What percentage of police killings occur during traffic stops in the U.S. vs other encounters?
Executive summary
Traffic-stop–initiated killings are a clear and recurring slice of U.S. police fatalities, but how large that slice is depends on who counts and how: mainstream data-collection projects report that routine traffic stops directly account for a modest share—roughly 7–10% of police killings in recent years [1] [2]—while some advocacy reports combine traffic stops with mental-health responses and other non‑gun incidents and conclude a much larger majority [3] [4]. The gap is not small noise; it reflects real methodological choices, gaps in official reporting, and differing policy arguments about what encounters to group together [5] [6].
1. Traffic stops as a discrete category: what the trackers say
Independent trackers that categorize killings by how the encounter began—Mapping Police Violence and media summaries drawing on its work—counted roughly 86 deaths in one recent year that began with a traffic stop and calculated that such stops made up about 7% of all police killings that year [1] [2], and Mapping Police Violence’s longer-term archive shows hundreds of traffic-stop fatalities since 2017 [2] [5]; other outlets report “more than a hundred” traffic-stop–begun killings in 2023 specifically [7], illustrating year-to-year variation and the influence of classification choices.
2. Why some reports claim a much larger share (and what they group together)
Advocacy and some research reports present a broader framing—combining traffic stops, responses to mental-health crises, and encounters where the person was not reportedly threatening anyone with a gun—and report that these combined categories account for a plurality or majority of police killings (the 2024/2025 Police Violence Reports and summaries cite figures like 64% or similar phrasing) [3] [4]. Those higher percentages do not contradict the narrower 7% figure so much as change the question: they assert that many killings originate from low‑threat contexts, of which traffic stops are one type [3] [4].
3. Data gaps and methodological fault lines that drive the divergence
There is no single, complete federal dataset that cleanly records the “type of encounter that began” every police killing, so projects rely on media reports, local records, and imperfect government releases—an approach Mapping Police Violence explicitly uses because official agencies often fail to provide basic data [5]. The Stanford Open Policing Project shows how ubiquitous and under‑tracked traffic stops are—tens of millions annually—so even small per‑stop risks can translate into significant absolute numbers [6], and that scale makes classification decisions consequential for headline percentages.
4. What independent and official analyses agree on despite differences
Even where percentages differ, multiple sources converge on two findings: many fatal encounters begin with what officers or trackers describe as non‑violent or routine situations (traffic stops, welfare checks, mental‑health calls) [4] [7], and Black and brown people are disproportionately represented among those killed in traffic‑stop encounters (Mapping Police Violence and The Guardian report racial disparities in traffic‑stop fatalities) [2] [5]. These shared facts are central to policy debates over whether to remove armed officers from traffic enforcement or to redesign responses to behavioral-health calls [2] [7].
5. How to interpret the best available answer to the headline question
Answering “what percentage of police killings occur during traffic stops vs other encounters” requires specifying the measure: if the question means “killings that began with a traffic stop,” the most-cited independent tallies put that near the single‑digit range—about 7% in a recent year, with roughly 86 deaths counted in that category by one tracker [1] [2]. If the question allows a broader category of low‑threat responses that includes traffic stops and mental‑health or non‑gun incidents, some reports estimate a majority share (e.g., figures reported around 64% by advocacy reports), but those figures reflect different grouping and assumptions, not a direct contradiction of the narrower statistic [3] [4]. Given persistent reporting gaps and definitional choices in the sources, the most defensible public statement is: standalone traffic-stop–initiated killings account for a measurable but minority share (roughly 7–10% in recent snapshots), while broader groupings that put traffic stops together with other low‑threat encounters produce substantially larger percentages [1] [2] [3] [4] [5].