Has Consolidation Actually Reduced the Number of Body Shops?

Short answer: No. US collision repair establishments went from 34,510 in 2015 to 35,422 in 2024, up 2.6 per cent, straight through the heaviest acquisition period the trade has had. But all private establishments grew 27.4 per cent over the same years, so collision repair fell 19.4 per cent in relative terms.

By Anthony Braswell for Quorum Industries LLC, The Autobody Directory · Updated 2026-09-01 · How this was written, and what the machine may not do

How this page was produced: Researched, drafted and checked with AI assistance under human direction, and signed off by the named author. How this site is written

Both come from the same file, and anyone giving you one without the other is selling you something. The figure that cuts the other way is in that file too: average annual pay in the industry rose 47.2 per cent across those years against 43.4 per cent for all US private industry.

The count, from the file it lives in

[REPORTED] The Quarterly Census of Employment and Wages counts establishments with payroll covered by unemployment insurance. For NAICS 811121, private ownership, national, annual averages:

Every row was pulled on 2026-09-01 from www.bls.gov, one file per year. Establishments rose by 912, or 2.64 per cent, over nine years, and employment rose 11.8 per cent.

The five rows above are a sample, so here is the whole decade. All ten annual files were pulled, not only the five shown. The count fell in three of the nine year-over-year steps and rose in the other six: down 29 in 2017, down 4 in 2020, down 162 in 2024. That is the agency's arithmetic and not ours, because QCEW publishes the change itself: oty_annual_avg_estabs_chg reads -29, -4 and -162 in those years and oty_annual_avg_estabs_pct_chg reads -0.1, 0.0 and -0.5. The decade low is 2015 at 34,510 and the decade high is 2023 at 35,584. Anybody who tells you the count has never fallen has not opened the file, and anybody who tells you it collapsed has not either.

The definitional break people cite here does not exist, and we checked before relying on it. The claim in circulation is that NAICS 2022 folded automotive glass replacement into 811121, so part of the gain is reclassification rather than new shops. The Census concordance says otherwise: in the 2017 to 2022 NAICS concordance 811121 maps to 811121 and 811122 maps to 811122, each with no change flag, and the 2022 NAICS structure file still lists 811122, Automotive Glass Replacement Shops, as a separate six-digit code. The QCEW files used here carry a separate national private row for 811122 in every year from 2015 to 2024. No reclassification adjustment is owed to the 2.64 per cent, and none is made.

That rules out exactly one thing. It is inconsistent with the strong collapse story, which says consolidation has been closing collision locations in bulk; at that scale the count would fall. It is entirely consistent with consolidation, because acquisition transfers an establishment without closing it. A shop bought on Monday is the same establishment in the same QCEW cell on Tuesday, and QCEW cannot see the owner.

The number the flat line is usually quoted without

[REPORTED, our computation from the same files] The same file carries industry code 10, all industries, private ownership. Comparing it to 811121 needs no deflator, because it is the same instrument in the same years.

Collision repair grew 2.64 per cent. The private economy it sits inside grew 27.39 per cent. Collision repair's share of all private establishments fell from 0.3741 per cent to 0.3014 per cent, a relative decline of 19.4 per cent. On the full ten-year series that share fell in every single year from 2016 to 2024, with no exception and no year of recovery.

That does not rescue the collapse story; nobody's body shop closed because a software company opened an office. It demolishes the complacent reading just as firmly. Holding steady while the denominator grows by more than a quarter is a sector that stopped adding locations. The 19.4 per cent is the whole of the relative decline on this instrument, with no reclassification adjustment behind it, for the reason given above.

The boundary on that comparison, and the test it survives. QCEW's all-industry growth is itself partly a very large post-2020 increase in very small employers, and neither series is adjusted for that. So cut the window at 2019, before that surge: from 2015 to 2019 collision repair establishments grew 1.14 per cent, all private establishments grew 7.68 per cent, and collision repair's share fell 6.07 per cent in relative terms. Same file, same instrument, four pre-pandemic years. The gap is not an artefact of 2020. What the comparison establishes is still narrow: on one instrument, over one window, collision repair added locations at roughly one tenth the rate of the economy around it.

