Transit safety incidents

Collisions, assaults, fires and derailments on U.S. public transit: every reported major event since 2014, by time of day, weather, light and place, from the National Transit Database.

Assaults on transit workers
Major events, 2025
16,495
individually reported incidents
Injuries, 2025
16,174
261 of them serious
Deaths, 2025
503
Assaults on transit workers, 2025
892
NTD flags these separately

Sixteen years, with a denominator

Counts alone rank agencies by size. FTA's threshold-adjusted time series carries the service each agency ran alongside the incidents it reported, which is what turns a count into a rate.

Deaths per hundred million vehicle revenue miles

2008 to 2024, complete years only

American transit killed 195 people in 2008 and 345 in 2024. That much is a count, and counts rise when systems grow.

This did not grow. Service fell over the same period, so the rate rose faster than the count: from 5.4 deaths per hundred million vehicle revenue miles to 9.65. Something got worse that is not explained by there being more transit.

Revenue miles are the denominator rather than boardings, and the choice matters here more than anywhere else on the site. This span covers 2020, when agencies kept running service that nobody rode. Divided by boardings the rate would appear to double in a single year, and what it would be measuring is the ridership collapse.

Most of these deaths are not passengers. The chart below shows who they are.

Injuries per hundred million vehicle revenue miles

Same denominator, same years

Who died, year by year

Suicide is recorded as an event, and is not a measure of how safely a system is run

Passenger deaths per hundred million revenue miles: 1.27 in 2016–2019, 2.49 in 2022–2024. Passengers killed in homicides per hundred million revenue miles: 0.39 in 2016–2019, 1.19 in 2022–2024.

Passengers include people boarding and leaving. FTA redrew its pedestrian categories in 2016, and reporters move deaths between them and "other" from year to year, so they are shown together.

Why the comparisons are split by mode

National rates per hundred million vehicle revenue miles, 2020 to 2024

Mode Revenue miles Events per 100M miles Injuries per 100M miles Deaths per 100M miles Agencies
Bus 8.37B 672 706 6.18 444
Heavy Rail 3.16B 852 599 24.47 15
Demand Response 3.15B 248 227 1.21 440
Vanpool 801.9M 39 15 0.75 104
Light Rail 539.6M 1,274 892 44.48 22
Commuter Bus 368.1M 208 199 4.35 92
Bus Rapid Transit 59.9M 2,350 2,677 11.68 19
Trolleybus 43M 1,060 1,053 6.97 5
Streetcar 27.3M 3,896 2,104 21.95 26
Ferryboat 23.4M 3,263 3,215 4.28 32
Publico 23.1M 82 91 0 1
Hybrid Rail 14.1M 976 530 98.98 3
Monorail / Automated Guideway 9.6M 5,843 5,457 31.24 6

The same measure differs by an order of magnitude from one mode to the next. A van carrying one booked passenger door to door reports far fewer events per mile than a streetcar sharing a street with traffic, and a railway reports events on its own track that a bus network never encounters. A single table ranking every agency together would rank them by what they run rather than how safely they run it — so the tables below compare each agency only with others running the same mode.

Heavy and light rail deaths include suicides and trespassers, who account for most of the difference from bus. The breakdown of who died is further down the page.

Highest reported rates: Bus

Events per 100M revenue miles, 2020 to 2024. 189 agencies running at least 7.44M revenue miles in this mode.

# Agency Per 100M miles Events Deaths
1 2,479 530 1
2
Albuquerque ABQRIDE
Albuquerque, NM
2,010 343 0
3 1,508 1,572 4
4 1,447 331 1
5 1,345 601 11
6 1,330 2,525 19
7
El Metro
Laredo, TX
1,286 108 0
8 1,270 5,205 18
9 1,266 2,982 14
10
Sun Tran
Tucson, AZ
1,245 471 2

Highest reported rates: Heavy Rail

Events per 100M revenue miles, 2020 to 2024. 14 agencies running at least 5.87M revenue miles in this mode.

