You Can Afford the Business. But Are You Qualified to Run It?
redactedA quarter of businesses with employees sit behind a licence screen. How that line was drawn, and how far it moves.
Part three of five.
Part one gave a number. 92,986 businesses in the United States pass every test I set. Part two showed every cut.
One cut removes a quarter of all businesses with employees. It is the licence cut, and more of it rests on my judgement than any other.
So in this part I test it. I move every line and show what changes.
Some of it held. Some of it did not. I would rather show you where it is soft than have you find it.
About a quarter of businesses with employees in the United States are in industries our screen flags for a licensed profession or trade. That is 1,601,343 businesses, or 1,505,350 without six industries swept in by their grouping. This piece shows how we drew that line, tests how firm it is, and measures what sits behind it. SBA loans to those businesses go bad less often. The businesses are smaller. Fewer change hands with SBA loans, but that gap is less sure.
1. A quarter of businesses with employees are in industries our screen flags
The second fork in this series asks whether a licence gets in a buyer's way. For 1,601,343 businesses, our screen said it may. That is 26.3% of every business with employees in the United States, and 13.8% of the jobs. So the screen flags a bigger slice of the businesses than of the jobs. The base is the 6,077,257 businesses that pieces 1 and 2 count. It leaves out membership and religious organisations, holding companies and funds.
Those businesses are in 69 industries that the screen flags as built around a licensed profession or trade. Law, medicine, dentistry and accounting. Insurance and real estate brokerage. Plumbing and electrical contracting. Pharmacies, funeral homes and salons.
In most of these industries the licence has to be held by an owner, or by a named licensed person the business employs. That is our reading for 84% of the businesses. Where a named person holds the licence, a buyer has to keep that person or hire another.
The other 16% are different. 10% are salons, other personal-care shops and some health practitioners' offices. There the shop and its staff are licensed. 6% are in six industries where, in our judgement, no owner needs a licence. They are flagged only because of the group they are filed in. Without those six the count is 1,505,350, or 24.8%. So, on our reading, not every business the screen flags is there for an owner's or a named person's licence.
We call the 69 industries the wall. The word is shorthand for what our screen flags. It is not a ruling on any state's law. We read no statute for this piece, and section 4 shows where the screen is soft.
Piece 1 promised that this piece would document the screen that draws the line. So it does three things. It shows how the screen works, and who decided each step. It tests how far the line moves. And it measures what is behind the wall: how safe, how profitable, how big and how often sold with an SBA loan.
redactedIn short
1,601,343 businesses are in 69 industries our screen flags as built around a licensed profession or trade. That is 26.3% of businesses with employees, or 24.8% without six industries swept in by their grouping. We call that group the wall. This piece documents the screen that draws that line, tests it, and measures what is behind it.
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. Three steps, and who decided each
The screen has three steps. None is pure measurement. It matters which parts are measured and which are ours.
Step one, a hand list checked by a survey: which occupations are licensed
We did not start from a survey. We started from a list of 47 occupations that we understood to need a licence. Then we tested each one.
The test is the Current Population Survey. It asks whether a worker holds a currently active professional certification or licence, and whether a government issued it. The question tells people to leave out business licences, such as a liquor licence. So the survey sees personal licences only.
We used the 11 monthly files forredactedThey hold 495,667 records of people in work. Weighted, 24.0% of workers hold a certification or licence. 21.6% hold one that a government issued. The Bureau of Labor Statistics (BLS) publishes the same two figures forredactedOur figures from the files match the published ones.
Those are records, not people. A household is interviewed four months running, so one worker can be more than one of them.
An occupation stayed on the list if at least 25% of its workers report a government licence. The full version gives two exceptions. Five failed and were struck out: private detectives, two kinds of engineer, hazardous-materials removal workers and security guards. We then added 13 that cleared the same bar, among them architects, morticians and social workers. The list ends at 55 occupations. The bar took some names off our list and let others on.
One of the five struck out was within chance of the line. Hazardous-materials removal workers came in at 17.3%, on 78 records. Section 4 puts them back as a test.
So the list is ours, and the survey is a filter on it. It is not every licensed occupation. The survey can measure 366 occupation groups. 92 of them clear the 25% line, and 39 of those are not on our list. School teachers, construction managers, financial advisers and analysts, and police officers are among the largest. The line alone would add occupations that our list leaves out. Section 4 adds them as a test.
Step two, judged: whose licence it has to be
No survey says whether a licence must sit with the owner or only with the staff. So we read each occupation on the list and wrote down which it is. 31 went on what we call the owner list. 22 went on as staff licences, such as nurses and truck drivers. Two are tied to premises, in child care.
redactedRead the last column with care. In 13 of the 31, our reading is that the licence must sit with the owners themselves. Law and medicine are the cases we are surest of. In 15, a named licensed person must be in charge: a pharmacist, a master electrician, a broker of record, a funeral director. A buyer could in principle hire that person. In three, the shop is licensed and so are the staff: hairdressers, other salon workers and massage therapists. So, on our reading, a licence on this list does not always mean a licensed owner.
That column is our reading of how each licence generally works across the United States. We checked no state's statutes for it, and the rules do differ by state. It is the weakest step in the screen. The reading is also applied to a whole group of industries at a time, which sits badly in places. Section 10 lists them. Two rates are within chance of the 25% line at their number of records: surveyors and dispensing opticians. Those margins are worked on records, and one worker can be several records, so the true margins are wider. Podiatrists have 26 records, too few to measure, and stay on the list on our reading alone.
Step three, measured inside limits we set: which industries are built around them
We flagged an industry for an owner's licence when one of the 31 occupations cleared two bars at once. The bars use what BLS publishes: how many employees in each occupation work in each industry. The occupation had to be at least 5% of the industry's jobs. And it had to be at least 5 times as common there as in all jobs. So it must hold enough of the industry's jobs, and be much more common there than in all jobs.
