Safety is supposed to be a ratchet. Structures improve, airbags multiply, scores go up and stay up. Mostly that holds — across 1,815 pairs of consecutive model years for the same vehicle, the rating held or improved 98.6% of the time.
The other 1.4% is 25 cars that got worse, and they are worth looking at because of where the losses landed.
| Vehicle | Years | Overall | Frontal | Side | Rollover |
|---|---|---|---|---|---|
| Ford Transit | 2023→2024 | 4→3 | 4→3 | 5→5 | 4→1 |
| Kia Niro | 2024→2025 | 5→4 | 4→4 | 5→5 | 4→4 |
| BMW X3 | 2024→2025 | 5→4 | 5→4 | 5→5 | 4→4 |
| Hyundai Santa Fe | 2023→2024 | 5→4 | 5→4 | 5→5 | 4→4 |
| Hyundai Tucson | 2021→2022 | 5→4 | 5→4 | 5→5 | 4→4 |
| RAM 1500 | 2021→2022 | 5→4 | 4→4 | 5→5 | 4→3 |
| Volkswagen Passat | 2021→2022 | 5→4 | 4→3 | 5→5 | 4→4 |
The side column never moves
Read that table vertically and one thing stands out immediately. Every single side impact score is 5→5.
Not one of the 25 declines involved side protection. The losses came entirely from frontal impact and rollover — which are precisely the two tests where the industry has not run out of room, and where cars still differ from one another.
Side impact was solved with curtain airbags and reinforced pillars, and once hardware solves a problem it does not un-solve itself between model years. Frontal and rollover depend on structure and geometry, and those change when the car changes.
Three ways a score falls
The first is that the test got harder. NHTSA revises protocols periodically — new dummy types, new measurement criteria, new thresholds — and a stricter exam produces a lower mark on identical hardware. Nothing about the car changed at all.
The second is a redesign. A new generation is a new structure, and a structure optimised for one crash mode can give ground in another. This is the case where the older car may genuinely be better in that specific test.
The third is quieter and catches people out: ratings are published per trim and drivetrain, not per model. A different configuration being submitted for testing changes the published number without anything being redesigned.
The Transit is the one to look at
Most of the declines are a single star on the overall figure. The Ford Transit is different in kind — its rollover score went from four stars to one.
That is not a testing artefact. Rollover resistance is close to a geometry calculation: track width against centre of gravity height. A tall van with a high roof and a load area above the axle line starts from a bad ratio, and a change to the body configuration being tested moves the number a long way.
For a commercial van carrying weight high in the load bay, that is a number with real consequences, and it is invisible in the overall rating unless you open the components.
What to do when you see a drop
Do not assume the older car is safer, and do not assume the drop is meaningless either. Find out which of the three causes applies.
If a whole cohort of vehicles dropped in the same model year, the protocol changed and the comparison across the boundary is not valid — the earlier score is an easier exam, not a better car.
If one model dropped while its competitors held, something specific to that vehicle changed, and it is worth reading which component fell.
Either way the rating describes a factory car. Structural repairs after a collision are never re-tested, and a rebuilt front end does not inherit the score the model earned. A free VIN check tells you whether the car in front of you is still the one that was rated.
Frequently asked questions
Can a car's crash rating get worse?
Yes, though it is rare — 25 times out of 1,815 consecutive model-year pairs in our data, or 1.4%. The Ford Transit fell from four stars to three between 2023 and 2024, and its rollover score dropped from four to one in the same step.
Why would the same car score lower a year later?
Three things cause it. NHTSA revises test protocols periodically and a stricter test produces lower scores on unchanged hardware. A redesign changes the structure, sometimes for the worse in one specific test. And ratings apply to a specific trim and drivetrain, so a different configuration being submitted can move the published number.
Should I buy the earlier model year if it scored higher?
Check what changed before assuming the older car is safer. If the drop came from a stricter test protocol, the older car would likely score the same as the newer one under current rules — its higher number is just an older exam. If the drop followed a redesign, the difference may be real.
Which crash test never dropped?
Side impact. In every one of the 25 declines we found, the side score stayed exactly where it was — almost always at five stars. The falls came from frontal impact and rollover, which are the two tests where cars still differ from each other.
Denis Kataev is a serial solopreneur and the founder of ForCar. With 15+ years in software engineering and 10 years in SEO, he builds data-driven products end to end — backed by a sharp eye for design. At ForCar he mines proprietary vehicle datasets, turning raw numbers into buying advice you can actually trust.
Keep reading
All articles →
Sudden Acceleration Reports Tripled in 2010. Then They Faded.
Unintended acceleration is the most contested failure in motoring. The complaint record does not settle the argument, but it does show something clear: reports of it follow the news as much as they follow the cars.
Every 250kg Cuts Injury Claims by About 9%. Someone Pays.
The protective effect of mass is usually argued from anecdote. Put measured kerb weights against claim frequencies and it becomes a clean line — along with the cost, which lands on somebody else entirely.
Even the Sticker Recalls Say "Risk of a Crash"
Some recalls sound trivial. A wrong number on a sticker, a missing line in the manual. We went looking for the harmless category and could not really find one, which says something about how the threshold works.
Mark ForCar as a preferred source and our data shows up first in your Google results. One click, no account.