34 Industries Passed Every Test. Four Questions Split Them.
redactedThe 34 industries that pass every test, by name. And four questions that set them apart. Part four of five. In part one I gave a number. 92,986 businesses pass every test I set. I named five of the 34 industries they sit in, and promised the rest. Here are all 34. I put four questions to each one. How safe have its loans been? How often does it change hands with an SBA loan? Is it growing? And what does it cost to finance? No industry leads on all four. The one that comes closest is a special case, and I say why. Checking this piece turned up a weakness in my own safety test. A reviewer then caught an error in my first attempt to measure it. Both are in the piece. 34 industries pass every test in this series. They hold 92,986 businesses with 20 to 999 staff. This piece names all 34 and puts four questions to them. How safe have their loans been? How often do they change hands with an SBA loan? Are they growing? And what do they cost to finance? No industry leads on all four. Hotels come closest, and hotels are a special case. 1. 34 industries, named Piece 1 counted 92,986 businesses that pass every test we set, in 34 industries. It named five of them and promised the rest. Here they are. That is the count on the licence screen as built. Section 11 gives two more that a fuller matching would add: sports and recreation instruction, and software publishers. redacted† Below the all-industry rate of 7.13%, but by no more than chance could explain on that many loans. The first column counts businesses with 20 to 999 staff. The other two cover businesses of every size. The charge-off rate is on loans of every purpose, and each change-of-ownership loan counts as one sale. Among sales reporting 20 or more jobs, the 34 change hands at 1 in 457 a year, as piece 1 said, or 1 in 529 counting each borrower once. What the 34 share is the screen. Our licence screen flags none of them. Each holds at least 1,000 businesses with 20 to 999 staff. And SBA-guaranteed loans to each have been charged off less often than the all-industry rate of 7.13%. Charged off means the SBA wrote the loan off as a loss, and its file of 30 June 2026 still shows it so. Section 3 says what that leaves out. Together their loans were charged off at 5.60%, on 66,697 loans. For 18 of them, the gap to the average is more than chance could produce. The other 16 are marked in the table. For those, the gap could be chance. One of the 18, commercial printing, is clear only on loans made since fiscalredactedSection 3 explains. They are a mixed lot. Six are in construction, and hold 15,183 businesses. Four are in administrative and support services, with 14,434. Five are repair shops and personal services. Four are manufacturers, four are shops and three are wholesalers. And they are concentrated. The five largest hold 36,573 of the 92,986, or 39%. 18 of the 34 hold fewer than 2,000 businesses each. So a few big industries carry much of the count, and many of the others are far smaller. In short 34 industries pass every test, and they hold 92,986 businesses. Their loans were charged off at 5.60%, against 7.13% for all industries. 18 are below that rate by more than chance. For the other 16 the gap could be chance. The five largest hold 39% of the businesses. That is the count on the SBA's file of 30 June 2026, and section 3 gives five others. What these numbers count. Every sale in this series is an SBA-financed sale. The Small Business Administration (SBA) releases loan-by-loan records for its two lending programmes, 7(a) and 504. They are the only free, public, deal-level record of business sales that we know of. Our best reading of Census data is that the SBA finances about a tenth of business sales. So every count of sales here is a floor, and "sells" means "sells with an SBA loan". Industries lean on SBA lending to different degrees, and nothing public that we know of shows what share of each industry's sales it finances. Where that matters most, we say so again. 2. What separates one from another Every one of the 34 has already passed the size bar and the risk bar. So those bars no longer tell them apart. Four further questions do. 1. How safe, exactly? The loans of some are charged off a third as often as the loans of others. 2. Does it change hands with an SBA loan? A business that is never sold cannot be bought. 3. Is it growing? A screen built on past loans cannot see where an industry is heading. 4. Can it be financed at its price? A 7(a) loan is $5 million at most, and the 7(a) rules have just tightened. One result is worth knowing first. Among the 34 we found no link between safety, size or growth and how often an industry sells with an SBA loan. 34 industries are too few to rule out a modest one. We measured each pair on a scale where 0 means no connection and 1 means they move in lockstep. Among the 34, charge-off rate against sale rate comes to −0.05. Businesses with a bench against sale rate comes to −0.17. Growth in larger locations against sale rate comes to −0.03. With 34 industries, a link has to pass about 0.34 either way to be more than chance. In short Four questions separate the 34: how safe, how often sold with an SBA loan, growing or not, and what it costs to finance. Among the 34 we found no link between safety, size or growth and how often an industry sells with an SBA loan. 34 industries are too few to rule out a modest one. 3. How safe, exactly? All 34 are below 7.13%. They are not equally far below it. And the rate does not mean the same thing for each. Hotels and motels have the lowest rate, 2.36% on 7,706 loans. Then come sheet metal work at 3.66% and other plastics products at 4.31%. Golf courses and country clubs are at 4.41%, and site preparation contractors at 4.67%. These are the industries on the list whose loans were charged off least often. At the other end, five industries are within a quarter of a point of the line. They are electrical equipment wholesalers (6.94%), car parts stores (7.01%), other specialty trade contractors (7.04%), masonry contractors (7.10%) and other support services (7.10%). They are below the line, but only just. For those five and 11 more, the gap to the average could be chance. A stricter test that kept only the 18 clear cases would leave 53,552 businesses, not 92,986. And no two neighbours in the order differ by more than chance. So read the list as a rough sort, not a ranking. One of the 18 needs a note: commercial printing. The loan file changed its printing codes around fiscalredactedOur rule counts a loan only where its code is in the 2017 industry list that Census uses. For commercial printing that leaves out 900 loans made under seven older codes. Add those older loans back and its rate is higher. With them its rate is 6.27% on 2,312 loans, not 5.67% on 1,412. That is still below the line, but within chance of it. Counted