SBA Lending Deserts Start With a Math Problem

• 

  SBA lending deserts by county

How many SBA loans does the typical American business community receive? Depending on who is counting what, the answer for a single year runs from under two per 1,000 businesses to more than ten.

Ask how well SBA lending reaches a community and you will get an answer that sounds precise and is not. The Small Business Administration reported about 84,400 7(a) and 504 loans in fiscal 2025. Divide that by the 8.4 million employer establishments counted by the Census Bureau's County Business Patterns, and you get roughly 10 loans for every 1,000 establishments. Use SBA's final 7(a) count and remove approvals that were cancelled or never funded, and the figure falls below 8. Count every business, including the tens of millions that have no employees, and it drops to about 2.

None of those answers is wrong. Each answers a different question. The differences matter a great deal when someone draws a map of "lending deserts," the counties where SBA-backed credit is scarce relative to the businesses that might use it. A county can look starved or well served depending on what goes into the fraction.

This article lays out the arithmetic behind SBA lending deserts: what belongs in the numerator, what belongs in the denominator, and why "rural" means at least three different things in federal lending. It then turns to the USDA programs that serve the same rural counties, to the 2026 changes that are reshaping the map, and to a practical method for measuring a single county. The point is not to settle where the deserts are. It is to make sure that when someone tells you where they are, you know what question was asked.

The denominator: employer establishments are a fraction of all businesses

The standard county-level source for counting businesses is the Census Bureau's County Business Patterns, known as CBP. Its 2023 edition, released in June 2025, counted 8,361,342 establishments with paid employees nationwide. About 7.15 million of them had fewer than 20 employees. For anyone building a county map, this is the natural denominator: it is published for every county, it is built from administrative records rather than a sample, and it counts the kind of Main Street business that SBA lending is meant to reach.

But CBP leaves out most of the business population. It does not count self-employed individuals, and it omits several categories of employer, including most government employment and most agricultural production. The Census Bureau's Nonemployer Statistics fill part of that gap. They counted 30,427,808 nonemployer establishments in 2023, which Census reports as 78.4 percent of all establishments in the country. The 2024 edition, released on September 29, 2026, shows nonemployer businesses up 5.1 percent from the year before.

So the market for small-business credit can be described as 8.4 million employer establishments or as nearly 39 million businesses of all kinds. SBA borrowers include sole proprietors and startups with no payroll, which argues for the larger number. But nonemployers rarely seek the loan sizes SBA programs are built around, which argues for the smaller one. Neither choice is neutral, and a rate computed on one cannot be compared with a rate computed on the other.

Two more limits apply. First, timing: CBP counts establishments as of March 2023, while the loans being compared run through fiscal 2025 and 2026. Second, the series itself is in flux. The Census Bureau says it has no planned release date for the 2024 edition of CBP, because it is working out how to publish while complying with a Commerce Department order that prohibits noise infusion, a technique long used to protect confidentiality. According to an advocacy coalition that follows Census data policy, tiered access to the data is under consideration. For now, 2023 is the latest county-level count of employer establishments, and when 2024 arrives, small counties may be harder to measure than before.

The numerator: three counts of the same year

If the denominator is slippery, the numerator is worse, because SBA's own fiscal 2025 figures appear in at least three versions.

On September 30, 2025, the last day of the fiscal year, SBA announced it had approved 84,400 7(a) and 504 loans for $44.8 billion. That figure comprised 77,600 7(a) loans for $37 billion and 6,750 504 loans for $7.8 billion. SBA's year-end activity report, as relayed by CT Acquisitions and an independent analyst, later put 7(a) at 78,078 loans and $37.29 billion.

Then there is the public loan file. According to an analysis of SBA's FOIA data as of March 31, 2026, there were 65,154 fiscal 2025 7(a) loans worth $32.43 billion once approvals that were cancelled or never funded were removed. Those removed approvals numbered 12,924 and totaled about $4.85 billion, or 16.6 percent of the original count. The arithmetic closes exactly: 78,078 minus 12,924 is 65,154. A later extract, as of June 30, 2026, shows 64,096 according to InterObservers, a data publisher, as further cancellations worked through.

