Eighteen Trains An Hour Cross This Station Empty
Penn Station sits under Midtown Manhattan and handles more than 350,000 rail passengers a day, more than any other station in the country. Three railroads use it: NJ Transit from the west, the Long Island Rail Road from the east, and Amtrak running the length of the Northeast Corridor. It is about to be rebuilt at a cost measured in billions, and one decision inside that rebuild will shape how the region travels for the next fifty years.
This is the evidence on that decision, drawn from the federal study of the station, the engineering study of the Hudson River tunnels, and census records of where people live and work.
Trains can stop here, or they can keep going
Almost every commuter train at Penn Station terminates. It arrives, empties, sits at the platform, fills again, and leaves the way it came. Railroads call this a turn. The alternative is through-running: the train continues across Manhattan and out the far side, picking up passengers as it goes, the way the London Underground or the Paris RER work. A rider from New Jersey could reach Queens or Long Island without changing trains.
Penn already runs trains straight through. It just runs them without passengers. An NJ Transit train arrives from New Jersey, discharges everyone, and continues east through the station to a storage yard in Queens. An LIRR train arrives from Long Island, discharges, and continues west to a yard on Manhattan's West Side. Railroads call an empty move like this a deadhead.
The Federal Railroad Administration counted these movements in a study of the station published in June 2026. Its table of morning rush-hour traffic includes a column for commuter trains running through with passengers aboard. That column reads zero.
| Morning rush, trains per hour | Hudson | East River |
|---|---|---|
| NJ Transit arriving from New Jersey | 20 | — |
| NJ Transit returning to New Jersey | 11 | — |
| NJ Transit continuing east to the Queens yard, empty | — | 9 |
| LIRR arriving from Long Island | — | 19 |
| LIRR returning to Long Island | — | 10 |
| LIRR continuing west to the Manhattan yard, empty | 9 | |
| Commuter trains running through in service | 0 | 0 |
Federal Railroad Administration, New York Penn Station Service Optimization Study, Phase I, Table 4, morning peak baseline. The report defines both yard moves as non-revenue deadhead service. The nine westbound LIRR movements are the difference between arrivals and departures through the tunnels, absorbed by the Manhattan yard; the evening table reverses the pattern.
Check the arithmetic behind “three times capacity”
The case for enlarging Penn Station rests on a number. The Regional Plan Association, a century-old civic group that advises on regional infrastructure, reported in April 2026 that the station runs at roughly three times the capacity it was designed for. That figure drives the argument for demolishing a block south of the station and building more platforms.
Here is where the number comes from. The Pennsylvania Railroad rebuilt Penn in the 1960s as two basement levels, selling the air rights above to Madison Square Garden. The rebuilt station was designed for 200,000 rail passenger trips a day. Today it carries more than 350,000 rail riders, and carried 455,000 before the pandemic.
A separate figure, roughly 600,000, counts everyone who walks through the complex daily. The engineering study that publishes it spells out who that includes: subway riders, office workers, Madison Square Garden crowds, and other pedestrians who are not railroad passengers at all. Divide that pedestrian count by the rail design figure and Penn runs at three times capacity. Divide the rail counts by the same figure and the multiple falls to 1.75 today.
Design capacity and the 2019 and pedestrian figures come from Doubling Trans-Hudson Train Capacity at Penn Station, a feasibility report published October 2024. Current rail ridership comes from the FRA Phase I report, June 2026. Penn Station is genuinely crowded and its platforms are genuinely too narrow; the question here is only what the three-times figure measures.
A train that turns blocks the track twice as long
The time a train spends stopped at a platform is its dwell. Amtrak, the LIRR, and NJ Transit set dwell standards for Penn jointly, through a body called the Tri-Venture Council. Those standards were last revised on December 6, 2013. They allow an NJ Transit train that turns around 22 minutes at the platform, a train running through to the Queens yard 12 minutes, and an LIRR train dropping passengers before heading to the Manhattan yard 6 minutes.
Crew habits do not explain that gap. A turning train must empty completely and then fill, one after the other, because there is no room to do both. Penn's platforms are 17 to 23 feet wide, all but one of them, which is the narrowest set among comparable major stations worldwide. A train running through empties and fills at the same time, at the same door.
Track capacity follows directly. Counting the safety interval between trains, a turn ties up its track for 24.5 minutes, or about 2.4 trains an hour. A through train in passenger service ties it up for 17.5 minutes, or 3.4 trains an hour. That is 40 percent more service on the same tracks, rising toward 69 percent if through trains hit the 12-minute standard.
