There is no single perfect safety number. NHTSA's 5-star ratings work within a class — they tell you whether one midsize sedan is safer than another midsize sedan. They do not tell you whether a 5-star compact is safer than a 4-star pickup. Different question, different answer. The honest method is layered: each layer answers something the previous one couldn't.
The short version: Ratings within class. Weight across classes. Dummy injury margins for the actual measured loads. Side-impact and rear-seat tests for specific scenarios. Death rates only as context. No single number does all seven jobs.
Within the same size and weight class — and within roughly ±250 lb of curb weight — NHTSA frontal star ratings are designed to be directly comparable. NHTSA itself cautions that frontal ratings should only be compared between vehicles inside this band, because the test puts every car against the same fixed barrier at the same speed.
Worked example: a 2024 Toyota Camry (~3,300 lb) and a 2024 Honda Accord (~3,200 lb) are both midsize sedans within 100 lb. Their NHTSA stars are directly comparable. Compare a 2024 Civic (~2,900 lb) against a 2024 Highlander (~4,500 lb) and you are outside the band by a factor of six — the stars stop being a fair comparison.
When you are cross-shopping a sedan against a midsize SUV, raw curb weight starts to matter. A 300-lb gap is normal. A 1,000+ lb gap materially changes who absorbs more energy in a real two-car crash. The lighter vehicle absorbs a larger change in velocity, and that delta-v is what the seatbelt and airbag have to manage.
The National Bureau of Economic Research found that being hit by a vehicle 1,000 lb heavier than yours increases the baseline probability of fatality in that crash by roughly 47%. Modern vehicle safety tech has narrowed that gap, but the underlying physics has not changed.
Star ratings round the result. Inside every NHTSA frontal test, an instrumented crash dummy records exactly how much force hit the head, chest, and neck. Two cars can both earn 5 stars while one quietly puts twice the force into the dummy's chest.
Two 5-star sedans of similar weight: one records chest compression at 30% of the federal limit, the other at 70%. Both pass. The second car used more than twice the injury-margin budget. SafeCarCompare's SafeScore surfaces this gap as a single number from 1 to 100.
Vehicle weight is not the only variable. Crumple zones live in a narrow vertical band — the bumper beam, frame rails, and engine bay. When the other vehicle's bumper sits substantially higher than that band, it rides over your crumple zone and pushes directly into the cabin. Engineers call this override / under-ride.
The IIHS Vehicle Compatibility research targets a bumper zone of 16–20 inches above the road. Pickups and body-on-frame SUVs often sit at 22–28 inches — measurably increasing fatality risk for occupants of the struck car, independent of mass.
Side-impact tests come closer to a cross-class comparison than fixed-barrier frontal tests. IIHS's updated side-impact test uses a 4,200-lb moving deformable barrier at 37 mph — designed to represent a modern midsize SUV striking the side of another vehicle. Every car gets hit by the same simulated impact, so a small sedan and a large SUV can be evaluated against the same scenario. It is still one specific scenario, not a complete cross-class score.
The 2022 IIHS moderate-overlap front update also exposed a gap most automakers had been quietly leaving — rear-seat occupant safety. Vehicles that aced the driver-side test sometimes performed badly when injury risk was measured for a back-seat adult occupant. If you carry passengers in the rear seat, this test result matters.
IIHS publishes driver death rates per million registered vehicle years by make and model. This is real-world data and worth knowing — but it is not a clean measure of vehicle design. Death rates mix the vehicle itself with driver behavior, miles driven, road type, geography, vehicle age, and who tends to buy each model.
A frequently cited example: the Toyota RAV4 and the RAV4 Hybrid are structurally nearly identical and have identical NHTSA crash-test results. The hybrid's real-world death rate is roughly 40% lower. That gap is not the structure — it is the driver mix. Useful information, but not a fact about the car.
SafeCarCompare deliberately does not show driver death rates as a primary safety score for this reason. We surface the underlying crash-test injury readings instead, where the variable is the vehicle, not the driver.
All six steps above assume a crash already happened. Half of the safety story is not crashing in the first place. A car that avoids a collision entirely will protect you better than a heavier car with poor visibility, slow brakes, or weak driver-assist tech.
As of 2023, AEB is standard on virtually all new vehicles sold in the U.S. — but on older models or budget trims it may be optional or absent. SafeCarCompare's Prevention Score rates these features on the same scale as structural safety so you can see both halves of the picture.
Fatal crash rates vary 3–5× between counties. Rural counties typically see the highest rates — driven by higher speeds, longer emergency response times, and less median separation on roads. None of the seven steps above account for where you actually drive. High-fatality areas benefit more from strong rollover resistance, side-impact ratings, and crash-avoidance technology.
SafeCarCompare shows injury margins from NHTSA crash-test data — beyond star ratings. Enter any two vehicles to see head, chest, and neck injury margins side by side.
Compare vehicles on SafeCarCompare →
SafeCarCompare — Vehicle safety data beyond star ratings