Pay, which goes the other way

[REPORTED, our computation, same files, same years] Average annual pay in NAICS 811121 went from $45,935 in 2015 to $67,639 in 2024, up 47.2 per cent. Across all US private industry over the identical period on the identical instrument, it went from $52,876 to $75,850, up 43.4 per cent. Collision repair pay grew 3.8 percentage points faster and closed part of the gap to the wider labour market, from 86.9 per cent of the all-private average to 89.2 per cent. Both are nominal, so no deflator was applied.

That cuts against the pay-suppression story, with a limit in the same paragraph: this measure divides establishment payroll by establishment employment, including managers, estimators, office and parts staff. We are not claiming that a body technician's pay rose 47.2 per cent. Payroll per covered employee did.

How big an American body shop actually is

[REPORTED] The size distribution comes from Census County Business Patterns 2022, file cbp22us.txt inside cbp22us.zip, 778,618 bytes, read 2026-09-01. NAICS 811121, all legal forms:

Fewer than ten employees: 26,499, or 76.4 per cent. Fewer than twenty: 32,398, or 93.5 per cent. Fifty or more: 215, or 0.6 per cent. One hundred or more employees: thirty-five establishments in the entire United States. Thirty-four of those are printed; the thirty-fifth is the one establishment Census withholds, and the next paragraph is why we are entitled to place it.

The instrument control, because a table you cannot check is not evidence. The seven published classes sum to 34,664 against a published total of 34,665, and Census prints "N" rather than a number in both of the top two classes, 500 to 999 and 1,000 or more. The control that places the missing establishment: across all 2,003 all-legal-form industry rows in that file, 778 rows have every one of the nine size classes published, and every one of those 778 sums to its stated total exactly, off by nothing. Not one row of the 2,003 over-sums, and every row that falls short has at least one suppressed class. The classes partition the total, so the shortfall is the suppressed count, and the one establishment 811121 is short has to sit at 500 or more employees. The 2023 file behaves identically: 757 fully published rows, all exact, and the same shortfall of one.

A second control, on the population. CBP 2022 gives 34,665; QCEW gives 35,136 for 2021 and 35,584 for 2023, so CBP sits below both, by 1.3 per cent against 2021 and 2.6 per cent against 2023. The two programmes use different reference periods and different treatment of small employers and they still land within three per cent of each other. Two agencies, one answer near thirty-five thousand establishments.

Which release this article runs on, and why it is not the newest one. County Business Patterns 2023 is published, at the same path as 2022, and for NAICS 811121 it gives 35,029 establishments and 246,720 employees, with 31 establishments at 100 to 249, 6 at 250 to 499 and the same single suppressed one above them: thirty-eight at a hundred or more. CBP 2024 returns HTTP 404 at that path, so 2023 is the current release. Every CBP figure in this article is nevertheless taken from the 2022 files, because the metropolitan and state denominators below come from cbp22msa.txt and cbp22st.txt and a numerator must not be divided by a denominator from a different year. The 2023 size distribution, and the shape it shares with 2022, is set out in who owns the body shop you are standing in.

National share is the wrong measure

Nobody drives fifty miles for a collision repair. The relevant market, for a consumer and for an insurer assembling a direct repair network, is a metropolitan area, and a company holding one per cent of the country can hold a quarter of a city.

[REPORTED] Classic Collision publishes its own store list at classiccollision.com/wp-json/wp/v2/wpsl_stores, which returned an X-WP-Total header of 348 when we read it on 2026-09-01. Its robots.txt reads User-agent: * with an empty Disallow and a Crawl-delay: 10, both observed. Against CBP 2022's 34,665 establishments, that is 1.00 per cent of the United States. Classic is the company measured in this article for one reason, and it is not a finding about Classic: it is the only large consolidator whose public store data carries a city, a state and a ZIP for almost every location. Crash Champions publishes 662 location URLs and no geography at all. That 662 is a count of location pages in the company's own sitemap on 2026-09-01 and is not a store count: the only independent store count found for the company is Moody's, which records 648 stores as at the second quarter of 2025, and the two are set side by side in who owns the body shop you are standing in.

