1 in 31 Used-Car Recalls Exists Because the First Repair Did Not Hold - and 15.9 Million Vehicles Were Called Back a Second Time

Key facts

• 170 of 5,250 car and light-truck recall campaigns since 2010 (3.24%, about 1 in 31) require previously repaired vehicles to come back for a new remedy.
• Those campaigns cover 15,873,272 vehicles (3.31% of the 479,339,224 vehicle-slots in the full set) and reference 175 distinct earlier campaigns.
• Ford: 65 of 641 recalls, 10.14% - more than three times the all-industry rate, and over three times General Motors (3.04% on 494) and Chrysler and Stellantis (3.17% on 567) at comparable volume.
• Toyota 0.32% (1 of 317), Mazda 0.00% (0 of 102).
• The Ford Maverick appears in 11 of the 170 campaigns, the most of any nameplate; the Ford F-150 is next at 10.
• Electrical and software is the largest component group at 50 of 170. Air bags account for 8, so this is not the Takata story.
• Median gap between the original recall and the re-fix: 18 months (n=190 pairs; range 1 to 118 months).
• 27 of the 170 are linked to two or more other campaigns on the same defect, and 42 of the referenced earlier campaigns are themselves re-fixes.

What does it mean when a recall repair gets recalled?

It means the manufacturer told NHTSA, in the remedy section of its Part 573 defect report, that the fix it previously offered is not sufficient and that cars which already received it must be brought back. The language is explicit and standardised. Ford's 2021 Explorer campaign 21V537000 reads: "This recall is an expansion of previous NHTSA recall numbers 16V-245, 19V-435, and 20V-675. Certain vehicles previously repaired will need to return for the new remedy." BMW's 22V119000 reads: "This recall supersedes and expands previous NHTSA recalls 17V-683 and 19V-273. Vehicles previously repaired under the prior recalls will need to come back for the new recall remedy."

That is a different event from an ordinary recall, and it is a different event from a recall that simply adds more vehicles. It is the manufacturer withdrawing a repair it has already performed on cars in customers' hands.

How many car recalls are really re-fixes of an earlier recall?

170 of 5,250, or 3.24%. Restricting to cars and light trucks matters: NHTSA's vehicle label also carries motorcycles, school and transit buses, heavy trucks, motorhomes and trailers, and this analysis excludes them.

The median and the mean disagree in an instructive way. A re-fix campaign is about twice the size of an ordinary recall by median but the same size by mean. So re-fixes are not driven by a handful of giant campaigns; the whole distribution sits higher. The largest single campaign in the set is 5.78% of its vehicle total and the top three together are 14.60%, which is what rules out a one-filing artefact.

Which manufacturers have to redo their recall repairs most often?

Ford, at 10.14% of its car recalls, three times the all-industry rate. The table is restricted to manufacturer groups with at least 50 car and light-truck recalls in the period.

The check that could have killed this ranking, and what it returned. A rate computed over a count can simply track the count: file more recalls, and you have more chances to reference an earlier one. Across these 18 groups the rank correlation between recall volume and re-fix rate is Spearman r = 0.3719 - positive and real, so some of the spread is a volume effect and the bottom of the table should not be read as a clean bill of health for small filers.

But volume does not explain the top. General Motors (494), Chrysler and Stellantis (567) and Volkswagen Group (531) all file recalls at Ford's order of magnitude and sit at 3.04%, 3.17% and 1.88%. Ford is at 10.14% with 641. Three same-sized control groups at a third of the rate is what licenses reading the Ford figure as a property of Ford rather than of its paperwork volume.

Which used models have been through a redone recall most often?

These are campaign appearances, not failure rates per car on the road. A nameplate that sells in volume and stays in production has more opportunities to appear, and the Maverick's 11 come from a model line launched for 2022 - a short, dense run rather than a long one.

The deepest chains are worth seeing directly. Ford's 25V467000, filed July 2025 on the Bronco Sport and Escape, replaces three earlier campaigns at once: "This recall expands and replaces previous recall numbers 22V-859, 24V187, and 25V165. Vehicles previously remedied under 22V-859, 24V187, and 25V165 will need to have the new remedy performed." A 2020 Escape owner who had the fuel-injector work done in 2022 was called back in 2024, again in 2025, and again that July.

What kind of defect most often needs a second repair?

Electrical and software faults, by a clear margin.

Electrical system at 50 of 170 is 29.4% of the set. That is consistent with how a software remedy works: a calibration can be shipped, found incomplete, and shipped again, where a replaced mechanical part usually stays replaced.

