Quick Note on Ridership Estimation and the STOPS Model
The Federal Transit Administration (FTA) has a model for ridership estimation for public transit lines, called STOPS. It’s designed to be simplified enough that transit agencies can compute rough projections with it without having a large permanent in-house team dedicated to it (the MTA has a team of about 12). At the Transit Costs Project, we’ve taken to using it to do estimates for potential extensions under various assumptions on speed, frequency, and service pattern. For example, the Effective Transit Alliance’s statement in support of QueensLink for the budget vote later today quotes a figure of 105,000 expected daily riders, with 75,000 of them new and 30,000 existing at the Rockaway stations. I’d like to both go over this figure in this post and explain why at the same time, STOPS must be used with caution, especially in New York.
The QueensLink ridership estimate
At the program, Franklin Tang is responsible for the use of STOPS. Using the tools of the model, he computed that under our assumptions of QueensLink service, total northbound boardings at all stations south of Rego Park-63rd Avenue, the branch point from the Queens Boulevard Line, are 52,634 per weekday. Every trip must be paired between north- and southbound travel, so this means a total of 105,268 trips using QueensLink. These numbers include stations that already have service on the A train, including Howard Beach and the Rockaways; one of the points of QueensLink is to give these areas faster service to Midtown than they get on the A, which takes the long way around via Downtown Brooklyn and Lower Manhattan. In a No Build scenario, we expect these preexisting stations to have about 30,000 total trips, taking into account some expected redevelopment around the Aqueduct.
The business case released by QueensLink, released using the same analysis, gives a total of 75,000 riders/day. This is not the net figure – the similarity is coincidental. Rather, it includes things it shouldn’t (namely, boardings at Rego Park, nearly all preexisting) and excludes things it should (namely, the southbound alightings), as it imitates MTA methodology, which for reasons I don’t quite get counted the Rego Park boardings in its estimate for QueensLink ridership.
The figure people should use in subsequent analysis is the higher one, which is the most comparable to how transit ridership of new projects is usually computed. Riders diverted from other lines still benefit from the project, especially in a case like QueensLink, where the diverted riders would enjoy an improvement in trip time to Midtown of about 10-15 minutes each way. Second Avenue Subway’s ridership, which we compare QueensLink’s favorably with in the ETA analysis, is based on estimates for total ridership, not just new trips, and includes diversions from the 4/5/6 and from crosstown buses in addition to new trips.
The problem of STOPS
STOPS is designed for quick-and-dirty calculations and is known to choke on New York. FTA intends for it to be used elsewhere in the United States, where I am told it gives more accurate numbers; the MTA’s in-house team can do more precise calculations by itself, without STOPS.
Case in point: we tried estimating the ridership of a Utica Avenue subway extension with STOPS, and got bad numbers. “Bad” here does not mean the numbers were low and I’d like them to be higher, but that the model gave nonsensical results. Total northbound boardings on the new stations amounted to 58,000; total southbound alightings only amounted to 25,000. In effect, the model thinks that if a Utica Avenue line opens, then 33,000 people will permanently emigrate from Southeast Brooklyn every day. We’ve rerun the numbers and at this point I don’t believe we made an error; rather, STOPS has some errors, perhaps with stations close to one another, that lead to random results that don’t quite average out correctly. Thus, I recommend caution when using it as it is still only good enough for a first-order estimate.
Pedestrian Observations 
I realize Midtown isn’t Stamford but people can have destinations … and origins… other than Midtown. I’m not going to dig out current ridership numbers. In nice round numbers, in 2019 Jay St.-Metrotech had twice as many passengers as 50th and 8th. Canal Street which isn’t Midtown and isn’t “Wall Street” had as many.
Not if you are going to all stuff going on around Jay St……… Unless you are clueless railfan who likes to take silly subway trips. I seem to remember some clueless railfan taking a silly trip from Queensboro Plaza involving two transfers, one with hike, to get to Jay St. Instead of taking the other train which has a cross platform transfer to trains going through Jay St. It would probably be a transfer at West 4th from a Sixth Ave local to an Eighth Ave. train. Though I suppose it could be foamer froth fest of Far Rockaway via Queens Blvd via 8th Ave to Lefferts Blvd.
