The Capacity Case That Was Never Modeled

Empirical, with normative framing.

The "three times design capacity" framing is rhetorically powerful but does not appear in subsequent quantitative analysis. It functions as a stage-setting claim that primes the reader to accept that the station is already at its limits.

The 600,000 passengers and 1,345 trains figures are widely cited and broadly defensible. The "three times design capacity" claim derives from comparisons of current ridership against the 1910 Pennsylvania Railroad design capacity, which itself was sized for a different operational paradigm — terminal turnbacks of long-distance trains, no commuter saturation. The comparison is technically true but analytically misleading: design capacity in 1910 reflected expected dwell times and train compositions that bear no relationship to modern commuter operations.

Tokyo Station handles comparable peak intensities through coordinated through-running across multiple operators using standardized rolling stock and unified dispatching.[2] Zurich Hauptbahnhof handles roughly comparable train volumes with a smaller platform footprint through Takt-based scheduling.[3] Neither station treats its 19th-century design as a binding constraint on 21st-century operations.

The "three times design capacity" framing is precisely backwards. Penn Station is not failing because it is overcrowded relative to its 1910 design. It is failing because its 1910 design assumed terminal operations that are operationally obsolete. The diagnosis is not "the station is overwhelmed" but "the station is being operated according to a paradigm that international peers abandoned decades ago."

Modeling assumption presented as derived requirement.

This is the load-bearing quantitative claim of the entire report. The 66-tph and 63-tph through-running figures are evaluated against this 90-tph anchor and found wanting. Without this anchor, the headline conclusion does not hold.

The 90-tph figure is presented as if it were a derived demand requirement, but it is in fact a tunnel-capacity ceiling. The report conflates the engineering capacity of the tunnels with the operational requirement to use that capacity. No demand analysis is presented showing that 90 peak-direction trains will actually be required, or that such a service plan would be operationally optimal once Gateway is complete. The 24+24=48 NJ-side and 21+21=42 East-side figures are tunnel maxima, not demand-driven targets. Real-world systems generally operate below their tunnel maxima for reliability reasons.

International peer systems generally do not size station throughput to tunnel maxima. The London Crossrail/Elizabeth Line tunnel was designed for 24 trains per hour but operates at 22 to 24 in peak with reliability margins.[4] The Paris RER A operates at 30 trains per hour in the central tunnel section, well above what the Penn tunnels would carry, but the central stations are sized to handle that throughput because they were built for through-running.[5] The relevant question is not "can Penn Station match tunnel maxima" but "what is the peak service requirement implied by demand projections under various network configurations."

The 90-tph anchor is an arbitrary target dressed as a derived requirement. A demand-driven analysis would start from projected ridership patterns under various service configurations, derive the peak service plan necessary to serve those patterns at acceptable load factors and reliability levels, and then evaluate whether through-running, terminal expansion, or hybrid configurations can deliver that service. The report skips this entire analytical step.

A demand-anchored target of 70 to 80 peak-direction trains per hour, operating with 95%+ reliability and tunnel-maximum utilization across the 6–10 AM peak period rather than just the 8–9 AM peak hour, would deliver greater total person-throughput than the 90-tph peak-hour anchor. Under this target, through-running configurations move from being "72% of tunnel capacity" to being "fully sufficient with margin."

Modeling assumption, presented as derived parameter.

These numbers are the second load-bearing element of the analysis. They drive the 66-tph and 63-tph headline figures.

No methodological provenance is given. The report does not cite a simulation, a peer review, an empirical observation, or an international benchmark. The parameters are simply asserted. This is the same pattern that the October 2024 WSP/FXC feasibility study followed in Appendix B, where dwell-time assumptions were similarly asserted without simulation backing.[6] The Amtrak FOIA responses from August 2025 and February 2026 confirm that no operational simulations underlie the 2024 study's dwell figures.[7][8] The April 2026 RPA report extends those figures with longer assumed dwells and provides no additional methodological foundation.

