The dead are not the passengers

The transit death the public imagines is a crash that kills the people on board. It is the rarest kind. Of the 3,569 people killed, 277 were riding the vehicle. Nearly four in ten were on the right-of-way or took their own life; a quarter were out on the road, in another car or on a bike or in a crosswalk, when a transit vehicle reached them.

Deaths on transit are mostly not passengers. Injuries, increasingly, are mostly not passengers either - they are the people paid to be there. Agencies logged 203 assaults on transit workers in 2014 and 643 in 2025. The direction is not in doubt. The exact multiple is, and the reason is drawn on the axis below.

Events carrying the FTA transit-worker-assault flag · Worker injuries are operator plus other transit-employee injuries on those events · 2026 excluded as a partial year

Events have climbed for a decade. On its own that means little: agencies also run more service than they did, and more service means more chances for something to happen. The test is whether events outgrew the service. Between 2022 and 2024 service grew +7.9% and events grew +16.5%. So the rate rose too, from 2.97 to 3.21 events per million miles. This is not a story about buses simply driving more.

Every "worst transit agency" list is really a list of the biggest transit agencies. MTA New York City Transit reports more events than anyone because it moves more people than anyone. Divide each agency's events by the miles it actually drove and the ranking comes apart: New York falls to #29, and the agencies at the top are New Orleans Regional Transit Authority, Minneapolis, Philadelphia, Indianapolis. Each line below is one agency. Where a line falls steeply, size was doing the talking.

The 14 agencies that top either list · Both ranks are taken among the 108 agencies clearing the 15M vehicle-revenue-mile service floor, so the columns compare like with like · Rows are spaced evenly and the true national rank is printed beside each name; the vertical gaps are not to scale · Rank direction and the printed numbers carry the meaning, so the line color only repeats it · Rate uses 2022-2024, the only years the service denominator covers

At the other end, Foothill Transit runs a scheduled bus system at 0.18 events per million miles - some 70 times better than New Orleans Regional Transit Authority. A couple of vanpool operators post lower numbers still, but a vanpool is not a bus, and comparing them would be the same mistake as comparing raw counts. Read every rate beside the mode mix on the agency's own page: see one.

Shade a map by raw event counts and you have drawn a population map: New York, Illinois, California. Shade it by events per mile of service actually driven and a different country appears. Louisiana reports more than 29 times as many events per mile as the quietest state. The national figure is 3.21.

Quartile classes, darker = more events per mile of service · 50 states have at least one agency with a service denominator · A state is the sum of its agencies: the events file carries no state of its own, so this map exists only through the NTD ID join · State mode mix matters - a state whose transit is mostly street-running light rail will sit higher than one running mostly suburban bus

The point of a rate is that it is answerable at your stop. Every agency below has its own page: what it reports, who gets hurt on it, how its worker assaults have moved, and where it sits against every other agency in the country per mile driven.