Every 7(a) and 504 loan the SBA disclosed across three decades — 1.23 million approvals, geocoded to the county, resolved to the industry, and traced from origination to payoff or charge-off.
How much SBA lending has actually happened, and who got it?
Everything below is the evidence for those three sentences. You can read only this box and still follow the argument. Underlined terms open a plain definition — the full glossary has them all.
Four FOIA extracts, 1,233,253 loan records, one as-of date. The coverage is not symmetric between the two programs, and every comparison in this report is built to respect that.
7(a) coverage has a ten-year hole. The extracts cover FY1992–FY1999, FY2000–FY2009 and FY2020–FY2026. The FY2010–FY2019 file is not in this dataset, so nothing bridges FY2009 to FY2020 — and the only fully seasoned post-crisis cohorts sit in that gap.
504 records run through FY2026. FY2026 is a partial year — approvals from 1 October to 31 December 2025 only.
Any head-to-head comparison between the programs in this report is therefore computed on the matched window, FY1992–FY2009, and labelled as such. Plates 01–07 describe that window. Plate 08 adds the modern 7(a) era, FY2020–FY2026, which is a structurally different programme and is only 23.7% resolved.
How the numbers here were derived.
Loan status codes: PIF paid in full · CHGOFF charged off · CANCLD cancelled before disbursement · EXEMPT active, outcome not yet disclosed · NOT FUNDED approved but never funded. Performance rates use resolved loans only (PIF + CHGOFF) as the denominator; cancelled, not-funded and still-active loans are excluded.
7(a) is a working-capital guarantee written in volume at small ticket sizes. 504 is a fixed-asset debenture written in fewer, larger, longer, better-collateralised pieces. They are not variants of one another, and their performance says so.
The only apples-to-apples comparison this dataset supports. Both programs, same 18 fiscal years, same as-of date, both books effectively fully seasoned.
| Measure | 7(a) | 504 | Spread |
|---|---|---|---|
| Loans approved | 1,008,922 | 109,820 | 9.2× more 7(a) |
| Gross approvals | $189B | $52B | 3.6× |
| Average loan | $186,985 | $472,945 | 2.5× larger 504 |
| Median loan | $80,000 | $350,000 | — |
| Jobs supported | 5,692,193 | 2,146,580 | — |
| Jobs per loan | 5.6 | 19.5 | 3.5× more per 504 |
| Capital per job | $33,142 | $24,196 | — |
| Resolved loans | 879,607 | 87,397 | — |
| Charge-off rate (count) | 20.2% | 12.8% | 1.58× higher 7(a) |
| Charge-off rate (dollars) | 10.8% | 12.5% | — |
| Dollars charged off | $17B | $5.07B | — |
| Cancelled before funding | 123,705 | 17,769 | — |
Across the same eighteen years, 504 charged off at 12.8% of resolved loans against 7(a)'s 20.2% — while writing loans 2.5× larger and supporting 2.7× more jobs per loan.
Between FY1992 and FY2007, 7(a) approvals quadrupled while the average loan shrank by 42%. The program did not simply grow — it changed shape, moving from relationship-underwritten credit into high-throughput, low-documentation channels. The charge-off record follows that shift almost exactly.
Bars show how many loans were approved each year; the line shows the average loan size.
Look for the bars rising while the line falls — more loans were made, but each one got smaller.
Bars: loans approved. Line: average gross approval, right axis. Volume peaks in FY2007 at the same moment average ticket size bottoms.
Stacked bands show what share of loans each year went through each paperwork process.
Watch the fast, low-paperwork bands grow from a sliver to most of the chart over time.
Share of 7(a) approvals by processing method. Express and low-documentation channels go from nothing to the majority of the program in fifteen years.
Volume, average size and charge-off rate by processing method. The relationship between ticket size and loss is inverted from what most credit intuition expects — because channel, not size, is doing the work.
| Processing channel | Loans | Avg size | Charge-off |
|---|---|---|---|
| SBA Express Program | 388,214 | $52,074 | 24.4% |
| Preferred Lenders Program | 219,801 | $412,798 | 17.6% |
| 7a General | 148,104 | $316,022 | 14.0% |
| Low Documentation Program | 144,730 | $69,546 | 15.9% |
| Certified Lenders Program | 50,546 | $292,200 | 9.5% |
| Community Express | 39,475 | $27,822 | 42.5% |
Horizontal bars show how many loans fell into each loan-size bucket.
