Amtrak should get the New Haven Line

Today, Amtrak is the owner or contracted operator of the entire 450-mile Northeast Corridor, save the roughly 50 miles from New Rochelle, NY, to New Haven, CT, known to most as the New Haven Line. The states of New York and Connecticut happened to purchase their respective portions of that line before Amtrak was authorized to acquire rail right-of-way in the mid-1970s. Today, the Connecticut Department of Transportation (CTDOT) contracts with Metro-North Railroad to dispatch and maintain the New Haven Line.

There would be an argument for continuing this arrangement if the New Haven Line were the tightest-run portion of the Northeast Corridor, but the evidence decisively paints the opposite picture. It’s time to turn over the last state-controlled Northeast Corridor segments to America’s national rail operator. Amtrak is certainly plagued by internal dysfunction causing inefficient operations and high construction costs. Still, it is very hard to see even them failing to improve upon CTDOT and Metro-North, whose persistent apathy to commonsense optimizations makes the famously stodgy Amtrak look downright modern.

CTDOT and Metro-North maintain an official view that the New Haven Line is too curvy for trains to travel meaningfully faster than they do today. This claim fails to account for schedules steadily getting slower as signaling systems and trains have gotten better and the overhead wire replacement project has concluded. The travel times in the latest Metro-North New Haven Line schedule, whose speedups have been widely advertised, still do not match those of 2015. For example, New Haven-Grand Central train 1565 was timetabled at 1:50 in 2015 and is currently scheduled to take 1:59 calling at the exact same stations.

Today, off-peak Grand Central-New Haven zone express trains achieve similar average speeds to those of peak-hour North Jersey Coast Line zone express runs linking New York Penn Station with the Shore. There should be no contest; M8 multiple units are more powerful than the locomotive-hauled sets that ply the Coast Line. Moreover, New Haven-Grand Central trains enjoy a longer average interstation distance and encounter less congestion. The two-track line linking New York Penn Station to New Jersey sees 450 trains per day, nearly the volume that Grand Central handles with four approach tracks and more than double the highest traffic on the quad-tracked New Haven Line.

Today, the maximum authorized speeds on much of the New Haven Line are unreasonably low. While trains take the famous Jenkins Curve on CTDOT trackage west of Bridgeport station at just 30 mi/h, Amtrak allows 40 mi/h on an equally tight curve in New London. The speeds listed for these two areas on Open Railway Map are easily verified using train ride footage, a stopwatch, and Google Earth. While Amtrak is reportedly petitioning for a speed boost on the Elizabeth S-curve, Connecticut has been silent about similar changes on its territory.

One of the main reasons for the New Haven Line’s slow speeds is that Metro-North lags Amtrak, let alone many of its overseas counterparts, in automating basic tasks like tie replacement. On most of the Northeast Corridor, much of the maintenance takes place during the night and on weekends. Astute viewers will note that the video I linked above was recorded on a Thursday, and Amtrak certainly has room to improve productivity. Still, it generally manages to restrict track outages to a much lesser duration than the rolling 24/7 closures that define travel on the New Haven Line.

All this slowdown is not facilitating high maintenance standards. On the contrary, despite the Metro-North network encompassing the New Haven Line receiving plenty of money over decades to clear out accumulated infrastructure work, about 1/3 of the ties and significant portions of other components remain in poor condition. Wooden ties can be counted on to last 10 — 15 years, which calls for a replacement of 5 — 10% per year. Rail and concrete ties can legitimately last 30 years or more, equating to 3 — 4% replaced per year. MTA documents clearly show that Metro-North has been lagging Amtrak crews for years.

There are good first-principles reasons to consolidate the Northeast Corridor’s infrastructure under Amtrak. While renouncing direct local control of an asset may seem counterintuitive, in transportation, scale matters. Amtrak already owns or operates an order of magnitude more electrified rail route than does CTDOT and has experience with mechanized track work.

It is also hardly the case that New York and Connecticut would completely lose their leverage; the Senate confirms the Amtrak board and president. Moreover, the change of guard could be traded for concessions such as Hartford Line capital improvements or a reactivation of Cedar Hill yard to simplify Metro-North yard moves and free up valuable New Haven land.

New Jersey shouldn't rest on transit service

While Governor Murphy delivered good news to NJ Transit riders this week, announcing that fares will not rise at least until mid-2023, attention still needs to be paid to transit service west of the Hudson River. To discuss fare policy without addressing finances is a dead end. Finances hold particular importance in New Jersey, where capital money meant for infrastructure has been used to run buses and trains—a problematic practice that slows investment in infrastructure. The amount transferred this year has increased, and revenue is below budget.

Notable progress has been made in stabilizing NJ Transit service and finances. Transit appears to be somewhat higher priority to policymakers than during the Christie Administration. The Main-Bergen Lines now see at least hourly service all week, and redesigns of Newark-area and Camden-area bus routes around many of the principles I lay out in this post is underway. To rest on laurels now would be a lamentable end given the potential out there.

NJ Transit has significant untapped potential to improve service delivery and cost effectiveness. To start, it should spread the service it runs more evenly during the day than it historically has. Peaky, rush hour-focused service like what NJ Transit has been running recently is often quite costly. Publicly available data shows that NJ Transit delivered 14% less total suburban rail service in 2019 than 2009, primarily due to off-peak service cuts, while spending 2% more in inflation-adjusted terms to run it. Currently, many engineers work upwards of 13 hours per day, drawing associated overtime pay, while only operating one train into New York or Hoboken and another out.

