Two agencies counted last year’s cyclist deaths. They disagree.
NHTSA says 1,103. IIHS says 1,075. Both are drawing on the same federal crash file, and the gap is a lesson in reading any safety statistic.
NHTSA's 2024 fact sheet reports 1,103 cyclists killed on American roads. The Insurance Institute for Highway Safety reports 1,075 for the same year. Both ultimately rest on the same source, the federal Fatality Analysis Reporting System.
The two also disagree about where people died. NHTSA puts 30 percent of deaths at intersections. IIHS puts it at 39 percent.
Neither of them is wrong
The likeliest explanation for the count is category. NHTSA reports on “pedalcyclists,” a group that explicitly includes tricycles and unicycles, and which since 2022 also includes riders of motorized bicycles — a definitional change NHTSA flags prominently in its own document, because those riders were previously counted as motorcyclists. IIHS reports on “bicyclists.” A 28-person gap is entirely consistent with two organisations drawing the boundary of “cyclist” in slightly different places.
The intersection figures diverge for a related reason. NHTSA reports three buckets — at an intersection, not at an intersection, and a third “other” category covering bike lanes, shoulders, driveways, sidewalks and medians, which takes 9 percent. Its own footnote says some of those “other” locations may in fact have been at intersections but were not distinguished in the collected data. Fold that residual differently and 30 percent becomes something closer to 39.
Why we are writing about a 28-person discrepancy
Not to score a point against either organisation. Both publish their definitions, both flag their own caveats, and the gap is the ordinary consequence of counting a messy thing carefully.
The reason it is worth an article is that this is what well-collected safety data looks like from the inside. It is the best national dataset that exists on this subject, produced by serious people to a consistent standard, and two competent readings of it still differ by 28 deaths and nine percentage points. Any statistic quoted in a headline has this underneath it.
The practical rule that follows: when a cycling safety figure is quoted at you, the useful question is not whether the number is right but what was counted. Does it include e-bikes. Does a rider struck on a bike lane at a junction count as being at the junction. Was the denominator population, or trips, or miles, or nothing at all.
The standard we are trying to hold ourselves to
We intend to publish aggregate close-pass data from the Rydiq cohort, and when we do, the same scrutiny should apply to us. That means stating what counts as a pass, what the sensor can and cannot resolve, how many rides and riders are behind a figure, and which roads are over-represented because of who happened to be riding.
It is easy to publish a number. The part that makes it worth anything is the paragraph explaining what would make it wrong.
Sources: NHTSA, Traffic Safety Facts: Bicyclists and Other Cyclists, 2024 Data (DOT HS 813 822, July 2026); IIHS,Fatality Facts: Bicyclists (2024 data).