The gap, in other words, is the difference between approving a loan and funding it. Press releases and year-end reports count approvals. The public loan file, once cleaned, counts loans that actually reached a borrower.

The 504 program shows an even wider gap that is harder to explain. SBA reported 6,750 loans in fiscal 2025. A vendor analysis of the public file reports 3,799. The likeliest reason is that 504 loans are funded only when SBA's debenture is sold, often months after approval, so a filter on funded loans would remove many recent approvals. That explanation is plausible but unconfirmed, and the vendor's claim of a collapse in 504 lending since 2021 should be treated as an artifact until someone reproduces it from the raw file.

Fiscal 2026 changes the picture again. According to American Banker, SBA approved 56,130 7(a) loans for $32.4 billion in fiscal 2026, while 504 dollar volume fell 9 percent to $7.1 billion. Against the final fiscal 2025 figures, 7(a) volume fell about 28 percent by count and 13 percent by dollars, and the average loan grew from about $478,000 to about $577,000. Lumos Data, a vendor, found that through nine months of fiscal 2026, 7(a) approvals were down 33.4 percent by count and 20.9 percent by dollars from the same period a year earlier. American Banker reports that Lumos also found that the autumn 2025 government shutdown pulled loans forward into late fiscal 2025, which exaggerates the year-over-year drop.

Behind the volume decline is a shrinking lender base. Lumos counts 1,141 lenders making 7(a) loans as of July 2026, down 18.9 percent from 1,407 a year earlier and a 30-year low. Eric Rabinovich of SBA Loan Lab told American Banker that deals have become objectively harder to close because SBA's credit box has narrowed.

One year, six rates

Combine the numerators and denominators and you get this. Every figure below is simple division of published totals. None is a county result.

Numerator (fiscal 2025)	Per 1,000 employer establishments (8.36 million)	Per 1,000 all businesses (about 38.8 million)
7(a) and 504 approvals, SBA press release (84,400)	10.1	2.2
7(a) approvals, year-end report (78,078)	9.3	2.0
7(a) loans net of cancelled or unfunded approvals (65,154)	7.8	1.7

The same year's lending can be described as 1.7 or 10.1 loans per 1,000 businesses, a sixfold range. For context, fiscal 2024's 70,242 7(a) loans work out to 8.4 per 1,000 employer establishments, and fiscal 2026's 56,130 work out to 6.7. On dollars, fiscal 2025's $44.8 billion works out to about $5,400 per employer establishment, and fiscal 2026's roughly $39.5 billion in combined 7(a) and 504 approvals to about $4,700.

Two lessons follow. First, any county ranking needs to state its numerator and denominator in the first chart, not in a footnote. Second, the national benchmark itself is soft. A county rate of 6 per 1,000 is a desert against a national rate of 10 and perfectly typical against a national rate of 6.7.

What earlier lending-desert research found, and what exists now

The idea of measuring SBA lending against local business counts is not new. In 2014, the National Community Reinvestment Coalition published "Small Business Lending Deserts and Oases," an analysis of all 3,219 counties for 2012. It found that SBA's 7(a) program guaranteed about 44,300 loans that year, about 14.6 per 10,000 businesses nationwide, on a base of roughly 30.8 million businesses. NCRC sorted counties into quintiles, with the bottom group averaging about four loans per 10,000 businesses and the top group about 36. Rural counties generally received fewer loans, though NCRC cautioned that its broader bank-lending data omitted smaller banks that are not required to report.

NCRC's national rate works out to about 1.5 loans per 1,000 businesses. On a comparable all-business base, 7(a)-only rates for fiscal 2025 run from about 1.7 to 2.0 per 1,000. The two periods use different business counts and different loan definitions, so treat that as the same order of magnitude, not a trend.

