How the vehicle recall figures are made

What a recall is in NHTSA’s file

NHTSA’s recalls file has a row for each make, model, model year and component a recall names. Rows are grouped by the campaign number NHTSA assigns, and a campaign is what these pages count as one recall: a campaign is counted once, however many makes, models, years and components it names. A campaign that names two makes is counted under each make, so a make’s count is of campaigns that name it, and the counts of different makes are not added.

A recall is counted by the date NHTSA received the manufacturer’s report (the Part 573 report). That is not the date owners were told, the date the defect was found or the date the vehicles were built. The newest year in the file is partial, so a year is drawn and ranked only once it is complete: 2025 is the latest complete year, and 2026 is held out.

Each row of the file also carries NHTSA’s “do not drive” field, Yes or No. These pages report it as NHTSA records it and do not judge the risk it describes.

The types, and why type I counts as a vehicle recall

NHTSA labels each campaign with a type, printed here as it is: V, E, T, C, I or X. V, E, T and C are vehicle, equipment, tire and child restraint. A vehicle recall here is a campaign of type V or type I, because NHTSA’s own campaign file calls the type I campaigns “Vehicle”, so they are counted with type V; the pages do not decode the other labels beyond that. The file holds 26,508 campaigns for vehicles, 26,455 of type V and 53 of type I, and 2,872 of type E, 789 of type T, 251 of type C and 11 of type X.

“Potentially affected”

“Potentially affected” is the manufacturer’s estimate of how many vehicles or items a campaign covers, repeated on each of its rows, and the file does not revise it once a remedy starts. A vehicle can be in several campaigns, so the estimates from different campaigns overlap. For that reason no page adds them into a number of recalled cars, and none puts an estimate in a title.

The VIN question

A campaign names makes, models and years, and sometimes a production window. Whether one car is in a campaign, and whether its remedy has been done, depends on its VIN, which this site does not hold. That answer is NHTSA’s recall lookup’s, and the manufacturer’s. These pages tell a reader which campaigns the file names for a make, never that a particular car is or is not recalled.

Complaints, investigations and ratings

These pages cover one NHTSA file: the recalls file, which lists each campaign a manufacturer has reported. They do not cover NHTSA’s complaints, its investigations or its safety ratings. A complaint is what a person reported to NHTSA, unverified, so counts of complaints reflect awareness and how many vehicles are on the road, not how many defects there are. An investigation is NHTSA’s inquiry, not a finding that a defect exists. Ratings are given per vehicle variant, ties are common, and the rules changed in 2011, so they cannot be averaged or ranked here.

The check against an independent value

The first question for every figure here is what independently produced value it is checked against. Today there is none.

What the figures are held to is the file read two ways: the campaigns of 2025 counted in the by-year table equal the campaigns the year’s list holds (they agree), and the table of makes names the same year (it does). That is one file checked against itself, and it cannot catch a fault in the file. NHTSA’s own campaign file is a second output of the same system, so agreement with it is not independent either.

The independent value would be NHTSA’s annual recall report, and these figures have not yet been checked against it. They are from the file retrieved on November 3, 2026.

The license, and NHTSA as the source

NHTSA’s record for its recalls on the Department of Transportation’s data hub names the license “Public Domain U.S. Government”. The government’s page on its works adds:

You cannot use government materials in a way that implies endorsement by a government agency, official, or employee. You also cannot use federal government trademarks or federal government agency logos without permission.

Every page names NHTSA as the publisher and carries no NHTSA or Department of Transportation seal or logo, and none claims to be official.

What this data cannot show

• A recall names makes, models and years, not vehicles: whether one car is in a campaign or has been repaired is the answer of NHTSA’s own VIN lookup.
• “Potentially affected” is the manufacturer’s estimate for a campaign, and a vehicle can be in several campaigns, so no page adds the estimates into a number of cars.
• A count of recalls is not a measure of safety: it reflects how many models a maker sells, how it divides a defect into campaigns and how readily it reports.

Sources

• NHTSA Office of Defects Investigation — Recalls flat file — National Highway Traffic Safety Administration, Office of Defects Investigation · U.S. Government work · Updated: Daily