The first bar does the work. Inside the industries we named, every pairing that clears it also clears the second. The lowest is 18 times. Dentists are 11.1% of the jobs in dentists' offices, which is 117 times their share of all jobs.
An industry not flagged that way is flagged for licensed staff if the 22 listed staff occupations hold at least 20% of its jobs. Home health care, trucking and nursing homes fall there. Not all of those workers hold a licence. 27.1% of truck drivers report one.redactedWhere two or more listed occupations clear the bars in one group, the share and the multiple shown are each the largest among them. All jobs means BLS's own count of all private jobs, 131.8 million. The screen was built on a smaller base, 123.3 million, because the file it read gives schools and hospitals only with public employers mixed in. On that base physicians' offices read 37 times, pharmacies 100 and therapists' offices 127. The other 11 rows differ by 3% to 6%. No industry changes class between the two.
Three limits on this step are ours, not BLS's.
• 19 of the 31 occupations count only inside industries we named. Lawyers count only in legal services, accountants only in accounting, insurance agents only in insurance. That stops an engineer employed by a utility from flagging the utility.
• We set 18 industries by hand. They hold 503,704 businesses. Nine were put behind the wall, and those nine hold 17.8% of it. Real estate brokerage, the second-largest industry there, is one.
• BLS counts employees. It leaves out the self-employed, and owners and partners of unincorporated firms. So the partners in a law partnership are not in the lawyers' 37.2%. Where only the owner must hold the licence, the share of jobs can be low, and the rule can miss it.
The rule works on BLS's industry rows. The two BLS files we read hold 295 industry rows for private employers. 262 of them are at the finest level. 14 of those are part-groups of BLS's own making, and our loader merged them into ten broader rows. That leaves 248 rows at the finest level.
The rule flags 26 of those rows for an owner's licence and 17 for licensed staff. Most rows are flagged for neither.
We carried those flags to the 952 industries in the Census firm counts, matching by industry code. Where BLS has no finer row, an industry takes the flag of the broader group. So an industry can be flagged through the group it sits in. The full version gives the detail.
In short
The screen has three steps. We named a list of licensed occupations, and a survey struck five from it. We judged who must hold each licence: an owner in 13 occupations, a named person in 15. BLS job counts then show which industries are built around those occupations, inside limits we set. No step is pure measurement. Step two is our reading, and it is the weak one.
3. 69 industries, read four ways
1,601,343 businesses, in 69 industries. Here are the largest. The full version lists all 69.
redacted† Behind the wall only because of the group it is filed in. Section 4 explains.
These are small firms. 96,335 businesses behind the wall, or 6.0%, have a bench of 20 to 999 staff. Past the licence fork it is 12.3%. The average business behind the wall has 10.6 jobs, against 21.3 past the fork. On either measure, the firms behind the wall are smaller.
Two industries lift that second average: temporary-help firms and professional employer organisations. They carry 6.8 million jobs on their books. Without them it is 19.4.
Real estate brokerage is the extreme. Its 143,248 businesses average 2.7 jobs each.
redactedThe wall is not one thing. On our reading, 707,735 of these businesses, or 44%, are in industries where an owner must hold the licence. 636,079, or 40%, are in industries where a named licensed person must be in charge. 161,536, or 10%, are in industries where the shop and its staff are licensed. Those are salons, barbers, other personal care, and one group of health practitioners' offices. So, on our reading, the owner's licence is only part of the wall.
The last 95,993, or 6%, are in six industries where, in our judgement, no owner needs a licence. They are behind the wall only because of the group they are filed in. Section 4 deals with them, and with the salons, which arguably belong with the licensed-staff industries.
6.0% of businesses behind the wall have a bench of 20 to 999 staff. Past the licence fork it is 12.3%. Behind the wall, the firms are smaller.
In short
The wall holds 69 industries, led by law, real estate brokerage, medicine, dentistry and insurance. The firms are small: 6.0% have a bench of 20 to 999 staff, against 12.3% past the fork. On our reading, an owner must hold the licence in 44% of the wall. 40% needs a named licensed person. 10% is a licensed shop with licensed staff, and 6% is there only through its group.
4. Where the line is firm, and where it is soft
A screen with judgement in it has to be tested. We tested six things. Each time we rebuilt the whole count.
1. The industry bars
redactedThe owner line moves little. Lower the share-of-jobs bar to 3% and 95,197 businesses join the wall, which is 5.9% of it. Raise it to 10% and 66,321 leave. Of those, 61,365 are then set aside for licensed staff, and 4,956 clear the fork. So most of what leaves the wall is still set aside.
The concentration bar moves less still. Change it from 5 times to 3, or to 10, and nothing moves at all. At 20 times, 4,956 businesses leave. None of the five settings, on either bar, changes the 34 industries that pass every test. In these tests the bars can move the edge of the wall, but not the list that passes.
The staff bar matters more. At 10% of jobs, 94 more industries are set aside, and one of the 34 drops out. At 50%, three more industries pass, and the count rises to 108,971. Where this bar sits does change which industries pass.
2. The list and its line
Step one rests on a 25% line and on a list we chose. We moved both.
redactedA lower line sets more industries aside. Drop the line to 20% and security guards come back onto the list. Three industries, with 11,270 businesses, are set aside.
A higher line lets trucking and related industries through. Raise it to 30% and surveyors, opticians and truck drivers drop off. 13 industries in trucking, moving, towing, delivery and waste collection clear the fork. They hold 178,575 businesses. One of them, solid waste collection, passes every test. The rest clear the fork but do not pass every test.
What if the doubtful occupations had stayed on the list? One struck-out occupation sat within chance of the 25% line: hazardous-materials removal workers. Keep them on the list, and four industries with 10,482 businesses join the wall. The 34 do not change. Private detectives, struck out at 18.6%, are the other doubtful case. Put back, they move no industry. Neither doubtful case changes what passes.
At 40%, 11 occupations drop off, among them accountants, electricians and plumbers. The wall shrinks to 1,403,303 businesses. 38 industries pass every test, holding 105,768. A line that high leaves a smaller wall and lets more industries pass.
