The eight-year rise and fall of a train service that crossed Penn Station—and what it shows about plans to let commuter trains continue through the station instead of ending there.
New York has one of the world’s largest passenger-rail networks, but it doesn’t operate as one system. The Metropolitan Transportation Authority (MTA) runs Metro-North Railroad, which serves the northern suburbs and Connecticut, and the Long Island Rail Road (LIRR), which serves Long Island. NJ TRANSIT runs commuter trains in New Jersey. Amtrak owns Penn Station and controls much of the track immediately east and west of it. Each organization has its own schedules, fares, crews, equipment rules, and operating authority.
Most commuter trains treat Penn Station as a terminal: they arrive, unload, and later reverse direction or leave the station without passengers. Through-running means that a train enters Penn from one side, stops briefly, and continues out the other side carrying passengers. Regional rail goes further by organizing those through services as a coherent network, with coordinated routes, fares, information, and responsibility for the complete trip.
From September 2009 through January 2017, the “Train to the Game” provided a limited real-world test. Selected Sunday trains carried football fans from Connecticut on Metro-North’s New Haven Line, crossed Penn Station, and continued to Secaucus Junction in New Jersey. Riders then changed to a separate train for the last 2.5 miles to the Meadowlands sports complex, now home to MetLife Stadium. LIRR and subway riders could join the same event market at Penn.
The surviving agreements and performance records show that a passenger train could cross Penn under the authority of several railroads. They also show why the service remained fragile. It used schedule space that had not been reserved for it permanently, equipment borrowed from NJ TRANSIT, crew changes at Penn, and agreements that could be reconsidered each year. The pilot therefore offers a useful distinction for today’s Penn Station debate: physical through-running was feasible, but a durable regional service required stronger institutional and operating commitments.
The idea emerged during a period of renewed interest in making the New York metropolitan area’s separate railroads function more like one network. On May 5, 2006, Eliot Spitzer—then New York’s attorney general and a candidate for governor—spoke to the Regional Plan Association, a nonprofit civic-planning organization, about closer transportation cooperation among New York, New Jersey, and Connecticut.
Spitzer used a trip from New Brunswick, New Jersey, to Mineola, Long Island, to explain the objective. Such a trip crosses the territory of NJ TRANSIT and the LIRR and ordinarily requires a change of trains at Penn. A “one-seat ride” would allow the passenger to remain on the same train. His broader point was that the region might gain useful new travel options by coordinating existing rail lines before building an entirely new network.
“We should undertake several important regional initiatives in partnership with New Jersey and Connecticut. Enabling what the experts call regional transit interoperability could make possible a one-seat ride from New Brunswick in New Jersey to Mineola on Long Island.”Eliot Spitzer, Regional Plan Association Regional Assembly, May 5, 2006
After taking office, Spitzer appointed Elliot G. Sander to lead the MTA. Sander argued that changing travel patterns—more trips across suburbs, state lines, and traditional railroad territories—required cooperation among the MTA, Amtrak, NJ TRANSIT, and the New Jersey Sports and Exposition Authority, which then oversaw the Meadowlands complex.
“The MTA, along with the LIRR and Metro-North, is working with Amtrak, NJ Transit and the New Jersey Sports and Exposition Authority on launching a pilot regional rail interoperability project.”Elliot G. Sander, Crain’s Breakfast Forum, June 5, 2007
The proposed football service was deliberately narrow. It would give selected Metro-North New Haven Line riders a direct train through Penn to Secaucus for early Sunday Giants and Jets games. Riders coming from Long Island or the subway would reach the same service at Penn. Sander presented the pilot as a practical test that could inform broader commuter service later.
The Permanent Citizens Advisory Committee to the MTA (PCAC), the state-created organization that supports the MTA’s three rider councils, placed the experiment within a broader policy agenda. Its 2009 report Going the Distance argued that true regional integration has three parts: trains and buses that connect, fares that work across agencies, and information that lets passengers understand the complete trip. The Meadowlands pilot advanced the first part, but its later problems appeared where the fare and information systems remained divided.