Local share, one company, with the working shown

[REPORTED, our computation] Numerator: Classic Collision's 348 store records, of which 346 carry a five-digit ZIP. The two that do not are Oregon City, Oregon and South Denton, Texas, whose published descriptions omit it, and they are excluded from every metro figure below. ZIP to county came from the Census ZCTA-to-county file, 47,863 rows, taking the county with the largest land-area overlap. County to area came from the BLS county-MSA-CSA crosswalk, which lists 3,251 counties of which 1,881 belong to one of 929 areas. Denominator: CBP 2022's metropolitan file cbp22msa.txt, 859 area rows summing to 32,094 establishments, the other 2,571 falling outside any area. All read 2026-09-01. Numerator and denominator are four years apart, a 2026 store list against a 2022 Census count, and nothing was adjusted for that.

One per cent of the United States and 28.6 per cent of Anchorage. A factor of twenty-eight, same company, same day, same data. Both numbers are one company's own store list set against every establishment the Census counts, so both are floors for consolidation and neither is anybody's market share.

Classic's eight Anchorage-area shops are six in Anchorage Municipality and one each in Palmer and Wasilla, both Matanuska-Susitna Borough.

These are floors, and every qualification travels with them. 340 of the 348 shops mapped to an area. Six of the eight that did not fell in counties belonging to no area, which is correct for the rural locations at Marianna, Crockett, Livingston and Kilmarnock and an instrument failure for Fairbanks, whose ZIP 99701 has more land area in the Yukon-Koyukuk Census Area than in Fairbanks North Star Borough. The other two are the records with no published ZIP. The control on the mapping: running the same 346 ZIPs through the Census 2023 delineation workbook instead of the BLS crosswalk maps 339 rather than 340, the single difference being Dayton, Tennessee, and returns every figure in the table above unchanged. More importantly, every numerator is one company's shops against every establishment in the area, counting nothing owned by Caliber, Crash Champions, Gerber, CollisionRight, Quality Collision Group, VIVE, Kaizen or any local multi-shop operator. The true single-market consolidated share is higher, by an amount we did not measure. And the other way: an establishment count weights a four-employee shop the same as a forty-employee one, so these are shares of locations and not of capacity, revenue or repair volume, none of which we measured.

[REPORTED, our computation] The state view needs no geographic inference, because both sides carry a state. CBP 2022 counts 49 collision repair establishments in Alaska, from cbp22st.txt. Classic ran 10 shops in the state when we read its list: 20.4 per cent of an entire state, on the same four-year vintage gap and the same one-company floor as every figure above. The 51 state rows in that file sum to exactly 34,665, the published national total, which is an additive control on the parse.

Where a national platform comes from

[REPORTED] Classic's own store data puts 67 shops in Texas, 56 in Florida and 45 in Georgia: 168 of 348, or 48.3 per cent of the company in three states. Ohio has 2, Maryland 1.

[REPORTED, trade press, publicly served portion of a partly subscriber-gated page] On 20 September 2021 CollisionWeek reported that Classic Collision "announced its expansion into Texas with the acquisition of all 45 ProCare Collision centers", that ProCare "was backed by the private equity firm Kinderhook Industries, LLC, operated in the greater Austin, San Antonio, and Houston, Texas markets", and that Classic "Now operates 150 collision repair centers in ten states." The rest sits behind a login we did not attempt.

Forty-five out of 150 is thirty per cent of the company arriving in one purchase, from one private equity portfolio to another. [REPORTED, our computation] Five years later the residue is visible: of Classic's 67 Texas shops, 19 are in the Houston area, 17 in Austin and 12 in San Antonio, so 48 of 67 sit in exactly the three markets ProCare operated in. We are not saying Classic Collision did anything improper. The point is descriptive: a national platform in this trade is often a regional one with a national website, and its largest state one transaction.

The merger filing threshold, and what it is not built to catch

[REPORTED] Hart-Scott-Rodino notification is denominated in dollars per transaction. Section 7A of the Clayton Act, 15 U.S.C. 18a(a)(2), requires it where an acquisition results in holding voting securities and assets of the acquired person "(A) in excess of $200,000,000 (as adjusted and published for each fiscal year...)" or "(B)(i) in excess of $50,000,000 (as so adjusted and published) but not in excess of $200,000,000 (as so adjusted and published)" together with a size-of-person test.

