimum loan counts
By loan size band	Yes, using the gross approval amount
By industry	Yes, using the NAICS code
By state or SBA district	Yes, using project state and district office
By 7(a) subprogram, such as Express or Preferred Lender	Yes, with code mapping from the data dictionary
Franchise versus independent	Yes, using the franchise code
Bank versus credit union versus nonbank lender	Only by proxy, from which identifier is present
Who initiated the cancellation	No
Why it was cancelled	No
When it was cancelled	No

Two cautions apply to any lender-level analysis. The lender name in the file reportedly reflects the institution a loan is currently assigned to, so mergers can move even cancelled loans to a successor's name. And recent years need to be reported with the pending share alongside, because a lender with many loans still in COMMIT status will look better or worse once those resolve.

With the September 30, 2026 data now public, the full picture for fiscal 2025, including how many of the 5,491 pending loans eventually funded, and a first look at fiscal 2026 can be computed directly.

How to measure it yourself

The calculation is simple enough for any analyst with SBA's public files and a spreadsheet tool capable of handling large CSVs. We recommend the following steps.

Use the right files and date them. Download SBA's 7(a) loan-level file covering fiscal 2020 to the present and the matching 504 file from data.sba, along with the data dictionary. Record the snapshot date, currently September 30, 2026, next to every result, because the same calculation on an earlier file gives different answers for recent years.

Count four groups for each approval year. For each fiscal year, count all approvals and sum their gross approval amounts. Then count loans that have funded, identified by a first disbursement date; loans coded CANCLD with no disbursement date; and loans still pending, coded COMMIT in the 7(a) file or NOT FUNDED in the 504 file. Report anything that does not fit, such as cancelled loans with a disbursement date, separately rather than forcing it into a group.

Report a floor and a ceiling. The cancellation rate, cancelled loans divided by all approvals, is the floor. Cancelled plus pending, divided by all approvals, is the ceiling. Show both by count and by dollars, and show how old the cohort is, because a two-year-old cohort has had far less time to resolve than a ten-year-old one.

Keep 504 separate. Show 504 results on their own, and treat NOT FUNDED as a timing category rather than a failure, given how long 504 projects take to complete.

Reconcile the totals. A year's total approvals in the loan-level file should land close to SBA's year-end activity report. Small differences are expected, because the loan-level snapshot is taken later.

What this means

For anyone reading SBA headline numbers: an approval figure is a commitment count. It overstates the credit that reached businesses by roughly a tenth in a normal year and by a sixth or more in fiscal 2025. Year-over-year comparisons should stay on one basis.

For borrowers: an SBA approval is a conditional promise with a clock attached. Closing conditions, appraisals, environmental reports and, since March 2026, ownership eligibility can all stop a loan after approval. A borrower whose lender is slow to close carries real risk that a purchase agreement or construction window will lapse first.

For lenders: speed and pull-through vary widely between institutions. A lender's approval volume and its funded volume can tell different stories, and the gap is measurable from public data.

For policymakers: program authority is managed on net approvals. When a sixth of a year's approvals never fund, the gross headline overstates the program's reach, and capacity tied up in loans that will never close is capacity unavailable to borrowers who would.

What this analysis can and cannot say

It can say that SBA's headline volume figures include loans later cancelled, by SBA's own definition. It can say that, by an independent analysis of SBA data, fiscal 2025 had the highest 7(a) cancellation rate in at least fifteen years, at 16.6 percent as of March 2026, with a ceiling of 23.6 percent. It can describe the mechanisms that cancel a loan and the documented forces acting on the program in 2025 and 2026.

It cannot attribute the cancellations to any cause, because the data contain no reasons. It cannot give a reliable 504 cancellation rate for fiscal 2025. It cannot give a final fiscal 2025 rate, because the most recent published analysis predates the latest data. And it cannot rank lenders, industries or states by cancellation rate until the September 2026 files are analyzed.

Where the published record goes wrong

This series documents where commonly repeated claims fail against the data. For this topic:

• SBA's 77,600 and 78,078 fiscal 2025 figures are approvals, not funded loans. SBA's own definition of gross approval includes loans cancelled in later years.

• The widely cited "16.6 percent of fiscal 2025 loans never funded" overstates the certainty and mislabels the measure. The 12,924 loans were cancelled before first disbursement as of March 31, 2026. Another 5,491 were pending, so the share not yet funded at that date was up to 23.6 percent, and the final cancellation rate could be higher than 16.6 percent.

• The 65,154 net fiscal 2025 figure is net of cancellations, not a count of funded loans, because it still includes pending approvals.

• Fiscal 2026 gross 7(a) dollars, $32.4 billion, equal fiscal 2025 net dollars almost exactly. Comparing them suggests a flat year. On a gross basis, fiscal 2026 was down 13 percent by dollars and 28.1 percent by count.

• SBA's fiscal 2025 combined count appears as 84,400 in two SBA releases and 85,000 in its annual report.

• SBA's estimate that the shutdown blocked $5.3 billion of lending describes approvals that never happened, not approved loans that were cancelled. Lumos's $1.7 billion is a different measure: volume pulled forward into September 2025.

• SBA's fiscal 2026 release describes the shutdown as both 43 days long and as halting lending for nearly two months.

• Approval-to-funding medians differ by source, 11 to 21 days in one analysis and 16 to 21 days in another, because they cover different loan types and years.

• A vendor's fiscal 2025 504 net count of 3,799 against SBA's 6,750 does not show a 44 percent cancellation rate. 504 debentures fund after project completion, and the figure has not been replicated.

  
  

Sources:

• U.S. Small Business Administration, news release 25-83, "Trump SBA Delivers Record Capital to Small Businesses in FY25," September 30, 2025

• U.S. Small Business Administration, news release 26-06 on the shutdown's impact, November 13, 2025, and news release 26-02, October 21, 2025

• U.S. Small Business Administration, news release 26-22 and 2025 Annual Report, January 20, 2026

• U.S. Small Business Administration, news release 26-113, "SBA Fuels Main Street Growth with $100 Billion Delivered in FY26," November 2, 2026

• U.S. Small Business Administration, loan program performance tables and definitions

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

• U.S. Small Business Administration, 7(a) and 504 Activity Reports, fiscal 2025 year end, data.sba

• U.S. Small Business Administration, Fiscal Transfer Agent notice, "7(a) Loan Program: Reporting of Fully Undisbursed Loans," March 21, 2022

• U.S. Small Business Administration, Procedural Notice 5000-872764

• U.S. Small Business Administration, Policy Notice 5000-1103, April 10, 2009

• 13 CFR Part 102, Subpart A, Appendix A

• U.S. Government Accountability Office, decision B-338157, July 1, 2026

• Congressional Research Service, Small Business Administration 7(a) Loan Guaranty Program, R41146, and 504/CDC Loan Guaranty Program, R41184

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

• NAGGL, "A Note from NAGGL's President," July 30, 2025, and summary of the FY27 budget request, April 9, 2026

• Shane Pierson, The State of SBA Lending, mid-2026 edition

• Coleman Report, "Coleman Benchmark Data Shows SBA 7(a) Loans Fund in 21 Days," August 4, 2026

• Lumos Data, "SBA 7(a) Program Performance in FY2026," July 15, 2026

• American Banker, reporting by John Reosti, March 2, October 9 and November 2, 2026

• Live Oak Bancshares, second quarter 2026 earnings call