Air bags at 8 is the number that settles what this is not. The best-known repeat-recall story in the US is the Takata inflator campaign, where inflators replaced under one recall were later recalled again. No campaign in this set of 170 is a Takata inflator replacement. The phenomenon measured here is a separate and much broader one.

Is this getting worse?

We computed the trend and we are not publishing it as a finding. The share of car recalls carrying return language runs 0.00% in 2010, 2011 and 2012, 1.52% in 2013, 0.00% again in 2014, and then climbs to 4.31% in 2022, 6.81% in 2024 and 12.50% in 2025.

A rate of exactly zero across four separate years is not believable as a real-world rate. Recall repairs failed before 2013. What changed is the writing: the instruction "vehicles previously repaired will need to return for the new remedy" became standard text in Part 573 filings. The series therefore measures documentation practice, and the only sound conclusion from it is that 170 is a floor and the true historical count is higher. Anyone turning this into a "recall repairs are failing more often" headline is reading the filing convention, not the engineering.

What a VIN check can and cannot tell you here

Can: a VIN check, including Zilocar's, shows that recall campaigns exist against a vehicle and how many, alongside accident and damage records, odometer readings and rollback indicators, theft records, junk and salvage auction records, ownership history, sales-listing history and valuation. For this story the useful output is the campaign numbers themselves.

Cannot: no vehicle history report shows recall remedy status, whether a software update was actually flashed to a particular car, or whether the remedy that was performed has since been superseded. That information sits in the manufacturer's Part 573 filing at NHTSA and in the franchise dealer's service record for that VIN. NHTSA's own VIN recall lookup shows open recalls; it does not flag that a closed one was later replaced.

So the sequence that actually works on a used car is: pull the VIN report to get the recall count and the campaign numbers, read those campaign numbers at NHTSA to see whether any says it supersedes or replaces an earlier campaign, and ask the dealer to confirm from the service history which remedy version the car received. A car showing a clean repaired status on a campaign that has since been superseded is exactly the case this data describes, and it is invisible on every report on the market, ours included.

Method

Source. NHTSA Office of Defects Investigation Part 573 recall flat file, FLAT_RCL_POST_2010.txt, from https://static.nhtsa.gov/odi/ffdd/rcl/FLAT_RCL_POST_2010.zip, public domain. Downloaded 2026-10-23 07:13 UTC; server Last-Modified 2026-10-23 07:05:42 GMT; zip member timestamp 2026-10-23 07:02:24; ZipFile.testzip() returned None. 246,638 rows, 29 tab-delimited fields on every row. Field positions are those documented in NHTSA's RCL.txt layout.

Freshness asserted before computing anything. The newest Part 573 report received date (RCDATE, field 16) in the file is 2026-10-20, three days before the run. A stale-but-parseable file is the known failure mode with NHTSA flat files, so freshness is asserted rather than assumed.

Universe. 246,638 rows collapse one-per-CAMPNO to 15,345 distinct campaigns, of which 13,712 carry RCLTYPECD='V'. "V" means vehicle, not car: it includes motorcycles, school and transit buses, heavy trucks, motorhomes and trailers. Campaigns were restricted to cars and light trucks by a two-part screen - the make must be a US passenger-car or light-truck brand, and for makes that also sell heavy, medium-duty or two-wheel product the campaign is dropped when its model or component strings name a heavy or non-car vehicle. Campaigns spanning both car and non-car makes are excluded as mixed. 5,250 campaigns pass. The screen is our judgement and every campaign that passes ships in the frozen dataset so it can be redone or rejected.

Re-remedy definition. A campaign counts when two conditions hold together: its own defect, consequence, remedy or notes text references a campaign number other than its own that exists in the file, and the surrounding text states that previously repaired, previously remedied or already repaired vehicles must return, need the new or an additional remedy, were remedied or repaired incorrectly, were previously inspected but not replaced, or that this campaign supersedes, replaces or reissues the earlier one. 170 campaigns satisfy both.

The reference regex had to be widened, and this would have cost 40% of the set. NHTSA's text writes campaign numbers inconsistently, sometimes hyphenated and sometimes not, within a single sentence: Ford's 25V467000 contains "22V-859, 24V187, and 25V165". A \d{2}V\d{3} pattern silently misses the hyphenated form. The pattern used accepts an optional hyphen or space and optional trailing zeros.

Reversal checks run, and the results either way.