Achieved dwell times at peer through-running stations are substantially shorter:

• London Elizabeth Line at central core stations operates with station dwells of 45 to 60 seconds (45 seconds typical; 60 seconds at major interchange stations such as Paddington and Liverpool Street), with 24 trains per hour throughput.[4]
• Paris RER A at Châtelet-Les Halles operates with peak station dwells of approximately 60 to 105 seconds, with 90 seconds as an approximate peak-stress midpoint.[5]
• Munich S-Bahn Stammstrecke central stations operate with a scheduled core dwell of 30 seconds, supporting 30 trains per hour.[9]
• Zurich HB S-Bahn through-platforms operate with station dwells of approximately 55 to 60 seconds.[3]

RPA's assumed station dwell time of 12 minutes for existing-width through-running (excluding the separate 2.5-minute interlocking/safety separation that RPA adds on top) is approximately six to twelve times longer than achieved dwell performance at international peer central through-running stations.

The mechanism RPA identifies for the long dwell — that boarding passengers cannot descend until alighting passengers have completely cleared the platform — is a constraint imposed by Penn's narrow platforms and by current passenger management practices, not an immutable feature of through-running operations. International systems with comparable or higher passenger volumes do not face this constraint because their platforms are wider, their vertical circulation is differently configured, or their station design separates the two passenger flows.

The dwell-time framework conflates Penn-specific narrow-platform constraints with through-running as an operational concept. Through-running at international peer systems operates with dwell times that are a fraction of what RPA assumes. If Penn's platforms and vertical circulation are inadequate to support shorter dwells, that is an argument for investing in platform widening and vertical circulation enhancement, not an argument against through-running. The report uses Penn's current physical limitations as the parameter set for evaluating an alternative that is specifically designed to address those limitations.

Under a wider-platform configuration with modern vertical circulation, signaling, and rolling stock with high-capacity doors:

• A 3-minute revenue-to-revenue dwell (achievable at peer systems with comparable passenger volumes) plus 90-second interlocking clearance yields capacity of 13 to 14 trains per track per hour.
• Even allowing for Penn-specific complexity factors (Amtrak intercity train mix, longer commuter consists than RER trains), a 5-minute dwell plus 2-minute interlocking clearance yields 8 to 9 trains per track per hour, double the RPA figure.

Engineering constraint, conditional on current platform widths and current vertical circulation.

Internally coherent given the parameter set. The report acknowledges that Platform 10, at 33 feet wide, can support pre-boarding similar to subway operations.

Platform widening is the standard international solution to this exact constraint. London's Elizabeth Line constructed new platforms 30+ feet wide specifically to enable simultaneous boarding and alighting at high throughput.[4] The October 2024 feasibility study evaluated a wider-platform configuration (Concept 2) and found it passed track geometry, constructability, and fire-life safety screening.[6] The RPA report acknowledges this configuration but evaluates it under unfavorable assumptions about track count reduction.

The "narrow platform" constraint is a constraint of Penn's current configuration, not of through-running as an operational paradigm. The report's analytical move is to take the most binding current constraint, freeze it as a parameter, and use it to argue against an operational reform whose entire point is to relax that constraint. If platform widening is feasible (as the 2024 study found), then the question is not "can through-running work with current platforms" but "what is the optimal platform configuration to support through-running, and what does that cost relative to terminal expansion?"

Evaluate platform widening as a capital project on the same time horizon as Penn South. The 2024 feasibility study's Concept 2, which decks over alternating tracks to widen platforms to 30 feet while retaining 12 through-tracks, should be evaluated under modern dwell-time assumptions rather than the assumed 10-to-12-minute dwells. Under achievable peer-system dwells, 12 wide platforms supporting 8 to 10 trains per hour each yields 96 to 120 peak-direction trains per hour, exceeding even the RPA 90-tph anchor.

Modeling assumption with embedded probabilistic logic.

This argument is internally inconsistent with the report's own demand analysis. Two paragraphs earlier, the report acknowledges that "the most effective utilization of through-running would be for relatively short runs making stops within the metropolitan core and then turning back" and that station pairing would be demand-driven. Yet the connectivity argument then reverts to a random-pairing logic ("the odds that the branch you started on is scheduled to continue") that no operational planner would actually deploy.

No real-world through-running system uses random branch pairing. The Paris RER pairs branches based on demand patterns, not stochastically.[5] London Thameslink pairs Bedford-Brighton, St Albans-Sutton, and similar high-demand combinations.[10] The Munich S-Bahn pairs branches based on travel-pattern analysis.[9] The "100 branch-to-branch combinations" framing is a strawman that bears no resemblance to how through-running pairings are actually planned.

