Penn Station's Missing Model

The claim RPA asks the public to accept

RPA's report is strongest when it inventories Penn Station's real constraints. Narrow platforms matter. So do reverse-peak service and divided railroad control. Those constraints belong in any serious operating model.

The report then makes a larger claim than its published evidence can support. It tests revenue-to-revenue through-running under the current 17-track arrangement and under a 10-track wider-platform concept. The federal program points to a hybrid configuration instead. RPA's two screens show a shortfall against its own 90-train benchmark; they do not establish the geometry Penn Station requires.1

RPA describes its revenue-to-revenue dwell calculations as an "independent assessment" that may differ from other organizations' assumptions.1 That disclosure is important. A screening assumption can identify a risk. Irreversible geometry requires more: observed passenger behavior, rebuilt-station conditions, and a train graph the public can inspect.

The missing operating record

The report does not provide the materials needed to reproduce its result. The missing record includes the train plan, platform and track assignments, reverse-peak timetable, yard and non-revenue moves, dwell distributions, delay and recovery inputs, and the outputs of any simulation or calculation workbook.

The related agency record does not fill the gap. The October 2024 railroad feasibility study said that "recent detailed rail operations simulations" informed its practical capacity assumptions.2 In February 2026, Amtrak reported that its searches found no simulations or modeling data in its possession and stated that no simulations were performed for the work at issue.34 NJ TRANSIT separately certified that it held no responsive RTC model for the requested through-running analysis, although federal and railroad records show personnel discussing Penn-specific RTC work for the Service Optimization Study.5 The MTA withheld analytical models, technical assumptions, drafts, and analyses under the deliberative-process exemption.6

A consultant may have run a model. The public has not received the model, workbook, or train plan needed to reconcile the studies with the records responses.

The official planning record recognizes the operating logic of through-running. The MTA/WSP 2021 white paper and the May 2021 Community Advisory Committee briefing identify fewer turning conflicts and shorter platform occupation as potential benefits. They also identify the hard constraints: track loss from platform widening, recovery needs, and institutional control.78 The unresolved question is quantitative: how much useful capacity remains when a reconstruction model includes those constraints?

The arithmetic behind the conclusion

RPA organizes Penn Station operations around five movement types. That vocabulary matters because its capacity hierarchy follows from the dwell, clearance, and platform-use values RPA assigns to each movement.

RPA recognizes the strong AM/PM directional imbalance and reverses its boarding and alighting assumptions between the peaks. The missing step is the passenger-flow analysis that converts that asymmetry into fixed dwell and clearance values for a reconstructed station.

The benchmark, reverse peak, and mixed service

RPA adds 48 peak-direction trains from the west to 42 from the east and treats 90 trains per hour as the station-capacity target.1 That number has a planning rationale: Gateway adds trans-Hudson capacity, and Penn Station Access adds eastern demand. The public report still does not show why summed tunnel capacity must become the scheduled station throughput in one peak hour.

In practical terms, tunnel room is not the same as a workable Penn Station schedule. Before a benchmark fixes the station design, agencies should show the train plan and reliability margin that make the benchmark necessary.

The strongest official objection to broad through-running arises from the interaction among reverse-peak service, mixed intercity and commuter operations, platform allocation, and remaining peak-direction capacity. The 2024 feasibility study found that one full-reconstruction concept could pass track-geometry and fire-life-safety screens, yet fail once agencies preserve those service requirements.2

The untested hybrid

RPA tests the present platform arrangement and a wider-platform concept that removes nine tracks. Neither test answers the federal design question now before Amtrak and USDOT.

The federal base scope treats passenger circulation as a design variable. It includes platform renovation and a consolidated public concourse, with additional vertical circulation.10 The 2022 Master Plan likewise evaluated alternatives with materially larger public circulation areas. Those figures do not decide train capacity. They do show why today's passenger circulation cannot be frozen as the final operating condition.

The 2010 Moynihan technical memorandum also found that platform-clearance performance varied by platform rather than failing uniformly across the station.13 It is a historical baseline, not a current survey. Its continuing value is methodological: the reconstruction requires a new platform-by-platform model, not a station-wide average carried forward as a permanent limit.

The connectivity substitution

RPA counts ten branches on each side of Penn Station, producing 100 possible branch-to-branch combinations. From that matrix, it argues that any one train is unlikely to serve a rider's desired pair. It illustrates the point with a Hempstead-to-Gladstone trip.1

This substitutes combinatorics for origin-destination analysis. Railroads build trunks around demand and line capacity, not random branch pairings. A credible plan may choose a few strong cross-core markets and keep timed transfers elsewhere. Counting every theoretical pairing misses that design choice.

Yards and implementation require a common comparison

RPA is right that through-running changes yard requirements. A short cross-core service may need layover or turnback capacity near its outer terminal. An integrated regional network also needs storage and recovery space.

The comparison is uneven. RPA presents peripheral yard acquisition as a visible burden of through-running while treating terminal-expansion storage as a background condition. Gateway planning has long included major storage facilities in New Jersey. The Penn South operating concept also assumed nearby storage.1415

Fleet and power constraints are substantial. Labor and governance may be harder still. They should be modeled as implementation choices, then compared with the delivery risks of a permanent terminal expansion.

The missing bridge between RPA’s earlier and current analyses

RPA's earlier plans associated through-running with different infrastructure, service, and fleet assumptions. Its April 2026 report reaches a different result for two within-footprint cases. Those positions may be reconcilable, but RPA has not published the technical bridge between them.

That history matters only because RPA now asks the public to accept a different capacity result. The report should show whether the change comes from infrastructure assumptions, operating assumptions, or the narrow within-footprint cases it selected.

Advocacy context raises the disclosure standard

RPA leads the Build Gateway Now coalition. FOIA productions show that RPA administered Amtrak-funded coalition management and research.20 Funding records identify $1.825 million in Amtrak funding to RPA from FY2015 through FY2024, including restricted Gateway/Amtrak funding and Build Gateway Now payments.22 The 2024-25 Amtrak campaign proposal included through-running research and a proposed report in the funded work plan. It said that work would present full through-running as disruptive and Penn South as the practical path to a version of through-running.20

This funding history does not show that Amtrak dictated RPA's conclusion, nor does it establish that the conclusion is wrong. It does show that the report emerged from an advocacy relationship in which Amtrak funded RPA's Gateway coalition work and included through-running research in the work program. That relationship raises the disclosure standard for the report's decisive operating assumptions.

The operating record required before the design is fixed

RPA documents real constraints at Penn Station. Agencies still need a public operating model before they design the federal reconstruction around RPA's capacity conclusion.

The agencies now face a different problem from the one RPA tested. They are advancing reconstruction with at least limited through-running. That requires a disclosed hybrid operating model before agencies set permanent geometry. If the Service Optimization Study cannot publish enough information to test the operating basis, agencies should not use it as public justification for the station's final geometry.