The pilot did not provide a one-seat ride all the way to the stadium. NJ TRANSIT supplied the physical train—a locomotive and passenger cars—while Metro-North crews operated it from Connecticut to Penn. At Penn, an NJ TRANSIT crew took over and operated the same train to Secaucus. Amtrak controlled the tracks and train movements on the portion between New Rochelle, Penn, and New Jersey.
At Secaucus, stadium-bound riders had to leave the pilot train. It stopped on the station’s upper level, which carries the Northeast Corridor, the main rail route linking Boston, New York, Philadelphia, and Washington. Riders moved through the station to lower-level tracks running roughly perpendicular to the upper level. There they boarded a Meadowlands Rail Line train operating between Hoboken Terminal and Meadowlands Station, with Secaucus as its only intermediate stop. The final ride took about 10 minutes. A future Gateway Program proposal called the Secaucus Junction and Bergen Loop would connect the lower-level network to the Northeast Corridor, but that connection did not exist during the pilot and is separate from the unbuilt Meadowlands loop discussed later in this article.
The trip began on Metro-North’s New Haven Line, with selected stops in Connecticut and Westchester County before the train reached New Rochelle, where Metro-North’s route meets Amtrak’s line toward Penn Station.
Stops differed by gameMetro-North-operated segmentAt New Rochelle, the train left the normal Metro-North route into Grand Central and entered Amtrak’s Hell Gate Line. That line crosses the Bronx and Queens, uses the Hell Gate Bridge, and reaches Penn through the East River tunnels. Metro-North engineers and conductors therefore needed additional training and formal permission to operate under Amtrak’s rules on Amtrak-controlled track.
About 35 minutes to PennAmtrak-controlled trackAt Penn, the passengers remained aboard while the operating crew changed. The same physical train shifted from Metro-North operation to NJ TRANSIT operation. The planned stop lasted about 13 minutes, followed by an eight-minute trip through the Hudson River tunnels to Secaucus.
About 13 minutes at PennSame train, different crewMeadowlands passengers got off at Secaucus, but the original train did not turn toward the stadium. It continued southwest as a regular NJ TRANSIT train and then followed its scheduled branch toward Trenton, Dover, or Long Branch. The event service was therefore an extension of an existing train, not a dedicated train whose final destination was the Meadowlands.
Passengers leave the trainOriginal train continues westThe actual change of trains occurred when riders moved from the upper-level Northeast Corridor platforms to the lower-level Meadowlands Rail Line platforms. The two sets of tracks crossed at different elevations and were not physically connected, so the Meadowlands train was a separate service rather than a continuation of the pilot train.
Transfer between station levelsA separate Meadowlands trainThe service depended on two principal documents: a 2008 operating agreement between Metro-North and Amtrak, and a 2009 memorandum of understanding between Metro-North and NJ TRANSIT. A memorandum of understanding records how organizations intend to work together, but it does not by itself create a permanent service obligation. Together, the documents assigned equipment, crews, fares, access charges, and responsibility for problems. They also show that the pilot was fitted around existing trains rather than given its own permanently protected schedule.
“All trains operated … will be extensions of trains regularly scheduled … and, accordingly, there will be no increase in the number of NJ TRANSIT trains operating via the North River tunnels.”Metro-North–Amtrak Agreement Regarding Football Game Service Pilot, November 2008
The pilot did not add another NJ TRANSIT train through the two North River Tunnels under the Hudson. Instead, it extended trains that were already scheduled west of Penn and used spare Sunday capacity in the four East River Tunnels between Penn and Queens. The railroads reviewed the schedule each year, and dispatchers retained flexibility to move other trains first. In practical terms, the pilot had no permanently reserved train slot of its own.