that way, 17 industries are clear, holding 51,542. So with its older loans counted, commercial printing drops out of the clear group. What counts as charged off? The count itself needs a note, and a reviewer of this piece found it. Our charge-offs come from the SBA's file of 30 JuneredactedThat file dates a charge-off only while the loan is still marked charged off. Once such a loan is closed as paid in full, the date is gone. The SBA's file of 31 March 2026 still dated 3,566 loans that it marked paid in full. So our count leaves out a loan that was charged off and then closed as paid in full. Count a charge-off that either file dates, and the rate is higher. Then 7.89% of the loans were charged off, not 7.13%. That needs the two files paired loan by loan, and the full version gives the method. We cannot say what paid in full means after a charge-off: repaid, settled or only closed. The SBA says that after a charge-off it refers all eligible parties to the Treasury for further collection. The dropped ones are mostly small. They are 17.3% of the charge-offs on loans of $50,000 or less, and 0.5% of those on loans over $2 million. So the June count leaves out a far bigger share of the charge-offs on small loans than on large ones. The March file also shows the status of open loans. The SBA's data dictionary says it withholds those, and the June file does. One March status is PURCH(NOT C/O). We read it as a loan the SBA has purchased under its guarantee and not charged off. How many purchased loans are there, and what becomes of them? There were 4,694 of them, beside 37,252 dated charge-offs. 82% had been sold on the secondary market, against 27% of all loans. By June, 331 were charged off and 78 were marked paid in full. The SBA's own guidance gives three ways such a loan is resolved: returned to regular servicing, paid in full, or charged off. Not every purchased loan ends as a charge-off. So we ran the risk test on five ways of counting a failed loan. The risk test on five ways of counting a failed loan • The SBA's 30 June 2026 file, as this series counts: All-industry rate 7.13%, Industries that pass 34, Businesses with a bench 92,986 • Its 31 March 2026 file, read as the June file reads: All-industry rate 7.03%, Industries that pass 32, Businesses with a bench 90,573 • The March file, every dated charge-off: All-industry rate 7.78%, Industries that pass 34, Businesses with a bench 103,248 • A charge-off dated in either file: All-industry rate 7.89%, Industries that pass 36, Businesses with a bench 105,417 • The March file, dated charge-offs and purchased loans: All-industry rate 8.76%, Industries that pass 37, Businesses with a bench 107,943 Each row applies the same test to its own count. That is 300 or more loans, a rate below that count's all-industry rate, and 1,000 or more businesses with a bench. The second row keeps a March date only where the March status is CHGOFF or CLSLN. The June file still dates all but 89 of those. The fourth row needs the two files paired loan by loan. Not every purchased loan ends as a charge-off, so the fifth row is an upper reading. The full version gives the method. Most of the list passes on each of these counts. 30 of the 34 pass on all five, and all 18 of the clear ones are among them. Four pass on some counts and not others: masonry contractors, car parts stores, other support services, and bars. And the list can grow as well as shrink. Four industries outside the 34 pass on at least one: new car dealers, recyclable material wholesalers, computer systems design, and drywall contractors. So the list runs from 32 to 37 industries. Read the 34 as the count on one file, read one way. The businesses move more than the industries: from 90,573 to 107,943. New car dealers are 11,361 of that. On the June count they miss the line by one loan in 390. So a big industry that only just misses the line can move the number of businesses a long way. And a sixth count is possible. Add the 2,666 loans the March file marks as in liquidation, and 37 pass, holding 108,108 businesses. Electrical equipment wholesalers then fail as well, by one loan, so 29 of the 34 pass on all six. That count is a looser upper reading. By June, 197 of those loans were charged off and 122 were marked paid in full. On the either-file count the 34 together are at 6.13%, not 5.60%. And 19 are clearly below the line, not 18. Their own rates rise by up to 1.4 points. Custom computer programming goes from 5.42% to 6.79%. Hotels barely move, from 2.36% to 2.37%. So the wider count lifts these rates unevenly, from a good deal to hardly at all. The rest of this piece keeps the June count, as pieces 1 to 3 do. We re-ran the comparisons below, and those in section 8, on the either-file count. None changes direction. With purchased loans counted, some things do change: findings in section 8, and which industries sit above their like loans. Each is said where it arises. Safe against what? A charge-off rate depends on the kind of loan as much as the kind of business. As a rule, a 7(a) loan may run more than 10 years only where it pays for property or long-lived equipment. And long loans fail less often. How much less is hard to read off the file. An earlier draft of this piece got it wrong, and a reviewer caught it. The file rewrites a loan's length after the loan fails. It then shows roughly the months that were left. So a failed 25-year loan looks like a shorter one. The file's own counts show the rewriting at work. Of the loans recorded at exactly 25 years, 15 of 50,701 were charged off. Of those recorded at just over 21 to just under 25 years, 601 of 3,499 were. So almost no failed loan is still recorded at the round figure, and far more are recorded just under it. So we rebuilt each failed loan's first length from its dates, two ways. One takes the longest standard length that the dates allow. The other takes the shortest. The truth should lie between them. The two are far closer on loans of 20 years or more than on shorter ones. So we split loans at 20 years, and nowhere else. redactedLike loans are other industries' 7(a) loans in the same band of size, and on the same side of the 20-year line. There are six bands of size. A rebuilt length is an estimate. Hotels' shorter loans rest on 32 to 43 failures. The full version gives the method. On either rebuild, long loans fail far less often than short ones: about 2% against about 8%. And about 88% of hotel loans run 20 years or more, where about 16% of all loans do. So hotels' 2.36% is mostly a long-loan rate. Set beside other long loans, hotels did no better. Their long loans were charged off at 2.1% to 2.2%. All long loans ran at 1.9% to 2.4%. That is level. The SBA's other programme gives a cleaner test. 504 loans finance property and long-lived equipment, or debt on them. And that file keeps each loan's first length. So here a loan's length can be read straight off the