In contrast, train crews on systems with frequent service all day spend much more of their time moving paying passengers. While many overseas counterparts to New Jersey’s suburban rail lines see at least three trains per hour seven days a week, around here, only the Northeast Corridor’s frequency approaches that. Many rail stations still get not a single weekend train, and the Meadowlands Line only gets served on game days despite American Dream having opened months ago. That’s a shame; the region is densely populated and has a wealth of high-quality regional rail infrastructure. That the Toronto suburbs—which resemble American sprawl much more than Manhattan—can fill a single bus route with 40,000 daily trips suggests strongly that a baseline of frequent, consistent transit service can spur a large ridership increase in relatively densely populated New Jersey.

Next, the agency should trim bus routes that parallel rail lines. Carrying a passenger one mile is much more expensive on a bus than a train. Accordingly, buses should generally run where trains don’t; the 712 is a good example of what NJ Transit buses should aspire to. To be fair, rail lines with long interstation distances, such as the Northeast Corridor, or nearly parallel job corridors that rail stations just miss—as is the case with the Coast Line and NJ-35, should continue to get supplementary bus service. However, many current bus routes run right on top of rail lines no more frequently than the trains. That kind of service is of no use to anyone but the absolute lowest-bar rider. Concentrating buses onto fewer, frequent, nonduplicative routes is one of the main reasons why Canadian metro areas boast significantly higher transit mode shares than North Jersey.

As contactless ticketing continues to roll out, NJ Transit should work with labor to reduce train crew sizes. Although one-person train operation now seen on overseas rail and subway systems will require many infrastructure investments, some reduction in train crew sizes is already possible. The new ticketing system supports proof-of-payment ticketing, in no small part because it’s already in place on the Hudson-Bergen Light Rail, Newark Light Rail, and RIVER Line. Expansion of proof-of-payment, wherein all passengers must validate their own tickets before boarding, to regional rail would allow most fare collectors to be moved up to conductor and engineer roles, particularly on trains that serve no or few low-platform stations. Many suburban rail riders already use an electronic ticket and would see little change. Expansion of frequent baseline service to the whole NJ Transit suburban rail network will require a considerable increase in the engineer and conductor pool, so there is no reason to fire anyone.

As NJ Transit adds service, certain trains should not come back. For example, as painful as this is to say as a former rider, the direct Hoboken-Bay Head trains should stay dead. They carried few riders while gumming up much fuller New York trains. Instead, NJ Transit should add Main-Bergen service and Bay Head shuttles so that transfers add less time.

New York area railroads must also increase their maintenance productivity and reverse a long trend of sacrificing midday frequency and speed in its name. Overseas systems of similar age that have suffered similar ups and downs perform much of their maintenance at night. While the PATH downtown tubes see only three to four trains per hour off-peak, riders on comparable subway and regional rail systems that dispense with midday work can expect consistent service levels of ten trains per hour or more.

One policy that I hope does not gain traction around here is free transit. Even in the depths of 2020-2021, fare revenue comprised around $370 million of the $2.6 billion NJ Transit operating budget. To date, no recurring sources to replace even that much income have been identified, much less enough to offset that from today’s rebounded ridership. In surveys such as this one by TransitCenter, low-income riders clearly express that service comes before fares in determining whether they will ride. To a woefully underserved metro area like New Jersey, a cut to service of even 10% is anathema. Moreover, fare revenue tends to respond much less to political winds than state or federal subsidies. While the farebox recovery ratio should not be the sole metric by which we judge transit service, there is nearly zero overlap between the transit systems that lead the world in providing quality service to a broad cross-section of society and the ones that don’t charge a fare.

While NJ Transit appears to have moved beyond the crisis of the late 2010s, transit service in the region remains in a problematic state. Both agency leaders and the officials to whom they answer need to use the tools at their disposal to remedy that.

Blessing in disguise

In a turn of events, an interview between Jim Cameron and Connecticut Department of Transportation (CTDOT) Commissioner Giulietti suggests that CTDOT may be forced to rethink its highly problematic procurement of new railcars for its diesel lines. Cameron writes, “The [CTDOT] request to the tiny rail car industry for new cars proposals brought a dismal response.” Upon reading the request for proposals (RFP) that Giulietti approved, the reason why is clear.

Despite the claim that the railcar industry is diminutive, the same builders that gave CTDOT’s procurement short shrift build plenty of rail cars yearly, including for the New York City Subway. The only reason that the railcar industry appeared small to CTDOT is because of unnecessary constraints specified in its RFP. Page ten contains what is likely the biggest offender: specification of the old Federal Railroad Administration rule that a car body must not deform when an 800,000-pound force is applied to its end. Basically no other country imposes this criterion on its passenger rail cars, and barely any manufacturers build to it.

Comically, the RFP does not even make reference to the correct subheading in the regulation! Whereas it claims “structural requirements” are found in 49 CFR Part 238, Subpart D, the actual regulation has been rearranged.

Worse, by making no mention of the FRA’s alternative compliance avenue—basically the Euronorm—finalized about three years ago, the RFP constrains carmakers to the old rule. Other US agencies are thankfully catching on to the benefits of the new rules; Metrolink is welcoming the first of its Eurospec diesel multiple units (DMUs). Despite a claim by a retired Metro-North manager, there is no reason why European trains cannot deliver level boarding at Northeast Corridor stations; plenty of them have floor heights of 48” above the rail, which is level or nearly so with any American high platform.

If the new cars were on their way, riders would have little to celebrate. The CTDOT vision for the branch lines is outdated by a century. The cars on order would be unpowered and require a diesel locomotive to haul them, which is about the worst-performing setup out there. Concentrating all the motive power into one locomotive increases maintenance costs and failure rates. Reliable coaches matter little when the machines that haul them go haywire routinely. While diesel locomotives struggle to travel 20,000 miles between failures, the unusually complex and heavy M8 electric multiple units (EMUs), which were built to the old FRA rules, routinely exceed tenfold that. Worst of all, diesels, especially diesel locomotives, have never and likely never will match the acceleration and braking of electric multiple units, which would consign inland Connecticut riders to large trip time penalties—likely a minute per stop.