The often-cited $22,287 cost was only a 2007-dollar example of what Metro-North might pay NJ TRANSIT to rent three complete trains for one service day. It did not include the entire cost of running the service. Amtrak charged separately for track access and electricity, while training, inspections, ticket handling, and disruption response created additional expenses. The surviving records do not contain enough final invoices to calculate the full public subsidy per rider.
| Measure | What the agreements show |
|---|---|
| Sample charge for NJ TRANSIT equipment | $22,287 per service day in 2007 dollars for three complete trains, maintenance allocated to those trains, and technician support. |
| Average fare revenue in 2009 | About $7,525 per game. |
| Fare revenue compared with the sample rental charge | About 34 percent in 2009 and roughly 10 to 34 percent in the reported years. |
| Amtrak track access and electricity | Separate charges based on how far each locomotive and car moved, including empty repositioning trips. |
| Other costs billed or assigned separately | Crew training, equipment testing, ticket inspection, repairs, substitute transportation, permits, and added operating support. |
| Charge for using scarce track capacity | Amtrak waived this charge for the pilot, so the accounts did not price the value of the schedule space used. |
The table should therefore be read as a comparison of documented components, not as a complete profit-and-loss statement. Even on that limited basis, the reported fare revenue never covered the sample equipment-rental charge.
Metro-North officials said the trains ran well, but they also described the service as a major undertaking. NJ TRANSIT supplied electric locomotives and up to ten double-deck passenger cars. These were “push-pull” trains: a locomotive powered the train from one end, while a control cab at the other end allowed it to reverse direction without moving the locomotive. NJ TRANSIT leased the equipment to Metro-North for the part of the trip east of Penn.
Metro-North employees could not simply take their usual train onto another railroad. They had to learn the NJ TRANSIT equipment, qualify on Amtrak-controlled territory, and demonstrate knowledge of the common operating rulebook used by northeastern railroads. The trains also had to pass tests for power supply, signals, braking, and inspections on every part of the route.
Penn Station was the point where the institutional handoff became visible. On a westbound trip, Metro-North operated the train into Penn and an NJ TRANSIT crew took over on the platform. The reverse occurred eastbound. Passengers stayed in the same cars, but the railroad legally responsible for the train changed. Amtrak’s Penn Station control center still decided how the train moved through the station and surrounding tunnels.
The original plan provided three passenger trains toward New Jersey before a game and three return trains afterward. It also required six empty repositioning trips to place the equipment where it was needed. Later games were difficult to serve because NJ TRANSIT had to return the borrowed trains to their normal locations in time for the Monday morning commute.
Passengers could buy a single ticket labeled for the Meadowlands, but the fare system was not truly unified. Metro-North, the LIRR, and NJ TRANSIT still set and divided their own fares. At Secaucus, employees manually checked MTA-issued tickets because the gates for the Meadowlands train could not read them. Penn’s departure boards also did not fully display the eastbound football service, so riders depended on announcements and staff instructions.
Daniel O’Connell, Metro-North’s director of operations planning, summarized the first season for the Metro-North Railroad Commuter Council (MNRCC), the official rider advisory council for the railroad. He said the trains had run smoothly, customer satisfaction was good, and the service had carried about 6,000 passenger trips. He also called the operation “a huge effort” and said riders needed clearer instructions at Secaucus. The train movement worked; the passenger experience around it was less complete.
Responsibility during a disruption was divided by territory. If a train failed, the railroad controlling that segment had to arrange rescue or substitute transportation. The Amtrak agreement covered the first several seasons, then renewed one year at a time and allowed either party to withdraw with 30 days’ notice. The same flexibility that made a pilot possible also made it easy to end.
The clearest comparison is Metro-North’s Yankees–E. 153rd Street station in the Bronx, which opened four months before the Meadowlands pilot. Both were new event services in 2009, but only the Yankees service rested on permanent infrastructure controlled by a single railroad. Regular Hudson Line trains stopped at the station. Special Harlem and New Haven Line trains reached it through the Mott Haven wye, a triangular track junction that lets trains switch among Metro-North’s three main lines. Additional shuttles ran from Grand Central.
Metro-North’s first-year report to the MNRCC shows how broadly that station drew riders. The railroad recorded about 500,000 rides after opening, including roughly 450,000 trips to or from baseball games. Of the game riders, 35 percent came from the Hudson Line, 30 percent from the New Haven Line, 25 percent from the Harlem Line, and the remainder from Grand Central shuttles. Familiar trains and parking charges of at least $28 at the stadium also made rail an attractive choice.
Within months, Yankees service had become part of Metro-North’s normal operating repertoire. The railroad could add trains, adjust game schedules, and communicate with passengers through systems it already controlled. The Meadowlands train remained a special exception whose equipment, schedule, and customer information depended on agreements with two other railroads.