The survey also asks a stricter question: is the licence required for the worker's job? On that stricter measure one listed occupation falls below 25%: accountants, at 23.1%. Take them off the list and the accounting group, with 120,490 businesses, leaves the wall. 36 industries pass, holding 96,791. So this question takes accounting out of the wall, and more industries pass.
Truck drivers only just clear that measure. Their margin is 0.3 points, at 25.3%. If they fell too, 37 industries would pass, holding 97,961. A fraction of a point decides whether a further industry passes.
Last, the 39 occupations we left off. How they are added matters. 34 of them match an occupation in the BLS job counts. Added as staff licences, which need 20% of an industry's jobs, they set aside 31 more industries with 74,804 businesses. 23 of those are in finance and insurance, and no industry that passes every test is among them. Added this way, they set more aside but leave the list that passes alone.
Added to the owner list, which needs 5%, they put 53 more industries behind the wall. Those hold 344,764 businesses. 245,640 of them are in seven building and land-development industries, flagged through construction managers. One of the 34 would drop out, leaving 33 industries and 91,902 businesses. So a longer owner list would shorten the list that passes, not lengthen it.
3. The grouping
BLS publishes most industries in groups, and the rule flags a whole group at once. That sweeps in some industries that, in our judgement, have no licensed owner. Tax preparers, payroll firms and bookkeepers are filed with accountants. Translators, pollsters and a catch-all of other professional services are filed with veterinarians.
Six industries flagged only because of the group they are filed in
• Bookkeeping and other accounting services: Businesses 44,726, With 20 to 999 staff 1,552
• All other professional and technical services: Businesses 20,848, With 20 to 999 staff 787
• Tax preparation services: Businesses 19,491, With 20 to 999 staff 488
• Payroll services: Businesses 4,434, With 20 to 999 staff 1,054
• Market research and polling: Businesses 3,785, With 20 to 999 staff 458
• Translation and interpretation services: Businesses 2,709, With 20 to 999 staff 112
By our own reading, those six should not be behind the wall. They hold 95,993 businesses, which is 6.0% of everything behind the wall. 6.97% of their loans were charged off, against 7.13% for all industries. So their loans look like the average industry's.
Move the six past the fork, and one more industry passes every test. It is bookkeeping and other accounting services, with 1,552 businesses. The 34 becomes 35, and 92,986 becomes 94,538. But that industry passes by three loans. 96 of its 1,376 loans were charged off, and 99 would put it over the line. So the extra industry only just passes.
Another 16 industries, holding 54,206 businesses, may have been swept in the same way. Title offices sit with lawyers. Cemeteries sit with funeral directors. Moving those as well adds no further industry to the ones that pass.
The grouping cuts the other way once. BLS publishes 14 part-groups, and our loader merged each into a broader row. One of them is grocery and alcohol wholesalers. On BLS's own row, the listed staff occupations hold 20.2% of its jobs, just over the staff bar. Nearly all are truck drivers.
Taken alone, that row would set aside 11 industries with 31,259 businesses. None of them passes every test. So the merging let those industries through, but it did not change the list that passes.
4. The reading
Step two was our judgement, so we tested three parts of it.
First, the salons. Suppose salon, barber and massage licences are read as staff licences. Then 168,340 businesses move from one set-aside group to the other. Nothing past the licence fork changes, because both groups are set aside.
That is the 161,536 businesses in the shop-and-staff group above, plus 6,804 podiatrists' offices. BLS files podiatrists in the same group as massage therapists, so they move with them.
Second, the owner column. Our own notes hedge seven of the 13 owner readings. Take the four most hedged: therapists, veterinarians, civil engineers and architects. Read those as a named person, and 148,049 businesses change column. The owner share of the wall falls from 44% to 35%.
Read all seven that way, and the owner share is 23%. That adds dentists, optometrists and accountants. What is left is law, medicine, chiropractic and podiatry. Nothing crosses the fork. These readings change how the wall is split, not what gets past it.
Third, the named person. A named licensed person can sometimes be hired. Suppose a buyer always could. Let those 38 industries through, and 39 industries pass every test, not 34. They hold 117,832 businesses, not 92,986. On that assumption, more industries pass.
The five that join are plumbing, heating and air-conditioning contractors, electrical contractors, insurance agencies, pharmacies and real estate brokerages. Electrical contractors pass by four loans.
5. The hand-set industries
We set 18 industries by hand, and they hold 503,704 businesses. Drop that step and let the rule decide. Five property industries join the wall, four of them lessors, which rent property out. Four securities and investment industries leave it. Pharmacies, fuel stations, schools and tobacco stores, with 115,471 businesses, fall back to not classified. The rule does flag pharmacies, but their 2017 code matches no BLS row.
With the hand step dropped, the wall becomes 1,649,337 businesses. The 34 that pass do not change. So in this test the hand step moves the edge of the wall, not the list that passes.
We also dropped the 19 limits on where an owner occupation counts. Then nine more industries, with 4,919 businesses, join the wall. The 34 do not change.
On other BLS figures, more industries join. On BLS's own count of all private jobs a tenth joins: land subdivision, with 4,555 businesses. That makes 9,474. With BLS's private rows for hospitals read as well, an eleventh joins: specialty hospitals, with 308. That makes 9,782. So the number that joins depends on which BLS figures the rule reads.
6. The choice to set aside
The largest effect comes from a choice, not a measurement. We set aside the 481,079 businesses with licensed staff or premises, although on our reading the owner need not hold the licence. And we set aside 438,800 that we could not classify.
Let both groups through, and 45 industries pass every test instead of 34. They hold 139,684 businesses, not 92,986. The largest additions would be schools, child day care, nursing homes, assisted living and fuel stations. Setting those groups aside is what keeps the count lower.
Two cautions on that larger number. 87% of the larger school locations are nonprofit. And the unclassified group is exactly that. For most of it, we have not tested for a licence at all.