[REPORTED] The Federal Trade Commission published the 2026 adjustments at 91 FR 2133, "Revised Jurisdictional Thresholds for Section 7A of the Clayton Act", FR Doc 2026-00877, published 16 January 2026, DATES 17 February 2026. The adjusted figures are $133.9 million for the original $50 million and $535.5 million for the original $200 million.

Two numbers circulate in commentary on this industry and both are genuine, so here is which is which. $133.9 million is the floor: below it a transaction is not reportable on size-of-transaction grounds at all. $535.5 million is the level above which the size-of-person test drops away. Anyone citing $535.5 million as the reporting threshold for a body-shop purchase has cited the wrong one.

What follows, as our own arithmetic. The average US collision repair establishment had 7.26 employees in 2024 and 76.4 per cent have fewer than ten. These estates were built a shop or a handful at a time, in transactions of a few million dollars. And the test is written per seller: the statute asks what, "as a result of such acquisition", the buyer "would hold an aggregate total amount of the voting securities and assets of the acquired person". Buying a hundred shops from a hundred different sellers in one metropolitan area is a hundred separate tests against a hundred separate sellers, and not one of those tests is the market. A regime that reviews transactions cannot see a strategy. We did not obtain the value of any individual shop acquisition, because the standard formula here is that terms were not disclosed, so we have not established that no deal in this sector was reportable.

[REPORTED] Not reportable is not the same as untouchable, and the statute says so in terms. Section 7A(i)(1) provides that any action by the agencies, or "any failure of the Federal Trade Commission or the Assistant Attorney General to take any action under this section shall not bar any proceeding or any action with respect to such acquisition at any time under any other section of this Act or any other provision of law." The absence of a premerger filing settles nothing about whether a purchase was lawful. What the arithmetic establishes is narrower than "no law applies": it is that nobody has to tell them.

The 1963 decree, and why nobody can tell you whether it is alive

[CONSENT] United States v. Association of Casualty and Surety Companies, et al., Civil Action No. 63 Civ. 3106, S.D.N.Y., Final Judgment entered 27 November 1963 by McLean, District Judge. The document DOJ publishes, read 2026-09-01, records that the parties consented "without admission by any party with respect to any issue herein" and that it was entered "before the taking of any testimony herein, without trial or adjudication of any issue". We do not claim it is an adjudicated finding of misconduct, and it is regularly described as one.

Section IV(A) enjoins each defendant "from placing into effect any plan, program or practice which has the purpose or effect of" five listed things. The fifth is:

"fixing, establishing, maintaining or otherwise controlling the prices to be paid for the appraisal of damage to automotive vehicles, or to be charged by independent or dealer franchised automotive repair shops for the repair of damage to automotive vehicles or for replacement parts or labor in connection therewith, whether by coercion, boycott or intimidation or by the use of flat rate or parts manuals or otherwise."

That clause is normally quoted with everything after "the repair of damage" removed, and those thirty words are the interesting part, because they reach parts and labour and name flat rate and parts manuals. Four things travel with it, and a clause quoted without its limiters is not a quote. Subsection (B), immediately after, provides that nothing in Subsection (A) prohibits "the furnishing to any person or firm of any information indicating corrupt, fraudulent or unlawful practices" by an appraiser or a repair shop. Section II binds each defendant and "its officers, directors, agents, servants, employees, committees, successors and assigns", and also "all other persons in active concert or participation with any defendant who shall have received actual notice of this Final Judgment", so successors are bound outright and a non-defendant is reached only on those terms. Section VI provides that nothing in Section IV waives "any rights or immunities that defendants may have under" the McCarran-Ferguson Act. Section VIII retains jurisdiction including "for the modification or termination of any of the provisions thereof".

[DISPUTED] Whether the judgment is still in force cannot be established from the public record. DOJ's case page, read 2026-09-01, shows the 1963 Final Judgment, a comment invitation on "the possible termination of this judgment" with a deadline of 2 September 2019 extended to 2 October 2019, and the line "Updated October 14, 2020". No termination order, no withdrawal, no statement of retention. Queried through the Federal Register API on 2026-09-01, "Association of Casualty and Surety Companies" returns count 0 and "63 Civ. 3106" returns count 0. The instrument control, because a zero from an unproved instrument is not a finding: the same endpoint in the same session returns 6,858 documents for "Antitrust Division" and 3 for "judgment termination", among them 60 FR 49421, "United States v. International Business Machines Corporation; Proposed Final Judgment Termination". The zeroes are real.