A long bar near the bottom (small loans) means the program mostly makes small-dollar loans.
41% of all 7(a) approvals were under $50,000.
Horizontal bars show the rate for loans grouped by how many years they were written for.
Longer bars mean higher ; compare short-term bars to long-term ones to see which lasts better.
Charge-off rate by original term. Short-term 7(a) paper — largely Express lines — carries four times the loss rate of 20-year real-estate paper.
The 504 debenture is never the whole deal. A certified development company takes the subordinate position, a private third-party lender takes the senior, and the borrower puts up equity. The file records the private side from FY2009 onward — and it shows the program mobilising more bank capital than it issues.
Bars show how many debentures were approved each year; the line shows the average project size.
Rising bars mean more loans overall, and a rising line means each project is getting bigger.
Bars: loans approved. Line: average gross debenture, right axis. FY2026 covers October–December 2025 only.
Three bars compare total debenture dollars, total private lender dollars, and the combined project debt.
A longer private-lender bar than the SBA bar shows the program pulls in more private money than it lends itself.
Third-party lender dollars are only recorded in the file from FY2009. Across the 121,120 projects where both sides are captured, every debenture dollar sits beneath $1.29 of private senior debt.
Top certified development companies by loan count, with the number of states each operates in and its charge-off rate on resolved loans.
| Certified development company | Loans | Debentures | States | C/O |
|---|---|---|---|---|
| CDC Small Business Finance Cor | 11,589 | $8.74B | 15 | 7.3% |
| Florida Business Development C | 9,273 | $5.93B | 21 | 13.0% |
| Empire State Certified Develop | 8,251 | $6.04B | 10 | 7.9% |
| Mortgage Capital Development C | 7,960 | $7.84B | 23 | 6.7% |
| Mountain West Small Business F | 6,537 | $3.91B | 16 | 8.2% |
| Florida First Capital Finance | 6,356 | $4.83B | 14 | 16.6% |
| Granite State Economic Develop | 6,319 | $2.82B | 10 | 8.3% |
| WBD, Inc. | 5,273 | $3.63B | 11 | 8.3% |
| Small Business Growth Corporat | 5,213 | $3.59B | 19 | 12.5% |
| Business Finance Group, Inc. | 4,561 | $2.81B | 11 | 7.7% |
60,838 504 loans — 27% of the file — are still coded active. A 20-year debenture approved in 2021 has neither paid in full nor charged off, so recent cohorts show artificially low resolved charge-off rates. Vintages after roughly FY2015 should be read as incomplete, not as evidence of superior credit. They are shaded accordingly in the next plate.
Every loan is geocoded to its project county — 99% of records matched. A tenth of counties absorbed three-quarters of the loans. Read this as a map of where capital went, not as a map of unmet demand: population is the denominator here, and Plate 07 sets out why that distinction matters.
7(a), FY1992–FY2009, against 2020 census population.
| State | Per 100k | Loans | Capital |
|---|---|---|---|
| 1Rhode Island | 964 | 10,576 | $1.23B |
| 2New Hampshire | 896 | 12,336 | $1.31B |
| 3Vermont | 840 | 5,399 | $769M |
| 4Montana | 769 | 8,342 | $1.34B |
| 5North Dakota | 628 | 4,890 | $755M |
| 6Utah | 569 | 18,629 | $2.37B |
| 7Idaho | 520 | 9,558 | $1.42B |
| 8Wyoming | 502 | 2,898 | $488M |
| 9Maine | 470 | 6,399 | $851M |
| 10South Dakota | 469 | 4,155 | $641M |
Same measure, bottom of the distribution. The spread between first and last is 7.3×.