Today, Metro-North ridership into Grand Central hovers around half of its pre-pandemic level, and they don’t use their M8 EMUs efficiently at all—sometimes having them lay over for hours in Midtown Manhattan. Were it to rationalize consist lengths and turn its trains around in reasonable windows, it would likely have enough cars to furnish decent service to the inland branches. The price that CTDOT was prepared to pay for one unpowered car would be enough to electrify about a mile of route assuming it could stick to the cost Amtrak incurred on its New Haven-Boston overhead line project. The sum it is prepared to commit for about seventy new trailers would wire most of the combined mileage of the Danbury Branch, Waterbury Branch, and Springfield Line.

The silver lining is that change may be afoot, by default, at the Connecticut Department of Transportation. Hundreds of senior staff have retired or will in the coming months. It is imperative that the Governor look for more enlightened transportation experts as he sets up for a likely second term.

Four Officials Who Cannot Build Rail

This afternoon, I livetweeted an Eno Center webinar that was ostensibly about how Amtrak and CTDOT were going to realize the CONNECT NEC 2035 program with major gains in productivity during track outages.

Spoiler alert: not one specific was given about how they were going to do this. The webinar largely consisted of self-congratulation. The four presenters, Richard Andreski, Nicole Bucich, Dennis Newman, and Mitch Warren, all of whom are longstanding members of Northeast Corridor top brass, claimed they had worked hard to reconcile the needs of multiple stakeholders and arrive at a workable plan to cut travel times on the existing Northeast Corridor. This claim is belied by the price tag of $117 billion—about threefold the cost of the entire Milan-Naples high speed rail spine to deliver a fraction of the end-to-end trip time cut. Moreover, a remark from Bucich that went something like, “We try not to use the word prioritization because we have so many stakeholders...we don't have answers yet," suggests that little actual critical thinking has gone into this plan at all.

As I previously tabulated, the Northeast Corridor has for nearly twenty years received well over what its management claims its steady-state cost to be, let alone what it should be. I would understand a claim that there is a $7 billion state-of-good-repair backlog. There is no reason why it should have grown from $9 billion in 2010 to multiple $10 billions today.

A five-minute Google search will turn up reports showing a large resemblance between the Northeast Corridor and many other routes such as the West Coast Main Line or East Coast Main Line, not to mention Japanese trunks. They show that every factor which the Northeast Corridor’s management claims is unique to its rail line (presence of more than one operator, mixed traffic, aging infrastructure, funding swings) is anything but. The Northeast Corridor is virtually alone among these in enjoying little to no improvement to its service offering in recent decades.

It is only fair to expect that an official charged with changing the course of a singularly stagnant piece of infrastructure would not claim there is nowhere to learn from outside a country singularly poor at running trains. Amtrak could realize most of the 56 minutes of Washington-Boston travel time gain that it says will cost $117 billion virtually overnight by slashing scheduled dwells. Sixteen-car Tokaido Shinkansen trainsets typically reverse at Tokyo Station in ten minutes or less, during which time they are cleaned.

The culture in my field has its issues, but I am certain that anyone who said something like “I have nothing to learn from any professional outside America,” would meet instant and widespread scorn. This is an area where Americans made and make valuable contributions. I personally know of many people at US transit agencies who know what they are doing and challenge themselves to get better. The mentality of those folks bears no resemblance to that on show at this webinar. No one who says, I don’t use the word “prioritization” belongs in charge of a $117 billion program.

Northeast US Rail Infrastructure Spending and Value

In the Northeastern United States, much talk of rail infrastructure concerns rehabilitation work to atone for the deferred maintenance of the mid-20th century. In Connecticut, where I am finishing my Ph.D. in an unrelated field, riders have had to put up with decades of slow speeds and tracks out of service around the clock, supposedly in service of plant upkeep. Yet the conditions do not seem to be improving. Travel times on the New Haven Line have slowed considerably since I moved up here in 2015, let alone 2000. Moreover, the investments proposed in both the Time For CT and CONNECT NEC 2035 mostly consist of maintenance with some expansion projects. Most media coverage leads uneducated observers to believe that the state really does lack the money to keep up its railroads. However, examination of data on similar work reveals that Connecticut and the greater New York area as a whole already forward comparatively generous sums for rail network upkeep.

A few years ago, the Union International des Chemins de Fer (UIC), the international standard setter of railroads, published a report on maintenance and renewal expenditures on European rail lines. In 2015, the eight countries from which data was collected spent 15.3 billion euros on 55,175 route miles (circa 88,000 track miles). That is circa $250,000 per track mile in 2020 dollars, which is a similar figure to that cited in an earlier publicly available Amtrak OIG report. It is reasonable to take this figure as a good estimate of the steady-state cost of a unit of railroad. The European countries evaluated are not known for low costs of living, and the average usage of the railroads, just under 40 train-miles per track mile per day, mirrors that of the Northeast Corridor. The Amtrak OIG report also discusses Amtrak’s own estimate of the cost to keep the Northeast Corridor at a steady state: $316,000 per track mile. Generally speaking, the money European railroads spend on their plant buys them work practices that confine maintenance to the wee hours when little to no demand exists. When maintenance requires disruption outside those windows, the schedule changes are generally communicated ahead of time in considerably greater detail than the vague announcements that are accepted as normal here.

