So this study does not count complaints. It asks a different question, one that needs no sales figures at all: when owners of a given vehicle complain to NHTSA, what are they complaining about? Specifically, what share of a model-year's complaints describe a fire.
That ratio is immune to the volume artifact. A vehicle that sells in tiny numbers and one that sells in millions can be compared directly.
Key findings
- Across the 2015 to 2023 model years, 1.41% of NHTSA complaints describe a fire. That is
the baseline every figure below is measured against.
- Kia complaints are 3.09x as likely to describe a fire as the average vehicle's (4.36%,
from 925 fire complaints out of 21,205). BMW is second at 2.64x, Hyundai fourth at 1.66x.
- **Kia and Hyundai together are 11.3% of all complaints in the cohort but 26.2% of all the
fire complaints.** Their combined fire share is 3.27%, or 2.32x baseline.
- The single worst model-year is the 2015 Toyota RAV4 at 15.68%: 61 of its 389 complaints
describe a fire, eleven times the baseline.
- On crashes the picture inverts. Tesla's complaints are 2.63x as likely to describe a crash
(12.43%, from 2,263 of 18,206), with Toyota second at 9.54%. The 2015 Toyota Corolla is the worst single model-year at 34.78%.
What this measures, and what it does not
This is the part most coverage of complaint data gets wrong, so it is stated plainly.
This measures the composition of a vehicle's complaints, not its risk. A 9.33% fire share for the 2015 Kia Soul means that of the complaints owners filed about that vehicle, 9.33% described a fire. It does not mean 9.33% of 2015 Kia Souls caught fire, and it does not support a claim that any vehicle here is "X times more likely to catch fire."
NHTSA complaints are unverified consumer reports. They are subject to publicity effects: a widely reported recall drives more owners to file, which can move a vehicle's numbers in either direction. Treat this as a map of where owner-reported fire experience concentrates, and a prompt for further reporting, not as an incidence rate.
Highest fire share, 2015 to 2023 model years
Baseline 1.41%. Minimum 100 complaints per model-year.
| Model-year | Fire share | Fire / total complaints |
|---|---|---|
| 2015 Toyota RAV4 | 15.68% | 61 / 389 |
| 2018 Hyundai Kona | 14.91% | 17 / 114 |
| 2022 Jeep Gladiator | 14.88% | 25 / 168 |
| 2016 Kia Rio | 13.89% | 15 / 108 |
| 2017 Volvo XC90 | 11.97% | 14 / 117 |
| 2018 Toyota RAV4 | 10.98% | 47 / 428 |
| 2015 Kia Sportage | 10.77% | 21 / 195 |
| 2016 Lincoln MKC | 10.48% | 11 / 105 |
| 2017 Toyota RAV4 | 10.03% | 40 / 399 |
| 2021 RAM 2500 | 9.87% | 15 / 152 |
| 2020 Kia Sorento | 9.78% | 18 / 184 |
| 2015 BMW X3 | 9.62% | 15 / 156 |
| 2015 Kia Soul | 9.33% | 112 / 1,200 |
| 2015 BMW 328i | 9.14% | 18 / 197 |
| 2020 Kia Sportage | 8.87% | 18 / 203 |
The 2015 Kia Soul and the 2015 Kia Optima (8.16%, 113 of 1,385) are the two entries with the largest samples, which makes them the most reliable figures on the list.
Fire share by brand
Brands below 3,000 complaints or fewer than 10 model-years in the cohort are excluded from this table because the ratio is not stable at that sample size. Volvo, for example, computes to 3.85x baseline on 534 complaints across 3 model-years, which is arithmetically true and journalistically useless. The full file flags every excluded brand rather than hiding it.
| Brand | Fire share | vs baseline | Fire / complaints | Model-years |
|---|---|---|---|---|
| Kia | 4.36% | 3.09x | 925 / 21,205 | 52 |
| BMW | 3.72% | 2.64x | 119 / 3,201 | 22 |
| Hyundai | 2.34% | 1.66x | 584 / 24,965 | 61 |
| Toyota | 1.81% | 1.28x | 342 / 18,929 | 61 |
| Jeep | 1.73% | 1.23x | 631 / 36,526 | 51 |
| RAM | 1.71% | 1.21x | 345 / 20,177 | 25 |
| Dodge | 1.30% | 0.92x | 124 / 9,515 | 26 |
| Chevrolet | 1.22% | 0.86x | 455 / 37,415 | 97 |
| Subaru | 1.20% | 0.85x | 195 / 16,215 | 38 |
| Chrysler | 1.15% | 0.82x | 137 / 11,908 | 19 |
The crash picture is a different set of brands
The same method applied to complaints that describe a crash produces almost no overlap with the fire table. Baseline 4.73%.
| Brand | Crash share | Crash / complaints |
|---|---|---|
| Tesla | 12.43% | 2,263 / 18,206 |
| Toyota | 9.54% | 1,805 / 18,929 |
| Mazda | 7.76% | 256 / 3,300 |
| Nissan | 6.57% | 1,174 / 17,869 |
| Subaru | 6.06% | 983 / 16,215 |
Worst single model-years: 2015 Toyota Corolla (34.78%), 2016 Toyota Corolla (33.77%), 2015 Toyota Prius (25.36%), 2016 Toyota Camry (25.20%), 2018 Tesla Model S (22.73%).
Toyota and Tesla dominating a crash-share table is consistent with complaint clusters around driver-assistance behaviour rather than around structural safety, and the concentration in specific model-years rather than across a marque supports that reading. It is a lead worth reporting out, not a conclusion this data can close.
Method
- Source: NHTSA ODI consumer complaints, the agency's full public flat file, published by the
National Highway Traffic Safety Administration at static.nhtsa.gov/odi/ffdd/cmpl/FLAT_CMPL.zip. The file layout and field definitions are documented in NHTSA's flat file specification. Anyone can download it and reproduce every number below.
- Scale: 2,228,766 rows scanned, 1,052,748 retained after dropping test records (14,750) and
duplicate submissions (13,317). Complaints from filing year 2015 onward.
- Cohort: 2015 to 2023 model years, 917 model-years, 408,941 complaints. Model-years before
2015 are excluded deliberately. Their complaints arrive years after sale and skew hard toward dramatic events: the 2002 Pontiac Grand Prix computes to a 49% fire share on 116 complaints, which is a selection effect, not a finding.
- Threshold: a model-year needs at least 100 complaints to be ranked.
- Casualties: death and injury figures are collected but never headlined. They are
user-entered and unreliable in aggregate. Values at or above 99 are sentinels and single complaints reporting more than 8 casualties are not credible; both are zeroed while the complaint itself is retained. Treat any casualty total as a floor.
Get the data
The full dataset, all 917 model-years with fire share, crash share and raw counts, plus the brand rollup with thin-sample flags, is free to use and republish with attribution and a link to this page.
Email [email protected] and we will send the file, or run a specific model, model-year or brand cut for you. Tell us the deadline and we will work to it.
Source data is public, so nothing here has to be taken on trust: the underlying NHTSA complaint file is linked in the Method section above, and the aggregation rules are stated in full.
Zilocar is a US vehicle-history service. AutoIntel Systems LLC, Delaware.
