Methodology
How every model year is scored
Deterministic arithmetic over federal records. No editorial judgement enters the score, and every figure on a model page can be reproduced from the inputs shown beside it.
The exposure window
A 2013 model has had thirteen years to accumulate complaints; a 2023 has had three. Comparing raw totals rewards nothing but youth. Every model year is therefore scored on complaints filed in its first 60 months on the road, and nothing outside it counts.
Model year Y is treated as going on sale 1 September Y−1. That rule is applied uniformly to every vehicle. Model years that have not yet completed 60 months are marked provisional, shown with their partial totals, and excluded from every verdict.
Severity weighting
Complaints are weighted by outcome, taking the highest applicable weight per complaint — they do not stack.
| Crash or fire | 4.0 |
|---|---|
| Injury or death | 3.0 |
| Stalling or loss of power | 2.5 |
| All other | 1.0 |
Stalling is not a structured field in NHTSA's data. It is derived by pattern-matching the complaint narrative, which matches roughly 12.7% of records.
The within-model index (MYI)
Each model year's severity-weighted total is divided by the median across that model's own mature years. A value of 2.0× means that year drew twice the weighted complaints of a typical year of the same car. This is what drives the years-to-avoid verdict, and it is the most reliable figure on the site because a model is only ever compared with itself.
Recalls and investigations
Recall campaigns are counted and weighted, with Park It / Park Outside campaigns counting triple. Open ODI defect investigations are counted as distinct investigations, not as records. Both are matters of public record: a manufacturer has filed a defect report, or the regulator has opened a file.
Cross-model comparison is withheld
Model Year Index does not currently publish a cross-model or segment-relative complaint rate. Comparing complaint counts between different vehicles requires knowing how many of each were built, and no free public dataset provides US production volume by model year.
We tested a proxy derived from recall campaign sizes, including a filtered version that excludes supplier-wide campaigns spanning many models. Calibrated against published sales figures, the proxy misstates relative volume by up to 50× — it ranks the Ford Mustang above the F-150 and the Toyota Corolla near the bottom. A segment ranking built on it would look authoritative and be wrong.
It is therefore withheld pending licensed production-volume data. Every score currently published rests on the within-model index, recalls and investigations only. When volume data is in place this page will say so and the change will be dated.
Limitations
- Complaints are unverified reports submitted voluntarily by owners. They are allegations, not findings. Reporting rates vary between models and are influenced by publicity.
- A recall campaign is a matter of record; a complaint is not. The two are never presented as equivalent.
- Coverage starts at model year 2010. A model needs at least 150 complaints across all its model years to be included, counted before the 60-month window is applied. The totals shown on model pages are the smaller in-window figures, so a listed model can display fewer than 150. Individual model years with fewer than 10 complaints are dropped entirely, because stray misfiled records would otherwise distort that model's median.
- No score is generated by a language model. The pipeline is arithmetic.
Data vintage
August 9, 2026 · 314 models · 2,233,015 complaint records · 3,406 recall campaigns
Source data is refreshed monthly from NHTSA and the DOE. See sources for the exact endpoints and retrieval dates.