Or you could think about three things at once and decide while it’s an amusing transit geek circle jerk there might be better places to spend money…. There’s a reason the LIRR was desperate to abandon it…. Does this “faster” involve changing from a local to an express. It’s going to be a Queens Blvd local isn’t it.
Yer gonna use STOPS to guesstimate ridership and then explain that STOPS doesn’t work. Okay.
”I realize Midtown isn’t Stamford”
Adirondacker when people suggest building HSR from DC-Phila-NYC (~32.8M) to Boston (~5M), passing through Stamford/Bridgeport (~900k):
’Stupid radians think Stamford is the center of the universe! That little town and the rest of SW Connecticut are sideshows. While would you send high speed trains there instead of a major destination?! Don’t they know the purpose of a train is to move people so you need to go where the people want to go!?’
Adirondacker when people suggest building a faster shorter route to Midtown Manhattan, the largest business district on the continent, with more people traveling there and back each day as the entire population some metro areas:
’Oh hey, look at these interesting secondary destinations people might want to travel to instead….’
Why do people need three ways to get to Midtown when there are places, within the city, that don’t have trains? Other that to whip up some railfan froth over a line that never should have been built and was abandoned for lack of interest in 1962 ? The people who are currently using the Queens Blvd line might be a bit peeved to discover they no longer have a train so people who have other trains can have a third route.
The railfannery gets as far Rego Park where the wundertrain apparently uses it’s Heisenberg Compensators to be in the same place at the same time as an existing Queens Blvd train. That does allow for intriguing possibilities like changing to Broadway train from the wundertrain and vice versa while in motion without involving a platform. Assuming it’s on the local tracks and Broadway trains continue to run on Queens Blvd once the Queens Blvd line is connected to Second Ave. Considering what happens on Queens Blvd when the express train goes down Second Ave is well outside of the scope of railfans working up a batch of foam over a line that was abandoned for lack of interest in 1962.
None of that changes that workers in Queens work many places not-Manhattan and the ones that do work in Manhattan can work places not-Midtown. Like Wall Street the country’s third largest business district. Which means it has more jobs worth-a-train-ride than Boston or Philadelphia or Atlanta or San Francisco or Denver. Or Stamford. Or change that someone who lives in Queens Village and works in Rego Park doesn’t give a shit if the train goes to Sixth Ave or Broadway or Eighth Ave. Or where it wanders in Brooklyn or if it terminates at the World Trade Center. Which isn’t in Midtown.
Why do people need three ways to get to Midtown and which Queens Blvd trains are you going to cancel to do it?
A book from the early 1900’s described Paris’ goal for expanding its new Metro was to make every building in the City within 500m walking distance of Metro entrance. It would appear from the first few minutes of this video, that Paris followed this principle for half a century.
https://youtu.be/wy7lNJ7jrYg
Walking distance from buildings to metro entrances is metric that can be computed, using open source data. Open Street Maps provides precise locations of metro entrances and buildings almost anywhere in the world. Python and the Osmnx package provide the means to extract this information and calculate walking distances. This is more universal metric for comparing metro access because it does not require US Census block data for evaluation.
I calculated the walk-to-metro distance for various cities back in 2024. Paris shows that 83% of all building entrances were within 500 m walking distance of a Metro entrance. The currently accepted figure for “walk-to-subway” distance is 800 m or 1/2 mile. Approximately 97% of all buildings in the 20 arrondissments are within 800 m of a Metro entrance.
The comparable figure for NYC (5 boroughs) was 43%. The rationale I’ve heard for justifying subway expansions is real estate development. I have a fairly parochial view for measuring an expansion’s benefit.
It’s the number of new/additional riders it attracts. However, not buildings are equidistant to a metro entrance. The benefit metric should be altered to (new) rider-walking miles eliminated. Thus, a new extension that brings buildings from 4 miles distant would be 4 times more valuable than one that brought buildings from 1 mile distant, given the same number of buildings for each extension. An addition time factor can be added, when comparing extensions that require 5 or 10 years for completion.