For part of it we could have, with files that were to hand. Only two of the 81 unclassified industries meet both the size test and the risk test: sports and recreation instruction, and software publishers. So only those two could pass, whatever class the rest were given.
Both of those unclassified industries clear the fork when BLS's rows for them are read. Take the first. BLS publishes private schools and hospitals in a separate file, and our screen did not read it. It has ten private rows: seven kinds of school and three of hospital, with 8.7 million jobs.
Run the rule on them, and 19 of the 81 get a class. The three hospital industries are flagged for licensed staff. The 16 school and training industries clear the fork, with 81,089 businesses. In BLS's rows for them, no listed owner occupation reaches 5% of jobs, and listed staff hold at most 11.4%. Pilots come nearest, at 4.5% of jobs in technical and trade schools, the row that holds flight schools. So no listed occupation is common enough in the school rows to flag them.
Software publishers were missed for a different reason: their code changed. The industry list was revised in 2022, and their code moved fromredactedtoredactedOur firm counts use the older list. So our counts and BLS's rows carried different codes for the same industry.
BLS has a row for them under the new code, with 639,460 jobs. The rule flags nothing there. Accountants are the largest listed occupation, at 1.5% of its jobs. BLS withholds the count for four listed occupations in that row: pharmacists, registered nurses, psychologists and miscellaneous healthcare support workers. So the rule flags nothing on the counts it can see. It cannot test the withheld ones.
So a fuller matching gives 36 industries, not 34. Sports and recreation instruction has 1,497 businesses with a bench. 6.55% of its 1,665 loans were charged off. Software publishers have 2,261 businesses with a bench. 6.10% of their 525 loans were charged off. Both rates are within chance of the line. So both pass the risk test, but neither by a clear margin.
The 36 hold 96,744 businesses, which is 4.0% more than 92,986. The unclassified group cannot add more than these two, because no other industry in it meets both tests. That is as far as this group can move the count. This series keeps 34 and 92,986, the count on the screen as built.
The 2022 industry list reaches further. Of the 62 industries left unclassified after the school rows, 53 have a BLS row under their 2022 code. 48 of them clear the fork, with 245,118 businesses. Four are flagged through the pharmacy group, and fuel dealers are flagged for licensed staff. On their newer codes, most of those industries clear the fork.
Three of the four are swept in: cosmetics, food supplement and other health stores.
Nearly every pharmacist in that group works in a pharmacy, and BLS gives pharmacies a row of their own. So those stores are flagged for pharmacists who nearly all work somewhere else.
Optical goods stores are the fourth, and a closer call. Dispensing opticians are on our list, and BLS has no row for those stores alone. There is no row to test them on by themselves.
For nine industries, with 80,949 businesses, BLS has no row that holds their jobs. Seven are in forestry, fishing and hunting. In farming, forestry, fishing and hunting, BLS's survey covers only logging and support activities for crop and animal production. The other two are online and direct sellers, which the 2022 list spread across retail. So the rule has no BLS figures to test them on.
Matching everything moves more businesses past the fork. With every match, 3,882,242 businesses would be past the fork, not 3,556,redacted,123 would have a bench, not 438,125. The count that passes stays at 36. These further matches add no industry to the ones that pass.
Schools, the two kinds of fuel station and tobacco stores are set by hand to licensed premises. Drop the hand step as well, and the rule lets all four through. Schools clear on BLS's private row. The other three clear on BLS's rows for their 2022 codes. So it is the hand step, not the rule, that sets them aside.
Two of the four meet the size and risk tests: schools, and fuel stations with convenience stores. Schools have 11,069 businesses with a bench, and those fuel stations 3,416. That makes 38 industries and 111,229 businesses. Schools are most of that addition. And, as above, 87% of the larger school locations are nonprofit.
We keep the hand step. It rests on the permits those premises need, on our understanding. So that larger count is a test, not the count we use.
Take the licence fork away altogether, the wall included, and 60 industries pass, holding 212,086.
In short
The owner line holds against its two bars: none of five settings changes the 34 that pass. It is softer elsewhere. Across the tests that move the line, the wall itself runs from 1,403,303 to 1,946,107 businesses, or 23% to 32% of those with employees. A lower staff bar, or a longer owner list, leaves 33. A 40% licence line gives 38. Let the named-person industries through and 39 pass, holding 117,832. The owner share of the wall is 44% on our reading, and 23% if every reading our notes hedge goes the other way. The largest effect is still our choice to set aside licensed-staff, licensed-premises and unclassified industries. Without it, 45 industries and 139,684 businesses would pass, not 34 and 92,986. A fuller matching to BLS's rows would add two industries, sports and recreation instruction and software publishers. That makes 36, holding 96,744 businesses. And 95,993 businesses are behind the wall only because of how BLS groups them, in our judgement.
5. SBA loans behind the wall go bad less often
Now to what sits behind the wall. Start with the measure this series has used throughout: how often SBA loans are charged off.
SBA loans behind the wall go bad less often. The SBA approved 479,007 loans under its 7(a) programme in fiscal years 2010 to 2019 that were not later cancelled. We call them seasoned loans, because they are old enough for losses to show. 7.13% were charged off, meaning written off as a loss. Past the licence fork it was 7.53%. Behind the wall it was 5.29%. So the wall's side has the lower rate.
That gap is more than chance. We call a gap more than chance when chance gives one that wide in fewer than 1 try in 20. We dealt the wall's label at random among the 821 industries on the two sides that have any seasoned loans, 200,000 times. Each time, random industries stood in for the wall. The wall came out that far below in under 1% of tries. Chance alone would rarely give a gap this wide.
It holds without dentists and veterinarians. Without dentists' offices it is about 1.3% of tries. Without dentists and veterinarians it is about 2.0%. Each is still under the line we set.
The gap also survives the obvious objections. The wall's rate is the lower in each of the ten approval years, and in each of five bands of loan size. Counted in dollars, the SBA charged off 2.30% of what it approved behind the wall, against 3.18% past the fork. Give the wall the other side's mix of years and loan sizes, and its rate is 5.25%.
