Kia’s Death Count Rose 331% in a Decade. Chevrolet’s Dropped by Half. America Barely Noticed.

I ran a computation nobody publishes, one that takes every fatal crash in the FARS database from 2014 to 2023, groups the vehicles by make, splits them into two production eras (2000s model years and 2010s model years), and counts the dead. What falls out is the single clearest picture of which automakers absorbed America’s road carnage over the past two decades and which ones quietly walked away from it.

Chevrolet’s 2000s model years produced 19,569 deaths in that fatal crash window; its 2010s model years produced 9,098, a 54% reduction that translates to roughly 10,400 fewer people dead in vehicles wearing a bowtie. Ford dropped 48%. Honda cut 39%, Toyota shed 31%, and General Motors killed off Pontiac entirely, whose 2000s-era vehicles had accumulated 2,600 deaths compared to a ghostly 34 in the 2010s. Saturn went from 519 to zero, because the brand no longer exists to build anything that can crash. Even Dodge managed a 47% cut.^[1]

Kia went the other way, violently. Its 2000s model years contributed 462 deaths to the FARS database; its 2010s model years contributed 1,992, a 331% increase that represents a 4.3-fold multiplication in the number of people killed in Kia-badged vehicles who then appeared in America’s federal fatal crash records.^[1]

Hyundai climbed 65%, from 2,390 to 3,932 deaths across the same era split, and Volkswagen rose 21%. Nissan, which had the third-highest body count of any brand in the 2000s at 6,007, managed to increase that to 6,660 in the 2010s while nearly every competitor was shedding thousands. Mercedes-Benz, a brand that trades on safety cachet and charges a premium for it, went from 234 to 386 deaths on a small but growing base.^[1]

The obvious rebuttal lands immediately, and it deserves to land at full force: Kia sold roughly 150,000 vehicles a year in the United States in the mid-2000s, and by the late 2010s that figure exceeded 600,000. A brand that quadrupled its fleet should, all else equal, roughly quadruple its body count, meaning the per-vehicle death rate may have improved considerably while the aggregate toll climbed, which is the population paradox in its purest form. A safer-per-unit vehicle can still produce more total deaths if enough units exist.^[2]

The paradox does not, however, neutralize the finding. What the data reveals is the result of a massive fleet recomposition that nobody tracks at the brand level: roughly 15,000 annual deaths disappeared from General Motors brands between these two production eras, and many of those drivers migrated to Korean showrooms. When a buyer who would have purchased a 2005 Pontiac G6 (908 FARS deaths, death rate 1.64 per 100 million VMT) instead purchases a 2015 Kia Optima (611 deaths, rate 0.58), that specific individual almost certainly improved their odds.^[1] But the Optima absorbed millions of buyers across the decade, and the aggregate effect is 1,530 more people dead in Kia vehicles that were built in the 2010s than in the 2000s.

Nissan’s trajectory is arguably more troubling because it cannot be explained by fleet growth. Nissan’s US market share was essentially flat between the 2000s and 2010s, hovering around 8%, yet its model-year death count rose 11% from 6,007 to 6,660, an increase of 653 additional deaths without a meaningful increase in the number of Nissans on the road. The Altima at a 2.88 death rate, the Sentra at 2.13, and the Maxima at a staggering 5.11 form a sedan trilogy that consistently kills above its class median, and those models did not get meaningfully safer across production eras even as their competitors retooled, re-engineered, and drove their numbers down.^[1][3]

Early 2020s model year data is telling despite its thin exposure window. Toyota leads the 2020s brand tally at 919 deaths, followed by Chevrolet at 858, Nissan at 613, Ford at 555, Kia at 547, and Honda at 475, and the pattern that emerges when you compare these shares to actual market share is that Kia and Nissan are running at a higher proportion of youngest-vehicle deaths than their sales volumes would predict while Ford, having exited nearly all sedan lines, is underrepresented in the death column for the first time in decades.^[1]

Real limitations are embedded in this analysis, and they deserve specificity. The FARS observation window runs from 2014 to 2023, which means 2000s model year vehicles had up to 14 years of fatal crash exposure while 2010s model years had as few as one year. Older vehicles accumulate more deaths partly because they have had more time to crash. That exposure bias favors the 2000s era across all brands equally, so brand-to-brand comparisons within the same era hold, but the absolute magnitude of the 2000s numbers is inflated relative to the 2010s. Chevrolet’s real safety improvement is probably even larger than 54% once you adjust for exposure time, and Kia’s 331% surge happened despite its 2010s vehicles having less crash exposure than its 2000s vehicles, making the raw increase a conservative estimate of the fleet-weighted effect.

The methodology is straightforward enough to replicate: FARS logs every fatal crash in the United States, each vehicle involved has a make, model, and model year, and the computation groups vehicles by make, buckets model years into eras (2000-2009 and 2010-2019), sums the associated deaths, and compares. No normalization, no fleet adjustment, no VMT modeling. The raw body count, by brand, by production decade, transparent enough that anyone with the public FARS query tool can verify every number in this article.^[4]

What you should do: Do not pick a car brand based on which direction its aggregate death count is trending. Pick a car based on the FARS per-model death rate for the exact vehicle you are considering. A Kia Telluride (rate 0.04) is extraordinarily safe despite Kia’s brand-level numbers. A Nissan Maxima (rate 5.11) will kill you at 9x the median regardless of what Nissan’s corporate trajectory looks like. The brand trend tells you where America’s road fatalities are migrating. The per-model rate tells you whether you personally should care.