What exists today, as far as we could determine, falls into other categories. SBA Loan Index publishes raw loan counts by metropolitan area and by congressional district, without a denominator. SparkMap, a mapping platform from the University of Missouri, shows combined 7(a) and 504 dollars per resident by county for 2019 through 2023. A syndicated series covered by local television stations ranked counties by 7(a) dollars per resident for fiscal 2023. A franchise-industry site computes franchise 7(a) loans per 1,000 CBP establishments, but only by state. LendingTree publishes average loan size per business by state.

Dollars per resident and loans per business are different measures with different weaknesses. A per-resident ranking rewards counties with a single large loan and few people. In the syndicated fiscal 2023 ranking, a small Kansas county topped the per-resident list on the strength of one loan, an illustration of how small numbers distort rankings. That is why the method section below emphasizes pooling years and setting size thresholds. We found no recurring public source that publishes both programs' loan counts per establishment for every county, split by urban and rural class.

Three meanings of "rural"

Any lending-desert story eventually becomes a rural story, and federal lending uses at least three definitions of the word.

The first is the USDA Economic Research Service's Rural-Urban Continuum Codes, or RUCC. The 2023 edition classifies 3,235 counties and county equivalents. Among the 50 states and the District of Columbia, 1,186 counties are metropolitan (codes 1 through 3) and 1,958 are nonmetropolitan (codes 4 through 9), with the nonmetro counties further sorted by the size of their largest urban population and by whether they sit next to a metro area. Another 89 are territorial. Codes 4 and 5 describe counties with an urban population of at least 20,000, codes 6 and 7 those with 5,000 to 20,000, and codes 8 and 9 those with fewer than 5,000. The 2023 edition raised the minimum size of an urban area from 2,500 to 5,000 residents, which moved hundreds of counties down the scale: counties coded 8, for example, roughly doubled to 466.

The second is SBA's own definition, which took effect with the fiscal 2027 fee notices. For loans approved from October 1, 2026 through September 30, 2027, SBA will waive both the upfront and annual fees on 504 loans to manufacturers, listed food-supply-chain businesses and businesses in rural areas, and will charge no upfront fee on 7(a) loans of $700,000 or less to those same categories. According to notices relayed by lenders and trade groups, SBA treats an area as rural if it lies in a county that the Census Bureau defines as at least 30 percent rural. That is a county-level test based on the share of the population outside Census-defined urban areas. Because many counties that ERS classifies as metropolitan contain large rural populations, SBA's rural set likely reaches well beyond the RUCC nonmetro counties. Some lenders were still describing the definition as pending as of early October.

The third is USDA's own test for the Business and Industry loan program, which looks at the project location, not the county. Eligible projects sit outside cities with populations over 50,000 and their contiguous urbanized areas.

These definitions can disagree about the same county, and a county that is rural for a fee waiver may not be rural for a USDA guarantee or for a continuum code. An article on lending deserts that says "rural" without naming one of them is making a claim that cannot be checked.

What independent evidence says about credit in rural places

The strongest recent evidence on rural credit gaps comes not from SBA data but from bank data. A September 2025 analysis by the Federal Reserve Bank of St. Louis used Community Reinvestment Act data on bank small-business loans of $1 million or less from 2019 through 2023. It found that bank small-business lending in persistent rural banking deserts fell 11 percent over the period, against 6 percent in persistent urban banking deserts. The bank's text describes the rural decline as about two and a half times the urban one, which does not match the chart's own labels, where 11 percent is about 1.8 times 6 percent. The analysis also found that persistent banking deserts declined less than areas that had gained or lost branches more recently, which cuts against a simple story that fewer branches automatically mean less lending.

Related work frames the scale of the problem. According to a Philadelphia Fed analysis summarized by the FedCommunities initiative, more than 1.1 million rural residents lived in census tracts without a bank branch within 10 miles in 2023, and on average they were about 20 miles from the nearest branch.

Survey data tell a similar story from the borrower's side. The Federal Reserve Banks' 2025 Small Business Credit Survey, which drew 6,525 responses from employer firms, found that 42 percent of firms that applied for financing received all of what they sought, 36 percent received part and 22 percent received none. Small banks had the highest full-approval rate at 57 percent. According to Third Way's analysis of the survey, 40 percent of rural small businesses received partial or no funding in 2024, and fewer than a quarter applied for an SBA loan.