The gap also holds inside the same lenders. 298 lenders made at least 30 loans on each side. Taken together, their rates are 5.44% behind the wall and 7.84% past the fork. At 214 of the 298, the wall's rate is the lower one. Half would be if the two sides were alike. So at most of these lenders, loans behind the wall went bad less often.
It survives a fuller count as well. These charge-offs are the ones the SBA's file of 30 June 2026 dates. Piece 4 shows that this file drops the date from a loan later marked paid in full. So the June count misses some charge-offs.
On the fuller count the rates are higher, and the wall is still the lower. Count a charge-off that either of two SBA files dates, and about 7.9% of all the loans were charged off, not 7.13%. The wall is then at 5.9%, against 8.3% past the fork. It is still the lower in each of the ten approval years and each of the five bands of loan size. In dollars it is 2.4% against 3.2%.
And it is still more than chance. Dealt at random as before, the wall came out that far below in about 1.1% of tries. Without dentists' offices it is about 1.7%. Without dentists and veterinarians it is about 2.6%. The fuller count moves the rates up and leaves the gap in place.
Inside the wall the order is the same. The rate is 4.6% where an owner must hold the licence, 6.1% for a named person and 8.7% for shop and staff. The owner group is lowest and the shop group highest here too.
On a still wider count the wall's rate is again the lower. The earlier of the two files, of 31 March 2026, also marks loans that, on our reading, the SBA has purchased under its guarantee and not charged off. Not all of those will end as charge-offs. Count them as well, and the wall is at about 6.6%, against about 9.3% past the fork. These rates are to one decimal, because they rest on pairing the two files loan by loan.redactedInside the wall the rates step up. Where, on our reading, an owner must hold the licence, 4.12% of loans were charged off. Where a named licensed person must be in charge, it was 5.39%. Where the shop and its staff are licensed, it was 7.92%. That last figure is above the rate for all industries. So not every kind of licence goes with safer loans.
Only one of those gaps is clearly more than chance. We dealt the three readings at random among the wall's 63 industries, 400,000 times. The owner group came out that far below the shop group in 1% of tries. It came out that far below the named-person group in about 1 try in 10. And the named-person group came out that far below the shop group in 8%. The clear gap is between the owner group and the shop group. The named-person group is within chance of both.
Two professions carry about half of the gap between the owner group and the named-person group. Dentists and veterinarians hold 29% of the owner group's loans. Without them its rate is 4.81%, not 4.12%.redactedIndustry by industry, SBA loans behind the wall still look safer. Leave out the six swept in, and 30 industries behind the wall have at least 300 loans. 22 of them are clearly below the all-industry rate. Clearly means by more than twice what sampling alone would produce.
Past the licence fork the share is much smaller. There the same is true of 37 of 179. Four of those 179 are finance and property industries that the series leaves out of scope. Without them it is 35 of 175. Piece 4 adds older loans for commercial printing, which makes it one fewer.
Not every industry behind the wall has safer loans. Five behind the wall are clearly above the rate: beauty salons, nail salons, other personal care, residential property managers and building inspectors.
The fuller count says much the same. On the fuller count of charge-offs it is 20 of the 30 behind the wall, and 37 of the 175 in scope past the fork. Beauty salons are then within chance of the rate. On either count, most of these industries behind the wall are clearly below the rate. Past the fork, most are not.
A caution about cause. This does not show that a licence makes a business safe. Dentists differ from restaurant owners in many ways besides a licence. What the data shows is narrower. Losses are low in a few professions, and high in salons and personal care.
These are SBA loans of every purpose. On acquisition loans alone the file is thin, because the SBA has flagged acquisitions only since fiscalredactedOf the acquisition loans approved in fiscal 2018 and 2019, 52 of 1,591 behind the wall were charged off, or 3.27%. Past the fork it was 314 of 6,271, or 5.01%. That gap is within chance: the wall came out that far below in 12% of tries. The flag is in steady use from January 2018, and five loans approved in December 2017 carry it. So it is on 7.3% of fiscal 2018's seasoned loans and 12.3% of fiscal 2019's.
4.12% of SBA loans were charged off where, on our reading, an owner must hold the licence. Where the shop and its staff are licensed, it was 7.92%. Not every licence goes with safer loans.
In short
5.29% of SBA loans behind the wall were charged off, against 7.53% past the licence fork. On a fuller count of charge-offs it is 5.9% against 8.3%. Either way the gap is more than chance. Inside the wall, the one clear gap is between the owner group, at 4.12%, and the shop-and-staff group, at 7.92%. The named-person group sits between them, within chance of both. None of this is proof of cause.
6. Higher in brokerage and the professions, on old data
Charge-offs are a lender's measure. Profit is the owner's, and the public data on that is weaker.
The IRS publishes income and receipts for corporations, by industry. We used its Corporation Source Book for tax yearredactedThat is the last year it splits each detailed industry by the size of the corporation. For later years, to 2022, its list of tables shows income by size for all industries together, and for S corporations by broad sector. We read the list, not each table, and have not used them.
The Source Book covers corporations only. The IRS has separate tables for partnerships and sole proprietors, which we have not used either. Those firms are not in these figures.
Few of the file's industry groups sit wholly behind the wall. At its finest level the file has 194 industry groups. We classed each one by the Census industries under its code. In 11 groups, every business is behind our wall. In four more, most are. So few groups stand for the wall in the comparison below.
With those limits stated, here is what it shows. Take corporations with assets under $25 million, which is about 99% of the returns in the groups we use. In the 11 wall groups, the median group's net income was 7.69% of business receipts. Median means the group in the middle. In 127 groups, every business is past the licence fork. 117 of them pass our filters, and their median was 3.55%. So the middle wall group kept more of each dollar it took in than the middle group past the fork.
redactedNet income as a share of receipts, the median group on each side
• Corporations with assets under $25 million: The 11 wall groups 7.69%, No-licence groups 3.55%, Wall groups above the no-licence median 10 of 11
• Corporations of every size: The 11 wall groups 5.84%, No-licence groups 4.69%, Wall groups above the no-licence median 7 of 11
The first row sets the wall groups against 117 no-licence groups, and the second against 125. The filters are applied again at every size, and more groups pass them.