file. Under that programme, hotel loans were charged off more often than the rest. Take the 504 loans approved in fiscal years 2010 toredactedof 2,744 hotel loans were charged off, or 2.59%. For every other 504 loan it was 1.50%. At the hotels' mix of length, size and year, the others would have run at 1.45%. That gap is more than chance. So what does the risk test in this series reward? Partly safe businesses, and partly long loans. The test stands as piece 1 defined it, and all 34 pass it. But hotels' low rate goes with the length of their loans. Set beside other long loans they did no better, and under 504 they did worse. The 34 as a whole hold up. Set each beside like loans: other loans in the same band of size, on the same side of the 20-year line. 31 to 32 of the 34 were charged off less often than those like loans, 15 to 17 of them by more than chance. So most of the list did better than their like loans, not just better than the average. Two or three were charged off more often than their like loans, but none by more than chance. They are car washes and car parts stores on both rebuilds, and bars on one. And which ones are above depends on the bands. Hotels themselves sit a little below their like loans. Their rate is 2.36%, against 2.55% to 2.88% for those loans. That is on charge-offs. Count purchased loans too, and most of the list still holds. Bars are then above their like loans by just more than chance on both rebuilds, and car washes on one. Hotels still sit a little below theirs: 3.96% against 4.20% to 4.83%. And 31 of the 34 are still below. Three other things point the same way for hotels. Half of the loans to buy a hotel are for $2.7 million or more. 1 hotel sale in 9 went through 504, which the SBA describes as financing for "major fixed assets". And the SBA's rules give as examples "storage units and hotels where the real estate fully secures the loan". Each is a sign of a loan that rests on property. In short The 34 run from 2.36% for hotels to 7.10% for other support services. 18 are clearly below the all-industry rate, and for the other 16 the gap could be chance. Hotels' low rate is mostly a long-loan rate: their long loans did no better than other long loans, and under 504 hotels did worse. On rebuilt loan lengths, 31 to 32 of the 34 were charged off less often than like loans, 15 to 17 of them by more than chance. The count behind all of this is the SBA's 30 June 2026 file's. On four other ways of counting a failed loan, 32 to 37 industries pass, and 30 of the 34 pass on every one. On a sixth, which adds loans in liquidation, 29 do. 4. Does it change hands with an SBA loan? This is the test in the title. It counts sales made with an SBA loan, and the 34 differ on it widely. redactedHotels are at the top: 1 hotel in 153 changes hands with an SBA loan in a year. For hardware stores it is 1 in 272, and for car washes 1 in 310. At the other end it is 1 in 4,839 for temporary staffing firms, and 1 in 5,060 for custom computer programming firms. So the top and the bottom of the list are very far apart. A fast industry among the 34 sells about 8 times as often as a slow one. Fast and slow here mean a tenth of the way from the top and from the bottom. Chance alone would give 1.2 times. So the gap is not chance. Much of it goes with how heavily each industry uses SBA loans at all. Rank the 34 by sale rate and by seasoned loans per business. Seasoned loans are 7(a) loans approved in fiscal years 2010 to 2019, of every purpose. The two orders line up at +0.73, on a scale where 0 means no connection and 1 means they move in lockstep. So among these industries, the heavier users of SBA loans tend to change hands with an SBA loan more often. Allow for that, and the gap shrinks but does not close. Count sales per seasoned loan instead of per business, and the gap falls from 8.2 times to 4.9. Chance alone would give 1.2. So about half of the gap beyond chance goes with SBA use, counted in times. On a ratio scale it is about 28%. Either way, much of the gap remains. A cleaner yardstick gives much the same answer. Seasoned loans include some loans to buy a business. So we also tried a yardstick that leaves those out: the other 7(a) loans approved in fiscal years 2020 toredactedOn that count the link is +0.64 and the gap falls to 4.6 times. That too is about half of the gap beyond chance, or about 30% on a ratio scale. Each loan counts as one sale, as in piece 1. But some sales take two loans. Nearly 1 in 5 of the 7(a) acquisition loans is one of two or more to the same borrower within 90 days. Counted once, the 34 had about 6,200 sales, not 6,825. They would change hands at 1 in 783 a year, not 1 in 710. So a count by loan shows somewhat more sales than a count by borrower. The order of the 34 barely moves. Services for the elderly and disabled move most, from 1 in 1,180 to about 1 in 1,670. So who is fast and who is slow stays much the same. The SBA's rules also treat one partner buying out another as a change of ownership. So some of these loans may be buyouts, not sales. The file cannot separate them. Half of the 34 change hands with an SBA loan less often than all in-scope industries together. In scope means past the licence fork, and not one of the 25 finance, property and gambling industries that pieces 1 and 2 leave out. Take all 741 in-scope industries together, and 1 business in 946 changes hands with an SBA loan in a year. 17 of the 34 are slower than that. So are most of the other industries. 464 of the other 707 are slower too, or 66%. Sales pile up in a few industries, so most industries sit below the pooled rate. Being slower than that rate is the normal case. That pooled rate leaves a few SBA-financed sales out. It counts only the sales the loan file places in one industry. With the 98 it cannot place, which pieces 1, 2 and 3 count, it is 1 in 942. So adding them barely moves it. The 34 lean on a few industries too. Hotels are 1,754 of the 6,825 sales, or 26%. Add general car repair, other specialty trade contractors and landscaping, and four industries supply 55%. Without hotels the 34 change hands at 1 in 903 a year. So a large part of the group's SBA-financed sales rests on hotels alone. Do the 34 sell more often with SBA loans than the rest? Against every other industry taken together, the 34 lead, but the lead is within chance. Set them against the other 707, and they sell 1.5 times as often: 1 in 710 against 1 in 1,056. To see whether that is more than chance, we dealt the 34 labels at random among the 741 industries, 200,000 times. We call a lead more than chance when it turns up in fewer than 1 try in 20. The 34 came out that far ahead in 11% of tries. That is too often to rule out luck. But the 34 are large industries, and most of the other 707 are small. So we also set them against the 46 other in-scope industries with at least 1,000 businesses with a bench. Those change hands