Data sources
Everything here comes from the Federal Transit Administration's National Transit Database, read through the USDOT open data portal. Nothing is estimated, modeled or filled in. Where NTD publishes no number, this site shows an em dash.
Files read
Each links to the dataset on the USDOT portal.
Service Data and Operating Expenses by Agency (Time Series 2.2)
2015 to the latest published yearBoardings, passenger miles, revenue miles and hours, vehicles, operating expense by function, and fare revenue, per agency per year. Everything on this site labeled cost per ride or subsidy per ride is computed from it.
Dataset ectq-t3k3
· last read September 17, 2026
· 22,105 rows
ok
Service Data and Operating Expenses by Mode (Time Series 2.1)
2015 to the latest published yearThe same measures split by mode and by whether the agency runs the service itself or buys it.
Dataset npsm-38gk
· last read September 17, 2026
· 37,206 rows
ok
Monthly Modal Time Series
January 2014 to one month agoMonthly boardings, service and safety events by agency and mode. Full reporters only. The only NTD publication that includes both major and non-major safety events.
Dataset 5ti2-5uiv
· last read September 17, 2026
· 192,186 rows
ok
Reporter Agency Information
Latest report yearAgency address, website, urbanized area, service area population, fleet size and reporter type.
Dataset ccvf-fykn
· last read September 17, 2026
· 2,914 rows
ok
Fuel and Energy
2022 onwardFuel consumed and miles driven on each fuel, by agency and mode.
Dataset 8ehq-7his
· last read September 17, 2026
· 3,759 rows
ok
Funding Sources and Capital Expenses
2022 onwardFederal, state and local funding, and capital spending. Four files, merged.
Dataset qpjk-b3zw
· last read September 17, 2026
· 6,799 rows
ok
What was spent on building, rebuilding and replacing, split by whether it expanded the system or kept it standing, and who funded operating against capital.
Dataset fphd-jyyj
· last read September 17, 2026
· 25,675 rows
ok
Capital Expenditures Time Series (TS3.1)
1992 to the latest published yearCapital spending by agency and year, going back three decades. A spreadsheet on FTA's own site rather than a portal dataset, which is why it is read differently from the rest. The fleet depreciation estimates are built from it.
File TS3.1
· last read September 17, 2026
· 36,440 rows
ok
FTA blocked the download; read from the copy committed with the site
Operating and Capital Funding Time Series (TS1.2)
1991 to the latest published yearFederal, state, local and directly generated money, split between running the service and building it, for every agency and year. The portal's own funding files begin in 2022; this is the same measure for thirty-four years, and it is what the funding page is drawn from.
File TS1.2
· last read September 17, 2026
· 44,561 rows
ok
FTA blocked the download; read from the copy committed with the site
Asset Inventory Time Series (TS4.1)
1992 to the latest published yearHow many vehicles each agency ran, how many met the ADA accessibility standard, and their average age, seating and standing capacity. The portal publishes the vehicle inventory for one year only.
File TS4.1
· last read September 17, 2026
· 66,375 rows
ok
FTA blocked the download; read from the copy committed with the site
Safety & Security Time Series (threshold adjusted)
2008 to the latest published yearIncidents, injuries and fatalities by agency, mode and year, with boardings and revenue miles in the same row, so a count can be turned into a rate. FTA adjusts this series for the reporting-threshold changes of 2010, 2015 and 2023; the unadjusted file jumps 34% in 2015 on the rule change alone.
File S&S Time Series
· last read September 17, 2026
· 24,659 rows
ok
FTA blocked the download; read from the copy committed with the site
Major Safety and Security Events
2014 onwardEvery reported major event on its own row: collisions, fires, derailments, assaults and suicides, with the time, usually a place, and counts of injuries and fatalities.
Dataset 9ivb-8ae9
· last read September 17, 2026
· 115,445 rows
ok
Revenue Vehicle Inventory
Latest report yearEvery revenue vehicle an agency owns, by type, with age distribution and lifetime mileage. The fleet-age charts and the past-useful-life shares come from it.
Dataset 6abt-uhgq
· last read September 17, 2026
· 23,482 rows
ok
Revenue Vehicle Useful Life Benchmark
Latest report yearHow many vehicles each agency reports as past their own useful life benchmark. Shown alongside the figure derived from age, because the two disagree and both are worth seeing.
Dataset nimp-626k
· last read September 17, 2026
· 17,703 rows
ok
Service Reliability
2022 onwardMechanical failures and the distance between them, by agency and mode.
Dataset amkt-4ehs
· last read September 17, 2026
· 3,731 rows
ok
Employees
2022 onwardHeadcount and hours worked, split between operators and maintenance, and between full-time and part-time.
Dataset uyv8-9jek
· last read September 17, 2026
· 4,018 rows
ok
The public-facing names agencies use. NTD reports legal names only, so the search index is built from NTD's own doing-business-as field, the provider names on agencies' GTFS feeds, and initialisms derived from legal names.
Dataset 2u7n-ub22
· last read September 17, 2026
· 14,323 rows
ok
How the figures are computed
- Cost per ride
- Operating expenses divided by unlinked passenger trips. A trip is a boarding, so a journey with one transfer counts twice.
- Subsidy per ride
- Operating expenses less fare revenue, divided by boardings. NTD does not publish this; it is computed here. It covers operating subsidy only, not capital.
- Fare recovery
- Fare revenue as a share of operating expenses.
- Current dollars
- Restated using CPI-U, U.S. city average, all items, annual averages, BLS series CUUR0000SA0, with 2024 as the base year.
- Passengers aboard
- Passenger miles divided by vehicle miles: how many people are on board on average at any moment. Reported by full reporters only, since smaller agencies do not file passenger miles.
- Average speed
- Vehicle revenue miles divided by vehicle revenue hours, stops and traffic included. Cost per mile is cost per hour divided by this.
- Capital spending
- What an agency spent that year, not what its capital cost it that year. NTD publishes no depreciation and this site does not estimate one, so capital is shown beside operating cost and never folded into it.
- Deaths per billion passenger miles
- Given twice: counting everyone killed, and counting riders alone. Most transit deaths are trespassers and suicides, so the two differ about eightfold and only the second is a measure of risk to a passenger.
- Safety rates
- Events per ten million boardings, from complete calendar years only. A partial year is left out rather than shown short.
- Zero-emission miles
- Miles run on battery, hydrogen or electric propulsion, as a share of all fleet miles. Compared in miles, not gallon equivalents.
What this site does not tell you
- An unlinked trip is a boarding, not a journey. A rider who transfers once counts twice. Agencies whose networks require transfers therefore look busier, and cheaper per ride, than those whose routes run direct.
- Agencies choose their own fiscal year. FY2024 means a different twelve months in different places, so year-to-year comparisons across agencies are close rather than exact.
- Reduced and rural reporters file less. Small agencies are not required to report every measure, so blanks on their pages usually mean "not required", not "zero".
- Purchased transportation moves costs around. An agency that contracts its service out reports the contract price, which bundles costs that a directly operating agency reports separately.
- NTD restates. Agencies correct earlier filings, and FTA republishes. Figures here match the portal as of the dates above and can change.
Reusing any of this
The FTA figures above are public domain and need no permission from anyone. What this site derives from them — subsidy per ride, safety rates, the depreciation estimate, the assumed-name index — is CC BY 4.0 and asks only that you say where it came from. The about page sets out both, with citation formats; /llms.txt says the same for machines.