Ten of the 11 wall groups are above the no-licence median. Five or six would be if the two sides were alike. We dealt the 11 labels at random among the 128 groups, 200,000 times. Ten or more came out above in under 1% of tries. Chance alone would rarely put that many above. Air transport is the one below, at −1.8%, a loss.
Include the largest corporations and the gap narrows. It is then 5.84% against 4.69% for 125 no-licence groups. Seven of the 11 are then above. That is within chance: 28% of tries. So for corporations of every size, we cannot tell the gap from chance.
The order inside the wall matters more than the average. Real estate brokerage, insurance, law, accounting and dentistry are at the top, between 10.6% and 15.3%. The two licensed trades, electrical and plumbing contractors, are at 5.1% and 4.2%. Personal and laundry services, the group that holds salons, barbers and funeral homes, is at 4.6%.
Pharmacies sit in a group of health and personal care stores. 44% of its businesses are behind the wall, and it is at 3.5%. So the figures are high for brokerage, insurance and a few professions. They are ordinary for the trades, and for the groups that hold salons and pharmacies.
The order does not follow our reading of the licence. Real estate brokerage and insurance agencies, at the top, are named-person industries on our reading. So are the two trades near the bottom.
Is that the licence, or the kind of work? The file cannot say. Most of the no-licence side is unlike the wall: 93 of its 117 groups mine, build, make or sell goods. But its 24 service groups have almost the same median, 3.50%. So the no-licence median is much the same with the goods groups left out.
Inside single sectors the comparison rests on very few groups. There the wall groups are all 15 in the chart.redactedake the sectors in turn. In professional services and construction, the wall groups are ahead. In professional services the four wall groups run from 6.2% to 12.0%. The five no-licence groups have a median of 2.01%. In construction the two licensed trades are at 4.2% and 5.1%, against 2.92% for four no-licence groups.
In transport and in repair, the picture is mixed. In transport the two wall groups split: water transport is above the no-licence groups, and air transport below. In repair and personal services the wall group sits between the other two.
Across the six sectors that have both kinds, 11 of the 12 wall groups are above their own sector's no-licence median. Six would be if the two sides were alike. Without the two groups marked †, it is nine of ten.
On a corporate tax return, officers' pay is a deduction, so net income is what is left after it. Professional firms differ in how much their owners take that way. So these are not the owners' earnings. The IRS withholds some cells to protect single firms. For the two contractor groups it withholds the $10 million to $25 million class, so their figures stop at $10 million. Its figures are estimates from a sample of returns, and it marks many of them as weak. The 15 groups in the chart cover industries with 90% of the wall's businesses. Property managers, investment advisers and pest control firms are in none of them.
In short
On IRS figures for tax year 2013, among corporations with assets under $25 million, the median wall group kept 7.69% of business receipts as net income. The median no-licence group kept 3.55%. Ten of 11 wall groups beat that median, where five or six would by chance. The high margins are in brokerage, insurance and a few professions. The licensed trades, and the groups that hold salons and pharmacies, are at 3.5% to 5.1%. With the largest corporations counted, the gap is within chance. The data is old and covers corporations only.
7. Smaller firms, and perhaps fewer SBA-financed sales
So the wall goes with safer loans and, on old data, higher margins in brokerage and the professions. Two more things go with it. The firms are smaller. And fewer of them change hands with an SBA loan, though that second gap is less sure.
Both sides of the wall
• Businesses with 20 to 999 staff: Behind the wall 96,redacted%), Past the licence fork 438,redacted%)
• Jobs per business: Behind the wall 10.6, Past the licence fork 21.3
• Share of seasoned SBA loans charged off: Behind the wall 5.29%, Past the licence fork 7.53%
• Seasoned SBA loans per 1,000 businesses: Behind the wall 59, Past the licence fork 78
• SBA-financed sales a year, all sizes: Behind the wall 1 in 1,694, Past the licence fork 1 in 970
• SBA-financed sales a year, 20 or more jobs: Behind the wall 1 in 731, Past the licence fork 1 in 583
Past the licence fork here means all 766 industries that clear it. That includes the 25 in finance, property and gambling that pieces 1 and 2 leave out of scope. Without them the charge-off and sale rows of the right-hand column read 7.60%, 1 in 942 and 1 in 567. The left-hand column holds 12 industries of the same kind, with 127,988 businesses. Without them those rows read 5.35%, 1 in 1,626 and 1 in 677. Another 919,879 businesses are in neither column: those with licensed staff or premises, and those we could not classify.
Size is the firm result. 6.0% of businesses behind the wall have a bench of 20 to 999 staff, against 12.3% past the fork. Deal the wall's label at random among industries, and the wall comes out that far below in under 1% of tries. That holds however the labels are dealt. For jobs per business it is under 3%, however we count. This gap is more than chance: firms behind the wall are smaller.
SBA-financed sales are less clear. Businesses behind the wall are 26% of the businesses and 17% of the SBA-financed sales. In a year, 1 in 1,694 changes hands with an SBA loan. Past the licence fork it is 1 in 970. On the face of it, businesses behind the wall change hands with an SBA loan less often.
But is that gap chance? Here the answer swings with how we deal and how we count, from 3% of tries to 13%. Only the low end counts as more than chance. Three things soften that gap.
The first is chance. Sale rates differ so much from one industry to the next that any 69 industries can look slow. We dealt the wall's label at random among all 835 industries, 200,000 times. The wall came out that far behind in 7% of tries. So on this deal the gap is within chance. Luck gives a gap that wide too often to rule out.
That deal ignores size, and the wall's industries are large. They average 23,208 businesses each, against 4,642 past the fork. Deal the labels within six bands of industry size, and it is 3% of tries. So chance is a weaker explanation than the first deal suggests.