National Transportation Database (NTD) disclosures show that, since 2015, NJ Transit, Metro-North (under which the New Haven Line is bundled), and the Long Island Rail Road have each spent well over what Europe spent on their infrastructure. Note that all three of those railroads spend both operating and capital money on their physical plant, so looking through two categories of NTD data is required. While NJ Transit has somewhat legitimate claim to having lean funding during the Christie years, this pattern generally holds since 1997, which is the earliest data present.

Just about every category of railroad construction work takes more disruption in Connecticut than elsewhere. In Paris, along the four-track approach to the Gare d’Austerlitz, the overhead wire and the poles that suspend it over 110 miles of track are being replaced during overnight outages over six years. The total cost is projected around $350 million. The vehicles being used to run the wire are expected to replace nearly a mile of it per nighttime window. Along the New Haven Line, replacement of 220 track miles of wire and selected poles took thirty years of 24/7 track outages and nearly $1 billion. Segments of track between Bridgeport and the Housatonic River were out of service for months to years just for wire to be strung up on the poles that were ready. Moreover, the scope of work came nowhere close to replacement of every pole.

Work Category	Metro-North experience	Comparable foreign experience
Overall capital expenditure	$682,000 per track mile (2015)	$250,000 per track mile (2015)
Track replacement	Daytime outages, rail and ties replaced separately. 5.3/440 rail miles on the New Haven Line are to be replaced in 2021	Both rails and ties replaced at night 500 m/hr
Catenary replacement	Daytime outages 30 years, $1 billion for 220 track miles	Night work, 6 years $300 million for 110 track miles
Ballast undercutting	Few miles per year, several weeks of daytime shutdowns	Overnight work, 100 m/night or more
Ballast tamping/regulation	Few miles per year, daytime shutdowns	Overnight work, 1 mile per night

Ballast undercutting and track replacement generally require the track to settle under the weight of passing trains, necessitating temporary speed restrictions after the maintenance forces have finished their job, no matter if done with manual labor or with machines. However, a day or two of slow speeds is superior to having the track completely out of service, as is common practice on the Connecticut-owned portion of the Northeast Corridor.

My photograph, 20 July 2020 13:13 near East Norwalk. Claw cars like this are a common sight on the New Haven Line, including during peak times on weekdays. Despite these extensive disruptions, crews have been replacing just over 1% of the Metro-North system’s ties per year. Note how removal of any one tie temporarily blocks an adjacent track. Most modern track renewal trains do not impose this constraint and generally perform tasks much faster.

Replacement of life-expired undergrade bridges seems to be the most disruptive activity taking place in Connecticut lately. Yet replacement of a pair of bridges in Port Chester started one-and-a-half years ago and is not due to wrap up until next spring. Just across the Sound, similar projects require weekends. Yes, in Port Chester, service has been maintained while elsewhere it is often shut down completely. However, traffic at Port Chester has been at times restricted to two of four tracks, and the bridges have been placed one track at a time. Moreover, just miles away in Stamford, the six-track Atlantic Avenue bridge was replaced one half at a time over a nine-day service reduction in 2019. Given that, it is hard to see how avoidance of shutdown maps to a two-year construction time in Westchester County. The costs and timelines for both the Stamford and Port Chester bridge replacements also far outstrip those for a similar undertaking in South London.

There is more than enough to conclude that Northeastern passenger railroads are shorting their riders, particularly in Connecticut. Saying we need $120 billion—a sum that should suffice for a new Boston-Washington high-speed line twice to thrice over—to save less than half the time of a dedicated route would is effectively saying, I cannot deliver reasonable value for money. The institutions need to absorb practices from elsewhere pronto, and the government needs to help them do it. Employers that (are required to) take years to hire staff and severely backload benefits are not going to be attracting the caliber of staff needed. The insularity pervading the consultant-written Time For CT report should dispossess anyone of the notion that consultants can drive needed change. Their outside perspective matters little when they are alums of the same railroads they are advising and when more-functional overseas rail institutions don’t use them. If someone like me with no money or networking skills can develop connections to those familiar with overseas practices, surely American railroaders can figure out knowledge transfer from their counterparts at a fraction of AECOM’s going rate. Above all, it’s time to stop portraying Northeast US passenger rail as starved when it is anything but that.

The "learning curve" argument

A Streetsblog writeup of the Eno Center’s inquiry into American transit project costs repeated a tired and sloppy claim: that we have to build more transit for costs to come down. For how often this line, that the United States must traverse a learning curve to build as much subway per dollar as the rest of the first world, gets repeated, the evidence that exists stacks up squarely against it.

If a learning curve were the problem, we would expect costs to be coming down in cities that have done the most tunneling. In the US, the two cities building the most underground rail recently are New York and Los Angeles. Yet, in both cities, the per-mile costs of projects have tended steeply upward. New York went from paying $500 million/mile in today’s dollars for the Archer Avenue subway, completed in 1988, to $2.4 billion/mile to extend the 7, completed in 2015. The final cost of the Second Avenue Subway’s first phase, completed at the turn of 2017, totals $2.7 billion/mile. The estimate for the long-awaited line’s second phase is $3.9 billion/mile—for a line whose tunnels are already partially built, and the estimate for the Amtrak Gateway Project—only 1/4 of which would be in a tunnel—totals some $3 billion/mile.

Similarly, in Los Angeles, the unit costs have trended upward in the last 30 years. The initial Red and Purple Line system built in the 1980s and 1990s cost an average of some $400 million/mile in today’s dollars. Since then, costs for each successive phase of the Wilshire Boulevard subway have increased even as the line has progressed away from the city center along a wide road that is not particularly densely developed. The innermost phase is to cost $700 million/mile—an already high figure, and the outermost phase is now projected at $1.4 billion/mile.

