7% of Car Recalls Reach Back to a Decade-Old Model Year. They Carry 28% of Every Vehicle Ever Recalled - and NHTSA, Not the Carmaker, Opens Them

Key facts

• 367 of 5,163 campaigns (7.11%) reach back to a model year that was already ten or more years old when the Part 573 report was filed. They cover 146,570,635 of 515,140,077 recalled units: 28.45%, a 4.00x over-representation.
• Remove every campaign whose component field mentions an air bag and the flip gets stronger: 4.39% of campaigns, 18.83% of units, 4.29x.
• On medians rather than sums - which no single giant campaign can move - a reach-back campaign covers 111,562 vehicles against 4,556 for the rest. 24.5x.
• The median recall is a near-new-car event: the median campaign's oldest affected model year is one year older than the filing year, and 32.19% of campaigns are filed in the same calendar year as the newest model year they cover.
• The regulator opens them. ODI influenced 24.35% of non-airbag reach-back campaigns against 7.57% of non-airbag recalls on cars under ten years old, with a clean monotone gradient from 0.93% at model-year-or-earlier up to 24.35%.
• A finding that failed its check, reported as failed: the per-manufacturer ranking of reach-back share collapses once airbag campaigns are excluded. Honda 14.58% to 5.88%, BMW 11.05% to 5.52%, Nissan 6.20% to 1.87%. No manufacturer claim is made in this article.

How old are cars when their recall is filed?

Mostly very young. Collapsing the flat file to one record per campaign and taking the earliest affected model year, the distribution is heavily weighted to cars that are new or nearly new - and the units run the opposite way.

Two thirds of campaigns land on cars no more than two years past their model year, and together those carry about a fifth of the units. The 7.11% that reach back a decade carry 28.45%. The median campaign size rises monotonically down the table, from 587 vehicles to 113,844 - a factor of 194.

Is this just the Takata airbag recall?

No, and that was the first thing tested, because it is the explanation that would have made the finding worthless. The Takata inflator campaigns are genuinely concentrated here: airbag campaigns reach back ten years 22.69% of the time against 4.39% for everything else, and nine of the twelve largest reach-back campaigns are inflator recalls.

But excluding them does not weaken the pattern. It strengthens it.

The non-airbag reach-back campaigns are a recognisable set in their own right: GM's ignition-switch campaigns of 2014 (14V400000, 5,877,718 vehicles across model years 1997-2008; 14V355000, 3,141,731; 14V252000, 2,440,524), Honda's 2023 fuel-pump recall (23V858000, 2,490,460 vehicles over model years 2013-2023), two Tesla software campaigns reaching back to 2012 model years (24V051000, 2,193,869; 23V838000, 2,031,220), and Ford's roof-and-pillar recall (24V031000, 1,889,110).

Are reach-back recalls only bigger because they cover more model years?

This is the mechanical explanation - more model years in a campaign means more cars in it - and it fails on its own terms.

Reach-back campaigns do span more model years: a median of six against two. But the gap between old and recent campaigns is widest, not narrowest, when span is held constant at a single model year.

A recall on a single model year that is already a decade old covers a median 16,900 vehicles. A recall on a single recent model year covers 755. Reweighting every reach-back campaign to the median size of its span peers leaves the actual unit total 5.5x higher than predicted.

The honest reading is a filter, not a defect rate: a small problem on a ten-year-old car never becomes a recall at all. By the time a defect is recalled a decade after the fact, the evidence has had to accumulate across a large population, which is precisely why the surviving campaigns are large. Note the single-model-year row rests on 25 campaigns, and 12 after airbags are removed, so it is the least certain number in this article even though it is the most striking.

Who opens a recall on a decade-old car?

The regulator, far more often than on a new car - and unlike the manufacturer ranking, this result survives every check applied to it.

NHTSA's recall file records whether a campaign was influenced by the manufacturer (MFR), the Office of Defects Investigation (ODI) or the Office of Vehicle Safety Compliance (OVSC). Excluding airbag campaigns throughout, the ODI share rises monotonically with the age of the oldest affected model year.

That is a 26-fold difference between the newest and oldest bands, and 3.2x between the under-ten and ten-plus groups. It is stable over time: 3.23x in 2010-2017 and 3.17x in 2018-2026. Within the reach-back group, the ODI-influenced campaigns are also the larger ones, with a median 211,438 units against 62,812 for manufacturer-initiated campaigns.