Data released by the MTA under New York’s Freedom of Information Law (FOIL) makes the difference in scale visible. For one early-season game beginning at 1:05 p.m., Metro-North counted 1,933 arrivals and 1,764 departures at Yankees–E. 153rd Street. That is 3,697 passenger movements; one person arriving and later leaving counts as two movements. The Meadowlands pilot averaged 647 movements per game in 2009—342 toward the game and 305 returning. The Yankees count was about 5.7 times as large. The comparison is not perfectly equal because the Yankees data covers the complete station operation, while the football figure covers only the New Haven Line pilot. That difference is itself important: the Yankees station connected directly to several established markets, while the pilot served one narrow corridor and still required a transfer.
One game, two service models
The same game count shows how quickly a permanent through station could disperse a crowd. Metro-North recorded 941 departures in one 15-minute period—more than half of all outbound movements—and 79 percent of departures within 45 minutes. Trains could arrive from one direction, load, and continue through the station rather than forcing every train to reverse from a track that ended at the platform.
The comparison does not prove that every event service should copy Yankee Stadium. It shows what permanent access and unified control made possible. Contemporary PCAC accounts called the new Yankees station a “rousing success” and the Meadowlands train both groundbreaking and unusually labor-intensive. The football pilot proved that several railroads could move one train through Penn. Its lower use and eventual cancellation reflect the limited service offered, not a failure of through-running as an operating concept. The February 2010 MNRCC minutes record both assessments at the same meeting.
The service began on September 20, 2009. For each game, the original plan scheduled three passenger trains toward New Jersey and three return trains, plus six empty trips to reposition the equipment. Across nine games in the first season, the pilot carried 5,826 passenger trips, an average of 647 per game.
The MTA’s 2010 financial crisis changed that model. At the same February meeting where Metro-North reviewed the Yankees and Meadowlands services, the railroad was preparing broader train cuts and shorter trains to reduce operating costs. The football service fell from six passenger trips per game to only two: one train toward the game and one train back. Ridership declined by 17 percent even though the scheduled passenger service fell by 67 percent. One inbound train still carried 401 passengers on November 28, 2010, showing that demand remained after the service was sharply reduced.
2009: 5,826 riders, six passenger trains per game. 2010: 4,837 riders, two passenger trains per game. 2011: 4,030 riders. 2012: 3,041. 2013: 2,572. 2014: 3,835. 2015: 2,689. 2016: 2,170.
With only one useful train in each direction, the service became risky for passengers. Missing the outbound train meant finding another way to the stadium; missing the return train could mean waiting for hours or piecing together a different route. The pilot continued in this reduced form until its final operation on January 1, 2017.
Super Bowl XLVIII—Super Bowl 48—did not use the Train to the Game pilot. The February 2014 event was not an early Sunday game covered by the pilot schedule. Spectators instead used ordinary rail services to Secaucus and then changed to NJ TRANSIT’s separate 2.5-mile Meadowlands train.
Planners expected about 10,000 to 12,000 rail passengers. Roughly 28,000 arrived by train, and an estimated 33,000 to 35,000 tried to leave by rail after the game. The critical failure occurred between Secaucus and the stadium, not at Penn. Meadowlands Station was a stub-end terminal, meaning the tracks stopped there and every arriving train had to reverse direction before departing. The concentrated crowd overwhelmed that terminal operation and the single rail approach, producing waits of several hours.
The line that opened in 2009 came from a broader alternatives study completed four years earlier. The 2005 final environmental impact statement examined seven possible rail alignments and selected a 9,500-foot route connected to the Pascack Valley Line by a wye, a triangular junction that lets trains enter the branch from either direction. The selected route became the present Meadowlands Rail Line.
The station was not sited on the assumption that it would always remain the end of the railroad. The environmental review says it was positioned and designed to allow a separate future extension to the Bergen County Line. That connection was never built. Had it been completed, trains could potentially have entered the Meadowlands from one side and continued out the other rather than making every movement reverse at the same terminal.