at 1 in 1,317 a year. Here the answer depends on how the gap is counted. The 34 came out that far ahead in 2.5% of tries. Counting a gap that wide in either direction, it was 5.8%. So counted one way the lead is more than chance, and counted either way it is just within. Piece 2 set them against the riskier industries alone, and got 4%, counted one way. So against other large industries the lead is a close call. Two kinds of business tip that scale, one on each side. Hotels carry much of the 34's lead. Without them the 34 change hands at 1 in 903, and each of those comparisons falls within chance. Restaurants hold up the other side. Two restaurant industries make 51% of the 46's sales. Without them the 46 change hands at 1 in 1,985. Take out hotels and restaurants both, and the 34 sell with an SBA loan 2.2 times as often as the other large industries. Chance gives that in about 1 try in 900, or about 3 in 1,000 counted either way. Luck alone would almost never give a lead that wide. The middle industry tells it more simply. The middle one of the 34 sells at 1 in 948 a year. The middle one of the 46 sells at 1 in 2,552. So on most counts the 34 lead other large industries. How firm that lead is depends on how hotels and restaurants are counted. Count only the larger businesses and the numbers get thin. In six years, 1,221 sales in the 34 reported 20 or more jobs. That is 1 business with a bench in 457 a year, as piece 1 said. Counted once per borrower it is 1,054 sales, or 1 in 529. Nine industries, holding 20,538 businesses, had fewer than ten such sales each. That is very little to go on for those industries. At that size the leads over all 707, over the 46 and over the riskier industries are each within chance. With hotels and restaurants both out, the 34 sell 1.6 times as often. Chance gives that in about 1 try in 23, or 1 in 11 counted either way. The usual caution applies. These are SBA-financed sales. A custom programming firm may well be sold with no SBA loan, and we would not see it. And five of the 34 are catch-all codes, such as other specialty trade contractors and other support services. They hold 11% of the businesses and 14% of the sales. The lender picks the code on a loan, so read those rates loosely. About 8 times is the gap in sale rates between a fast industry and a slow one among the 34. Chance alone would give 1.2 times. Passing every test says little about how often a business sells with an SBA loan. In short Sale rates among the 34 run from 1 in 153 a year for hotels to 1 in 5,060 for custom programming. About half of the gap beyond chance goes with how heavily each industry uses SBA loans, counted in times, or under a third on a ratio scale. 17 of the 34 change hands with an SBA loan less often than all in-scope industries together, as do 66% of the others. Hotels alone are 26% of the sales. Each loan counts as one sale, and all of this counts SBA-financed sales only. 5. Is it growing? A screen built on past loans cannot see where an industry is heading. So this piece adds a test the earlier ones did not have. It gives three answers, because there are three ways to count. Census counts business locations and jobs by industry every year. We compared 2023 with 2017 on three measures: locations with 20 or more staff, locations of every size, and jobs. Together, the 34 grew faster than all industries on larger locations and on jobs. Across all industries in those files, larger locations rose 5.3% and jobs 8.7%. For the group they rose 7.3% and 10.0%. On locations of every size the group grew slower: 5.1% against 6.4%. So the answer depends on the measure. redactedWhichever way growth is counted, most of the 34 grew. On larger locations, 24 grew and ten shrank. On locations of every size, 21 grew and 13 shrank. On jobs, 30 grew and four shrank, though masonry contractors lost only 35 jobs of 139,redactedof the 34 grew on all three, and they hold 46,630 of the 92,986 businesses. So fewer of them grew on every measure at once. The measures can disagree. Car washes have 14.9% fewer larger locations than inredactedBut they have 20.9% more locations of every size, and 5.9% more jobs. Hardware stores have 10.5% fewer larger locations, 4.4% more locations and 5.7% more jobs. In six of the 34, larger locations and jobs moved in opposite directions. So fewer large sites does not have to mean a shrinking industry. Only commercial printing is clearly down on every count. It has 15.3% fewer larger locations, 16.8% fewer locations and 15.9% fewer jobs. Machine shops and masonry contractors are down on all three too, by less. Set that beside the last section. Hardware stores, car washes and commercial printing are three of the five industries that change hands with an SBA loan most often. They are also the three that lost the most larger locations. Read hardware stores' fall loosely, though. Their count of larger locations moved by 7.6% to 20.0% in five of the six years. A wider cut leans the same way. Of the ten with fewer larger locations, six are among the ten that change hands most often. Luck alone would put about three there, and six or more about 1 time in 53. That is the strongest of three cuts, and we found it by looking. On jobs, four industries shrank and three of them are among the ten that sell most often. Luck gives that about 1 time in 15. On locations of every size there is no pattern: four of 13, where luck gives about four. We do not claim that one causes the other. Ranked across all 34, growth in larger locations and sale rate do not line up: −0.03. 10 of 34 industries have fewer larger locations than inredactedOnly four of those ten have fewer jobs. Fewer large sites does not mean a shrinking industry. In short On larger locations, 24 of the 34 have grown since 2017 and ten have shrunk. On jobs, 30 have grown and four have shrunk. Three of the five that change hands with an SBA loan most often have fewer larger locations than in 2017, but only commercial printing also lost jobs. One of the three, hardware stores, is on a count that swings from year to year. Growth depends on the measure. Across all 34, growth in larger locations does not line up with sale rates. 6. Can it be financed at its price? The last test is money. The loan file holds no prices, so we read size off the loans. That is a rough guide, and it can err in either direction. redactedSBA 7(a) loans to buy a business in the 34 run larger than usual, and hotels are why. The median loan to buy one is $984,800, against $675,000 for all acquisitions. The median is the middle loan by size. Without hotels it is $664,750, close to the figure for all acquisitions. Larger businesses borrow more. Where the sale reported 20 or more jobs, the median loan in the 34 was $1,686,000. A third of those loans were for $2.7 million or more. And 10.6% were at $5 million, the most a 7(a) loan can be. Most of their loans at the limit are hotels: 86 of the 125. Without hotels the median is $1,327,700, and 4.0% are at the limit. So with hotels left out, larger businesses still borrow more, but fewer of their loans reach the limit. Three things blur the link between a loan and a price. Some sales take two loans. Group the loans made to one borrower within 90 days, and the 34's median borrowing is $1,200,000 a sale, not $984,800 a loan. With 20 or more jobs it is $2,174,000. So a single loan can understate what the buyer borrowed. Many loans also pay for a building. 