| State | Per 100k | Loans | Capital |
|---|---|---|---|
| 1South Carolina | 132 | 6,778 | $1.52B |
| 2North Carolina | 147 | 15,331 | $2.87B |
| 3Tennessee | 156 | 10,784 | $2.08B |
| 4Alabama | 160 | 8,022 | $1.63B |
| 5Virginia | 171 | 14,735 | $2.74B |
| 6Kentucky | 183 | 8,237 | $1.31B |
| 7Indiana | 219 | 14,843 | $2.59B |
| 8Florida | 221 | 47,535 | $9.50B |
| 9West Virginia | 221 | 3,967 | $585M |
| 10Georgia | 235 | 25,126 | $6.91B |
| County | Loans | Capital |
|---|---|---|
| 1Los Angeles, CA | 44,505 | $10,368M |
| 2Harris, TX | 15,806 | $3,461M |
| 3Maricopa, AZ | 14,827 | $3,977M |
| 4Cook, IL | 14,562 | $2,147M |
| 5San Diego, CA | 13,213 | $3,061M |
| 6Orange, CA | 12,659 | $3,394M |
| 7Dallas, TX | 11,719 | $3,094M |
| 8Miami-Dade, FL | 9,204 | $1,241M |
| 9King, WA | 9,156 | $1,800M |
| 10New York, NY | 9,138 | $1,065M |
| 11Tarrant, TX | 7,616 | $1,962M |
| 12Salt Lake, UT | 7,391 | $1,051M |
| County | Loans | Capital |
|---|---|---|
| 1Los Angeles, CA | 9,907 | $9,573M |
| 2Orange, CA | 4,838 | $3,865M |
| 3San Diego, CA | 4,127 | $2,851M |
| 4Cook, IL | 3,352 | $2,330M |
| 5Maricopa, AZ | 3,192 | $2,394M |
| 6Miami-Dade, FL | 3,184 | $2,104M |
| 7Salt Lake, UT | 2,642 | $1,610M |
| 8Riverside, CA | 2,588 | $1,953M |
| 9San Bernardino, CA | 2,424 | $2,046M |
| 10Clark, NV | 2,366 | $1,584M |
| 11Alameda, CA | 2,313 | $1,887M |
| 12Santa Clara, CA | 2,186 | $1,968M |
353 U.S. counties received no 504 loan at all in thirty-four years. 2,198 of 3,142 counties — seven in ten — received fewer than twenty-five.
Sector concentration in SBA lending is extreme and stable. Restaurants, retail, auto repair, dentistry and childcare recur at the top of both programs across three decades — but the same industry can carry twice the loss rate under one program as under the other.
Each bubble is an industry sector: its left-right position is loan volume, its up-down position is , and its size is average loan size.
Bubbles high on the chart lost more money; bubbles that cluster apart by color show the two programs fund different, differently risky sectors.
Horizontal: loans approved. Vertical: charge-off rate on resolved loans. Bubble area: average loan size. Lime = 7(a), pale = 504. The two clouds barely overlap.
| Sector | Loans | Capital | Avg | Jobs/loan | C/O |
|---|---|---|---|---|---|
| Retail Trade | 154,739 | $27B | $175,839 | 4.5 | 24.7% |
| Other Services | 89,664 | $15B | $171,307 | 5 | 21.1% |
| Accommodation & Food Services | 86,834 | $22B | $251,845 | 12.3 | 24.6% |
| Professional & Technical Services | 76,896 | $10B | $134,150 | 5.6 | 19.0% |
| Construction | 74,751 | $9.58B | $128,218 | 6.9 | 23.7% |
| Manufacturing | 73,005 | $18B | $245,163 | 8.5 | 18.9% |
| Health Care & Social Assistance | 65,348 | $14B | $216,622 | 6.1 | 12.6% |
| Wholesale Trade | 52,044 | $11B | $207,654 | 5.1 | 22.7% |
| Administrative & Waste Services | 38,509 | $3.97B | $103,062 | 9.6 | 22.6% |
| Transportation & Warehousing | 25,605 | $2.94B | $114,713 | 7.1 | 26.5% |
| Arts, Entertainment & Recreation | 17,576 | $3.72B | $211,717 | 6.5 | 23.6% |
| Real Estate & Rental | 16,076 | $2.46B | $153,218 | 5.4 | 29.2% |
| Information | 13,110 | $1.78B | $135,887 | 6 | 24.6% |
| Finance & Insurance | 10,940 | $1.31B | $120,011 | 5.9 | 26.6% |
| Sector | Loans | Capital | Avg | Jobs/loan | C/O |
|---|---|---|---|---|---|
| Accommodation & Food Services | 28,468 | $26B | $910,584 | 16.1 | 19.4% |
| Health Care & Social Assistance | 27,981 | $17B | $597,136 | 10.6 | 6.7% |
| Retail Trade | 25,699 | $16B | $639,577 | 11.8 | 12.3% |
