39,345 Dead. NHTSA Calls It Progress.

NHTSA dropped its preliminary 2024 fatality estimate last month, and the press releases practically glowed. 39,345 people killed in motor vehicle crashes — a 3.8% decline from 2023’s 40,901, and the first time the count dipped below 40,000 since 2020.^[1] The fatality rate fell to 1.20 per 100 million VMT. Lowest since 2019. Eleven consecutive quarters of decline. NHTSA used the word “encouraging.”

So I ran the decomposition nobody asked for.

Our FARS data covers 187,058 deaths across 337 vehicle models from 2014 to 2023. Sort those deaths by the model year of the vehicle involved, and a pattern emerges that NHTSA’s press release conveniently skipped: vehicles built before the 2012 ESC mandate account for 134,491 of those deaths — 71.9% of the total. Model years 2000 through 2007 alone produced 79,681 fatalities, or 42.6% of every death in the dataset.^[2] A single model year — 2005 — generated 11,363 deaths. The entire 2020 model year? 2,873. A 4-to-1 ratio.

The math writes itself. Those 2000–2007 vehicles — built without mandatory electronic stability control, before the IIHS small overlap test existed, with crash structures designed when the Clinton-era NCAP was the only game in town — are hitting 20 years old. They’re rusting in driveways. Failing state inspections in the 15 states that still do them. Getting parted out at Pull-A-Part. Every year, another few hundred thousand of them vanish from the fleet, and the death toll drops a little.

That’s not progress. That’s attrition.

The Behavioral Floor

The post-2012 fleet — vehicles with ESC, better crash structures, increasingly standard forward collision warnings and AEB — still killed 52,567 people in our dataset. Strip out the aging deathtraps, and you’re still looking at roughly 5,257 deaths per year from modern vehicles alone. Scale that to the national toll, and even a fully modernized fleet would still be producing somewhere north of 25,000 annual fatalities.^[3]

Why? Because 20% of drivers in FARS fatal crashes tested positive for alcohol or drugs.^[2] That number — one in five — hasn’t materially changed in a decade. No amount of side-curtain airbags fixes a BAC of 0.18. Electronic stability control can’t compensate for 86 mph in a 35 zone. The vehicle improvements that drove fatalities down from 55,000 in the early 1970s are still working, but they’re approaching diminishing returns while the human problem holds steady.

Where America Actually Ranks

Convert NHTSA’s 2024 rate to the metric the OECD uses — deaths per billion vehicle-kilometers — and the US clocks in at roughly 7.5.^[4] That’s an improvement from the 8.5 recorded in 2021. It’s also still 1.9 times Germany (3.9), 2.3 times the UK (3.3), and 3.3 times Norway (2.3). America drives the same vehicles as these countries — the F-150 sells in Europe, the RAV4 is global, German automakers build half their cars here. The vehicle fleet isn’t the gap. The roads are wider. The speed limits are higher. The DUI enforcement is weaker. The seatbelt usage is lower.

Norway kills 2.3 people per billion vehicle-km. Even if the US magically replaced every pre-2012 vehicle tomorrow — deleting 72% of our FARS death toll — the remaining modern-fleet death rate would still be roughly double Norway’s. The ceiling on vehicle safety is real, and we hit it a while ago. Everything past this point is roads, laws, and the decisions people make at 11 PM on a Saturday.

Limitations

FARS captures only fatal crashes — the 39,345 deaths are a fraction of the estimated 6.7 million total crashes annually. A vehicle with declining fatalities may still have rising injury crashes. Our fleet estimates use VMT modeling, not odometer readings, introducing ±15% uncertainty for low-volume models. The pre-2012 death concentration partly reflects fleet composition (there were simply more old cars on the road during 2014–2023) rather than solely vehicle design deficiency. The international comparison uses 2021 US data against 2022–2024 averages for other nations, so the gap may be slightly narrower than presented.

The Counterargument at Full Strength

There is a legitimate case that NHTSA’s optimism isn’t misplaced. The 11 consecutive quarters of decline^[1] extend back to Q2 2022 — a period during which the average fleet age continued rising (now 12.6 years per S&P Global Mobility), meaning the “old cars dying off” explanation should have been slowing down, not accelerating. Something behavioral may actually be improving — perhaps post-pandemic driving normalization, increased seatbelt use, or a reduction in the extreme risk-taking that spiked during COVID lockdowns. NHTSA’s own data shows that the rate dropped even as VMT increased 1%, which you can’t fully attribute to fleet composition. This argument deserves respect. It may even be correct. But “we’re slowly reverting to 2019 levels after a pandemic-era spike” is restoration, not progress. In 2019, 36,096 people died. We’re still 9% above that.