44% of the 34's acquisition loans run over 10 years, the mark of property or long-lived equipment. And a buyer who puts in more cash borrows less. So a loan's size is only a rough guide to the price. The rules those loans now face On 1 October 2026 the SBA changed its rules for buying a business under 7(a). Piece 1 set them out. In brief: What changed on 1 October 2026 • Earnings against debt payments, standard acquisition: Before 1.15 times, Now 1.25 times • Independent report on earnings: Before Not required, Now Required at a $3m purchase price, property left out • Wait before a seller's note can be refinanced: Before 24 months, Now 36 months • When the lender may value the business itself: Before $250,000 or less financed, Now $350,000 or less purchase price SOPredacted, Appendix 15, effective 1 OctoberredactedWe read the text the SBA published on 25 September 2026, and two later copies of the file. One was saved on 1 October 2026, and one was re-posted on 7 OctoberredactedAppendix 15 reads the same in all three. "Before" is SOPredactedThe 1.25 test applies to a standard acquisition, which the rules call an Initial Acquisition, and to an owner buyout and an employee-ownership deal. An existing business buying another in its own line stays atredactedThe old 1.15 applied to loans over $350,000. The rules call the earnings report a quality of earnings report. The price that triggers it leaves out owner-occupied property. The report is not required for owner buyouts or employee-ownership deals. Nor is it required when a business is bought with owner-occupied special-purpose property, and the rules give hotels and storage units as examples. Two of those lines fall unevenly on the 34. Take the earnings report first. It is required when the purchase price is $3 million or more, with any owner-occupied property left out of the price. A $2.7 million loan is roughly what that price implies with 10% down. So we treat a loan that large as the mark of such a price. Loans that large are more common in the 34, until hotels are set aside. 19.6% of their loans are that large, against 11.1% of all acquisition loans. But most of the large ones are hotels, and a hotel bought with its building can be exempt. Without hotels the share is 9.8%. Even that share overstates it. 44% of those large loans outside hotels run over 10 years, so part of each may be property. And owner-occupied property is left out of the price that triggers the report. Loans that large which run 10 years or less are 5.5% of the 34's loans outside hotels. So the share with no sign of property is smaller still. At the small end, a lender may still value the business itself when the price is $350,000 or less. That implies a loan of about $315,redacted% of the 34's loans are that small. Bars are the extreme, at 45.8%. Some small loans are second loans, so the share of sales that small is lower: 13.6%. So only a modest share of these sales is small enough for that. The higher earnings test applies at every size. In short The median 7(a) loan to buy a business in the 34 is $984,800. Where the sale reported 20 or more jobs it is $1,686,000, or $1,327,700 without hotels. Loan size is a rough guide to price: some sales take two loans, and many loans also pay for a building. Since 1 October 2026, a standard acquisition under 7(a) must show earnings of 1.25 times its debt payments. 7. The industry at the top, and why it is a special case Hotels and motels have the lowest charge-off rate of the 34 on the count this series uses. They are second on size. And they are first on sales when hotels of every size are counted. No other industry is in the top five on both the rate and the sales. On the count this series uses, hotels have the lowest charge-off rate and the most SBA-financed sales of the 34. The median loan to buy one is $2.7 million, and 3 in 4 carry a franchise code. That makes hotels look like the answer. Seven things say otherwise. The low rate is a long-loan rate. About 88% of hotel loans run 20 years or more. Those loans were charged off about as often as other long loans. And under the SBA's 504 programme, hotel loans were charged off more often than other loans. Section 3 has the figures. The low rate is also a slow one. A charge-off is the last step. On the SBA's file of 31 March 2026, 128 hotel loans were marked purchased by the SBA and not charged off, beside 177 with a charge-off date. For all loans it was 4,694 beside 37,252. So for each charge-off, hotels had far more purchased loans than loans in general had. Counting both, hotels were at 3.96% against 8.76%. Not every purchased loan ends as a charge-off, so that is an upper reading. Hotels' rate is then somewhere between about a third and under half of the all-loan rate. Add loans in liquidation and hotels are at 4.63%. Sheet metal work is then lower, at 4.27% on 328 loans. The two are within chance of each other. So with loans in liquidation counted, hotels are no longer the lowest, but that gap is within chance. The loans are property-sized. The median loan to buy a hotel is $2,700,redacted% of those loans are for $2.7 million or more, against 11.1% of all acquisition loans. 14.4% are at the $5 million limit. So set beside all loans to buy a business, loans to buy a hotel are typically far larger. Most are franchised. 74.6% of hotel acquisition loans carry a franchise code. Across all acquisition loans it is 17.3%. A buyer is usually taking on a brand agreement too. Most of the SBA-financed sales are small hotels. 88% of the hotel sales do not report 20 jobs. Count only the sales that do, against hotels with a bench, and hotels are fourth of the 34, at 1 in 227 a year. Other specialty trade contractors, pet care and hardware stores are ahead. Counted once per borrower, hotels are third. So for larger businesses, hotels are near the top on sales but not at it. The count of larger hotels is a pandemic count. Census sizes a firm by its staff in mid-March, and March 2021 was a low point for hotels. Its count of hotel locations with 20 or more staff was 18,761 in 2019, 12,984 in 2021 and 18,173 inredactedSo the 8,104 in our table is very likely too low, and the 1 in 227 too fast. A count that is too low makes sales look more frequent than they are. Bars are in the same position. In 2021 their larger locations were 34% below 2019, and hotels' were 31% below. Nor is it clearly growing. Hotels have 0.4% fewer larger locations than in 2017 and 7.0% fewer jobs, though 3.8% more locations of every size. None of that makes a hotel a bad business. It makes it a different purchase: a property-sized loan and, usually, a brand agreement. The 2.36% is the lender's experience on long loans, and not necessarily the owner's. In short On the count this series uses, hotels have the lowest charge-off rate and the most SBA-financed sales of the 34. But about 88% of their loans run 20 years or more, and those did no better than other long loans. Buying a hotel means a property-sized loan and, usually, a brand agreement. Most hotel sales are small hotels, and the count of larger ones is a pandemic count. 