Dwell standards from the Tri-Venture Council table reproduced in the FRA Phase I report. Track occupancy adds the 2.5-minute interval used in the Regional Plan Association's own capacity model, April 2026. Platform widths and the international comparison come from the FRA report.
Half the congestion is scheduling, not steel
West of the platforms sits a dense tangle of switches known as A Interlocking, where trains from the Hudson River tunnels sort themselves onto 21 platform tracks. Because trains cross each other's paths there, they wait. The FRA counted 113 conflicting movements between eight and nine in the morning. Assign each tunnel a fixed group of platform tracks and a fixed destination beyond, and the same modeling drops that count to 51. No excavation is involved.
Punctuality points the same way. The LIRR runs 97 percent of trains on time. NJ Transit's Northeast Corridor line runs 81 percent. Amtrak's Northeast Corridor runs 76. The LIRR is the only operator at Penn working from a centralized plan that assigns trains to tracks in advance. The other two do not have one.
Penn also lacks a standard system for confirming a train is ready to leave. Dispatchers rely on informal word, and departures lose fifteen to thirty seconds each after the signal turns green. NJ Transit does not announce which track a train will use until it has already stopped, which keeps passengers upstairs when they could be waiting on the platform.
Conflict counts and punctuality from the FRA Phase I report. Amtrak's figure covers the Northeast Corridor for 2025, NJ Transit's covers its Northeast Corridor line for January 2026, and the LIRR's covers the system as of February 2026. Each railroad's number reflects its whole network, not Penn Station alone.
Fifteen of twenty-one tracks can already do this
Two tunnels run under the Hudson River to New Jersey. Four run under the East River to Queens. For a train to pass through the station it needs a platform track wired to both. Trace the connections and the answer is tracks 5 through 19 — fifteen of the twenty-one, 71 percent of the station. The Regional Plan Association and the FRA reach that same set independently, from different documents.
Electricity is not the obstacle either. The two railroads draw power differently: NJ Transit from overhead wires, the LIRR from an electrified third rail beside the track. Overhead wire covers all 21 platform tracks and third rail covers 5 through 21, so every through-capable track already carries both. What is missing is rolling stock that can use both, since the LIRR runs single-level cars to clear its tunnels while NJ Transit runs double-deckers. That is a purchasing decision, not a construction one.
The track running the other way is not spare
Approach capacity is the serious argument against through service, and it deserves a straight answer. The rail lines feeding Penn are not symmetrical. Between Harold Interlocking and Jamaica in Queens, four tracks generally give three to the busy direction and one to the other. Between Millburn and Kearny in New Jersey, three tracks split two and one. Floral Park to Hicksville on Long Island follows the same pattern.
A single reverse-direction track binds when demand runs that way, and here it does. Compare each town's daytime population against the people who live there. Jamaica, Hoboken, Paterson, and Bridgeport all shrink during the workday, meaning more residents leave for work than arrive. Through-running would add service on corridors where the supposedly spare track already carries real outbound demand.
Track counts from the Regional Plan Association, April 2026. Daytime change is daytime population minus resident population within each drawn area, from census transportation tables covering 2017 to 2021. These figures show that the listed towns send workers out. They do not show that those trips use these rail lines or fall in the reverse rush hour. Matching each town to observed reverse-direction train volumes would settle it.
A small market, and every trip in it changes trains
Census employment records pair each worker's home address with their workplace. Keep only the addresses within a mile of a commuter rail line, then count the jobs linking a New Jersey county to Nassau, Suffolk, Fairfield, or New Haven — people crossing from one side of Penn Station to the other. Those are the commutes a through-running timetable would serve directly and a terminating one cannot.
The total is a few thousand jobs. Set against the roughly 330,000 rail-served jobs commuting into the New York core, it is small, and the reason cuts both ways. A trip that requires changing trains at Penn today is a trip most people do not make by rail at all. The current count measures the timetable that exists as much as the demand that would appear under a different one.
Build more platforms, or use the ones already there
The rebuild of Penn Station is funded and moving. What has not been settled is whether it also expands south, taking a city block to add dead-end platforms, or reorganizes the trains inside the footprint the station already occupies. A fair test models both on the same terms and publishes the results. No operating agency has published one — and a third design, examined and cleared two years ago, never reached the comparison at all.
Expand South
Acquires and clears property south of the station, then builds additional dead-end platforms for trains that turn around.
- 1Adds capacity measured by how many trains can terminate.
- 2Keeps each railroad within its own territory and equipment.