Nobody should write that the decree is in force. Nobody should write that it was terminated. An injunction whose status neither a repairer nor a journalist can confirm is not doing much work either way. Note also what DOJ posts: not a court-stamped original but a commercial reporter's reproduction.

The measurement Massachusetts ordered twice and nobody produced

[REPORTED] The Massachusetts legislature has ordered the count twice, and the Report of the Auto Body Labor Rate Advisory Board, 22 December 2025, read 2026-09-01, reproduces both mandates. Section 108 of Chapter 182 of the Acts of 2008 told the first Special Commission that "the commission shall also report on the number of auto body shops in the commonwealth from 2000 until present, including the number of shops that have closed during that time period". Section 130 of Chapter 24 of the Acts of 2021, which created the body this report calls the 2022 Special Commission, required a study covering "(iv) the number of auto body shops in the commonwealth each year from 2008 to the present, including the number of shops that have closed during that time period".

[REPORTED] The board that reported in 2025 was never asked for it, and that has to be said before anything is built on its silence. The Auto Body Labor Rate Advisory Board was established by Section 292 of Chapter 238 of the Acts of 2024, and the list of data that section directs it to collect is labour rates in neighbouring states, auto body shop costs, total labour costs, inflation data, workforce data, vocational-technical school trends, insurance premiums, and "any additional information as requested by the advisory board". No shop count. No closures.

[REPORTED] Neither of the two ordered series appears anywhere in the 2025 report. The word "closed" occurs in it exactly twice, both times inside those quoted mandates and never in an answer. The report gives a current registration count, 1,497 shops registered with the Massachusetts Division of Standards, and no historical comparison and no closure count. A board member's own list of remaining questions still asks "What are the reasons for shop closures? Can some information be compiled about the causes?"

What that is worth, at its real strength and no higher. It is not proof that the series cannot be built, because this board was not ordered to build it. It is this. Two Massachusetts commissions were ordered to produce a shop count and a closure count, thirteen years apart, and neither series has been published. The board sitting in 2025 had the registration list inside the room: its fourteen members include the Legal Counsel to the Commissioner of Insurance and an actuary from the Attorney General's office as co-chairs, the Deputy Commissioner and General Counsel of the Division of Standards, which holds the register itself, and John Kwoka, Neal F. Finnegan Distinguished Professor of Economics at Northeastern University, as the statutory economist. It met ten times, it was authorised to request any additional information it wanted, one of its own members asked for exactly this, and the series is not in the report. So every published assertion about collision shop closures should be treated as unsourced until someone produces the document, including those that favour independent shops.

What the count cannot see, and what it undercounts

It cannot see owners: QCEW and CBP count establishments, so the 662 locations in Crash Champions' own sitemap contribute 662 of them exactly as 662 independent owners would. It cannot distinguish stability from churn either, because a flat count is equally consistent with three thousand closing and three thousand opening every year.

And it undercounts businesses exactly where the question is. [REPORTED, our computation] CBP 2022 counts 1,019 collision repair establishments in Massachusetts; the Division of Standards has 1,497 registered auto body shops, a gap of 478 or 47 per cent above the payroll-based figure. Both are right on their own terms: a payroll-based count requires employees covered by unemployment insurance, so an owner-operator with no payroll never appears, while a registration list may retain dormant registrants. We are not saying which effect accounts for how much of that gap. But it is roughly the size of the owner-operator segment, which is precisely the segment a consolidation question is about.

The negative control, stated in both directions

Ask what the record would look like if consolidation were not reducing shop numbers or suppressing pay, then ask how much of it looks like that.

Four things do. Location counts held, 34,510 to 35,422. Pay rose, faster than the private economy. The industry is still overwhelmingly made of very small establishments, 76.4 per cent of them under ten employees and thirty-five in the whole country at a hundred or more, and the two consolidators we counted shop by shop on 2026-09-01, Classic Collision at 348 shops and Crash Champions at 662 distinct location URLs in its own sitemap, against Moody's 648 stores for Crash Champions at the second quarter of 2025, hold 1,010 locations between them on the sitemap count, 2.9 per cent of US establishments, which is a floor and not the consolidated total. And nobody has produced a closure series.