The same companies that manage to dig at best a mile of subway under US cities end up tunneling much farther overseas. The obstacle is not that no one knows how to build cheap subways, it’s that they cannot do so in the current US political environment. Blue America already spends vast sums on transit infrastructure. The investments on which the effectiveness of these expenditures hinges are as much political as financial. They are investments into planners and construction supervisors within public agencies who can function mostly without consultant involvement. They are investments into the willpower to defend well-planned, logical projects from inevitable opposition, instead of skipping population and job centers in the name of inoffensiveness. They are likely investments into private sector job quality, re-unionization, and social programs so that the impetus for rail projects to create jobs can be subjugated to the need to move people cost-effectively.

Whether advocates realize it or not, the learning curve fallacy is the perfect cover story for those who don’t care about costs. They will always be able to claim “We’ll do better next time” if not pressed to live within reasonable means from the earliest stages of project development. Mankind has been building railroad tunnels for well over a century. The idea that the United States needs to learn from scratch how to build cost-effectively by building dozens of miles of overbudget tunnels defies common sense. That approach has also not drummed up the willpower to spend what we are being told is needed for the new Hudson River and Baltimore tunnels. The responsible thing to do is to retire this claim.

Be Glad ARC Was Cancelled

Recently, the cancelled Access to the Region’s Core (ARC) project came up in two different contexts. A few weeks ago, New Jersey freshman Rep. Tom Malinowski debated his challenger, State Senator Tom Kean, Jr. Both candidates were asked about the Gateway Project. In response, Kean cited his role in obtaining off-peak one-seat service to Manhattan from the Raritan Valley Line, and Malinowski fell back on the tired trick of dinging former governor Chris Christie over the cancellation of ARC. Neither candidate offered any acknowledgement of the greater New York construction cost bloat. Subsequently, ARC arose on Twitter, with Matt Friedman of Politico posing a perfectly logical question of why many transit advocates celebrate or are at least nonplussed over its demise. I will attempt an answer.

Although ARC would have added some trans-Hudson capacity, and improved somewhat on today’s situation, it would have done so in a highly suboptimal way. By the time ARC was funded, the project had evolved to terminate in a new cavern a block away from and deeper than the existing New York Penn Station. Finding out why the much better version connecting to Grand Central was scrapped—to avoid pissing off a small but apparently untouchable group of railroad managers—doesn’t take much reading between the lines. While later conversion of the deep terminal into a through station would still be possible, any transfer between the new and current stations would have entailed a long walk, making New Jersey-Queens or New Jersey-Connecticut travel tougher rather than easier. Subway transfers would also probably take longer. Conversely, whoever decided to tie the Gateway Project into the existing stub-end tracks (1 — 4) of Penn Station made the correct call.

There are some situations where a deep cavern station is unavoidable. New York Penn Station is not one of them. It has a set of terminal tracks, numbers 1 — 4, that only connect to New Jersey. Today, they do not handle the train volume they are capable of handling because of the station design. Today, any train leaving the terminal tracks westbound not only temporarily blocks eastbound trains into the terminal, it blocks all eastbound trains from New Jersey period. For this reason, increasing the use of the terminal tracks past a certain point requires a decrease in the throughput on the through tracks. In fact, the new tunnel proposed under Gateway, linking right to them, is exactly the fix that they need. Dedicating the terminal tracks to the new tunnels would keep the trains serving them out of the way of traffic on the through tracks. Note that while there would be fewer conflicts caused by reversals, they would not disappear entirely. For that reason, breaking the platforms out to the east to Grand Central is still necessary to realize the full potential of the new trans-Hudson tracks. In any case, the Gateway Project would provide some crossovers that could be used for interlining during maintenance and breakdowns, as shown in the below graphics from the Hudson Tunnel Project website.

The four-track terminal at Haussmann-St. Lazare on the Paris RER E handles up to 16 trains per hour. The combination of conversion of the terminal to a through station and installation of a modern signalling system should raise that to 28 trains per hour. If dedicated to the new Gateway tunnel with no other modifications, it is reasonable to expect New York Penn tracks 1 — 4 to handle somewhere around 20 trains per hour; today they handle nowhere near that. Extension of platforms 1 and 2 (serving tracks 1 — 4) to the West End Concourse and concomitant addition of new stairs to each would speed passenger circulation, allowing more and longer trains. Completion of the Central Concourse would disconnect the midpoint stairwells from the upper level and link them to an extended, much larger lower level hallway, which would almost certainly help downstream flow. Dedicating one or two through tracks to terminating Gateway trains would also add some more slots. On New York’s recent rail transit projects, a theme quickly emerges that new stations are a bigger contributor to the budget and schedule bloat than tunnels. For that reason alone, any sane analysis would have elevated a solution that ties into an existing station over one requiring a new cavern.

In the long term, connecting tracks 1 — 4 eastward to Grand Central would dramatically increase their capacity. Such a connection would eliminate the delays arising from reversals. Trains reversing at compact outlying terminals have fewer tracks to cross and thus can process faster than those reversing at city center terminals such as Penn Station or Grand Central. Furthermore, since alighting passengers would split roughly evenly between Penn Station and Grand Central, it would roughly cut in half the time needed to discharge passengers at either stop. The vertical access to each of platforms 1 and 2 can each currently move some 700 passengers per minute. It is not hard to see a capacity of 1,000 per minute if the platforms are elongated to reach the West End Concourse and the eastern stairwells widened and straightened. That should allow each of tracks 1 — 4 to handle a train discharging 800 — 1,000 people (about half a full load) and loading onward riders over 3 minutes. Adding 2 minutes between each train gives a total slot time of 5 minutes, which would permit 24 trains per hour over two tracks each way. Even this short window should be plenty to execute a crew change. For example, I saw one happen on a Tokyo area regional train during a normal enroute station stop lasting a minute, if that. Had the front cab not had a clear partition, I would have had no idea of the event as it added no delay whatsoever. My understanding is that SEPTA changes its crews in central Philadelphia similarly quickly.

