Methodology
Three kinds of number appear on this site, kept apart in the database and on the page:
- Published figures, stored exactly as RTD or the FTA published them, each tied to a logged load of a named source.
- Derived figures, calculated from published ones by the formulas below and rebuilt from scratch on every refresh.
- User-contributed observations, always labelled, never used in any published or derived figure.
System cost rates
For each NTD report year and mode (bus, light rail, commuter rail, demand response), RTD-operated and contractor-operated rows are added together, then:
- Cost per revenue hour = operating expense ÷ vehicle revenue hours
- Cost per revenue mile = operating expense ÷ vehicle revenue miles
- Cost per boarding = operating expense ÷ unlinked passenger trips
- Cost per passenger-mile = operating expense ÷ passenger miles
- Average fare = fare revenue ÷ unlinked passenger trips; farebox recovery = fare revenue ÷ operating expense
- Average trip length = passenger miles ÷ unlinked passenger trips
- Average load (riders on board) = passenger miles ÷ vehicle revenue miles
Fully-burdened cost
Operating expense leaves out what the vehicles, track and buildings cost. Fully-burdened cost adds depreciation and interest expense from RTD's Annual Comprehensive Financial Report, allocated to each mode by its share of operating expense. Only ACFR figures marked verified against the page are used. Until they are loaded, fully-burdened columns show "n/a" rather than an estimate. NTD capital spending is shown separately. It is not used here because it is cash spent in a year, which jumps around.
Route cost
RTD does not publish cost by route. It does publish each route's annual boardings and service hours, compiled for 2018–2024 by Greater Denver Transit.
- Hours factor = the mode's NTD vehicle revenue hours ÷ the sum of all route service hours RTD reports for that mode and year. Rail reports train-hours while NTD counts each car (factor around 3–4). Bus reports leave out layover time that NTD counts (factor around 1.25).
- Route cost (hours basis) = route service hours × hours factor × the mode's cost per vehicle revenue hour. In effect, each mode's NTD operating cost is shared among its routes by share of hours. This is the headline figure: operator time is the largest cost, and hours capture slow, congested routes better than miles do.
- Route cost (miles basis) = scaled hours × the route's scheduled speed from the current GTFS feed × cost per vehicle revenue mile. A cross-check. It runs higher for fast express and rail routes and lower for slow downtown ones.
- Cost per boarding = route cost ÷ route boardings. Subsidy per boarding = (route cost − boardings × average fare) ÷ boardings. Free routes (16th Street FreeRide, Free MetroRide) are given zero fare revenue. This is labelled loss per boarding on the site; annual loss = route cost − boardings × average fare.
- Worst value: within each mode and year, routes are ranked by loss per boarding, and those above the median are flagged. Because every service hour of a mode costs the same in this model, this is the same set as the bottom half by boardings per hour, so by construction about half of each mode is flagged. The "worst value" ranking lists flagged routes first, by annual loss, so the top of the list is where the most money goes to the least-used service. Table colours: loss per boarding is coloured by that within-mode rank; annual loss by its rank among the routes shown.
- Passenger-miles = boardings × the mode's average trip length. This is an estimate: express and regional riders travel further than average.
What this misses. Every route is charged the system-average rate. In reality, peak-only routes cost more per hour (split shifts, buses idle midday), and contracted routes have different unit costs from RTD-operated ones. Route years without a matching NTD year are not costed. The calibration table shows how much of RTD's total service the matched routes cover.
Scheduled service (GTFS)
From the current GTFS feed: each trip's revenue miles = the length of its shape, measured in PostGIS on the WGS84 ellipsoid. Its hours = last arrival − first departure. Annual totals = Σ trips × days the trip runs in the feed window × 365 ÷ window length. A route is peak-only if it has weekday trips, none starting between 10:00 and 14:59, and no weekend trips. Route type (local / limited / express / regional / airport) comes from RTD's naming conventions and can be corrected in the admin.
Security and fare evasion
NTD security events are federally reportable events only, shown per million boardings. Fare-check figures come from RTD news releases and dashboards, cited row by row. Estimated lost fares = non-payment rate × the mode's annual boardings × average fare. This is an upper bound, because some riders who didn't pay would not otherwise have ridden.
Robotaxi scenario
Every input is set by you. Vehicle-miles = boardings × trip length × (1 + empty-mile share) ÷ riders per vehicle trip. Operator cost = vehicle-miles × cost per vehicle-mile. Compared on net public cost: (bus cost − bus fares) against (robotaxi cost − rider fares).
- RTD side. The route's estimated cost is split using NTD operating expense by function for its mode and year: vehicle operations and vehicle maintenance always count; facility maintenance (stops, stations, garages) and general administration count only when ticked, because most of them continue when one route is cut. "Without security" removes a user-entered annual security amount as a share of RTD's total operating cost that year. Vehicle replacement and debt interest are not included until RTD's ACFR figures are transcribed.
- Accessible fleet. Cost per mile × (1 + accessible share × accessible premium).
- Verdict. Robotaxi better if its net cost is below the bus by more than the "too close" band, RTD better if above by more, otherwise too close to call.
- Rush hour. Cars needed at once = (boardings ÷ 365 × busiest-hour share ÷ riders per car) ÷ (60 ÷ minutes per ride), where a ride is trip miles × (1 + empty share) ÷ speed, plus 3 minutes for pickup and drop-off. Average daily boardings understate weekday peaks.
- Not modelled: traffic and curb space, and whether an operator would serve low-demand areas at all.
Conclusion page
The Conclusion page is an argument; its numbers come from the sources on this site, combined as follows.