| Manufacturing | 25,578 | $20B | $790,894 | 20 | 9.4% |
| Other Services | 18,806 | $9.13B | $485,614 | 8.5 | 9.8% |
| Professional & Technical Services | 18,044 | $9.78B | $541,954 | 11.3 | 5.8% |
| Construction | 14,742 | $7.61B | $515,977 | 14.9 | 10.5% |
| Wholesale Trade | 13,934 | $11B | $807,831 | 14 | 6.4% |
| Arts, Entertainment & Recreation | 5,789 | $4.90B | $845,966 | 14.3 | 16.3% |
| Real Estate & Rental | 5,129 | $3.77B | $734,482 | 11.9 | 12.5% |
| Administrative & Waste Services | 4,971 | $2.50B | $502,675 | 15.7 | 7.2% |
| Transportation & Warehousing | 4,494 | $4.06B | $903,894 | 16.1 | 7.4% |
| Finance & Insurance | 3,406 | $1.52B | $445,160 | 9.2 | 13.1% |
| Educational Services | 2,663 | $1.97B | $740,033 | 12.2 | 9.1% |
Where a sector appears in both programs, the 504 structure loses less — in fifteen of sixteen sectors. The gap is widest where a 7(a) borrower typically has little to pledge: transportation and warehousing (26.5% vs 11.8%), wholesale trade, administrative services. It closes and then inverts in exactly one place — accommodation and food services, where 504 charges off at 27.3% against 7(a)'s 24.6%. Hotels and restaurants are the one asset class where the building does not save the loan.
Six-digit NAICS, ranked by loan count.
| Industry (NAICS) | Loans | Avg | C/O |
|---|---|---|---|
| 1Full-Service Restaurants | 36,594 | $218,277 | 25.5% |
| 2Limited-Service Restaurants | 26,412 | $193,169 | 26.7% |
| 3General Automotive Repair | 17,577 | $152,718 | 17.5% |
| 4Offices of Dentists | 15,586 | $282,779 | 6.2% |
| 5Child Day Care Services | 12,275 | $262,131 | 16.3% |
| 6Beauty Salons | 11,819 | $88,033 | 23.9% |
| 7Offices of Chiropractors | 10,516 | $118,355 | 15.8% |
| 8Gasoline Stations with Convenience Stores | 10,476 | $545,316 | 21.2% |
| 9Landscaping Services | 10,395 | $87,988 | 18.7% |
| 10Drycleaning and Laundry Services (except Coin-Operated) | 10,369 | $199,177 | 21.3% |
| 11Beer, Wine, and Liquor Stores | 8,934 | $256,449 | 14.1% |
| 12Fitness and Recreational Sports Centers | 8,175 | $184,641 | 24.2% |
| 13Offices of Physicians (except Mental Health Specialists) | 7,929 | $232,642 | 9.7% |
| 14Electrical Contractors | 7,835 | $130,005 | 20.1% |
Six-digit NAICS, ranked by loan count.
| Industry (NAICS) | Loans | Avg | C/O |
|---|---|---|---|
| 1Full-Service Restaurants | 10,278 | $613,152 | 20.6% |
| 2Hotels (except Casino Hotels) and Motels | 10,080 | $1,479,979 | 21.9% |
| 3Offices of Dentists | 5,904 | $488,271 | 3.0% |
| 4Offices of Physicians (except Mental Health Specialists) | 5,835 | $679,911 | 6.3% |
| 5Child Day Care Services | 5,331 | $661,886 | 9.9% |
| 6General Automotive Repair | 4,387 | $373,673 | 7.2% |
| 7Limited-Service Restaurants | 4,252 | $558,803 | 12.9% |
| 8Offices of Lawyers | 3,608 | $522,405 | 5.2% |
| 9Gasoline Stations with Convenience Stores | 3,323 | $580,684 | 18.1% |
| 10Car Washes | 3,157 | $725,323 | 15.5% |
| 11Automotive Body, Paint, and Interior Repair and Maintenance | 2,349 | $477,451 | 7.5% |
| 12Veterinary Services | 2,300 | $535,545 | 2.9% |
| 13Fitness and Recreational Sports Centers | 2,262 | $847,344 | 11.7% |
| 14All Other Specialty Trade Contractors | 2,129 | $565,380 | 12.0% |
Five-year eras. Retail's share of 7(a) erodes steadily as professional, health and administrative services rise.
$23.0 billion was charged off across both programs. The variance is not random and it is not mainly about the borrower's industry — it is about vintage, channel, term and, above all, which institution wrote the loan.