8. Is buying a business safer than starting one? One claim hangs over all of this: that buying an existing business is safer than starting one. The loan file can test it, and a harder version of it too. The loan file has flagged loans to buy a business since the start of 2018, and loans to start one for longer. Take the loans approved from January 2018 to Septemberredacted% of the loans to buy a business were charged off, against 7.96% of the loans to start one. That is 43% lower, and the gap is more than chance. Loans charged off: buying a business against other 7(a) loans • Loans to buy a business: Loans 9,567, Charged off 4.55%, The same, at the buying loans' mix of size , , Of size and the 20-year line , • Loans to start a business: Loans 14,164, Charged off 7.96%, The same, at the buying loans' mix of size 6.24%, Of size and the 20-year line 6.52% • Loans to businesses at least two years old: Loans 45,838, Charged off 7.27%, The same, at the buying loans' mix of size 4.24%, Of size and the 20-year line 4.68% • All other loans: Loans 76,009, Charged off 7.60%, The same, at the buying loans' mix of size 4.74%, Of size and the 20-year line 5.26–5.27% The last two columns weight each kind of loan's charge-off rates by the buying loans' mix, with loans from any lender. The first uses six bands of size. The second adds the 20-year line, rebuilt as in section 3, and gives the range across the two rebuilds where they differ. Businesses at least two years old are those the lender recorded as existing, or as two years old or more. 15,973 loans in these months carry no answer, and 34 are to businesses under two years old. Both are counted only in all other loans. The two kinds of loan are not alike, though. A loan to buy a business is larger and longer than most. So we matched them. At the buying loans' mix of size, loans to start a business would have been charged off at 6.24%. Buying still comes out 27% lower. Add the 20-year line, rebuilt as in section 3, and it is 30% lower. On our seven matchings buying runs from 27% to 30% lower, and each gap is more than chance. All seven split length only at 20 years, or not at all. Where the loan reported 20 or more jobs the gap is smaller, 6% to 17%, and within chance on all seven. Now the harder version. Is a loan to buy a business safer than a loan to one that keeps its owner? Raw, it looks so: 4.55% against 7.27% for businesses at least two years old. Matched, most of that edge goes. At the buying loans' mix of size, those loans would have run at 4.24%. Add the 20-year line and it is 4.68%. Across the seven ways of matching, buying runs from 7% higher to 12% lower. On five of the seven the gap is within chance. On one buying is the safer, by 12%. One sits on the line. Count a charge-off that either file dates, and buying is the safer on two of the seven. That average hides two opposite results. 6 in 10 loans to buy a business are for more than $350,000. For those, buying did worse than like loans on every matching: 3.54% against 1.58% to 2.33%. For loans of $350,000 or less it did better on all seven: 6.05% against 7.57% to 9.55%. Every one of those gaps is more than chance. But none of the seven matchings holds the lender fixed, and lenders differ. 18 lenders made some small loans to buy a business, but rarely: under 2% of their loans of $350,000 or less. They hold 14,151 of the 35,650 small loans to existing businesses, and 14.39% of those were charged off. At all other lenders together it was 5.26%. Seven of the 18 hold 9 in 10 of those failures. Those 18 made 194 of the 3,854 small loans to buy a business. So these lenders hold a large part of the small loans to existing businesses, and made very few of the small loans to buy a business. Leave their loans out of the like loans, and on charge-offs dated so far the edge goes. Buying then runs at 6.05% against 5.22% to 5.94%: higher on all seven, within chance on six, and on the line on size alone. Leave out only the three with the most failures, and the gap is already within chance on all seven. So we ran all seven matchings again, inside the same lender. That uses 80% to 91% of the buying loans: those whose lender made a like loan. There too the edge for small loans goes. For loans of $350,000 or less, buying ran at 6.21% to 6.51%, against 4.86% to 5.62% for the same lender's like loans. That is higher on all seven, though not always by more than chance. Only at $150,000 or less are the two level. For larger loans buying did worse on every matching, as before: 3.65% to 3.68% against 1.67% to 2.14%. One caution on the small loans. Weight them the other way, at the lenders of the loans to existing businesses. The gap is then within chance on all seven, and runs both ways: 8.70% to 11.60% against 9.54% to 10.16%. That test is weak. Nearly half of its weight rests on the 194 buying loans the 18 lenders made. So for small loans the safe reading is no better, not worse. Larger loans did worse that way round too. The same matching cuts into the comparison with starting. Inside the same lender buying is 11% to 17% lower, not 27% to 30%. That gap is within chance on our test, though only just on four of the seven. Hold the sector fixed as well, and it is 5.50% against 5.77%, on the 39% of buying loans that can be matched. The jobs a loan reports point the same way. For loans reporting 20 or more jobs, buying did worse than like loans from any lender on all seven matchings: 4.60% against 2.61% to 3.11%. That rests on 82 charge-offs in 1,781 loans, and on a jobs figure the SBA does not check. With purchased loans counted that gap is within chance: 8.48% against 7.44% to 8.90%. Inside the same lender buying is the higher on all seven, on each count. On each count one of the seven, size alone, is more than chance. But these loans are young, and a charge-off is the last step. Section 3 described the loans the SBA has purchased and not charged off. Among the buying loans that are dated or purchased, 43% were still at the purchased stage in MarchredactedFor like loans to businesses at least two years old it was 46% to 48%. For loans over $350,000 it was 56% against 68% to 71%. So we counted purchased loans as well, on the March file. The