CRA data and SBA data should not be mixed. CRA captures loans of $1 million or less from banks large enough to report, whether or not the loans carry any federal guarantee. SBA's public loan file captures guaranteed loans from banks, credit unions, nonbank lenders and community development companies, up to the program limits. A county can look well served on one and underserved on the other, and nonbank lenders are a large share of SBA volume that CRA does not see.

The USDA layer: fewer loans, larger checks

For the rural half of the lending-desert story, SBA data are only part of the picture. USDA's Business and Industry guaranteed loan program is the main federal program for larger rural projects, and its scale is very different.

According to a study prepared by Summit for the National Rural Lenders Association and submitted to a House Agriculture subcommittee in September 2025, core B&I obligations from fiscal 2015 through 2024 totaled $13.4 billion across 2,821 loans. Annual counts ranged from 176 to 354. The average guarantee grew from $2.75 million in 2015 to $8.1 million in 2024. In fiscal 2024, B&I made 176 loans, about a quarter of one percent of the 70,242 7(a) loans SBA approved that year, while its average guarantee of $8.1 million compares with an average 7(a) loan of roughly $443,000. The study was commissioned by an industry group, which is worth noting when citing it.

The guarantee terms changed in fiscal 2026. A Federal Register notice published March 9, 2026 set the B&I guarantee at 85 percent for loans under $5 million, up from 80 percent, with the guarantee remaining at 80 percent for loans of $5 million to $25 million. The initial fee is 3 percent for both bands, and the annual fee on the standard program is 0.55 percent.

The practical implication for a county map is that the two programs answer different needs. In a small rural county, a count of SBA loans measures access to Main Street credit: a restaurant, a gas station, a veterinary practice. A single B&I project can bring in more guaranteed dollars than every SBA loan in the county over several years. A county with no SBA loans in a given period may still host a $10 million rural project, so a desert on one map may be an oasis on the other. The honest picture needs both layers.

2026 changes the map

Several shifts this year affect where SBA dollars go, whatever the baseline.

The volume decline described above is concentrated in smaller loans and among larger lenders. According to Lumos, the $350,000-to-$500,000 band fell sharply in fiscal 2026, and the 25 largest lenders cut volume more steeply than the program as a whole.

SBA's operating rules tightened. SOP 50 10 8.1 took effect October 1, 2026 for loans receiving an SBA number on or after that date. It splits ownership-change deals into four categories, requires a quality-of-earnings report for acquisitions of $3 million or more, and revises underwriting for both standard 7(a) loans over $350,000 and small loans of $350,000 or less. American Banker described it as the start of the fiscal year with an even tighter credit box. For 504 loans, the minimum debt service coverage ratio rose to 1.15, according to the community development company SomerCor.

At the same time, limits loosened at the top. Since July 4, 2026, a borrower can hold up to $5 million in 7(a) loans and $5 million in 504 loans at once, a combined $10 million. That helps larger borrowers more than the small rural businesses at the center of the lending-desert debate.

The fiscal 2027 rural fee waivers push in the other direction. For a rural borrower getting a 7(a) loan of $700,000 or less, the upfront fee disappears. Whether that brings more rural borrowers to SBA lenders, or whether tighter credit standards offset it, will not be visible in loan-level data until late 2027.

How to measure a lending desert in your county

You do not need to wait for someone else's map. SBA's public loan file records each loan's project county and state by name, along with the approval date and fiscal year, and County Business Patterns publishes establishment counts for every county. Here is a method that avoids the traps described above.

Count loans by project location. Use the project county and state, not the borrower's mailing address and not the lender's headquarters. A national lender's home county would otherwise look flooded with loans that were made elsewhere.

Exclude cancelled approvals. Count only loans that were actually made. Keep 7(a) and 504 counts separate, because 504 funds on a different schedule.

Pool five years. Use fiscal 2021 through 2025 and divide by five. Single-year county counts are mostly noise: a county with 300 establishments at the national rate of about 8 per 1,000 would expect fewer than three loans a year, so one extra or missing loan changes its rate by about 40 percent.