This brings us to QueensLink which would add 4 new stations. Only its proposed Parkside station lies more than 1/2 mile from existing subway stations. The proposed Ozone Park, Atlantic Ave, Jamaica Ave (nee Brooklyn Manor) would not provide many new rider-miles due to their proximity to existing stations on the A and J lines. Parkside is a little less than 1 linear mile from the Continental station on the Queens Blv line. The Q23 connects to buildings from north to east and south from the Parkside station to Continental.
A significant percentage of the area from northwest to southwest of the Parkside station lies within St. John Cemetery. That’s not likely to provide any new ridership. The LIRR Montauk Branch lies 420m (1400 ft) south of the Parkside station. The line is at grade with no crossings. It effectively blocks off any additional walk-to-subway riders south of its right-of-way. The conclusion is that the Parkside will offer very little additional ridership or building-miles due to the constraints of the cemetery and the LIRR.
I do not have benefit of the STOPS analysis. I’ll assume that number of additional passengers due to the Parkside station would be 1/4 of the bloated 105K total or 26.25K.
The question of operating cost has not been raised. The proposal includes extending the M train to Rockaway Park and the G train to Forest Hills. It’s easy cost to estimate because it’s the product of the number of car-miles (derived from the GTFS) and the service efficiency ($/VRM) that’s published by the NTD. The extra M train distance is approximately 10.5 miles (underestimate). There are 230 M train trips scheduled to and from Forest Hills.
The 2024 NTD tables report that NYCT’s heavy rail’s operating expenses were: $6,500,549,691 and it operated 380,426,893 vehicle-revenue-miles. This comes to approximately $17.08/VRM. The M train operates 8-car trains. This means the daily expense to extend service to Rockaway Park would be $330,130.76.
The G train extension from Court Sq to Forest Hills is approximately 6 miles. There are 236 G trains scheduled to arrive at Court Sq, when the M runs to Forest Hills. The M would be diverted to Rockaway Park and the G would provide additional service to Forest Hills. The G operates 5 car trains. The approximate daily cost for the extended G service would be $120,979.59.
This brings the total daily increased operating cost incurred by QueensLink to $451,110.35 or roughly $17.18 per additional passenger at Parkside. This does not include the portion of the ride from Rego Park into Manhattan.
Not all of NYC is created equal with respect to subway access. Manhattan is the outlier because 86% of all its buildings are located within 800 m walking distance of a subway entrance. This brings up the SAS.
It’s doubtful that any additional buildings came within 800 m of a subway entrance that did not already enjoy such access. There is hard data about how many more riders it attracted, thanks to NYMTC’s yearly Hub Bound Travel Report. The 2017 report shows the number of subway riders crossing the 60th threshold by subway line for both 2016 and 2017 the before/after SAS. The data is found in Appendix I, Table 5 on page I-10
https://www.nymtc.org/Portals/0/Pdf/Hub%20Bound/2017%20Hub%20Bound/2017_Hub-Bound-August_15_2019.pdf?ver=2019-08-15-080949-340#page=38
The 2016 pre-SAS data recorded 386,250 and 307,982 riders using the Lex express and local for a total of 694,232 daily riders. The 2017 post-SAS data recorded: 149,277; 330,963 and 247,372 riders on the SAS, Lex express and Lex local, for a total of 727,612 riders. That’s a net gain of 33,380. The figure pegs the SAS construction cost at $268,096 per new rider.
This result should challenge the assertion that providing alternate access to areas that already enjoy such access would draw even more subway use. This lesson is being ignored because the MTA will duplicate this experiment with phase 2 and expect a different result.
Is STOPS a rational tool to estimate ridership on CAHSR?
It’s not intended for intercity rail usage.
FTA intends for it to be used elsewhere in the United States, where I am told it gives more accurate numbers
“
I would love to see an anylsis of the models accuracy. Metro Transit has 2 projects, Gold Line and Northstar, where ridership was way, way, way off. 50% off in the case of Northstar ( paper napikin math including diffs in frequency ) and 80-90% short for the new Gold Line BRT. Feels liek quite a few others have been way too often.