The record does not establish that the Super Bowl failure ended support for the extension, and it cannot prove that the added connection would have prevented every problem in 2014. It does establish that planners recognized the operational value of preserving a second rail outlet. Such a connection could have reduced conflicts between arriving and departing trains and given dispatchers more ways to move equipment when demand exceeded the forecast.
The Super Bowl therefore supports a limited lesson. A journey can work well across most of the region and still fail at its final transfer or terminal. The event did not test whether trains could operate through Penn Station, because the pilot train did not run. It tested the capacity and resilience of the separate Secaucus–Meadowlands service.
The football trains used spare Sunday capacity in the East River Tunnels and extended NJ TRANSIT trains that already had scheduled slots through the Hudson River tunnels. A train slot is the reserved time and route that allows a train to move through a busy section of railroad without conflicting with other trains. The pilot’s slots were reviewed annually, and some planned tunnel work could still take precedence. The service had no permanent schedule right, dedicated operating budget, or long-term obligation after the initial agreement expired.
Going the Distance had warned that infrequent and frequently disrupted weekend service made regional travel unattractive. The pilot illustrates the problem. Lighter Sunday traffic created room for an experiment, but relying on leftover capacity also limited the games and times it could serve. Capacity that happens to be available is not the same as capacity reserved for a dependable public service.
The same NJ TRANSIT train crossed Penn with passengers aboard, but the operating responsibility changed there. Metro-North and NJ TRANSIT supplied different crews, Amtrak controlled the station and surrounding tracks, and separate agreements governed each part of the trip. Changes to Penn’s tracks and platforms can make through movement easier, but concrete and steel alone cannot create a regional service. The participating railroads also need one timetable, compatible equipment, coordinated staffing, and clear authority when something goes wrong.
PCAC’s earlier fare research made the same institutional point: a regional system should be judged by the passenger’s complete trip, not by whether each agency performs its own segment acceptably. The Meadowlands pilot created enough coordination to move a train across Penn, but no single organization was accountable for the journey from the first station to the stadium.
Passengers could purchase one ticket, yet they still encountered separate fare systems. Employees manually checked MTA tickets at Secaucus because NJ TRANSIT’s stadium-line gates could not read them, and the information shown at Penn was incomplete. A combined paper ticket simplified the purchase, but it did not remove the operational boundaries behind it.
Future through-running should appear to the rider as one service: a single, understandable fare; consistent information before and during the trip; and clear instructions when a connection is missed or a train is delayed. The passenger should not have to understand which railroad owns each track in order to complete the journey.
The sample NJ TRANSIT rental estimate represented only one part of the pilot’s cost. Amtrak access, electricity, employee training, and disruptions were handled separately. Responsibility for rescuing a disabled train changed at territorial boundaries. A permanent service needs a complete budget and a clear decision-maker before a failure occurs, rather than negotiations among agencies during the failure.
The pilot’s regional train ended its useful role for stadium passengers at Secaucus, where they changed from the upper-level Northeast Corridor to a separate lower-level train. The Super Bowl later showed how the final stadium segment and its terminal could be overwhelmed, although the through-running pilot itself was not operating. The unbuilt Bergen County Line extension would not have removed the transfer at Secaucus, but it could have given Meadowlands trains a second route through the complex instead of requiring every movement to reverse at the terminal. The complete trip is only as dependable as its most constrained connection.
Amtrak’s promise of “at least limited through-running” changes the central question at Penn Station. The issue is no longer whether a train can physically cross the station; the Meadowlands pilot demonstrated that it can. The question is whether Amtrak, the MTA, and NJ TRANSIT will turn that capability into a regular passenger service that can expand in stages.
In May 2026, the U.S. Department of Transportation and Amtrak selected Penn Transformation Partners, led by the construction firms Halmar and Skanska, to help plan and deliver the station project. The current concept keeps Madison Square Garden above the station and emphasizes a new train hall on Eighth Avenue, wider passenger areas, and easier movement between platforms and the street.
The most important commitment for railroad operations is Amtrak’s statement that the project will provide “at least limited through-running on the regional rail network.” That phrase does not require the separate Penn South expansion, which would extend the station into blocks immediately south of the existing footprint and demolish buildings including those on Block 780. A different strategy can begin inside the existing station: convert tracks and platforms in phases as specific pairs of routes become able to share equipment and operate through Penn.