When it comes to high-functioning regional rail, I am increasingly convinced that the main barrier is a small group of kings of the hill. Unlike New York, other cities around the world have a long record of executing new infrastructure in concert with operational changes that necessarily bridge institutional divides. A day or two before Thanksgiving in 2017, on board a New Haven-Grand Central Metro-North train, I talked to a senior Metro-North engineer who was off duty at the time. At one point, the discussion got to the football trains. He explained to me that he knew of scores of other engineers who would have been happy to run trains through Penn Station to New Jersey and that the Amtrak dispatcher that killed the football trains. The Japanese run private trains through into Tokyo on public lines; Paris got the SNCF and RATP to cooperate on the Paris RER; Italy got the FS and FNM to work together on the Milan Passante. America can dethrone the hilltop emperors within the MTA, NJ Transit and Amtrak who refuse to talk to each other.

Nor is the problem money, at least not in greater New York. This region found billions of non-federal money to shovel into East Side Access and new subways under the West Side and Second Avenue. It finds the money to pay for subway and regional rail conductors that technology made obsolete decades ago. Even as problematic as the Trump Administration has been on transit, some major projects in blue states got funding out of them. Yes, Christie should have contributed to the Gateway Project, which correctly plans to tie into New York Penn Station and actually sets up nicely for a connection to Grand Central, but ARC deserved to die. It is time to retire the hand wringing over that botched effort.

What to do about Gateway?

President-Elect Biden will have a Democratic House and probably a Republican Senate to work with come January. For all the buzz about Amtrak Joe moving into the White House, he is likely to want to accommodate the Senate to a large extent. That means that boosters of infrastructure, including the Amtrak Gateway Project, should plan around continued scarcity of federal funding. The Gateway Project, which would add a 2.7 mile new tunnel under the Hudson linking New Jersey to New York Penn Station, rehabilitate the tunnel that already exists, and quad-track a further 7.5 miles of surface railroad, is a worthy project. It is not worth the current $30 billion asking price. If it were, someone would already be building it. Here, I lay out my thoughts on what needs to happen to get the work done.

Buy the right trains for through running

Today, many trains turn back where they came at Penn Station, wasting capacity. Even famously stodgy Amtrak finds that conversion of those reversals to continuations without handling reverse-peak passengers would enable each platform track to process a train on average every 12 minutes. The Amtrak analysis is pessimistic that trains can discharge peak-direction passengers and load reverse-peak ones while still maintaining 12 minute slots. I am not. After the Moynihan Project is complete, the least-efficient through platform should allow a flow of over 700 passengers per minute. Even with 2 trains carrying around 1,500 people each arriving one after another (2 minutes or so apart) on each side of that platform, such a flow rate should handily clear alighting passengers and allow new boardings within 12 minute slots (which require a 10 minute dwell), or even less. At a train per track every 12 minutes, the 15 through tracks, 7 per direction, should handle at least 35 trains per hour. Crucially, LIRR trains that continue to the West Side Yard via the tracks that can’t reach New Jersey should process through significantly faster than that; they will not take on any reverse-peak passengers and discharge to platforms with even more capacity.

To run through with passengers and not just to yards, NJ Transit and LIRR will need trains that can use both overhead wire and third rail, like the M8 or Class 700. They should be single-level trains built under the new FRA crashworthiness rules to reduce the time that each stop adds. Even factoring in a Buy America premium, modified Class 700s are likely to cost significantly less than the price NJ Transit is prepared to pay for custom-built, overweight Bombardier Multilevels, which will have no ability to use third rail. Because of their sluggish acceleration and slow turnover of passengers, Multilevels provide no capacity advantage over single-level stock. NJ Transit should shelve any further orders for these cars and confine them to Hoboken service. It can apply the savings to electrification and platform projects so that as few trains as possible into Manhattan need diesel engines or low-platform capability.

Unlike today’s bastardized through running, where trains discharge passengers and run empty to yards, true through running will generate new passenger demand on top of using Penn Station more efficiently, generating political support for Gateway and setting up the institutional muscle memory that will maximize the project’s utility. Lighter trains that turn over passengers faster will be able to make more stops in the same travel times, simplifying complex service patterns to the benefit of currently underserved inner suburbs.

Cost control

The Gateway Project should cost nowhere near $30 billion, the current asking price, to build. No rail tunnel in the world, proposed, under construction, or built, has ever cost the $3.5 billion/mile that is being demanded for the new Hudson tubes. This estimate dwarfs costs incurred on even New York tunnel projects. Boring and outfitting the Second Avenue Subway tunnels cost around $500 million/mile, as did the Manhattan tunnels for East Side Access (digging here and outfitting here). Note that the $9.5 billion estimate for the new Gateway tunnel includes no work on any station, which was the source of most of the cost and schedule bloat on Second Avenue and East Side Access. Underwater construction does not account for this premium either; the Channel Tunnel cost $600 million/mile, Crossrail tunnels cost $400 million/mile, and the Fehmarn Fixed Link, which will be wide enough to carry a highway and rail line, is estimated at $800 million/mile. Gateway’s total per-rider cost far outstrips those of comparable projects, a selection of which I tabulate here.