A line for each program tracks the eventual of loans made in each year.
Look for a peak year where losses were highest — that tells you which of loans turned out worst, regardless of program.
Share of each origination cohort's resolved loans that ended in charge-off. Both programs peak on the FY2007 vintage. Shaded region marks 504 cohorts that are not yet seasoned enough to read.
Horizontal bars show the rate for loans grouped by dollar size.
Longer bars mean that size group loses money more often; compare small versus large loan bars.
Horizontal bars show the rate for loans grouped by dollar size.
Longer bars mean that size group loses money more often; compare small versus large loan bars.
Horizontal bars show the rate for each type of borrower entity (like corporation or sole proprietor), for both programs.
Longer bars mean that kind of borrower more often; compare across entity types and across programs.
Top lenders by loan count. Same program, same guarantee, same era — and a three-fold spread in charge-off rate between the best and worst large originators. Bank of America wrote the most loans in the file at an average of $62,500 each; Wells Fargo wrote 27% fewer loans for 2.7× the dollars.
| Lender | Loans | Capital | Avg size | States | Charge-off |
|---|---|---|---|---|---|
| 1Bank of America, National Association | 98,847 | $6.18B | $62,547 | 52 | 26.9% |
| 2Wells Fargo Bank National Association | 72,036 | $16B | $228,740 | 54 | 16.7% |
| 3JPMorgan Chase Bank, National Association | 51,674 | $5.31B | $102,735 | 48 | 27.1% |
| 4Citizens Bank, National Association | 43,692 | $2.07B | $47,267 | 31 | 19.9% |
| 5U.S. Bank, National Association | 42,516 | $8.01B | $188,288 | 45 | 18.1% |
| 6PNC Bank, National Association | 40,649 | $6.37B | $156,813 | 46 | 20.1% |
| 7Bank of Hope | 31,434 | $4.44B | $141,231 | 52 | 41.5% |
| 8Capital One, National Association | 21,236 | $1.14B | $53,895 | 51 | 42.1% |
| 9Zions Bank, A Division of | 20,622 | $3.78B | $183,133 | 46 | 18.6% |
| 10Manufacturers and Traders Trust Company | 19,098 | $2.73B | $142,970 | 34 | 13.8% |
| 11Readycap Lending, LLC | 18,427 | $8.95B | $485,742 | 51 | 32.0% |
| 12The Huntington National Bank | 16,880 | $3.17B | $187,522 | 33 | 15.7% |
| 13Truist Bank | 15,714 | $3.38B | $215,340 | 45 | 14.9% |
| 14KeyBank National Association | 14,394 | $2.24B | $155,501 | 32 | 12.9% |
| 15TD Bank, National Association | 14,082 | $3.16B | $224,323 | 40 | 14.2% |
Banks taking the first-lien position beside the debenture, FY2009 onward.
| Third-party senior lender | Projects | Debenture beside it | Avg |
|---|---|---|---|
| JPMORGAN CHASE BANK, NATIONAL ASSOCIATION | 5,452 | $4.12B | $754,975 |
| BANK OF AMERICA, NATIONAL ASSOCIATION | 3,410 | $3.13B | $917,549 |
| ZIONS BANK, A DIVISION OF | 2,916 | $1.95B | $669,139 |
| WELLS FARGO BANK NATIONAL ASSOCIATION | 2,760 | $2.63B | $953,601 |
| BMO BANK NATIONAL ASSOCIATION | 1,737 | $1.74B | $1,001,440 |
| FIRST-CITIZENS BANK & TRUST COMPANY | 1,427 | $1.23B | $865,383 |
| TRUIST BANK | 1,323 | $1.02B | $773,299 |
| U.S. BANK, NATIONAL ASSOCIATION | 1,164 | $1.10B | $942,756 |
| GLACIER BANK | 1,133 | $780M | $688,107 |
| PNC BANK, NATIONAL ASSOCIATION | 1,021 | $662M | $648,192 |
| COLUMBIA BANK | 977 | $846M | $866,383 |
| REGIONS BANK | 957 | $741M | $774,460 |
FY2020–FY2026 records, as of 30 June 2026. 317,155 funded loans and $171.4 billion — nearly matching the entire FY1992–FY2009 book in dollars on a third of the loan count. Everything the preceding plates describe about 7(a)'s shape has reversed.