(a) Expansion contamination, the live threat. 31 of the 170 campaigns also add new vehicles to the earlier recall's scope, and an expansion is not by itself a failed repair. Rather than drop them, all 31 were checked for explicit return language and 31 of 31 carry it; the four largest were read in full. Ford's 21V537000 is simultaneously "an expansion of previous NHTSA recall numbers 16V-245, 19V-435, and 20V-675" and a campaign where "certain vehicles previously repaired will need to return for the new remedy". Expansion and re-fix are not mutually exclusive, so the 31 remain in. Excluding them anyway would give 139 campaigns and 10,337,263 vehicles, and the Ford finding survives either basis.

(b) Volume confound on the manufacturer ranking. Rank correlation between a group's recall count and its re-fix rate is Spearman r = 0.3719 across the 18 groups with at least 50 recalls. The effect is positive and is disclosed above: small filers' low rates are partly a detection artefact. The top of the table is not, because General Motors (494 recalls), Chrysler and Stellantis (567) and Volkswagen Group (531) are the same size as Ford (641) and sit at roughly a third of its rate.

(c) Single-campaign concentration. The largest re-fix campaign is 5.78% of the set's vehicle total and the top three are 14.60%, against a median campaign of 9,856 vehicles. The finding is distributional, not one filing.

(d) A Takata check, because that is the obvious alternative explanation. Zero of the 170 campaigns involve a Takata inflator replacement, and air bags are only 8 of 170 component assignments. The set is not a restatement of the inflator saga.

(e) A trend was computed and refused. See "Is this getting worse?" above. Four separate years at exactly 0.00% identify the series as a record of filing convention, which makes 170 a floor and makes any trend claim unsupportable from this file.

Looser definition, for transparency. Admitting campaigns that reference an earlier campaign with supersede-family language but no explicit return instruction gives 233 campaigns and 32,492,865 vehicles. The article uses the tighter 170 throughout.

Corroboration outside the flat file. Campaigns 22V859000, 21V537000, 25V467000 and 22V235000 were pulled individually from NHTSA's public recalls API. Unit counts matched the flat file exactly (521,746; 631,176; 694,271; 594,894) and the remedy text matched the classification in every case.

Aggregation. Campaign counts are distinct CAMPNO. Vehicle counts are POTAFF (field 12) taken as the maximum across a campaign's rows. Manufacturer figures are grouped on a normalised MFGNAME (field 8) because NHTSA carries several spellings of one company and, in Toyota's case, separate regional distributors as their own filers; the raw spellings merged into each group ship in the dataset. Nameplate counts pair each model string with the make on its own row rather than the campaign's first make, which is why Lincoln Corsair appears as Lincoln and not Ford.

Known limits. · POTAFF is the manufacturer's launch-time estimate of units potentially affected. It is not a count of cars still on the road, still unrepaired, or still in the United States, and it is emphatically not a count of how many cars actually had a repair redone - the number that were already repaired when the second campaign opened is not in this file at all. So this analysis supports "how much exposure was re-opened" and cannot support "15.9 million cars got a bad repair". The car and light-truck screen is ours. 2026 is a partial year. Campaign-to-campaign date gaps use the earliest RCDATE on each campaign and are available for 190 of the referenced pairs. We did not contact NHTSA or any manufacturer before publishing.

Frozen dataset. All 170 campaigns with their filing date, filer, makes, models, model years, components, vehicles affected, the earlier campaigns each references, whether it also expands coverage, and the exact language matched: car-recall-repair-redone-already-repaired-vehicles-nhtsa-data-2026-10-23.csv

Sources

• NHTSA Office of Defects Investigation, Part 573 recall flat file, FLAT_RCL_POST_2010.zip - https://static.nhtsa.gov/odi/ffdd/rcl/FLAT_RCL_POST_2010.zip. Retrieved 2026-10-23 07:13 UTC. Last verified 2026-10-23.
• NHTSA public recalls API, campaign records for 22V859000, 21V537000, 25V467000 and 22V235000 - https://api.nhtsa.gov/recalls/campaignNumber. Last verified 2026-10-23.
• Frozen dataset for this article - https://zilocar-news.github.io/zilocar-rss/data/car-recall-repair-redone-already-repaired-vehicles-nhtsa-data-2026-10-23.csv. Last verified 2026-10-23.

Analysis by the Zilocar Data Desk. Zilocar runs US vehicle history reports by VIN, covering recall presence and count, accident and damage records, odometer and rollback indicators, theft records, salvage and junk auction records, ownership and sales-listing history, specifications, safety ratings and valuation. Recall remedy status, open investigations and per-unit software state are not available on any vehicle history report and belong to NHTSA and the franchise dealer.