The Federal Railroad Administration, the national rail-safety and investment agency, began an approximately 18-month Service Optimization Study in October 2025. That study should identify the first actual through route before the station design becomes difficult to change. It should specify which services would connect, how often trains would run, how long they would stop at Penn, and what happens when a late train loses its planned slot.
A “route pair” is the service created when a line arriving from one side of Penn is matched with a line leaving from the other. For example, a Metro-North route from Connecticut could be paired with an NJ TRANSIT route in New Jersey. Philadelphia used this approach when it joined formerly separate commuter railroads through its Center City tunnel. Planners compared ridership, trip length, service frequency, equipment, and yard access before deciding which lines should operate together. Penn needs the same route-by-route discipline rather than a single all-or-nothing conversion.
The first phase should be a hybrid operation. Some trains would continue to begin or end at Penn, while a smaller number of carefully selected routes would run through at dependable intervals. Amtrak’s 2014 study reached a similar conclusion: through-running is technically possible, but the amount of capacity it creates depends on the operating plan, and terminal and through services are likely to coexist. “Limited” should describe the first step in a published expansion sequence, not a permanent token service.
The timetable and organization matter as much as the construction. Britain’s 2018 expansion of the Thameslink network—a cross-London rail system that sends suburban trains through the city center—suffered severe disruption when infrastructure, schedules, staff training, and decision-making were not ready at the same time. Before Penn’s first route pair begins, the railroads need a tested timetable, enough compatible trains and qualified employees, and one chain of command for both normal service and emergencies.
The Meadowlands pilot established the minimum technical fact: one passenger train can cross Penn under the control of several railroads. It also showed the limits of treating that movement as an isolated exception. Crews and legal responsibility changed at Penn, the schedule relied on borrowed slots, fare and information systems remained divided, and stadium passengers still had to change trains at Secaucus.
A meaningful first phase should therefore identify the routes by name, reserve their schedule slots, provide useful service throughout the day, and publish the sequence for adding later routes. Each phase should work as a complete service on its own while preserving the ability to convert more tracks and platforms inside the existing station. The railroads should evaluate station construction against that operating plan before new structures lock in a less flexible layout.
Penn Station will become a regional-rail project only when “limited through-running” means the first enforceable stage of a larger program. The relevant test is not whether a ceremonial or special-event train can cross Penn once. It is whether ordinary passengers can rely on the first route pairs every day—and whether the railroads have committed to build from them.
The documents below include the original pilot proposal, customer materials, agreements, and annual ridership and revenue figures used in this case study. The 2005 Meadowlands rail alternatives analysis provides the planning history for the built Pascack Valley Line connection and the unbuilt future extension toward the Bergen County Line.
The integrated renderer is unavailable in this browser. The complete document is shown below using the browser’s PDF viewer.
Open the complete document in a browser tabThe 2007 Connecticut Department of Transportation report proposing the pilot as a strategy for Penn Station access. All 8 pages are displayed below; scroll inside the viewer.
The integrated renderer is unavailable in this browser. The complete document is shown below using the browser’s PDF viewer.
Open the complete document in a browser tabThe official customer brochure promoting the regional event service. All 7 pages are displayed below; scroll inside the viewer.
The integrated renderer is unavailable in this browser. The complete document is shown below using the browser’s PDF viewer.
Open the complete document in a browser tab| Year | Games | Riders | Average per game | Revenue |
|---|---|---|---|---|
| 2009 | 9 | 5,826 | 647 | $67,727 |
| 2010 | 9 | 4,837 | 537 | $56,230 |
| 2011 | 10 | 4,030 | 403 | $46,849 |
| 2012 | 10 | 3,041 | 304 | $35,352 |
| 2013 | 10 | 2,572 | 257 | $29,900 |
| 2014 | 12 | 3,835 | 320 | $44,582 |
| 2015 | 11 | 2,689 | 244 | $31,260 |
| 2016 | 11 | 2,170 | 197 | $25,226 |
| Total revenue | $337,126 | |||