Do Harder Crash Tests
Actually Predict Who Dies on the Road?
IIHS made crash tests dramatically harder in 2022β2023. But do the new results predict real-world death rates better? We found 16 vehicles that let us check.
Sources: IIHS crash test data (2024β2026), IIHS driver death rate study (2017β2020 model years)
π¬ The Natural Experiment
Most updated crash tests are on new vehicles β making it impossible to compare with older death rate data. But some vehicles kept the same generation across both eras. These are our test cases.
Where the IIHS tested the exact same platform under both the updated crash tests (2023+) and published real-world death rates (2017β2020 model years).
Vehicles that pass both updated tests have an average death rate nearly half that of vehicles that fail at least one.
π The Headline Finding
We split the 16 vehicles into two groups: those that scored "Good" on both updated tests, and those that scored "Marginal," "Poor," or "Acceptable" on at least one.
8 vehicles β below the national average of 38
8 vehicles β 86% higher than the passing group
π‘ What this means: The updated crash tests aren't just harder for the sake of it. They appear to discriminate between vehicles that actually kill more drivers vs. ones that don't. The old tests β where nearly everyone scored "Good" β couldn't tell the difference.
π All 16 Same-Generation Vehicles
Sorted by death rate. Death rate = driver deaths per million registered vehicle years (IIHS).
| Vehicle | Death Rate | Updated Front | Updated Side | Generation |
|---|---|---|---|---|
| BMW X3 4WD | 0 | Good | Good | G01: 2018β2024 |
| Volvo XC90 4WD | 4 | Good | Good | SPA: 2016+ |
| Toyota Camry hybrid | 19 | Good | Good | TNGA-K: 2018β2024 |
| Volkswagen Atlas 4WD | 20 | Good | Good | MQB: 2018+ |
| Ford Explorer 4WD | 22 | Good | Good | CD6: 2020+ |
| Subaru Forester | 30 | Good | Good | SGP: 2019β2024 |
| Jeep Wrangler 4-door 4WD | 39 | Good | Good | JL: 2018+ |
| Toyota Camry | 48 | Good | Good | TNGA-K: 2018β2024 |
| Subaru Outback β | 5 | Marginal | Good | SGP: 2020β2024 |
| Volvo XC60 4WD β | 5 | Good | Acceptable | SPA: 2018+ |
| Toyota RAV4 hybrid 4WD | 20 | Marginal | Acceptable | TNGA-K: 2019β2025 |
| Toyota RAV4 2WD | 35 | Marginal | Acceptable | TNGA-K: 2019β2025 |
| Volkswagen Jetta | 47 | β | Acceptable | MQB: 2019+ |
| Toyota RAV4 4WD | 56 | Marginal | Acceptable | TNGA-K: 2019β2025 |
| Toyota Corolla | 58 | β | Acceptable | TNGA-C: 2019β2025 |
| Nissan Altima π΄ | 113 | Marginal | Poor | CMF-CD: 2019β2024 |
βοΈ Controlling for Vehicle Size
One obvious objection: the "Good" group includes larger vehicles (BMW X3, Volvo XC90, Ford Explorer) that should have lower death rates simply because they're bigger. Is this just a size effect?
We normalized each vehicle's death rate against the age/gender-standardized death rate for its vehicle class from Farmer (2023), the IIHS study that adjusts for driver demographics.
These vehicles have death rates 30% below what's typical for their size class
40% higher ratio than the passing group β the signal holds even after size adjustment
| Vehicle | Farmer Class | Class Std DR | Actual DR | Ratio |
|---|---|---|---|---|
| BMW X3 4WD | Luxury SUV Midsize | 11 | 0 | 0.00Γ |
| Volvo XC90 4WD | Luxury SUV Large | 17 | 4 | 0.24Γ |
| Toyota Camry hybrid | Car Midsize | 60 | 19 | 0.32Γ |
| VW Atlas 4WD | Nonluxury SUV Large | 20 | 20 | 1.00Γ |
| Ford Explorer 4WD | Nonluxury SUV Large | 20 | 22 | 1.10Γ |
| Subaru Forester | Nonluxury SUV Small | 42 | 30 | 0.71Γ |
| Jeep Wrangler 4dr 4WD | Nonluxury SUV Midsize | 27 | 39 | 1.44Γ |
| Toyota Camry | Car Midsize | 60 | 48 | 0.80Γ |
| Subaru Outback | Wagon Midsize | 5 | 5 | 1.00Γ |
| Volvo XC60 4WD | Luxury SUV Midsize | 11 | 5 | 0.45Γ |
| Toyota RAV4 hybrid 4WD | Nonluxury SUV Small | 42 | 20 | 0.48Γ |
| Toyota RAV4 2WD | Nonluxury SUV Small | 42 | 35 | 0.83Γ |
| VW Jetta | Car Midsize | 60 | 47 | 0.78Γ |
| Toyota RAV4 4WD | Nonluxury SUV Small | 42 | 56 | 1.33Γ |
| Toyota Corolla | Car Small | 54 | 58 | 1.07Γ |
| Nissan Altima | Car Midsize | 60 | 113 | 1.88Γ |
Key insight: Even after controlling for vehicle size and driver demographics, the Nissan Altima's death rate is 1.88Γ its class average β nearly double what you'd expect for a midsize car. Combined with its Marginal/Poor updated test results, this is strong evidence that the car itself is part of the problem, not just its drivers.