Among businesses with a bench, the gap is within chance. The real figures are 1 in 731 against 1 in 583. The wall came out that far behind in more than a quarter of tries. Within size bands it was 11%. For firms of that size, we cannot tell the gap from chance.
The second is that three industries carry 44% of the gap in yearly SBA-financed sale rates. Law offices, real estate brokerages and physicians' offices hold 27% of the wall's businesses and 198 of its 5,672 sales. That is a far bigger slice of its businesses than of its sales. Without them the wall sells at 1 in 1,273 a year. That is within chance: 25% of tries on the first deal. So on that deal, without those industries, we cannot tell the gap from chance.
The third is that the wall's industries have fewer SBA loans of any kind. They have 59 seasoned loans per 1,000 businesses, against 78 past the fork. That gap is itself within chance, at 7% of tries. Count sales per seasoned loan instead of per business, and the wall has 6.0 sales per 100 loans against 7.9. On that measure the wall came out that far behind in about 1 try in 5. So measured against how much these industries borrow from the SBA, the gap in SBA-financed sales is within chance.
So read the sale rate as a lean, not a finding. A dentist buying a practice may borrow from a bank with no SBA guarantee, and we would not see it.
redactedThe range inside the wall is wide, and that part is not chance. 1 funeral home in 267 changes hands with an SBA loan in a year, and 1 pharmacy in 360. For law offices it is 1 in 15,870. For real estate brokerages it is 1 in 18,287. Behind the same wall, some industries change hands with an SBA loan far more often than others.
Across those 22 industries, a fast one sells with an SBA loan 13.8 times as often as a slow one. Fast and slow mean a tenth of the way from the top and from the bottom. Chance alone would give 1.2 times. So the spread is far wider than chance would give.
Here too the rate tracks how much each industry borrows from the SBA. Rank the 22 by each, and the two orders line up at +0.64, on a scale where 0 means no connection and 1 means they move in lockstep. The industries that borrow more from the SBA tend to be the ones that change hands more often with an SBA loan.
That comparison leaves out five finance and property industries, as pieces 1 and 2 do past the fork. Two of them, portfolio management and investment advice, share sales that the loan file cannot place in one industry. With all 27 the gap is 22.4 times.
We looked for a second measure and did not find a clear one. Census asks owners what became of their previous business. By sector, the share sold lines up only loosely with the share behind the wall. The link is −0.37 across 15 sectors, on the same scale. The minus sign means sectors with more behind the wall had fewer sold.
But a link that loose is within chance. Chance alone gives a link that strong about 1 time in 6, counting either direction. Counting only links that far below zero, it is about 1 time in 12. So this measure points the same way, but too weakly to tell from chance.
Professional services is half behind the wall and has one of the lowest sold shares, at 19.4%. Health care is more than half behind the wall and sits above the middle, at 29.5%.
Census Annual Business Survey, reference yearredactedIt asks owners about the previous business they started, and files each answer under the sector the owner is in now. The share is sold out of sold or closed. A survey estimate.
1 in 1,694 businesses behind the wall changes hands with an SBA loan in a year. Past the licence fork it is 1 in 970. Three large industries carry 44% of that gap in yearly rates.
In short
6.0% of businesses behind the wall have a bench, against 12.3% past the fork. That gap is firm. 1 business in 1,694 changes hands with an SBA loan in a year, against 1 in 970. That gap is less sure. Chance put the wall that far behind in 3% to 7% of tries, or 13% counting gaps either way. Three large industries carry 44% of it, in yearly rates. And the wall's industries have fewer SBA loans of any kind. Inside the wall, funeral homes and pharmacies sell most often with SBA loans. Law offices and real estate brokerages very rarely do.
8. What the wall is not
1. "A licensed business cannot be bought"
Sometimes it can. 481,079 businesses are in industries our screen flags for licensed staff or premises, not for a licensed owner. Inside the wall, 40% are in industries where the rule, on our reading, asks for a named licensed person, who can sometimes be hired. And the SBA financed 5,672 sales behind the wall in six years. We cannot see whether those buyers held the licence.
2. "The licensed side is the safe side"
On average, yes: 5.29% against 7.53%. But salons are behind the wall too. Beauty salons' loans were charged off at 7.76%, and nail salons' at 8.56%. Both are above the rate past the fork. The safety sits with a few professions, not with the word "licensed".
3. "Practices change hands all the time"
Not with SBA financing, as far as the loan file shows. 1 business in 1,694 behind the wall changes hands that way in a year. For law offices it is 1 in 15,870. Sales made without the SBA are in no public file that we know of. So behind the wall, sales with SBA financing are rare. Any others are out of our sight.
9. If a licence is in your way
If you are buying: four questions
• Who has to hold the licence? An owner, a named person in charge, or only the staff? Those are three different problems.
• If it is a named person, can you hire one? Then it is a hiring problem. We have not measured how hard that hire is.
• If it is the staff or the premises, the licence may not have to be yours. Expect approvals, and check.
• Check your own state. Our screen reads how licences generally work. It does not read state law.
If you are selling a business behind the wall
• On our reading, most buyers here must hold the licence or employ someone who does. Six industries, with 95,993 businesses, are there only through their group. In our judgement they need no such licence.
• With SBA financing, sales behind the wall run at 1 in 1,694 a year, against 1 in 970 past the fork. That gap is less sure than it looks. Funeral homes and pharmacies run far faster than law offices.
If you are lending or advising
• Do not treat "licensed" as one category. Where, on our reading, an owner must hold the licence, 4.12% of SBA loans were charged off. Beauty salons ran at 7.76%.
• The margin evidence is from 2013, covers corporations only, and sets the wall mostly against goods industries. Use it as a direction, not a level.
10. Assumptions, gaps and limits
Assumptions
• That our reading of each licence is right. Step two is judgement, and we checked no state's statutes. 15 of the 31 occupations rest on a rule that a named licensed person must be in charge. That person can sometimes be hired.
