NHTSA Says Speeding Deaths Are Back to Pre-Pandemic Levels. They're Not.

Somebody at NHTSA rounded up. By 557 bodies.

In its July 2026 anti-speeding campaign launch, the agency declared that an estimated 10,035 people died in speeding-related crashes in 2025, framing the figure as "a return to pre-pandemic numbers last seen in 2019."^[1] There is a problem with that sentence. In 2019, NHTSA's own Fatality Analysis Reporting System recorded 9,478 speeding-related deaths.^[2] The number that is allegedly a return to 9,478 is 5.9 percent higher than 9,478. A 5.9 percent overshoot containing 557 additional fatalities is not a return. It is a generous reinterpretation of what the baseline used to mean, performed in a press release designed to make the trend look better than it is.

The broader fatality picture tells the same story with larger numbers. NHTSA separately announced that Q1 2026 traffic deaths dropped 4.3 percent to 7,770 and the fatality rate fell to 0.99 per 100 million vehicle miles traveled, the lowest quarterly figure since 2014.^[1] Good news, genuinely. But IIHS reports that 39,254 people died on American roads in 2024, their most recent final-year count.^[3] In 2019, 36,096 people died.^[4] That gap is 3,158 lives, an 8.7 percent increase over the "pre-pandemic" year the agency keeps invoking as the target it has reached. Even with 2025's decline, the body count remains thousands above the real 2019 number. Calling this a return to normal requires ignoring the normal you started from.

NHTSA threads the rhetorical needle by switching metrics. When the rate looks good, they cite the rate. The 2025 annual fatality rate was 1.10 per 100 million VMT, identical to 2019's 1.10.^[1][4] On that measure, the risk per mile driven has genuinely returned to its pre-pandemic position. What the rate obscures is that Americans drove substantially more miles in 2025 than in 2019, so an identical rate applied to a larger denominator produces a larger numerator. More miles at the same risk per mile equals more dead people, and the agency knows this because they publish both numbers and highlight only the one that flatters the narrative.

To visualize the distortion: imagine a hospital that treated 100 patients in 2019 with a 10 percent mortality rate, killing 10 people, and in 2025 treated 130 patients at the same 10 percent rate, killing 13. Announcing a "return to pre-pandemic mortality rates" while three additional families plan funerals is technically defensible and substantively dishonest. NHTSA is doing the traffic-fatality version of this arithmetic, then attaching it to a campaign called "Speeding Catches Up with You" that asks drivers to slow down while the agency speeds past its own data.

This is not an abstract statistical complaint. FARS data from 2014 to 2023 shows which vehicles populate speeding-context fatal crashes. The Ford Mustang carries a fatality rate of 6.02 per 100 million VMT with 2,739 total deaths across the study period, a 21.9 percent impairment rate, and the highest absolute body count in the sports car class.^[5] That means 78 percent of Mustang drivers in fatal crashes were sober. The Chevrolet Camaro sits at 3.44 with a 23 percent impairment rate. These are performance vehicles being driven fast by sober people who die because the physics of a high-speed single-vehicle impact do not care about your blood alcohol content. A speeding campaign that claims the problem is solved when 557 extra people are dead is a campaign optimized for press coverage, not outcomes.

Meanwhile, the states are not all moving in the same direction. National Safety Council preliminary data through May 2026 shows California's fatalities up 25 percent with 242 additional deaths year over year. The District of Columbia is up 129 percent. Rhode Island is up 56 percent.^[6] A national average that declines while individual states spike is not a uniform success. It is a composition effect where large states improving mask smaller states deteriorating, and California's increase alone nearly offsets West Virginia's 45 percent decline.

What the numbers actually say: The rate per mile is back to 2019. The number of people killed is not, and claiming otherwise by approximately 557 deaths is the kind of rounding error that only happens to people who are already dead. If you drive a vehicle and want to know whether the roads are as safe as they were before COVID, the honest answer is: per mile, roughly yes; in total body count, no. Check your speedometer. Check your VIN at nhtsa.gov/recalls. And check the baseline before you accept the government's arithmetic on whether safety has been restored.

Source: NHTSA FARS 2014–2023 for vehicle-specific data; NHTSA preliminary estimates and press releases for 2025–2026 figures; IIHS for 2024 final-year count. The 2025 speeding fatality estimate of 10,035 is preliminary and subject to revision; if it revises downward, the gap narrows but the framing remains misleading unless it drops below 9,478. See methodology for caveats.