Thirteen Quarters of Fewer Dead. The Last Streak This Long Took a Recession.

Somewhere around the second quarter of 2022, American roads started getting quieter, though not in volume because people drove more miles. Quieter in the way that matters: fewer bodies at the end of the quarter. NHTSA published its latest projection in September 2025, and the headline number is striking enough on its own: 17,140 people died in motor vehicle crashes during the first half of 2025, an 8.2 percent decline from the same period in 2024.^[1] The fatality rate per 100 million vehicle miles traveled fell to 1.06, the lowest mid-year reading since 2014.

But that number has a deeper story, one that only emerges when you lay this streak next to every other sustained decline in the agency’s records going back to 1981. NHTSA tracks four major decline periods, and each one, until now, shared a critical feature that this one does not.

A Short History of Americans Dying Less on Roads

The first sustained decline ran from Q2 1981 to Q4 1983, eleven quarters long. Paul Volcker had jacked the federal funds rate above 20 percent to strangle inflation, unemployment hit 10.8 percent, and gasoline prices in real terms were still elevated from the second oil shock. Americans drove less because they could not afford to drive more, and deaths dropped because exposure dropped. When the economy recovered in 1984, the streak ended and fatalities climbed back.

Round two ran from Q4 1990 through Q2 1993, another eleven quarters powered by another recession. The early ’90s contraction was milder, but VMT growth stalled completely for the first time since the oil embargo era. Deaths fell for eleven quarters, then stopped falling the moment driving volume resumed its climb. Nobody celebrated structural improvement because it was understood, correctly, that the improvement was rented.

The third decline was the longest on record. Seventeen consecutive quarters, Q2 2006 through Q2 2010, powered by the Great Recession and a gallon of gas touching $4.11 in the summer of 2008. VMT dropped approximately 3.6 percent peak-to-trough, the sharpest sustained contraction in postwar American driving. Fatalities plunged from 42,708 in 2005 to 32,999 in 2010, a 23 percent collapse that perfectly tracked the economic collapse underneath it. By 2015, with the recovery in full swing, deaths had climbed back to 35,484. The improvement evaporated because it was never structural. It was a recession wearing a seatbelt.

Then came the pandemic spike, and COVID-era roads, emptier and faster and more reckless, produced a fatality rate of 1.34 in 2020 and 1.38 in 2021, numbers not seen since 2006. Deaths peaked at 43,230 in 2021, blowing past the pre-pandemic level of 36,355 in 2019 by nearly 7,000 bodies.^[1]

And then, starting in Q2 2022, deaths began falling again and have not stopped since.

What Makes This One Different

Thirteen quarters in, the current decline has already outlasted the 1981 and 1990 streaks. It is four quarters shy of matching the Great Recession’s record. But what separates it from all three predecessors is not duration. It is mechanism.

During the first half of 2025, vehicle miles traveled increased by 12.1 billion miles, a 0.8 percent rise over the same period in 2024.^[1] Americans are driving more than they were a year ago, more than they were two years ago, and they have been driving more, year over year, throughout most of this 13-quarter decline. Recessions, gas shocks, and pandemics cannot explain a decline that persists through increasing demand, because increasing demand is precisely what ended every prior streak.

The fatality rate tells the story more cleanly. At 1.06 per 100 million VMT in H1 2025, the rate has fallen below its pre-COVID level (1.11 in full-year 2019) and is approaching the post-recession floor of 1.08 in 2014.^{[1] [2]} Something happened to the denominator that previous declines never touched: each mile driven became genuinely less lethal, not because fewer miles were driven, but because the miles themselves got safer.

The Acceleration

The decline is not just persisting; it is accelerating. Q1 2025 saw a 5.9 percent drop in fatalities, and Q2 2025 saw a 10.2 percent drop.^[1] June 2025 specifically registered a 12 percent year-over-year decline in deaths, with all ten NHTSA administrative regions showing decreases. Thirty-eight states, the District of Columbia, and Puerto Rico projected fewer deaths in H1 2025 than in H1 2024.

Eleven states did not. Hawaii saw fatalities jump 47.8 percent, on a base small enough (46 deaths in H1 2024, 68 in H1 2025) that statistical noise could explain much of the swing. Kansas climbed 21.2 percent, Idaho 18.6, and Massachusetts, the northeastern outlier in a group otherwise dominated by Western and Mountain states, jumped 15.1 percent. Wyoming, Montana, New Hampshire, West Virginia, North Dakota, Utah, and Illinois round out the list, most with single-digit increases on small population bases.^[1]

Meanwhile, D.C. projected a 73.1 percent decline (26 deaths to 7), and Connecticut fell 32.5 percent. Mississippi, historically one of the most dangerous states per mile driven with a 2024 rate of 1.84, dropped 25.7 percent to 1.36, a step-function improvement concentrated in a state that had the most room to fall.