Read carefully, INFLUENCED_BY records who the filing says drove the campaign, not a finding of fault. What it supports is narrow and still useful: a recall on an old car is much more likely to be the end of a regulatory process than a manufacturer's own catch.

Does this still happen, or is it a story about 2014?

It still happens, at the same rate as a share of campaigns. Of the 311 campaigns filed in 2026 up to 23 October, 22 (7.07%) reach back ten or more model years - almost exactly the long-run 7.11%.

The units are the part that is period-specific. Those 22 campaigns carry 3,657,565 vehicles, 10.98% of 2026's 33,319,082 recalled units, well below the long-run 28.45%, because 2026 has produced no campaign on the scale of Takata or the GM ignition switch. The frequency is a stable feature of the file; the 28.45% unit share depends on a handful of historic campaigns and should not be quoted as a forward-looking rate.

Recent examples, all filed in 2026: Ford's transmission campaign 26V237000 (1,392,935 vehicles, model years 2015-2017, filed 14 April), Honda and Acura's frame and rear-suspension campaign 26V365000 (965,561 vehicles, model years 2014-2023, filed 4 June), Ford's body campaign 26V448000 (288,314 vehicles, model years 2016-2019), and Ford's evaporative-emissions campaign 26V369000 (257,868 vehicles, reaching back to the 2012 model year).

What a VIN check can and cannot tell you here

The practical sequence for an older used car is short: pull the VIN's written history to see what the car has been through and whether recalls are on record, then run the VIN through NHTSA's own free recall tool for open-campaign status, then ask a franchised dealer to read the manufacturer's repair history against that VIN before you buy. A Zilocar report costs $39.50 for a single report, or $29.50 for the first month and then $59.75 a month for 20 reports.

Method

Source. NHTSA Office of Defects Investigation recall flat file FLAT_RCL_POST_2010.zip, from https://static.nhtsa.gov/odi/ffdd/rcl/, downloaded 2026-10-29 at 07:10 UTC. Archive 15,022,318 bytes; FLAT_RCL_POST_2010.txt 312,574,896 bytes; member timestamp 2026-10-28 07:02; server Last-Modified 2026-10-28 07:05:14 GMT. Tab-delimited, no header row. Field positions pinned against the authoritative layout RCL.txt in the same directory (last updated May 2025), which documents 29 fields.

Freshness assert, run before any computation. The newest RCDATE in the file is 2026-10-23, six days before the run. A recall or complaint flat file from this source can silently regress to a years-old extract under a fresh timestamp, so the newest record date - not the byte count and not Last-Modified - is checked first and the run aborts if it is more than 30 days stale.

Universe, N = 5,163. The file carries 246,908 rows resolving to 15,361 distinct campaigns. It is collapsed to one record per CAMPNO before anything is counted, because POTAFF, RCDATE and INFLUENCED_BY are campaign-level fields repeated on every campaign-model-year row; a row count multiplies them. Restricted to RCLTYPECD = 'V' (13,724 campaigns) and then to the 20 passenger-car and light-truck Part 573 filers listed in the part573_filer column of the frozen CSV, giving 5,163 campaigns covering 515,140,077 units, filed 2010-01-04 to 2026-10-22.

Why the filer restriction matters. · RCLTYPECD = 'V' means vehicle, not car. The largest V filers in this file include Forest River (626 campaigns), Daimler Trucks North America (501), Jayco (266), Blue Bird Body Company (207), Altec Industries (206) and Keystone RV (195) - recreational vehicles, heavy trucks, buses and towable industrial equipment. A used-car claim computed on the unrestricted V basis would not be a used-car claim.

Age definition. Model-year age is the calendar-year difference RCDATE year minus earliest affected YEARTXT; age_newest_my uses the latest affected YEARTXT. The 117 campaigns whose only model year is the 9999 unknown sentinel are dropped. This is a calendar-year difference, not months on the road: a car sold in late 2015 as a 2016 model reads one year younger than its actual age, so the age bands are conservative by up to a year at the boundary.

Why YEARTXT and not BGMAN/ENDMAN. The manufacturing-date fields would give a sharper measure and are unusable. Their coverage across V campaigns collapses from 98.7% in 2018 to 9.1% in 2020, 1.3% in 2021 and 1.3% in 2023, recovering only to 59.0% in 2026. Any series built on them would measure which campaigns reported the field. YEARTXT is populated for 100% of campaigns in every filing year from 2013 onward and 99.15% overall.

