The rollover test in the government's five star car ratings has not caught a single vehicle in fifteen years, and that is the reason given for removing it

The test is being removed because nothing fails it.

That is the case the National Highway Traffic Safety Administration makes for dropping the Fishhook manoeuvre, the driven rollover test inside its five star car ratings, in a request for comments filed on Friday morning under docket NHTSA-2026-1651. Every light vehicle subject to the electronic stability control standard would stay on its wheels through the manoeuvre, the agency writes, as shown by no tip-ups in fifteen years of testing.

So the test still runs, and it has stopped telling anybody anything.

What survives, and what a buyer sees

Two things carry the rollover rating and only one of them is going.

The Static Stability Factor, adopted in 2001, is a measurement of how a vehicle is shaped rather than how it behaves when driven hard. The Fishhook manoeuvre, adopted two years later, is the driven half. The agency says the risk figure is mainly influenced by the first of the two, because most real world rollovers are tripped events, and that it will keep publishing the rollover rating on its website and on the Monroney label, the sticker in the window.

Dynamic testing does not end either. It moves entirely to compliance work under the stability control standard, using a different manoeuvre, and the agency argues that satisfies the mandate Congress wrote into the TREAD Act of 2000 and reinforced five years later. The Fishhook run would stop with model year 2027 vehicles.

The second test failed in the other direction

The side air bag proposal is not the same argument in different clothing, and the distinction is worth holding on to.

The Fishhook is being dropped for catching nothing. The out of position side air bag protocol catches roughly two vehicles a year, and the agency writes that not one of those failures has ever been connected to a real world incident in those models. It then states the reverse case: it found a crash in which an occupant was injured by a side air bag deployment in a vehicle that had passed every one of those tests.

The test that flags cars nobody was hurt in also passed the car somebody was hurt in.

The agency's explanation is repeatability. Small changes in how a crash dummy is positioned produce very different injury readings, it writes, and because the protocol allows positions to be adjusted case by case to find a worst case, results have been hard to repeat and reproduce. The consequence is stated plainly in the document: the agency has been unable to open recall proceedings on any failure the test has produced.

The paperwork behind it

The protocol asks manufacturers to submit data for up to 30 test positions per vehicle, with agency staff spot-checking somewhere between 30 and 40 percent of crash-tested models. From model year 2026 that submission had already been cut back to an attestation of pass or fail plus the internal report numbers.

Three dummies are used: a fifth percentile female and Hybrid III three-year-old and six-year-old child dummies, placed close to a deploying bag and measured against injury values the document describes as very similar to those in the frontal air bag standard.

The same filing proposes updates to the mid-term and long-term halves of the ten year plan the agency published in December 2024. Comments are due 60 days after publication, which the document sets for 17 August. It is signed by Jonathan Morrison, the administrator.