- 3Requires demolition and construction beside a station already being rebuilt.
- 4Leaves every turn at 22 minutes and every yard move empty.
Run Trains Through
Carries passengers on trains that already cross the station, using tracks wired to both rivers today.
- 1Keeps Midtown the largest destination in the timetable.
- 2Turns empty yard trips into trips carrying passengers.
- 3Cuts the time each train occupies a track from 24.5 minutes to 17.5.
- 4Fits within the station's existing footprint.
A third design passed the engineering tests, then disappeared
There was a third option. The October 2024 engineering study examined a layout drawn from the advocacy group ReThinkNYC, whose Regional Unified Network proposal was published in 2017: keep 12 of the 21 tracks, remove the other nine, use the recovered space to widen most platforms, and connect what remains so trains can run through. Fewer tracks, but each one usable far more often, because wide platforms let passengers board and alight at the same time. The plan beside this text follows the study's own drawing: twelve retained tracks and eleven platforms widened into the recovered space.
That layout passed the study's tests for track geometry, constructability, and fire and life safety. It failed on one thing only: the operating performance it produced under the service plan the study assigned to it. That plan reserved four tracks for frequent metro-style service and loaded the remaining eight with Amtrak growth, conventional suburban trains, Empire and Hudson Line service, express trains overtaking slower ones, and reverse-peak branch service — at dwell times of six to seven minutes and nine minutes between trains on a platform.
Under those assumptions it produced 32 peak-direction trains an hour under the Hudson against an imposed requirement of 48. That is a finding about a timetable, not about a building.
What The Layout Cleared
- ✓Track geometry.
- ✓Constructability, through selective structural work in phases.
- ✓Fire and life safety.
- ×Operating performance, under one assigned service plan.
What The Federal Study Tested
- 1Nine operating concepts across three station layouts.
- 2Platform widening only as a two-track change combining Platforms 5 and 6.
- 3A far more radical integrated station, rejected for requiring column removal beneath Madison Square Garden, Moynihan, and the subway.
- 4The 12-track layout appears nowhere in the published concepts.
The reason is scope. The federal study set out to find capacity through targeted, minimal work that fits inside the rebuild already underway and can be built in stages while trains keep running. Removing nine tracks is not minimal. So a design that had cleared the engineering screen was left out, and the earlier operational verdict was allowed to stand unexamined.
The study did not apply that standard evenly. One of its own concepts also failed simulation, on arrival variability and yard congestion. The FRA kept its physical design and proposed retesting it with more through-running, different turnback arrangements, and alternative routing. The 12-track layout, whose failure was likewise a product of dwell assumptions and service allocation, was never given that second look.
The plan follows the track and platform layout of Figure 5-31 in the Doubling Trans-Hudson Capacity feasibility study, October 2024, redrawn from the published vector, recoloured, and stripped of the structural-column notation that figure was drawn to carry. It is a concept layout based on, not a faithful representation of, ReThinkNYC’s Regional Unified Network proposal of 2017. Configuration, screening results, and service assumptions from the same study, Alternative 2, Design Concept 2. Concept structure and scope language from the FRA Phase I report, June 2026. Both studies measure peak-direction trains per hour under the Hudson, but they impose different requirements on that number, so the 32 that failed one test is not equivalent to the 32 the other treats as a gain.
The order of these decisions decides the outcome
The FRA's study found that Penn Station can move up to 32 trains an hour under the Hudson River, up from 24 today, inside the existing footprint, through longer platforms, more stairs and escalators, and a reorganized track layout. The agency takes no side on the outcome: it says the same design supports a timetable where trains terminate and a gradual shift toward running them through.
That neutrality is what makes the finding hold. A federal agency with nothing to gain examined the station and found the second option available without new land. Design work runs into 2027, and a track layout fixed before the operating analysis is finished forecloses whatever that analysis finds. Concrete is permanent in a way that a timetable is not.
Anyone can read these documents. The FRA's Phase I report, the Trans-Hudson feasibility study, and the Regional Plan Association's capacity report are all public, and the sources are listed below. Amtrak and its partners hold public briefings on the rebuild, and the federal study invites comment.
One question is narrow enough to answer and specific enough to be worth asking. The federal study evaluated nine operating concepts across dozens of scenarios. Was the 12-track layout among them? If it was, publish its track plan, timetable, dwell assumptions, train pairings, and simulation results. If it was not, explain why a configuration that had already cleared the engineering screen was left out of a study commissioned to reconsider how Penn Station operates.