• That our list of occupations is the right one. The survey struck five names from it. It could not measure one, and it put two more within chance of the line. We kept one on at 15.9%, as a premises licence. The survey did not write the list.
• That 25% is the right line for calling an occupation licensed, and that 5% and 5 times are the right bars for an industry. Section 4 shows what moves.
• That an industry group can be flagged as a whole. Six industries are behind the wall only through their group, and 16 more may be.
• That charge-offs on SBA loans say something about businesses that never borrow from the SBA.
• That margins from 2013 still rank industries the same way. The levels will have moved.
Where the reading sits badly
• Building inspection services are counted where an owner must hold the licence, because they are filed with engineers. Building inspectors themselves are on our staff list.
• Chiropractors', optometrists' and therapists' offices are read as owner-licensed. Mental-health practitioners' offices are set aside for licensed staff.
• Nursing-home administrators are on the list as a named person. Nursing homes are set aside for licensed staff all the same, because one administrator is a tiny share of the jobs.
• Tobacco stores and fuel stations are set aside for permits on the premises. Liquor stores and bars are not, though a liquor licence is a permit of the same kind.
• Podiatrists' offices are flagged only through massage therapists, who share their BLS group.
• Real estate appraisers are read as a named person. Surveyors, whose firms our notes describe the same way, are read as an owner.
Data gaps
• No public record, that we know of, of who must hold a licence, state by state and industry by industry.
• No margin data by detailed industry and size together after tax yearredactedFor later years the IRS's list of tables shows detailed industries without a size split, and income by size for all industries at once. For S corporations it shows income by size for broad sectors. For all corporations by sector and size, it shows only the spread of assets. It has separate tables for partnerships and sole proprietors. We read the list, not each table, and have not used them.
• No count of sales made without an SBA loan.
• No licence class for 438,800 businesses, seven in ten of them in retail. 84,356 of them are schools, training providers and hospitals. A BLS file that the screen did not read would class those. The 2022 industry list would class 53 more industries. For the last nine, with 80,949 businesses, BLS has no row that holds their jobs.
• No count of owners. The survey measures workers, and BLS's job counts leave out self-employed owners and partners. A licence that only the owner must hold shows up as a low share. Private detectives were struck from the list at 18.6%, and that may be such a case.
Limitations
• The screen left 81 industries unclassified. On a fuller matching to BLS's rows, two of them would pass every test. That gives 36 industries, not 34. Section 4 gives the detail.
• Licence rates are self-reported, and one household member often answers for the others. A worker may be wrong about who issued a credential.
• The 495,667 records are not that many people. The same worker can appear in several months, and our extract cannot tell them apart. A rate measured on 60 records may rest on far fewer people.
• No survey was taken in October 2025, during the federal shutdown. BLS's annual figures rest on the same 11 months.
• Occupations are measured in broad groups, which can hide a licensed one. Bus drivers sit in a group with taxi and shuttle drivers. Its rate is 22.9%, under the line. So bus operators are past the fork, while trucking is set aside.
• The survey asks whether a worker holds any government licence, not which one. 30.2% of accountants report one, and we read that as the accountancy licence.
• The IRS groups are broad. In four of the 15 wall groups, some businesses are not behind the wall. For 37 groups whose code is an IRS combination, we read the coverage from the title. For six broader groups we read it from the IRS's own list of industries.
• The charge-off comparison is between industries. It is not a comparison of like businesses.
• Charge-offs are counted on the SBA's file of 30 June 2026, which drops the date and the amount from a loan later marked paid in full. Section 5 gives the rates on a fuller count, and the chance check on the wall against the fork. The other chance checks were run on the June count only.
• Charge-off rates count loans. Section 5 shows the gap also holds in dollars, by year, by loan size and inside the same lenders. We have not tested it further than that.
• The acquisition-only comparison rests on 52 charged-off loans behind the wall.
• The sale-rate gap between the two sides is within chance on one deal of the labels, and outside it on another.
• Unless we say otherwise, a deal of the labels is counted one way: how often the wall came out that far below. Counting gaps in either direction gives a larger share. For sales on the first deal it is 13% of tries, not 7%. The sector link in section 7 is given both ways.
• The industry on a loan is the lender's choice. The industry on a firm is Census's. Every sale rate here divides one by the other.
• Sale rates count SBA-financed sales only, and a firm once in each industry it works in.
The full version is on acquidex.com. It carries every table, the list of all 69 industries, every source and the complete method.
This is part three of five. Part four names the 34 industries that pass every test, and what each one trades away.
________________________________________
Sources
Every count here is public, free to download and recomputable. The SBA figures are from the agency's own loan-by-loan releases.
• Census Statistics of US Businesses, 2021 firm-size tables: census.gov/programs-surveys/susb.html
• BLS Occupational Employment and Wage Statistics, May 2024: bls.gov/oes/
• BLS Occupational Employment and Wage Statistics, frequently asked questions, read 9 October 2026: bls.gov/oes/oes_ques.htm
• BLS Current Population Survey, certification and licensing, table 49, 2025 annual averages: bls.gov/cps/cpsaat49.htm
• BLS Current Population Survey, questions and answers on certifications and licences, read 9 October 2026: bls.gov/cps/certifications-and-licenses-faqs.htm
• Census Current Population Survey methodology, collecting data, read 9 October 2026: census.gov/programs-surveys/cps/technical-documentation/methodology/collecting-data.html
• 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
• IRS Statistics of Income, Corporation Source Book, tax year 2013: irs.gov/statistics/soi-tax-stats-corporation-source-book-us-total-and-sectors-listing
• IRS Statistics of Income, Corporation Complete Report, tax years 2014 to 2022; not used here: irs.gov/statistics/soi-tax-stats-corporation-complete-report
• Census Annual Business Survey, owner characteristics, reference years 2017 to 2023: census.gov/programs-surveys/abs.html
• Census County Business Patterns, 2017 to 2023: census.gov/programs-surveys/cbp.html