- Cost trend. NTD operating expenses, fares and unlinked trips, all modes, per report year. Cost per boarding = operating cost ÷ trips; fares cover = fares ÷ operating cost.
- Administration share. NTD general administration ÷ total operating expense, all modes (and bus alone).
- Arrests per officer. Arrests = sum of RTD's published monthly figures. Officers = average of RTD's published counts around the period (2025: May 2024 and December 2025; 2026: December 2025). Department cost per arrest = Transit Police and Emergency Management budget for the year ÷ arrests scaled to a full year. The department does more than arrests, so this is cost alongside that work.
- Stop coverage. 2020 Census block populations at each block's internal point inside RTD's district polygon. A block is covered if any stop in the current GTFS feed lies within 402 m (¼ mile) or 805 m (½ mile), straight line in UTM 13N. Frequent = a stop on a route with at least 120 weekday trips, both directions.
- CO₂. NTD fuel and electricity by mode × EPA tailpipe factors (10.21 kg/gal diesel, 8.78 kg/gal gasoline) and eGRID2023 RMPA CO₂e for electricity; ÷ NTD passenger-miles. Robotaxi per passenger-mile = kWh per mile ÷ (1 − empty share of all miles) × grid factor, one rider. Empty share 46% is Waymo's California figure from CPUC data. No upstream or manufacturing emissions.
- Owner economics (all inputs are assumptions you set; defaults from Tesla's published Cybercab price, subscription and Austin fare). Per car per year: trips = trips a day × days; car miles = trips × trip miles × (1 + empty miles per paid mile); fare = base + per-mile × trip miles; owner revenue = trips × fare × (1 − platform share); running cost = subscription + car miles × (kWh/mi × $/kWh + maintenance/mi) + trips × cleaning + insurance; depreciation = price × (1 − resale share) ÷ years; profit = revenue − running cost − depreciation.
- RTD buying rides. For the chosen routes (latest route-cost year): avoidable bus cost = route cost × (operations + maintenance + facilities + administration share), as on the robotaxi page with its defaults; RTD pays (fare − rider payment) per ride, where the rider payment defaults to NTD bus fares ÷ bus boardings; saving = avoidable cost − bus fares − RTD's payments. Cars needed = the larger of (rides ÷ trips per car a year) and the rush-hour cars from the robotaxi page; when rush hour sets the fleet, owner profit is recomputed at the lower trips per car. Bus diesel for those routes = their revenue miles ÷ NTD bus diesel miles per gallon (excludes deadhead, so understated).
- Sell-off. Each RTD parcel (Colorado public parcels, owner RTD) is put in one group, first match wins: kept (touches the downtown Denver box −105.006 to −104.984 W, 39.736 to 39.759 N, or the East Metro garage at 14100 E Colfax); rail (within 20 m of a light or commuter rail shape in GTFS); Park-n-Ride (within 150 m of an RTD Park-n-Ride); garages, depots and offices (within 150 m of one in the NTD facility inventory); other. Sale value = county assessor value × (1 − sale discount); fund = that × (1 − FTA share). Both are your assumptions. Police saved = the department's FY2027 proposed budget, all of it when no trains run, otherwise less the share you keep for rail. Loan payment = standard amortising payment. Revolving fund: each year, repayments on earlier loans are added to the cash, then whole cars are lent.
- Who is near service. Each 2020 block's population is split by its census tract's American Community Survey shares (2020–2024 estimates, via Esri Living Atlas): 65 and over (B01001), age 5–17 (B01001), living with a disability (B18101, civilian non-institutionalised). Then the same ¼-mile and frequent-service tests as above.
- Door-to-door services. Robotaxi fare for a trip = the calculator's base fare + per-mile fare × trip miles. School: each district's 2024-25 transport operating cost (CDE worksheet line 1) ÷ (pupils × school days × 2) gives cost per child per one-way ride; robotaxi = fare ÷ children sharing. Access-a-Ride: NTD demand-response operating cost and trips; trip miles = passenger-miles ÷ trips. Medicaid: Colorado's 2023 trips and cost per trip; trip miles as Access-a-Ride. Ambulance: Denver Health transports × the low-acuity share you set × (Medicare 2026 BLS non-emergency base + mileage rate × miles − robotaxi fare). Wheelchair trips needing accessible vans are not separated out.
- High school. Rides and cost = all-district totals × the high-school share you set (default: grades 9–12's share of Colorado enrolment, 31.9%), assuming the same cost per ride as the district average; robotaxi = fare ÷ teens sharing.
- Wheelchair trips. Trips a year default = RTD's stated ~$1M a month of accessible Access-on-Demand cost × 12 ÷ its $270 per trip. Automated accessible van per trip = robotaxi fare for the average paratransit trip × 1.5 + van price ÷ (trips per day × 360 × years) + helper cost per trip. Vans needed = trips ÷ (trips per day × 360).
- Bus automation retrofit. Whole pilot project cost ÷ 3 vehicles (PANYNJ and CDOT pilots), × RTD's 955 buses. Pilot costs include engineering and evaluation.
- Where the savings go. One plan from the calculators at their defaults: every bus route replaced (owner calculator, "every bus route"), Access-a-Ride and wheelchair rows from the door-to-door calculator, land sold with rail kept. Police is not added again (security sits in the administration share already counted). Per household = RTD's yearly saving ÷ housing units in the district's 2020 census blocks. Sales-tax share = saving ÷ RTD's 2027 proposed sales and use tax. Interest = 2025 ACFR interest expense. Workforce = amount ÷ budgeted full-time bus operators.
- Posts for X. Built from the same figures each time the page loads; length counts each link as 23 characters, as X does.