What Is Actually Killing Fewer People

NHTSA does not specify a cause in the September projection, and honest analysis should resist the temptation to attribute a complex, multi-system outcome to any single intervention. But several candidates are structurally plausible in ways that prior decline drivers (economic contraction, gas prices) are not.

Automatic emergency braking now ships on roughly 90 percent of new vehicles sold in the United States, a voluntary commitment that ramped sharply after the 2016 industry pledge.^[3] Fleet turnover means the share of AEB-equipped vehicles in the active fleet grows every year as older cars without the technology are scrapped. Electronic stability control, federally mandated since model year 2012, has saturated most of the fleet; IIHS estimated ESC prevents roughly 7,000 crashes per year once fully deployed.^[4] Lane departure warning, blind spot monitoring, and adaptive cruise control are now standard on vehicles across most price segments. In 2025, the median new car on an American lot carries more active safety technology than a 2015 Mercedes S-Class.

Infrastructure changes are harder to quantify but present: federal highway spending under the 2021 Bipartisan Infrastructure Law has funded road geometry improvements, rumble strip installation, cable median barriers, and pedestrian safety redesigns in thousands of projects across the country. None of these produces a sudden decline, but all of them compound over time.

The Regression Objection

The strongest case against this decline being structural is simple: the COVID spike was enormous, and what goes up must come down. The 43,230 deaths in 2021 represented a behavioral anomaly. Emptier roads, higher speeds, more impairment. The subsequent decline could be nothing more than a return to baseline. Under this reading, the 13-quarter streak is an artifact of how far the pandemic pushed the numbers up, not evidence that the numbers have found a new floor.

It is a legitimate objection, and it deserves to be met with data rather than dismissed with narrative. The pre-COVID fatality rate was 1.11 per 100 million VMT (full-year 2019), and the H1 2025 rate is 1.06, which is not a return to baseline but a breakthrough below it.^[1] If this decline were purely regression to the mean, the rate would have stabilized around 1.10–1.12. It did not. It kept falling, through the pre-COVID floor, past 2018’s 1.14 and 2019’s 1.11 and 2015’s 1.15, and it is now sitting at a level last seen in 2014. The regression theory does not explain rates below the pre-existing trend line.

What This Does Not Prove

The causation chain remains unproven at the resolution that matters. Deaths fell, VMT rose, the fleet is safer, and yet we do not know how much of the decline each factor contributed because NHTSA’s projection methodology does not decompose causes; it counts bodies. The VMT increase itself could be compositional: if the additional 12.1 billion miles are disproportionately highway miles (which carry lower fatality rates per mile), then the risk per mile on local roads, where most deaths occur, may not have improved as much as the aggregate rate suggests. Without road-type VMT breakdowns, the “each mile got safer” thesis is plausible but not confirmed.

State-level estimates use a methodology that distributes regional projections proportionally based on early notification data. For small states, this amplifies noise. Hawaii’s 47.8 percent increase represents 22 additional deaths: meaningful for 22 families, statistically volatile for policy conclusions. Wyoming’s 10.2 percent increase is 5 deaths on a base of 49. These are not the same kind of data as California’s 2.4 percent decline on a base of 1,826.

And the second half of 2025 historically produces more fatalities than the first. Q3 and Q4 carry summer driving, holiday weekends, and longer hours of high-speed travel. Whether the 13-quarter streak extends to 15 or 17 depends on data that has not been collected yet.

The Implication

For 44 years, the pattern was fixed: American traffic deaths declined only when Americans stopped driving, and the mechanism was privation rather than progress. Recessions and gas crises saved lives by accident, and prosperity killed people by putting them back behind the wheel. If the current decline is structural—and the rate data below pre-COVID levels suggests it is, at least partially—then this is the first time in the modern era that the United States has made its roads meaningfully safer while people kept using them, not by driving less but by dying less per mile.

But 17,140 people still died in six months. The rate of 1.06 per 100 million VMT is still three to five times higher than peer nations like Sweden (0.21), Norway (0.24), and the UK (0.33).^[5] Eleven states are going the wrong direction, and the technology that appears to be contributing most (AEB, ESC, lane-keeping) only protects the occupants of the cars that carry it, which means the safety dividend is distributed by fleet age: newer cars, richer drivers, safer miles. If you are driving a 2009 Camry with no active safety systems in rural West Virginia (fatality rate: 1.63 and climbing), the 13-quarter national decline is an abstraction. Your road is getting worse.

Check your VIN at nhtsa.gov/recalls. If your vehicle predates 2018, it likely lacks AEB entirely. If it predates 2012, it may lack ESC. The fleet is getting safer. The question is whether your car is part of that fleet.