GM Spent $600 Million Proving Safety Tech Works. FARS Says Its Fleet Has 19,082 Excess Deaths.

In May, General Motors published the eighth installment of its research partnership with the University of Michigan Transportation Research Institute, a study that examined 12 million MY2020–2024 GM vehicles matched against more than 700,000 police-reported crashes across 18 states. Automatic emergency braking reduced rear-end crashes by 57%, backing-tech suites cut backing crashes by 86%, and front pedestrian braking slashed injury-producing pedestrian collisions by 35%.^[1] Peer-reviewed methodology, real crash reductions, and a press release that described GM's "mission" of "zero crashes" as if the scoreboard were already tilting their way.

We ran a different calculation against a different scoreboard. For every vehicle in the FARS database with 50 or more deaths and measurable VMT, we computed how many fatalities you would expect if it performed at the death rate of its vehicle class (sedan, SUV, pickup, van, or sports car), then subtracted expected from actual to get excess deaths. Sum those excess deaths by parent manufacturer and you get a portrait of which companies built vehicles that killed more people than their category peers predicted. GM vehicles produced 19,082 excess deaths over the decade covered by FARS. Ford finished second at 13,093, nearly 6,000 behind. Toyota came in at 4,283, and Mercedes-Benz registered zero excess deaths across its entire portfolio.^[2]

Six of the top 15 highest-excess-death vehicles industry-wide carry a GM badge, led by Chevrolet's Silverado with 2,703 excess deaths above the pickup-class average and the Impala with 2,673 excess deaths above the sedan-class average, killing at a rate of 5.0 per 100 million VMT in a class that averages 1.16. Behind those two sit the Tahoe at 1,974 excess, the Trailblazer at 1,952, the S-10 at 1,162, and the Cobalt at 1,100, which is the car whose ignition switch GM knew was defective for over a decade before it recalled a single unit.^[3]

GM will rightly note that the UMTRI study covers MY2020–2024 vehicles and that FARS deaths overwhelmingly involve older models, because a median vehicle in a FARS fatal crash is roughly a decade old and the Silverado deaths in this dataset are dominated by GMT800/900-era trucks rather than the T1 generation with Super Cruise and AEB. A counterargument writes itself: "We fixed the problem, our study proves it, and you're measuring the past." Fair enough, as far as it goes.

But GM declared "zero crashes" as its active corporate mission while sitting atop the largest manufacturer excess-death count in the industry, and those older vehicles are not hypothetical: they remain registered, insured, and driven on American roads because GM sold them and the average car stays in the fleet for 12.6 years. Seventeen of those 72 models in the dataset carry excess deaths above zero, and just the Impala accounts for an estimated 656,250 vehicles still on the road, each accumulating miles at a death rate more than 4x the sedan-class average. A mission statement running years ahead of fleet reality does not erase fleet reality.

From a buyer's perspective, the contrast across manufacturers is stark. If you purchased a Hyundai-Kia product during the same FARS window, your manufacturer's total excess death count was 617, which is less than what GM's Cobalt alone contributed. For Mercedes-Benz, the figure was zero, meaning not a single model in the dataset kills at a rate above its class average after adjusting for miles driven. GM's 19,082 is 31 times Hyundai-Kia's entire portfolio overshoot, a gap that no single study can bridge no matter how well its methodology holds up.

Nothing about this calculation invalidates GM's UMTRI findings, because AEB almost certainly does reduce rear-end crashes by 57% in the vehicles that have it. Scope is what separates triumph from tragedy here: UMTRI examines GM's best five model years against a controlled crash set, while FARS examines everything GM put on the road for a decade weighted by how much each vehicle is actually driven. One is a clinical trial showing that a new drug works in a controlled population; the other is the epidemiological record showing that the same company's older drugs are still circulating through the bloodstream of the national fleet, and the body count has not caught up to the brochure.