How we turn crash-test results into one safety ranking
The score combines IIHS test results with a smaller contribution from class-level driver fatality rates. Vehicles need all required results to enter the ranking; awards do not add points.
Front, side, and headlight performance.
Including rear-passenger restraints and head protection.
The IIHS pedestrian front crash prevention result.
Real-world driver death rates for the vehicleβs class.
Start with the tests shoppers can compare
IIHS publishes many evaluations, but not every result is available for every current vehicle. The primary score uses four results with a clear, repeatable role in the comparison. If both an original and an updated test appear, we use the updated result.
| Core result | Share of primary score | Why it is included |
|---|---|---|
| Small-overlap front | 30% | A demanding frontal crash concentrated on one corner. |
| Updated moderate-overlap front | 30% | Includes protection for a rear-seat passenger. |
| Updated side | 30% | Uses a heavier, faster striking barrier than the original test. |
| Headlights | 10% | Measures how well drivers can see without excessive glare. |
One conversion for every result: Good = 100, Acceptable = 75, Marginal = 25, and Poor = 0.
Look past the overall grade
We select the latest documented build period for each test and keep its subratings together. Rear restraints come from the updated moderate-overlap test; rear head protection comes from the updated side test; front restraints come from the driver- and passenger-side small-overlap tests. Structure uses the lowest applicable structure grade across the three crash tests. Earlier-build variants are retained for reference but excluded from current scores and injury-measure tie-breaks.
Equipment follows a conservative baseline: standard pedestrian braking and the lowest headlight rating across applicable current trims. Better optional equipment is described on model pages without combining unverified packages. Build restrictions remain part of the result. Older day-only or night-only pedestrian protocols do not substitute for the combined pedestrian test.
Two vehicles can earn the same overall grade for different reasons. We therefore include four occupant-protection details that tell shoppers more about the passenger compartment and how the restraints controlled the dummies.
| Detail result | Main ranking weight | Rear-seat-priority weight |
|---|---|---|
| Rear passenger restraints and dummy kinematics | 10% | 20% |
| Rear passenger head protection | 5% | 5% |
| Structure and safety cage | 5% | 5% |
| Front occupant restraints and dummy kinematics | 5% | 5% |
The front-occupant result is the average of the driver and passenger restraint/kinematics ratings. We do not add another overall verdict here, because that would count the same test twice.
Add context without letting it take over
Pedestrian front crash prevention
IIHS rates each vehicle's automatic emergency braking against a pedestrian target: Good = 100, Acceptable = 75, Marginal = 25, Poor = 0. This contributes 10% of the score. Vehicle-to-vehicle crash prevention 2.0 is also relevant to the 2026 Top Safety Pick+ award; we display it separately without changing the published scoring weights.
What changed in September 2026. Until then this 10% was an award term (Top Safety Pick+ = 100, Top Safety Pick = 80). The award is decided mostly by the same tests that make up the other 80% of the score, so that term double-counted them. Replacing it with the pedestrian result removed the overlap. The current ranking reflects the extracted pedestrian ratings; corrections and new test results can change scores and ordering. Award status is still shown on every page, unweighted.
Vehicle-class context
The July 2026 model-specific release is available on our new death-rate study page. The new-car score retains its separately documented class reference and baseline; the model-specific release does not silently replace it.
Farmer (2023) reports driver death rates by vehicle class after standardizing for driver age and sex. This helps capture the broad effect of size and vehicle type, but it is not a death rate for an individual model.
Why a class rate and not a model rate. Raw model-level death rates carry heavy selection bias: they reflect who buys and drives each car as much as the car itself. The class rates we use are demographically standardized, which removes much of that bias at the cost of resolution. Model-specific rates appear on our used-car pages as context, never as an input to this score. Vehicle mass is not a separate input; its effect is already inside the class rates, and current research suggests the self-protection benefit of extra weight flattens above roughly 4,000 lb while harm to other road users keeps rising.
NHTSA stars. NHTSA 5-star ratings are shown on every model page for reference and carry no weight, because the current NHTSA tests overlap with IIHS tests while discriminating less between vehicles. We will revisit this when NHTSA publishes ratings under its upgraded test program.
Put the four pieces together
How equal scores are ordered
Two vehicles often earn the same score. When that happens we look at the granular injury measures IIHS publishes inside the three crashworthiness tests already in the score: small overlap front, the updated moderate overlap front test, and the updated side test. Every head, neck, chest, hip, thigh and leg result is averaged on the same Good 100, Acceptable 75, Marginal 25, Poor 0 scale, and the higher average is listed first.
This never changes the score itself, and it is deliberately narrow. Headlights, front crash prevention, seat belt reminders and child seat anchors are left out because they are scored elsewhere or vary by trim. Vehicles that match on both the overall score and the injury-measure average share a rank.
Main score55% primary tests + 25% occupant details + 10% pedestrian front crash prevention + 10% class context
The formula is deterministic: the same saved inputs produce the same score. Vehicles with identical scores and injury-measure averages share a rank.
See the exact class-context calculation
100 Γ clamp(1 β (β(class death rate Γ· 38) β 0.30) Γ· 1.70, 0, 1). The value 38 is the standardized average driver death rate per million registered vehicle years in the source study. We apply a square root adjustment to prevent class fatality rates from overshadowing individual crash test scores.
Unknown is not the same as safeβor unsafe
If IIHS has not published a result, we do not assume the vehicle would pass or fail. A partial record receives a coverage-adjusted provisional score so readers can inspect what is known, but it does not enter the numbered ranking.
Complete evidence record
All four primary results and all four selected occupant-protection details are available.
Provisional record
Vehicles awaiting tests remain fully searchable with all available data displayed, but are excluded from the main leaderboard until testing is complete.
How the coverage adjustment works
Evidence-adjusted component = quality of reported results Γ (0.70 + 0.30 Γ coverage). This prevents a partly tested vehicle from scoring as though every unknown result were Good.
What changes in the rear-seat-priority view?
The default ranking gives rear-passenger restraints and dummy kinematics 10% of the complete score. The optional rear-seat view raises that result to 20% and reduces the primary-test component from 55% to 45%.
Rear-seat-priority score45% primary tests + 35% occupant details + 10% pedestrian front crash prevention + 10% class context
Both views require the same complete testing record. The family view simply shifts weight to rear-seat protection without estimating untested data.
What this ranking cannot answer
- It is not a model-specific fatality prediction. Current vehicles do not yet have years of comparable real-world exposure.
- It does not measure danger to people outside the vehicle. See our separate analysis of driver and collision-partner death rates.
- It cannot choose the right vehicle for every family. Price, reliability, child-seat fit, visibility, and the exact trim still matter.
- It does not make incompatible tests comparable. Original and updated IIHS protocols are shown as different evidence.
Where the data comes from
- IIHS vehicle ratings and awards: model-year vehicle pages and the current Top Safety Pick index.
- Class-level driver death rates: Farmer (2023), standardized for driver age and sex.
- Model-year matching: a rating or award applies only to the model year and configuration identified by IIHS.