# Agency Per 100M miles Events Deaths
1 3,394 268 0
2 1,802 1,795 23
3 1,665 1,279 53
4 1,447 493 5
5 1,294 1,178 28
6
Baltimore MTA
Baltimore, MD
1,157 262 9
7 1,037 122 2
8 921 3,142 79
9
Los Angeles Metro
Los Angeles, CA
907 267 19
10 884 14,909 483

Highest reported rates: Light Rail

Events per 100M revenue miles, 2020 to 2024. 20 agencies running at least 3.93M revenue miles in this mode.

# Agency Per 100M miles Events Deaths
1 2,916 370 21
2
Houston METRO
Houston, TX
2,911 436 9
3 2,517 697 1
4 2,337 93 4
5
Metro Transit
Minneapolis, MN
2,292 450 13
6 1,723 261 8
7 1,563 762 36
8
Sound Transit
Seattle, WA
1,505 533 7
9 1,279 166 4
10
(St. Louis) Metro
Saint Louis, MO
1,156 302 6

Highest reported rates: Demand Response

Events per 100M revenue miles, 2020 to 2024. 31 agencies running at least 20.2M revenue miles in this mode.

# Agency Per 100M miles Events Deaths
1
(St. Louis) Metro
Saint Louis, MO
755 162 1
2 490 568 2
3 433 127 0
4
Capital Metro
Austin, TX
365 94 1
5
MTA Flint
Flint, MI
353 74 0
6 350 157 0
7 350 174 2
8
Pace Suburban Bus
Arlington Heights, IL
349 424 2
9 338 104 0
10 311 102 0

How to read these tables

The floor is service, not a round number. Each table admits agencies that ran enough of that mode to expect about 50 reported events at the national rate. Below that, one incident more or less moves a rate by several percent and the ordering is noise. The floor is set on miles rather than on events on purpose: a floor on events would admit a small agency with a high rate and turn away a small agency with a low one.

Read these as reporting tables as much as safety ones. An agency that records every slip and near miss sits above one that records only what it must, and nothing in the data tells them apart.

Commuter rail is not here, and neither is PATH. FTA's safety series excludes commuter rail, the Alaska Railroad, PATH's heavy rail and three hybrid-rail services, because they answer to the Federal Railroad Administration and report their safety data there.

Who died, 2025

The total is not a measure of risk to riders, and NTD gives the breakdown that shows why.

Of 337 deaths, 33 were passengers. Most of the rest were people on the track or the road, and 65 were suicides.

Set against distance traveled, that is 8.49 deaths per billion passenger miles counting everyone, and 0.83 counting riders alone — a difference of roughly eight times.

Both are true. They answer different questions: the first is how many people die in connection with transit, the second is the risk of being on board. Quoting the first as the second is the commonest way these figures are misread.

What kind of incident

Every major event NTD has published since 2014: 116,384 of them, each reported individually.

Events by year and kind

Incident level, not monthly totals

Where they happen

2025, by reported location

When, and in what conditions

The monthly file cannot answer either question. The incident file records the time of every event, and the weather and light for about nine in ten.

By hour of day

All reported events in 2025

The shape follows service, not danger: more vehicles are running in the afternoon than at four in the morning. Read it against how much service runs at each hour.

Light and weather

Conditions recorded at the time of the event

Weather at the time

Weather Events Injuries Deaths Injuries per event
Clear 9,476 9,701 212 1.02
Cloudy 1,045 1,013 24 0.97
Raining 771 733 14 0.95
Tunnel 445 231 151 0.52
Snowing / sleeting 280 246 5 0.88
Foggy / misting 65 58 1 0.89
Other 3 1 0 0.33

Monthly totals

The other half of NTD's safety reporting: counts, including both major and non-major events, back to 2014.

Events by type

Complete calendar years only

Rate, not count

Events and injuries per hundred million vehicle revenue miles, major and non-major

Monthly events

Every month since January 2014