America Has Had Three Great Death Declines. The First Two Ended Badly.

NHTSA dropped its annual numbers in April. Estimated 36,640 traffic deaths in 2025, a 6.7 percent decline from 2024's final count of 39,254.^[1] The fatality rate fell to 1.10 per 100 million vehicle miles traveled, the second lowest in the 50-year history of FARS, behind only 2014's 1.08. Q4 2025 marked the fifteenth consecutive quarter of year-over-year decline in traffic fatalities, an unbroken streak stretching all the way back to Q2 2022, when America was still reeling from the pandemic-era surge that had pushed annual deaths above 42,000 for two consecutive years.

Fifteen quarters, and Washington is celebrating like somebody cracked a code that forty years of highway safety budgets couldn't. Secretary Duffy credited law enforcement partnerships and affordable-car initiatives in the press release, because of course he did.^[1] What nobody mentioned is that this is the third time America has strung together a decline like this, and both previous streaks ended the same way: violently, with every saved life handed back within a couple of years.

FARS data goes back to 1975, and three multi-year decline streaks stand out on NHTSA's own quarterly chart.^[1] During the early-1990s recession, eleven consecutive quarters of decline ran from Q4 1990 through Q2 1993 while the economy contracted, unemployment climbed past 7 percent, and fewer Americans could afford to drive. By 1995 the economy had recovered, and deaths climbed right back up. A recession bought a pause, not progress.

Streak number two was the big one, stretching seventeen consecutive quarters from Q2 2006 to Q2 2010, powered by the Great Recession, gasoline above four dollars a gallon, and the ESC mandate that took effect via FMVSS 126 in 2007.^[2] Vehicle miles traveled cratered, and deaths followed. By 2011 the economy was thawing; by 2015, the surge hit with an 8.4 percent single-year increase in fatalities, the largest in decades, including Q3 2015's staggering 11.4 percent quarterly spike. Within six quarters, deaths had climbed to 37,806, fully erasing seventeen quarters of progress.

That brings us to the current run: fifteen quarters and counting, from Q2 2022 through Q4 2025, just two quarters from matching the Great Recession record. NHTSA projects it as the fifth-largest annual percentage decrease in FARS history.^[1] And there is one structural difference between this streak and the first two that should make every traffic safety researcher stop celebrating and start asking harder questions: vehicle miles traveled went up. Not by a little, but by 29.8 billion miles in 2025, a 0.9 percent increase over 2024.

Both previous streaks were recession dividends: fewer people driving, fewer people dying. This one is happening while Americans drive more than ever. That either means something genuinely changed in the engineering of the vehicles Americans drive, or it means the primary cause of the decline is something NHTSA has never bothered to measure.

Our cross-tabulation of FARS per-model fatality data points toward a candidate explanation that has nothing to do with enforcement campaigns or policy initiatives. Between 2014 and 2023, twenty-four vehicle models accounted for half of all deaths in the FARS dataset. Those twenty-four models are overwhelmingly sedans and pickups from the 2000 to 2008 model year cohort: Silverado, F-150, Accord, Civic, Camry, Altima, Impala, Malibu.^[3] Vehicles that were built before electronic stability control was standard, before side-curtain airbags were ubiquitous, before lane departure warnings existed as a concept outside of a Volvo brochure. They were the deadliest generation of vehicles in American history. They are finally, measurably, dying.

Model-year data makes the pattern unmistakable: sedans accounted for 70.6 percent of all fatal-crash deaths among MY2009 vehicles, with SUVs at 14.8 percent. By MY2022, the sedan share had collapsed to 40.7 percent, and SUVs had risen to 44.2 percent.^[3] Somewhere around MY2021, SUV deaths crossed above sedan deaths for the first time. The American fleet is turning over from vehicles with fatality rates of 2.0 to 5.0 per 100 million VMT to vehicles with rates below 0.5, and the national average is being dragged down not by better driving but by better metal.

That distinction matters enormously for what happens next. If this decline is primarily a fleet-turnover effect, it has a mathematical ceiling: once the deadly cohort is scrapped, the dividend plateaus and the curve flattens. That flattening could arrive as early as 2027 or 2028, when the last of the high-volume 2005-era sedans cross the twenty-year scrappage threshold. If, on the other hand, this is driven by behavior change or technology adoption, the decline could continue.

Monthly data offers one additional clue worth examining. Q2 2025 showed the sharpest decline of the entire streak, with June 2025 posting a 12 percent year-over-year drop.^[1] But July and August retreated to declines of just 1.7 and 2.5 percent, respectively. The summer months, when VMT spikes and younger drivers hit the road, barely budged. If this decline were genuinely behavioral, if Americans had simply become better drivers or more responsive to speed campaigns, you would expect the improvement to show up uniformly across all four seasons. The seasonal pattern looks more consistent with a fleet-composition effect: the vehicles that crash in summer (newer SUVs, trucks driven by younger demographics) are not the same vehicles being subtracted from the death toll by scrappage.

So where does that leave the celebration? The 1.10 rate is real and worth acknowledging. But the previous two decline streaks produced identical euphoria at identical milestones, and both ended with a reversion that killed thousands of additional people when the underlying economic driver reversed. The current streak may be structurally different because it is driven by physics, not macroeconomics. Fleet turnover is not cyclical; a scrapped 2004 Impala does not come back when GDP ticks up. That is a genuine reason for cautious optimism, provided nobody mistakes it for evidence that American drivers got safer or that NHTSA got better at its job.

What you should do: If you are driving a sedan from the 2000-2008 cohort, this article is describing your vehicle, so check your VIN at nhtsa.gov/recalls. These vehicles have the highest per-mile fatality rates in the FARS dataset, often two to five times the rate of a comparable current-model SUV. If your next vehicle purchase is between a used 2006 Malibu and a used 2016 Equinox, the mortality gap between those two options is larger than most people believe. America's fleet is turning over, and you want to be on the right side of it.