Standard international practice is to identify high-demand origin-destination pairs and pair branches accordingly, with through-ticketing and timed transfers handling lower-demand pairings. A Penn through-running plan would pair, for example, Hempstead-Secaucus, Babylon-Trenton (for express service), Port Washington-Newark, and similar high-demand combinations. The "Hempstead to Gladstone" example RPA uses to dismiss connectivity is exactly the kind of pairing that no operational plan would attempt.

A demand-weighted connectivity analysis that identifies the top 20-to-30 cross-Hudson O-D pairs by ridership, evaluates which of those pairs would benefit from one-seat rides under various pairing scenarios, and quantifies the connectivity benefit in passenger-minutes saved. This is the analysis the report would need to do in order to make a defensible connectivity claim either way.

Empirical/engineering claim with implicit normative framing.

The report compares yard land requirements unfavorably to expansion land requirements, but the comparison is incomplete. Terminal expansion at Block 780 displaces approximately 200 residents, multiple businesses, and a Gothic Revival church listed as eligible for the National Register, on a city block in central Manhattan.[11] Yard expansion at Sunnyside or in the Bronx involves industrial and railyard-adjacent parcels with very different displacement profiles. Comparing these on raw acreage alone is analytically misleading.

The acreage figures are reasonable order-of-magnitude estimates, but the report does not compare cost per acre (Manhattan core vs. Sunnyside vs. Secaucus), displacement impacts (residential vs. industrial), construction complexity (deep-cavern excavation under active tracks vs. yard construction on existing rail rights-of-way), or schedule risk.

Through-running implementations in international cities have generally relied on existing yard infrastructure plus modest peripheral expansion. Munich, Paris, and London did not require multi-acre new yard development to support through-running. The RPA framing assumes a maximalist yard requirement (full layover for all through-running trainsets) that does not match international practice.

The yard argument compares a maximalist through-running yard requirement against an idealized expansion footprint without accounting for the asymmetric costs of Manhattan core demolition versus peripheral yard expansion. Sunnyside Yards, in particular, is already a railyard; modest expansion or re-configuration is operationally and politically very different from clearing a Manhattan block. The comparison should be conducted on a cost, schedule, displacement, and risk basis, not on raw acreage.

Engineering constraint, addressable on a procurement timeline.

The constraint is real. Dual-mode rolling stock exists (Bombardier ALP-45DP, NJT Multilevel III dual-modes, various European examples) and is in active service.[12] Battery-electric multiple units with overhead pickup are emerging technology. Full power-system conversion is feasible but expensive.

International peer systems achieved through-running by either converging on a single power standard over decades (Munich S-Bahn, Paris RER), or deploying dual-mode equipment for specific corridors (various German and Swiss applications). Both paths have 10-to-30-year implementation timelines, comparable to the timeline for terminal expansion.

The report treats power-system incompatibility as a current constraint rather than as a procurement-cycle planning question. A 30-year capital plan that mandates hybrid rolling stock for all replacement procurements at NJT and LIRR, combined with selective electrification investments, would deliver fleet compatibility on a timeline that aligns with through-running implementation. This is a governance and procurement coordination problem, not a technical impossibility.

Operational recommendation.

This recommendation is significant. The report acknowledges that some through-running is operationally beneficial. But the report does not analyze a hybrid regime in which 30-50% of peak service is revenue-to-revenue through-running and the remainder is drop-and-go or load-and-go. Instead, the analysis evaluates only the binary "all through-running vs. all current operations" framings, then recommends marginal through-running expansion.

International peer systems generally operate hybrid regimes. London Thameslink runs revenue-to-revenue through-running on the central core but turns trains at peripheral terminals.[10] The Paris RER mixes through-running on the central trunk with terminal operations at peripheral stations. Munich and Zurich operate Takt-patterned schedules that combine through-running with timed connections.

The report's binary framing is the wrong question. The right question is what mix of operations maximizes capacity, connectivity, and reliability. A hybrid regime that uses revenue-to-revenue through-running for the highest-demand cross-Hudson O-D pairs, drop-and-go for lower-demand commuter flows, and selective terminal turnbacks where operationally optimal could deliver capacity exceeding either the all-current or the all-through baseline.