Project	Tunnel length (mi)	Stations	Cost (billion USD)	Estimated new daily trips	Cost per new daily trip
Gateway	2.7 new, 2.5 rehabilitated	2 expanded	$30	200,000	$150,000
Second Avenue Subway Phase I	2.0	3 new, 1 upgraded	$5.8	200,000	$29,000
Paris RER East-Ouest Liaison Express	5.0	3 new, 10 upgraded	$5.7	250,000	$23,000
Brisbane Cross River Rail	3.7	4 new, 2 upgraded	$3.8	150,000	$26,000
London Crossrail	13	10 new, 31 upgraded	$23	500,000	$47,000

The first phase of the Second Avenue Subway just added 200,000 daily transit journeys for $5 billion. That price tag should be generous for Gateway, which promises to add the same amount of daily rides to the Northeast regional rail network. It involves 2.5x as much new or rehabilitated tunnel as Second Avenue Phase I and requires no new underground stations—which were the source of most of the cost and schedule overruns under Second Avenue—to handle its projected travel volume. Yet right now the Gateway planners want six times the cost of Second Avenue Phase I.

If Gateway management just gets as much tunnel per dollar as the MTA managed under Second Avenue, the 2.7-mile new Hudson tunnel is a $1.3 billion project. Let’s say the rehabilitation of the existing tunnel costs as much; that still totals only about 1/4 of the current ask for the new tunnel.

The rest of the trackwork, which is modifications and expansion of the surface railroad, should be a cakewalk in comparison. The NICTD West Lake project is to cost $120 million/mile and includes several surface stations. This is high by rest-of-world standards; the Morley-Ellenbrook line in the Perth suburbs is to cost $70 million/mile. There is no excuse for the quad tracking of the 7.5 miles of railroad from North Bergen to Newark to cost more than $1 billion.

Completion of the Central Concourse and widening of other passageways in Penn Station is estimated at $340 million. It is reasonable to expect a $1 billion expenditure on the station to pay for that plus other improvements such as elongating platforms 1 and 2 to the West End Concourse and redoing other hallways and stairwells.

The lighter cream color denotes public circulation space on the lower concourse level of New York Penn Station. The freight elevator corridor just east of Eighth Avenue is also highlighted as public space, which is a probably a mistake or poor choice of color. Observe how little of the floor plate is open for circulation. Fix that, and the existing station footprint will accommodate the Gateway traffic.

Do not expand Penn Station

The Moynihan Station project will be fully complete in a matter of months, and will have added stairwells and escalators to several platforms that needed them. Amtrak banked on being able to process a train every 12 minutes at each platform in the presence of through running without the benefit of the Moynihan project. Those stairwells will cut the worst clearance times by 1.5 minutes and speed boarding passengers too. Completion of the Central Concourse, which Moynihan did not include and is currently expected to cost $340 million along with other hallway widening, would further speed platform clearance and upstairs flow. All of that should cut 2 minutes from the expected dwell time, giving a train every 10 minutes, or 6 trains per hour, on each track. Across the through running tracks in the station, that yields 42 trains per hour, more than the pre-pandemic combined train volume from Long Island and Connecticut. The four-track terminal at Haussmann-St. Lazare on the Paris RER E handles 16 trains per hour; it is reasonable to expect the four stub-ended tracks at Penn Station to handle at least that same volume coming through the new tunnel. Dedicating a couple more tracks to the new tunnel and extension of the platforms to the West End Concourse should add slightly more capacity.

Plot for a connection to Grand Central

Maximum capacity through the new Hudson tunnel requires a connection from Penn Station’s terminal tracks to Grand Central for two reasons: elimination of conflicts from train reversals and division of the passenger load, and thus the dwell time, over two stations rather than one. The project involves about a mile of tunnel and reconfiguration of Grand Central and Penn Station. A connection to the 100-numbered tracks that serve Westchester County is preferable over one to the 300- or 400-numbered tracks that (will) serve Long Island, since Long Island has a connection to New Jersey that properly fitted dual-power trains would be able to use while the Metro-North network does not.

Linking the 100-numbered tracks southward would require demolition of the elevators that currently serve all those platforms and replacement with elevators to each platform farther north. This work is not free, but it should not be as difficult as digging an entire new city block size hole. The combination of existing plans to route some New Haven Line trains to Penn Station, the increased capacity of a Penn Station-Grand Central tunnel, and just plain turning the trains faster would lighten the load on the upper level platforms. For that reason, I think it is plausible to convert some of those platforms to concourse space leading to elevators and stairwells to the lower-level platforms. Maximizing the connection also probably requires a flyover at Melrose just north of the Harlem River, which would take out the 144th Street bridge over the tracks but should cost low $100-millions.

The reforms that have to be made to build the Gateway Project for $5 billion or less will not only free up money for other projects like the Grand Central-Penn Station, they will help those projects come in at reasonable, if still high by rest-of-world standards, costs. In the presence of a sanely priced Gateway Project, it becomes reasonable to expect $2 billion if not $1 billion to buy the Penn-Grand Central connection plus another tunnel linking the northernmost Penn Station platforms to the Empire Line. It is also then becomes plausible for $5 billion or so to afford a tunnel from the East Side Access caverns to Hoboken via Lower Manhattan, giving Hoboken Division service its own trans-Hudson tunnel and supplanting loads of trans-Hudson buses.

The New York infrastructure sector needs to learn from foreign professionals how to achieve reasonable yields out of each dollar and as it does so advance “organization” and “electronics” components that will improve the utility of the “concrete” if it gets built. We have tried to do the exact opposite for as long as I have been following rail issues. You tell me whether it has worked.