Only 23.7% of the modern book has resolved, against 99.4% of the legacy book. Median time from approval to charge-off is 4.2 years and 44% of ultimate loss emerges by year five. Cumulative charge-off currently stands at 0.52% of balance and will rise by a large multiple. Use these rates to rank, not to price.
| Measure | FY1992–FY2009 | FY2020–FY2026 |
|---|---|---|
| Funded loans | 885,217 | 317,155 |
| Capital approved | $164B | $171B |
| SBA-guaranteed share | 72.8% | 75.5% |
| Average loan | $185,099 | $540,285 |
| Median loan | $80,000 | $200,000 |
| Median term | 7 years | 10 years |
| Jobs supported | 5,012,419 | 3,320,685 |
| Jobs per loan | 5.7 | 10.5 |
| Sold into secondary market | 25% | 37% |
| Median initial rate | not recorded | 9.00% |
| Share resolved | 99.4% | 23.7% |
| Charge-off rate (resolved) | 20.2% | 9.2% · unseasoned |
| Loss as share of balance | 10.66% | 0.52% · unseasoned |
Bars show how many modern-era loans were approved each year; the line shows the average loan size.
A climbing line means loans are getting bigger; compare the shape to the earlier era's chart to see how the program has changed.
FY2026 covers October 2025 through June 2026 only. The rate environment is visible in the loan-size line.
Top national 7(a) lenders by loan count, FY2020–FY2026. The pre-crisis roster of money-centre banks has been replaced by SBA specialists and non-bank lenders.
| Lender | Loans | Capital | Avg | States | C/O |
|---|---|---|---|---|---|
| 1The Huntington National Bank | 35,040 | $9.54B | $272,403 | 51 | 8.7% |
| 2U.S. Bank, National Association | 14,227 | $3.57B | $250,959 | 50 | 9.4% |
| 3TD Bank, National Association | 14,172 | $2.18B | $154,112 | 26 | 32.4% |
| 4Wells Fargo Bank National Association | 10,263 | $3.02B | $294,113 | 50 | 25.6% |
| 5Manufacturers and Traders Trust Company | 9,634 | $1.30B | $135,258 | 22 | 23.6% |
| 6Live Oak Banking Company | 9,024 | $13B | $1,385,691 | 52 | 2.6% |
| 7Readycap Lending, LLC | 8,680 | $3.79B | $436,195 | 51 | 10.9% |
| 8BayFirst National Bank | 8,659 | $1.61B | $185,719 | 52 | 29.6% |
| 9JPMorgan Chase Bank, National Association | 7,376 | $1.92B | $260,426 | 48 | 11.4% |
| 10Newtek Bank, National Association | 7,285 | $2.93B | $401,835 | 52 | 0.4% |
| 11Northeast Bank | 5,958 | $852M | $143,077 | 51 | 16.1% |
| 12Celtic Bank Corporation | 5,225 | $3.54B | $678,037 | 53 | 7.9% |
| 13United Midwest Savings Bank National Association | 4,846 | $1.63B | $336,743 | 51 | 26.0% |
| 14KeyBank National Association | 4,167 | $1.61B | $386,788 | 45 | 11.6% |
Low-documentation and Community Express are gone. Preferred Lenders is now the majority of the programme — which, on the FY1994–FY2005 evidence, is the better-performing structure.
| Channel | FY1992–2009 | FY2020–2026 | Modern C/O |
|---|---|---|---|
| Preferred Lenders Program | 21.6% | 55.2% | 6.1% |
| SBA Express Program | 38.6% | 37.0% | 12.2% |
| 7a General | 14.6% | 6.2% | 6.0% |
| Community Advantage Initiative | — | 0.7% | 28.3% |
| Working Capital CAPLine | — | 0.3% | 0.5% |
| International Trade Loans | 0.4% | 0.3% | — |
| Low Documentation Program | 14.1% | discontinued | — |
| Certified Lenders Program | 5.0% | discontinued | — |
The modern file records business age as a category. It discriminates sharply, and in the direction underwriting intuition expects.
| Borrower stage | Loans | Avg loan | Jobs/loan | C/O |
|---|---|---|---|---|
| Existing or more than 2 years old | 174,220 | $444,080 | 10.3 | 8.5% |
| New Business or 2 years or less | 52,532 | $468,063 | 8 | 13.8% |
| Startup, Loan Funds will Open Business | 51,565 | $542,539 | 11.6 | 9.5% |
| Change of Ownership | 34,978 | $1,150,703 | 13.9 | 4.5% |
Average loan size rose from $185,099 to $540,285, median term from seven years to ten, and jobs per loan from 5.7 to 10.5. The programme that spent the 2000s chasing volume at $50,000 a loan now writes half-million-dollar credit — and it has not yet been tested through a downturn.