Class standardized death rates from: Farmer, C.M. (2023). "Demographic adjustments to driver death rates." Traffic Injury Prevention. Table 1 (case-vehicle driver deaths), 2017-2020 MY.
𧩠The Interesting Cases
The outliers are where the real insight lies. They show us what crash tests measure β and what they can't.
π΄ Nissan Altima: The Strongest Evidence
The Altima scored Marginal on the updated front test and Poor on the updated side test β the worst combination of any vehicle in our sample. Its death rate of 113 is 3Γ the national average.
People often ask: "Is the Altima dangerous because of the car, or because of who drives it?" This data suggests it's both. The updated crash tests now independently confirm the Altima has genuine structural weaknesses β it doesn't just fail because IIHS moved the goalposts. The car objectively provides less protection for its occupants in a side impact than a Toyota Camry or Subaru Forester of the same era.
β Subaru Outback: The Paradox
The Outback scored Marginal on the updated front test β but has one of the lowest death rates of any non-luxury vehicle (5 deaths per million). How? Its real-world safety likely comes from crash avoidance rather than crash survival: standard AWD, excellent EyeSight driver assist, and cautious owner demographics.
This is the best evidence that crash tests measure crashworthiness, not total safety. A car that prevents crashes in the first place can have a low death rate even if its structure isn't best-in-class.
π Toyota Camry: Same Car, Different Drivers
Identical platform, identical crash test results (Good/Good). Yet the hybrid has a death rate 2.5Γ lower. This is a pure driver demographics effect β hybrid Camry buyers tend to be older and drive more conservatively.
This is why death rates aren't just a crash test score in disguise β they capture everything, including who buys the car and how they drive it.
ποΈ Jeep Wrangler: Good Tests, Dangerous Car
Good on both updated tests, but still has an above-average death rate. Why? Rollovers. The Wrangler's high center of gravity creates fatality risk that no front or side crash test evaluates. The tests measure what happens when a barrier hits you β not whether your vehicle tips over at highway speeds.
π What This Tells Us About Crash Tests
The updated tests are better, but crash tests alone can never tell the whole story.
What crash tests measure well
- Structural strength in specific impact scenarios
- Restraint system effectiveness (airbags, belts)
- Protection for different-sized occupants (new!)
- Side intrusion resistance against heavy vehicles (new!)
What crash tests can't measure
- Crash avoidance technology (AEB, BSM, lane keep)
- Rollover risk (center of gravity, stability)
- Driver demographics and behavior
- Vehicle size/weight advantage in multi-car crashes
This is why we rank vehicles using both crash tests AND real-world death rates. Neither alone tells the full story. The Subaru Outback proves crash tests aren't everything, and the Nissan Altima proves death rates aren't just about drivers.
β οΈ Caveats
- Small sample (16 vehicles) β This is a signal, not proof. A proper study would need hundreds of vehicles.
- Death rates reflect drivers, not just cars β The Camry hybrid/regular split proves this definitively.
- Generation spans aren't exact β Mid-cycle refreshes may alter some structural components. We selected only vehicles where the same fundamental platform spans both eras.
- Confounding variables β Vehicle size, weight, AWD vs. 2WD, driver age, and geography all affect death rates independently of crash performance.
- The bar moved, the physics didn't β A car scoring "Acceptable" on the updated side test isn't worse than when it scored "Good" on the old test. The test got harder β the car didn't change.
Explore More Safety Data
Sources: IIHS updated crash test protocols (2022β2023), IIHS driver death rates by make and model (2017β2020 MY), manufacturer model generation data. Analysis by Informed for Life.