Four things do not. The count held in absolute terms and fell 19.4 per cent in relative terms while the economy around it grew by more than a quarter. The pay base is establishment payroll, not technician wages. The national share is the wrong denominator, and one company alone reaches 40.0 per cent of Lufkin and 20.4 per cent of Alaska. And the count cannot see ownership, which is the one thing the argument is about.

The verdict is narrow. The strong collapse narrative is false on the best available federal data, and the strong reassurance narrative is false too. The count rose 2.6 per cent in nine years, which is a fact about locations, and it has been mistaken for a fact about ownership, competition and consumer choice, none of which it can speak to.

What we could not establish, and the walls we hit

• Firm counts as opposed to establishment counts, and entry and exit. Business Dynamics Statistics and the SUSB firm-size tables would answer this; we did not obtain them.
• Any consolidator's metropolitan share other than Classic Collision's. Crash Champions' 662 location slugs carry no city or state.
• Whether the two Classic records that publish no ZIP, at Oregon City and South Denton, fall inside a metropolitan area. They are excluded from every metro figure here.
• A total consolidated share of US establishments. Our two-company figure of 2.9 per cent is a floor, not an estimate.
• The status of the 1963 judgment. Neither a termination nor a retention order exists in any source we could reach.
• The value of any individual shop acquisition, almost never disclosed.
• A shop closure series for any state or the nation. Massachusetts ordered one twice and neither has been published, which is not the same as proof that none can be built.

Walls and tool failures, kept separate, because they are not the same thing. One real wall. The CollisionWeek article is subscriber-gated: the public page carries the three passages quoted above and then reads "Subscribers need to be logged in to see rest of this article." We did not attempt the login.

Everything else on this list is about our tools, not about the sites, and two of the entries did not survive retesting. www.bls.gov returns HTTP 403 and a bot-policy page to ordinary automated clients, while its robots.txt, read before any fetch, disallows only /scripts, /crs, /_private, /iisadmin, /srchadm, /advisory/members/ and /idcf, with nothing covering /cew/; the QCEW files here were read from that permitted path. www2.census.gov has a User-agent: * group with no Disallow line at all. The Federal Register document page for FR Doc 2026-00877 had served a CAPTCHA during drafting and on retest returned the notice cleanly, so it is not a wall and is not recorded as one; we cite govinfo.gov because govinfo publishes the paginated text. The Census 2023 delineation workbook list1_2023.xlsx had failed a zip CRC check during drafting and on retest downloaded intact, and it is the second crosswalk used in the metro control above. A tool refusing a URL is evidence about the tool, and a wall recorded once and never retested costs more than the wall.

REJECTED

Claims that did not survive verification, absent rather than softened.

• "84 per cent of collision repair shops have fewer than ten employees." Taken from NAICS sector 81, which contains barber shops, pet grooming and private households. The industry figure is 76.4 per cent, so the proxy overstated the case for independents by 7.6 points.
• Top-three-state concentration figures for Kaizen, VIVE and Quality Collision Group, not re-derived here from those companies' own location data.
• A six-company consolidated share of US establishments, which included components we did not re-verify.
• A breakdown of establishments by legal form of organisation. The Census file publishes single-letter codes we could not source to a Census document, so reading one as sole proprietorships is unsourced and cut.

Corrections

Each item was in an earlier internal draft.