Read together, the geography, the industry mix and the loss record describe a delivery system whose reach is determined less by where demand is than by where intermediaries already are.
The top 10% of U.S. counties absorbed 74% of all 7(a) loans and 76% of all 504 debentures. 353 counties received no 504 loan at any point in thirty-four years, and 2,198 of 3,142 received fewer than twenty-five. The Gini coefficient of 7(a) loans across counties is 0.813. Rhode Island borrowers received 964 loans per 100,000 residents against South Carolina's 132, under identical eligibility rules.
Those facts are not in dispute. The inference usually drawn from them is. Every figure above divides lending by population, which assumes each resident is equally likely to need business credit. Normalised instead by measured entrepreneurial activity, apparent access deficits largely close: across the 51 state geographies the correlation between 7(a) lending per capita and the rate of new entrepreneurs is -0.085, against new employer business formation 0.209, and against young-firm job creation 0.279. See the companion Kansas–Missouri analysis, where the same state ranks 29th of 51 on a population denominator and 15th on a formation denominator.
The defensible reading is narrower: capital is highly concentrated, and it tracks where businesses already exist rather than where people are. Whether that constitutes withheld credit or absent formation cannot be settled from lending data alone — and for most of the country the entrepreneurship data points to the second.
Over the matched FY1992–FY2009 window, 504 wrote loans 2.5× larger than 7(a), supported 3.5× more jobs per loan, and charged off at 12.8% against 7(a)'s 20.2%. Larger, longer, collateral-backed credit performed better — not worse.
The conventional read is that 504's real-estate collateral explains the gap. That is part of it. But the deeper difference is structural: a CDC underwrites a project with a private senior lender already committed beside it, so two independent credit judgements have to agree before the debenture issues.
Between FY1992 and FY2007 approvals rose 4.2× while average loan size fell from $248,591 to $143,487. The growth came almost entirely from streamlined channels: SBA Express wrote 388,214 loans at an average of $52,074, and Community Express wrote 39,475 at $27,822.
Those channels charged off at 24.4% and 42.5% respectively, against 14.0% for conventionally processed 7(a) General loans. The program bought reach with documentation, and the price appears in the loss column with a five-year lag.
The FY2007 7(a) cohort charged off at 37.1% — four times the FY1993 cohort. No industry, size band or legal structure in this dataset produces a spread that wide. When a loan was written matters more than almost anything about who it was written to.
504 shows the same curve at roughly half the amplitude, peaking at 27.1% on the same vintage. Both programs were underwriting into the same asset bubble; one was simply more exposed to it.
Among the fifteen largest 7(a) originators, charge-off rates on resolved loans range from 12.0% to 42.1% — a 3.5× spread on an identically structured, identically guaranteed federal product. Origination practice, not program design, is doing that work.
Volume and dollars also decouple sharply. Bank of America appears 98,847 times in the file for $6.18B; Wells Fargo appears 72,036 times for $16B. They are running different businesses under the same program name.
Full-service restaurants are the single most-funded industry in both programs — 36,594 7(a) loans and 10,278 504 debentures. Adding limited-service restaurants, hotels, auto repair, dentists' offices and childcare describes the recurring core of SBA lending across three decades.
Sector risk within that core is not uniform. In 7(a), real estate and rental charged off at 29.2% while health care charged off at 12.6% — the same program, a 2.3× difference in outcome.
Where the file records both sides of the transaction (FY2009 onward), $98B of debentures sat beneath $127B of private senior debt across 121,120 projects — a ratio of 1.295:1. The federal exposure is the subordinate slice of a much larger private commitment.
That leverage is the argument for the program and the reason its geography matters so much. Where no CDC operates, the private capital that would have been mobilised alongside it does not simply arrive by another route.
Records analysed: 1,233,253 · county match rate 99.4% (7(a)) / 99.1% (504) · population denominators from the 2020 Decennial Census · county boundaries from US Census / Atlas, Albers USA projection · dollar figures are nominal, not inflation-adjusted.
This is document 1 of 5 in the reading path. Back to the Research index.