same loans, with purchased loans counted as well • Loans to buy a business: Loans 9,567, Dated or purchased 7.60%, The same, at the buying loans' mix of size , , Of size and the 20-year line , • Loans to start a business: Loans 14,165, Dated or purchased 11.34%, The same, at the buying loans' mix of size 10.10%, Of size and the 20-year line 10.62–10.65% • Loans to businesses at least two years old: Loans 45,838, Dated or purchased 9.72%, The same, at the buying loans' mix of size 7.60%, Of size and the 20-year line 8.50–8.53% The March file alone, with no pairing. A loan counts where it has a charge-off date, or the status PURCH(NOT C/O) and no date. Not every purchased loan ends as a charge-off, so these are upper readings. Loans are counted on the March file. The last two columns are matched as in the table above, from any lender, with each purchased loan's first length rebuilt as well. With purchased loans counted, buying still runs lower than starting: 7.60% against 11.34%, and 25% to 29% lower on the seven matchings. Inside the same lender it is 6% to 13% lower, within chance on our test. Against like loans from any lender to businesses that kept their owner, buying is the safer by more than chance on five of the seven matchings, by 11% to 18%. On size alone the two are level. That is the small loans again. Over $350,000 buying ran higher on every matching: 7.35% against 5.28% to 7.12%. Inside the same lender buying did worse on all seven than like loans to businesses that kept their owner: 7.81% to 7.96% against 5.40% to 6.43%. For loans of $350,000 or less it is 8.16% to 8.54% against 6.39% to 7.38%: higher on all seven. For larger loans it is 7.56% to 7.63% against 4.73% to 5.82%: worse on every matching. Add loans in liquidation as well, and no direction changes. Larger loans are then at 9.33% to 9.41%, against 6.14% to 7.37%. The count with purchased loans can lean either way. Purchased loans are mostly loans that had been sold on the secondary market, and buying loans are sold more often: 57% against 46% to 53% for like loans. That leans against buying. But three months on, the larger purchased loans to existing businesses that had ended were paid in full as often as charged off: 6 and 6. Those to buy a business were mostly charged off: 12 against 3. Those are small numbers. So the fair reading is narrower than the claim. Loans to buy a business have been charged off less often than loans to start one. Matched on size inside the same lender, the gap falls to 11% to 17%, only just within chance on our test. At 20 or more jobs no edge is shown: inside the same lender buying ran level or a little higher. Against a like loan to a business that kept its owner, the answer turns on the lender. Set beside any lender's like loans, small loans to buy a business did better. Set beside the same lender's, they did no better, and loans over $350,000 did worse on every count. The SBA's own rules open on the point: "Change of ownership transactions include credit risks that are not present in other segments of 7(a) lending." And a franchise shows no clear edge 17.3% of the acquisition loans approved in fiscal years 2020 to 2025 carry a franchise code. In fiscal 2018 and 2019 it was 21.7%. Among those older loans, 4.62% of the franchised ones were charged off, on 2,079 loans. For independent ones it was 4.54%, on 7,493. That gap is within chance. It still is without hotels, which are 398 of the franchised loans: 5.23% against 4.62%. That is on charge-offs dated so far. With purchased loans counted it is 7.02% against 7.77%, still within chance. Add loans in liquidation and it is 7.84% against 9.40%. That gap is just more than chance, and just within it without hotels. Among all other loans approved in fiscal years 2010 to 2019, franchised businesses did worse: 8.86% against 7.02%. In short Loans to buy a business were charged off 43% less often than loans to start one: 4.55% against 7.96%. Matched on size inside the same lender the gap is 11% to 17%, only just within chance on our test. At 20 or more jobs no edge over starting is shown. Against like loans to businesses that kept their owner, buying shows no edge once the lender is held fixed. Loans over $350,000, 6 in 10 of them, did worse than the same lender's like loans on every count. A franchise shows no clear edge. 9. What the list corrects 1. "The safest of the 34 are the ones that sell" Among the 34 we found no link between the charge-off rate and the SBA-financed sale rate. It comes to −0.05, well within chance. But 34 industries, all picked for low charge-offs, are too few to rule a link out. The wider set shows one. Across the 175 in-scope industries with enough loans to rate, lower charge-offs go with more SBA-financed sales. That link is −0.37, far more than chance. So the link shows up across many industries, though we did not find it inside this short list. 2. "Fewer large sites means a shrinking industry" An industry can have fewer large sites and still more jobs. Car washes have 14.9% fewer locations with 20 or more staff than in 2017, and 5.9% more jobs. Hardware stores have 10.5% fewer and 5.7% more. Of the ten industries with fewer larger locations, only four have fewer jobs. So fewer large sites need not mean a shrinking industry. 10. If you are choosing among the 34 If you are choosing an industry • Passing the screen is the start. Half of the 34 change hands with an SBA loan less often than all in-scope industries together. So do two thirds of the rest. • Check all four tests. No industry leads on all of them. • Ask whether the industry is growing, and on which measure. Larger sites, all sites and jobs can disagree. • Know the size of the borrowing first. With 20 or more jobs the median is $2,174,000 a sale, or $1,775,000 without hotels. If you are selling in one of the 34 • If your buyer borrows under 7(a), in most cases the lender must now see earnings of 1.25 times debt payments, on past results. • If your price is $3 million or more, leaving out any property, the lender must in most cases also get an independent report on earnings. If you are lending or advising • 16 of the 34 are within chance of the average, and no two neighbours in the order differ by more than chance. Treat the list as a rough sort, not a ranking. • A low charge-off rate may reflect long loans as much as a safe business. Hotels' long loans did no better than other long loans. • Do not sort SBA 7(a) loans by length as the file records it. The file rewrites a failed loan's length. • Say which file a charge-off rate comes from, and how a failed loan was counted. On five counts, 32 to 37 industries pass this series' test. • Loans to buy a business have been charged off less often than loans to start one. Matched on size inside the same lender the gap is 11% to 17%, only just within chance on our test. At 20 or more jobs no edge is shown. Against the same lender's like loans to businesses that kept their owner, buying has been no safer at any size, and worse over $350,000. 