• The 1963 judgment was tiered as adjudicated. It was entered on consent, without admission and without adjudication of any issue, so it is tiered [CONSENT].
• The judgment was said not to bind successors unless a court had said so. Section II expressly binds successors and assigns.
• Section IV(A)(5) was quoted ending at "the repair of damage". It continues for thirty more words, quoted in full here.
• Section II was then read as binding nobody beyond defendants and their successors. It also reaches persons in active concert or participation with a defendant who have actual notice, and that clause is now in the sentence.
• Section IV(A)(5) was quoted without Subsection (B), which limits it. (B) is now quoted with it.
• The mapping rate was reported first as 339 of 348 and then as 342 of 348, the second on the belief that all 348 records carry a ZIP. 346 carry one. 340 map to an area on the BLS crosswalk and 339 on the Census delineation file, and the six metro figures are identical either way.
• The mandate reading "from 2000 until present" was attributed to Section 292 of Chapter 238 of the Acts of 2024. It is Section 108 of Chapter 182 of the Acts of 2008. Section 292 orders no shop count and no closure count at all, and the article now says so before drawing anything from the 2025 report's silence.
• CBP 2022's 34,665 was said to sit between the QCEW counts for 2021 and 2023. It sits below both.
• "The only year-over-year fall in the sample is 2024" was true of the five sampled years and false of the decade. All ten annual files were pulled and the three falling years are named.
• $535.5 million was cited as the size-of-transaction threshold. The floor is $133.9 million.

2026-09-01, cross-article audit. Four corrections made while reconciling this page against the other nine in the series.

• The glass reclassification caveat was false and has been removed. An earlier draft said NAICS 2022 folded automotive glass replacement, code 811122, into 811121, and used that to call the 2.64 per cent gain overstated and the 19.4 per cent relative decline conservative. The Census 2017-to-2022 NAICS concordance maps 811121 to 811121 and 811122 to 811122 with no change flag, and the 2022 NAICS structure file still carries 811122 as its own six-digit code. Both sentences are cut and no reclassification adjustment is applied anywhere on this page.
• Thirty-four became thirty-five, and a REJECTED entry was withdrawn. This page reported thirty-four US establishments with 100 or more employees, and separately rejected the inference that the one establishment missing from the published size classes sits in a suppressed class. The rejection was wrong. Of the 2,003 all-legal-form industry rows in cbp22us.txt, 778 have all nine size classes published and every one of the 778 sums to its stated total exactly; no row over-sums; every short row has a suppressed class. The classes partition the total, so the missing establishment is at 500 or more employees and the figure is thirty-five. Fifty or more becomes 215. The same test on the 2023 file gives thirty-eight, which is the figure carried in who owns the body shop you are standing in.
• County Business Patterns 2023 is published and this page does not use it. An earlier draft named CBP 2022 without saying a newer release exists. It does, at the same path, giving 35,029 establishments and 246,720 employees; CBP 2024 returns HTTP 404. This page stays on the 2022 files so that the metropolitan and state numerators and denominators sit in one vintage, and now says so.
• The Crash Champions 662 is labelled as what it is. It is a count of location pages in the company's sitemap on 2026-09-01, not a store count, and it now travels with Moody's independent 648 stores at the second quarter of 2025.

Related

• Who owns the body shop you are standing in
• The shops are profitable, the balance sheets are not
• Fewer crashes, costlier repairs, and what that does to a shop
• What Boyd's SEC filings say about consolidation

Sources

• BLS, QCEW national area slices, NAICS 811121 and code 10, private, 2015 to 2024, at www.bls.gov, one file per year, plus the county-MSA-CSA crosswalk. Read on 2026-09-01.
• Census, County Business Patterns 2022: us, state, metro, plus the 2020 ZCTA to county file and the 2023 delineation workbook used as the second crosswalk. Read on 2026-09-01.
• Census, County Business Patterns 2023, us file, read for the current-release comparison only; the 2024 equivalent returns HTTP 404. Census, 2017 to 2022 NAICS concordance and 2022 NAICS structure, for the 811121 and 811122 codes. Read on 2026-09-01.
• FTC, 91 FR 2133, 16 January 2026, via govinfo, and 15 U.S.C. 18a. Read on 2026-09-01.
• DOJ Antitrust Division, case page and 1963 Final Judgment, plus Federal Register API phrase queries and controls. Read on 2026-09-01.
• Massachusetts, Auto Body Labor Rate Advisory Board report, 22 December 2025. Read on 2026-09-01.
• Classic Collision store records, the Crash Champions sitemap, and CollisionWeek on the ProCare acquisition, 20 September 2021, public portion only. Read on 2026-09-01.

General consumer information: not legal, insurance, or financial advice. Requirements, coverage, and practices vary by state, policy, and manufacturer.

More guides