Divide by CBP 2023 establishments for the county, and multiply by 1,000. The result is annual loans per 1,000 employer establishments. Compare it with the national range of roughly 6.7 to 9.3 for 7(a) alone, and say which end you are using.

Watch the name matching. The loan file lists counties by name, and names vary. Suffixes such as county, parish and borough appear inconsistently, and saint can be abbreviated. Virginia's independent cities share names with surrounding counties, Connecticut's counties were replaced by planning regions in Census data in 2022, and Alaska's boroughs and census areas have changed. A loan that fails to match is a loan that disappears from the numerator, which understates the rate.

Do not rank small counties. Below roughly 500 establishments, a rate is more a statement about chance than about access. Report such counties as a group, and shrink their rates toward their state or rural-class average before comparing.

Name your rural definition. Say whether you mean ERS nonmetro, SBA's 30 percent rural counties or USDA's 50,000-population test, and do not switch between them.

Add the USDA layer. For rural counties, a lending-desert finding on SBA data should be checked against USDA guarantee activity before it is called a desert.

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:

• SBA's fiscal 2025 7(a) volume appears as 77,600 loans in the September 2025 press release, 78,078 in the year-end report and 65,154 in the cleaned public loan file. They measure approvals, final approvals and loans actually funded.

• SBA's fiscal 2025 504 count of 6,750 differs from a vendor's 3,799. The difference likely reflects when 504 loans are funded, and the claimed collapse should not be repeated until it is reproduced.

• SBA described fiscal 2025 as 84,400 loans in September 2025 and as more than 85,000 later. The final 7(a) figure plus the 504 count gives about 84,800.

• "Small businesses" and "establishments" are not interchangeable. The commonly cited 33 million figure counts firms including nonemployers, while County Business Patterns counts 8.4 million employer establishments. Rates differ by a factor of four depending on which is used.

• Some published summaries put the fiscal 2026 B&I guarantee at 90 percent for loans under $5 million. The Federal Register notice sets 85 percent. The 90 percent rate applies to isolated Alaska projects off the road system.

• The St. Louis Fed's text says rural banking-desert lending fell about two and a half times as much as urban lending, but its chart shows declines of 11 percent and 6 percent.

• Per-resident dollar rankings are not per-business rates, and a county can top them on a single loan.

• "Rural" has three federal definitions, and SBA's new one is a county test based on the share of residents outside urban areas, not the USDA continuum code.

• One widely used reference site lists 23,848 lenders completing at least one 7(a) loan in fiscal 2024. Industry data counts about 1,100 active 7(a) lenders in 2026, down from about 1,400 a year earlier.

Sources:

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

• U.S. Small Business Administration, news release on fiscal year 2025 results, September 30, 2025

• U.S. Small Business Administration, 2025 Annual Report

• U.S. Small Business Administration, Information Notices 5000-881796 and 5000-881797, FY 2027 504 and 7(a) program fees

• U.S. Small Business Administration, Policy Notice 5000-879058 and SOP 50 10 8.1

• U.S. Census Bureau, County Business Patterns, 2023

• U.S. Census Bureau, Nonemployer Statistics, 2023 and 2024

• USDA Economic Research Service, Rural-Urban Continuum Codes, 2023

• USDA Rural Development, FY 2026 OneRD guarantee fee notice, Federal Register, March 9, 2026

• Federal Reserve Bank of St. Louis, Small Business Lending Trends and Banking Deserts, 2019-23, September 2025

• Federal Reserve Banks, 2026 Report on Employer Firms: Findings from the 2025 Small Business Credit Survey

• FedCommunities, reporting on Federal Reserve Bank of Philadelphia research on rural banking deserts

• National Community Reinvestment Coalition, Small Business Lending Deserts and Oases, 2014

• Summit, USDA B&I Guaranteed Loan Program: Economic Assessment 2025, prepared for the National Rural Lenders Association

• Third Way, 8 Findings: Small Businesses in Rural America

• American Banker

• NAGGL