The U.S. Has Two Official Counts of Its Road Dead. They’re 3,535 Bodies Apart.
NHTSA says 39,254 people died in motor vehicle crashes in 2024. CDC’s National Center for Health Statistics says 42,789.[1][2] That is not a rounding error but 3,535 human beings who exist in one federal database and vanish from another.
Both numbers are official, both are correct, and they measure entirely different realities that nobody distinguishes when the press releases drop.
FARS, NHTSA’s Fatality Analysis Reporting System, counts deaths on public roads, from police crash reports, within 30 days of the incident. That single number drives every headline and every policy decision. NCHS counts every death where a motor vehicle was the cause, using death certificates filed by medical examiners. Its window stretches to one year after the crash. It includes parking lots, driveways, private roads, farm paths.[3]
The 3,535 in the gap are specific kinds of dead. The TBI patient who survived 46 days in the ICU before her brain swelled shut. An elderly man backed over in a Costco parking lot. The teenager who rolled a side-by-side on private land. The toddler run down in a family driveway. Every one of them died because of a motor vehicle, and NHTSA does not count a single one.
From this gap flow dueling narratives. For 2025, NHTSA estimates 36,640 deaths and announces a 6.7 percent decline.[1] The National Safety Council estimates 37,810 and announces a 12 percent decline.[4] Both held press conferences and claimed credit for progress, and neither acknowledged they were not measuring the same universe of dead.
Even the fatality rate, supposedly the objective yardstick, disagrees. NHTSA reports 1.10 deaths per 100 million vehicle miles traveled. NSC reports 1.14.[2] Scale that four-hundredths gap across 3.3 trillion miles and you get 1,320 additional expected deaths in one version of America.
NSC, to its credit, publishes a disclaimer on its data-details page: “NSC national motor-vehicle fatality estimates are not comparable to NHTSA figures.”[3] It also concedes that its own 2024 preliminary estimate of 44,680 missed the NCHS final of 42,789 by 4.4 percent, the widest miss since 2002. Nobody reads the footnotes, and journalists grab whichever number fits the lede.
None of this is fraud, and both methodologies have defensible rationale. But when two federal-level institutions cannot agree on how many Americans a known cause of death killed in a single calendar year to within four digits, ask the harder question: if we cannot count the dead, what exactly are we optimizing?
Next time a headline says traffic deaths hit a “record low,” check the denominator, the time window, and who qualifies as dead. Because 3,535 families buried someone last year, and the agency responsible for preventing their deaths never recorded that they happened.