The NHTSA overall star rating is built from three separate tests. I pulled them apart across 2,337 rated model years to see which one cars actually do well in.
| Test | Average stars | |
|---|---|---|
| Side impact | 4.98 | Near-perfect |
| Frontal impact | 4.38 | |
| Rollover | 4.09 | Worst |
| Overall | 4.68 |
Side impact is effectively solved
4.98 out of 5. Across more than two thousand model years, almost every car takes full marks.
This is genuinely surprising, because a side impact is the harder physics problem. There are a few inches of door between the occupant and the striking vehicle, against several feet of engine bay in a frontal collision.
The reason the scores are so high is that the industry attacked this specific problem. Side-curtain airbags went from optional to universal, door beams got stronger, B-pillars went to high-strength steel, and the seat structures changed. Twenty years of engineering aimed at one test, and the test stopped discriminating between cars.
Frontal impact is 0.6 stars behind
4.38 against 4.98. In a scale where 81% of cars take five stars overall, a 0.6-star gap between two components is substantial.
Frontal collisions are harder to keep improving because the physics are less forgiving: the whole mass of the car has to be decelerated, and there's a limit to how much energy a crumple zone can absorb before the deceleration itself becomes the injury. Cars also got heavier over the same period, which works against the frontal result.
Rollover is where the overall score gets lost
4.09, and rollover is the worst of the three components in 38% of rated models — against 11% for frontal.
That makes rollover the single most common reason a car misses five stars overall. And unlike the other two, it isn't really a structural engineering problem. Rollover resistance is a function of how tall the vehicle is relative to its track width, which means the fix is to build a lower car — exactly what the market has spent twenty years not asking for.
It's also the component that has improved least over time, which we looked at in crash scores improved everywhere except rollovers.
What to read from this
Overall stars hide the interesting part. Two cars with the same overall rating can be four stars in different places. If you drive a lot of motorway miles, the frontal number matters more to you than the average does.
A tall vehicle's four stars is usually rollover. That's a shape trade-off you accepted when you chose an SUV, not a sign of a poorly built car — the pattern behind Jeep's brand average, which we covered in Jeep averages 4.09.
Stars describe the model as tested, not the car as sold. Equipment varies by trim, and a vehicle with an unrepaired airbag recall doesn't perform to its rating at all. A VIN check shows whether the safety campaigns on that specific car were ever completed.
Frequently asked questions
Are side impacts more dangerous than frontal ones?
The physics are harder — there's only a door between the occupant and the striking vehicle — but cars score better in side tests, averaging 4.98 stars against 4.38 for frontal, because twenty years of engineering was aimed specifically at that test.
Which crash test do cars score worst in?
Rollover, at 4.09 stars. It's the worst of the three components in 38% of rated models, compared with 11% for frontal impact, making it the most common reason a car misses five stars overall.
Why doesn't rollover resistance improve?
It isn't a structural problem. Rollover resistance depends on vehicle height relative to track width, so improving it means building lower vehicles — the opposite of what the market has been buying for two decades.
Does the overall star rating tell me enough?
Not really. Two cars with the same overall score can lose their stars in different tests. If most of your driving is motorway miles, the frontal component is more relevant to you than the combined figure.
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 →
Two Thirds of 2022 Cars Failed the Seat Belt Reminder Test. By 2026 It Was 18%.
A rating that measures a chime sounds trivial until you look at what it is for. Nearly half of people killed in crashes are unbelted, and the cheapest available intervention is a car that refuses to shut up.
54,604 Complaints Describe Something That Almost Happened. The Power Train Leads.
Safety data counts what happened. A large set of complaints exists specifically to report what did not, and those reports arrive from systems that rarely appear in crash statistics at all.
The Passenger Side Was Tested Fifteen Years After the Driver's. It Scores Better.
For over a decade, one side of the car was crash tested and the other was assumed to match. When the assumption was finally checked, the results said something unexpected about how safety engineering responds to being measured.
Mark ForCar as a preferred source and our data shows up first in your Google results. One click, no account.