11. Assumptions, gaps and limits Assumptions • That the licence screen draws the right line. Piece 3 tests it. Let through the industries where, on our reading, only a named licensed person is needed, and 39 industries pass, not 34. They hold 117,832 businesses. • That loans of every purpose rank industries fairly for acquisitions. The risk test uses all 7(a) lending from fiscal 2010 to 2019, because acquisitions were not flagged before 2018. • That a rate on all loans is the right yardstick. It rewards industries that borrow long. On rebuilt lengths, two or three of the 34 were charged off more often than like loans. With the lender held fixed as well, five were, none of them by more than chance on charge-offs dated so far. • That our rebuilt loan lengths are near the truth. The file does not give a failed loan's first length. We give two rebuilds, and the truth should lie between them. • That the SBA's 30 June 2026 file is the right count of charge-offs. It drops the date from a charged-off loan later marked paid in full. Counting those, the all-industry rate is 7.89%. The same 34 pass, and two more join them. Three of those 36 turn on a single loan: car parts stores and other support services stay in by one, and recyclable material wholesalers join by one. Motor vehicle parts wholesalers miss by one. • That loans approved in fiscal 2010 to 2019 are a fair sample. 49% of the charge-offs the June file dates are from March 2020 or later, so the pandemic is in the record. • That one sale rate describes an industry. It counts businesses of every size, it counts each loan as a sale, and it counts SBA-financed sales only. • That larger locations are the right measure of growth. Locations of every size give a different direction for 13 of the 34, and jobs for six. • That a $2.7 million loan marks a $3 million price. More cash down makes the loan smaller. Property in the loan, or a second loan, breaks the link the other way. Data gaps • No sale prices. The loan file records loans. • No record of what secures a loan. We read property from a term of more than 10 years. • No record of a failed 7(a) loan's first length. The file shows roughly the months that were left. • No record, in the June file, of a charge-off later closed as paid in full. Nor of the status of any open loan, such as one the SBA has purchased and not charged off. • No definition of the March file's status codes beyond the five in the SBA's data dictionary. We read PURCH(NOT C/O) and LIQUID as purchased and in liquidation. • No count of sales made without an SBA loan. • No count of how many buyers compete for a business. • No loans approved afterredactedThis series treats them as too young to judge. Limitations • 16 of the 34 are below the average by less than chance could produce. With commercial printing's older loans it is 17. • A loan counts only where its code is in the 2017 industry list. 27,138 of the 479,007 seasoned loans carry a code that is not, most of them approved in fiscal 2010 toredactedPlacing those we could by their titles, no industry joins or leaves the 34. • The figures for like loans in sections 3 and 8 rest on loan lengths we rebuilt, and on the bands we chose. We give a range, and the plain figures beside them. • The 34 sell more often than other in-scope industries. Whether that is more than chance depends on the comparison, and on how hotels and restaurants are counted. • Each loan counts as one sale. Counted by borrower, the 34 had about 6,200 sales, not 6,825. Some may be partners buying each other out. • The industry on a loan is the lender's choice. The industry on a firm is Census's. Every sale rate divides one by the other, and five of the 34 are catch-all codes. • A sale rate partly measures how much an industry borrows with an SBA guarantee at all. • Firms are sized by their staff in mid-March 2021, mid-pandemic. Bars and hotels are the clearest undercounts: their larger locations were 34% and 31% below 2019. • About 4,400 of the 92,986 are likely nonprofit, mostly in services for the elderly and disabled and in golf and country clubs. Piece 2 sets that out. • Sales reporting 20 or more jobs are few. Nine of the 34 had fewer than ten in six years, or 11 counted once per borrower. • The 1,000 bar rests on the mid-March 2021 sizes too. Testing laboratories miss it by 10 businesses, with 2.90% of their loans charged off. • The list moves with the file and with the count. On the SBA's file of three months earlier, read as the June file reads, 32 pass. On five counts in all, 30 of the 34 pass every time, and 29 on a sixth that adds loans in liquidation. • The comparison of buying with like loans turns on what is matched. With size and length matched, small loans to buy a business look safer. With the lender matched as well, they do not. The file names the bank a loan is now assigned to, which may not be the bank that made it. And inside one lender the cells are thin. The line at $350,000 is one of our six band edges. The file cannot say what a loan paid for, what secured it, or how the loans still open will end. • The licence screen missed two industries. It did not read BLS's separate file of private schools, and it missed a 2022 change of industry code. With both read, sports and recreation instruction and software publishers would pass every test. That would make 36 industries and 96,744 businesses. The first holds 1,497 businesses with a bench, with 6.55% of its loans charged off, and changes hands with an SBA loan at 1 in 1,821 a year. The second holds 2,261, with 6.10%, at 1 in 3,324. Piece 3 sets that out. On the either-file count of charge-offs as well, 38 industries and 109,175 businesses would pass. Software publishers pass that count by one loan. • The rules table describes rules that took effect on 1 OctoberredactedEvery loan in this piece was made under the old ones. The full version is on acquidex.com. It carries every table, every source and the complete method. This is part four of five. The last part is the map: where SBA loans buy businesses, and 15 attempts to explain why. ________________________________________ Sources Every count here is public, free to download and recomputable. The SBA figures are from the agency's own loan-by-loan releases. • SBA 7(a) and 504 loan-level FOIA releases, 7(a) as of 30 June 2026; 504 as of 31 MarchredactedNot linked, because the download path moves between releases • SBA SOP 50 10, lender and development company loan programs, SOPredactedwith technical policy updates, published 25 September 2026, effective 1 October 2026: sba.gov/document/sopredactedlender-development-company-loan-programs • Census Statistics of US Businesses, 2021 firm-size tables: census.gov/programs-surveys/susb.html • Census County Business Patterns, 2017 to 2023: census.gov/programs-surveys/cbp.html