Cleanup and dedup. Deduplication is on CAMPNO. Component, make and model values are unioned across a campaign's rows, so is_airbag_campaign is true if any row of the campaign names an air bag. Filer attribution uses MFGTXT (the Part 573 filer), not MAKETXT: a single campaign routinely spans several makes, and MAKETXT carries spelling variants that split one manufacturer across rows.

Aggregation. Campaign counts and POTAFF sums by age band; medians reported alongside every sum, because a handful of historic campaigns dominate the sums.

Reversal checks, all reported either way.

1. Airbag exclusion. Headline flip moves from 4.00x to 4.29x - it strengthens.
2. Span control. Holding model-year span at exactly one year, reach-back campaigns still carry a median 16,900 units against 755; span-matched reweighting leaves actual units 5.5x above prediction.
3. Outlier removal. Dropping the twelve largest reach-back campaigns leaves 3.19x.
4. Period stability. 4.25x (2010-2017), 3.41x (2018-2026), 4.29x (2022-2026).
5. The check that failed. The per-manufacturer reach-back ranking does not survive the airbag exclusion: Honda 14.58% to 5.88%, BMW 11.05% to 5.52%, Nissan 6.20% to 1.87%, General Motors 10.24% to 5.69%, Mazda 22.55% to 10.67%, while Mercedes-Benz moves from 3.09% to 3.26%. The raw ranking was largely one supplier's inflator recall distributed across the brands that fitted it. No manufacturer-level claim appears in this article, and the per-filer table is deliberately not published as a ranking.
6. ODI finding re-tested. The regulator-influence gradient survives the airbag exclusion (3.22x), is monotone across five age bands, and is stable at 3.23x (2010-2017) and 3.17x (2018-2026).

Corroboration. The twelve largest reach-back campaigns were independently re-queried against https://api.nhtsa.gov/recalls/campaignNumber. 12 of 12 matched POTAFF exactly; 10 of 12 matched the report-received date exactly and two differed by one day, a date-rendering difference rather than a data defect.

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, still in the United States, or ever actually defective. This file can support "how much exposure was opened" and cannot support "how many dangerous cars are out there" or any model-level worst-car ranking.
• Because POTAFF is exposure rather than survivorship, the 28.45% unit share overstates the share of currently circulating used cars affected, since the oldest cohorts have had the most time to be scrapped.
• The 28.45% figure depends on a handful of historic campaigns. As a share of campaigns the pattern is stable (7.07% in 2026 against 7.11% long-run); as a share of units it is not, and should not be quoted as a forward-looking rate.
• INFLUENCED_BY records who the filing says drove the campaign, not a legal finding.
• NHTSA republishes this file continuously, so a later download will not reproduce these counts exactly. The frozen CSV is the reproducible artefact.

Frozen dataset. All 5,163 campaigns with every field used here - campaign number, report date, earliest and latest model year, span, both age measures, units, filer, makes, component, airbag flag and reach-back flag - are published at https://zilocar-news.github.io/zilocar-rss/data/recalls-on-older-used-cars-nhtsa-data-vin-check-2026-2026-10-29.csv. Every number in this article can be recomputed from it, including the checks that failed.

Sources

• NHTSA Office of Defects Investigation recall flat files, https://static.nhtsa.gov/odi/ffdd/rcl/ - FLAT_RCL_POST_2010.zip, retrieved 2026-10-29 07:10 UTC. Last verified 2026-10-29.
• NHTSA ODI recall file layout, https://static.nhtsa.gov/odi/ffdd/rcl/RCL.txt - field definitions, last updated May 2025. Last verified 2026-10-29.
• NHTSA recall API by campaign number, https://api.nhtsa.gov/recalls/campaignNumber - used to cross-check 12 campaigns. Last verified 2026-10-29.
• NHTSA VIN recall lookup, https://www.nhtsa.gov/recalls - the authority for open-recall status on one vehicle. Last verified 2026-10-29.
• 49 CFR Part 577, Defect and Noncompliance Notification, https://www.ecfr.gov/current/title-49/subtitle-B/chapter-V/part-577 - owner notification requirements. Last verified 2026-10-29.
• NMVTIS, https://vehiclehistory.gov - the authority for legal title brands. Last verified 2026-10-29.