Obsessing over the wrong project

It’s not as if I have much to add to Ben Kabak’s writings on the LaGuardia Airtrain, but if there are any doubts that Andrew Cuomo is a terrible transit governor, his ongoing love affair with this terrible project should erase them. The latest twist is a letter rehashing debunked arguments in favor of the proposal from people who would benefit from its construction regardless of its transportation merit.

I would be more understanding if Cuomo were doubling down on an overpriced but useful project like Amtrak Gateway or Second Avenue Phase II. The next question becomes, if he is so great at building infrastructure, why can’t he figure out how to dig new tunnels, be them under East Harlem or under the Hudson, for the same per-mile cost as the Second Avenue Phase I ones? Even reactivation of the LIRR Rockaway Beach Branch would be more useful than linking LaGuardia along a highway corridor to Willets Point, which will even in the best-case scenario get nowhere near the service that Jamaica does and never match the utility of a one-seat ride to Manhattan on the N. Instead, Mr. Cuomo has spent his energy on just about the least useful project possible. At $1 billion per mile, the asking price is now at least tenfold any remotely justifiable figure. At this point, there is no path to good New York transit that runs through this governor, and anyone who is so unwilling to cross Cuomo that they would rather continue making feeble defenses of a thoroughly debunked project has no contribution to make either.

In comparison, dare I say that I see quite a bit to be hopeful about in the new crop of leftist city, state and federal officials that have sprung up in the past five years, even though their grasp of urbanism is underdeveloped. In the spirit of full disclosure, I have donated a total of around $300 to Justice Democrats including Rep. Ocasio-Cortez since 2017. When it comes to transit, it needs to be said that many prominent American progressives are pushing a misguided vision. In service of the laudable goal of extending the reach of transit to the poor, they prioritize fare-free transit while ignoring that fare policy changes and subsidies would accomplish that—and already do to some extent—while preserving fare income that insulates agencies from political swings and incentivizes good performance. It’s also not hard to see that many loud progressive voices don’t see automobile use as a key problem.

On housing, it is even worse. Most recent housing proposals out of the American left center around mythical bottom-up, community plans for public housing. They fundamentally misdiagnose the main issue with housing in America as one of speculation when it is actually one of insufficient supply. They stubbornly claim that enough community input will smooth over most controversy when it is well chronicled that community input channels invariably favor the wealthy and those opposed to change. Ironically, their ideas, if realized, would further empower the exact people responsible for the supply shortage: homeowners who have the social clout, money, and time to fight needed reforms and dominate “community-based” plans with constantly shifting arguments in service of their selfish goals.

Still, I think there remain compelling reasons for knowledgeable housing and transit advocates to continue dialogue with young left-leaning Democratic candidates and signs that doing so will bear fruit. Despite other NIMBY statements, Rep. Alexandria Ocasio-Cortez is a cosponsor of a fairly sweepingly YIMBY House bill. She has expressed at least some support for linking the Astoria Line to LaGuardia. A number of YIMBY city council candidates have sprung up.

In general, the incremental, both-sides approach embraced by 1990s Democrats prioritizes half-measures designed to minimize controversy at the expense of projects that trade short-term turmoil for higher benefits. For all the unrealistic talk on the left of bottom-up planning and pastoralism—which deserves to be criticized, most Clinton Democrats have never seen a new veto point that they didn’t like. All these checkpoints mostly serve to enrich middlemen and elevate the loudest voices, which are almost always unrepresentative of the majority. In comparison, it is hard for me to see how the latest Democrats’ comparative boldness would not be generally good for infrastructure projects.

Moreover, one of the key factors emerging from research into New York’s high construction costs is the importance of a healthy back bench of internal planners and engineers. In lieu of maintaining this kind of state capacity, Third Way-style Clinton Democrats have tended to outsource to consultants that in turn tend to donate heavily to their campaigns. Recent leftist candidates’ fundraising strategies tend heavily toward individual contributions and avoid large contributions from construction and consultant firms that would be expected to favor the status quo. Granted, they have attracted donations from and collaborated with unions that would likely mount at least some opposition to needed work rule reforms, but in the long run, there are plenty of expansion proposals to go around for more-efficient construction gangs to tackle. Despite certain left elements favoring bottom-up planning and pastoralism, general endorsement of another New Deal and Medicare for all, to me, indicates at least an implicit embrace of an increase in state capacity.

While a political culture that is unafraid of using the public sector for bold ends will not transform mass transit on its own, transit will necessarily continue to stagnate under the 1990s New Democratic ideology wherein projects are de facto judged almost only on inoffensiveness. Transit projects in particular depend on the most disruptive elements, such as downtown bus lanes or city center tunneling, coming to fruition. For all of Cuomo’s self-promotion as a doer, he lies firmly in the make-no-waves incrementalist bin. A handful of property owners is apparently too much trouble to bother with extending the Astoria Line to LaGuardia, and redoing work rules and design procedures is unthinkable. Generally speaking, among the left-leaning candidates gaining ground around New York, I see a willingness to anger powerful groups in service of a greater goal that is absent in most Clintonists. I say that counts for a lot, especially if advocates can prevail over some of the more fantastical planks of their urbanism platform.

Time to Revisit Overhead Electrification under Park Avenue?

One of the commonly cited excuses to not build the tunnel needed for through running between the ex-New York Central network feeding into Grand Central and the ex-Pennsylvania/Jersey Central/Lackawanna network feeding Penn Station is the need for trains to handle both third rail and overhead wire. I have written on this blog that cars that do so and can fit every single New York-area regional rail tunnel have existed for decades. On top of that, recent discussion with a knowledgeable consultant suggests to me that modern overhead line equipment (OLE) could potentially fit into the Park Avenue tunnels with today’s trains.

Per an old Congressional hearing transcript, the distance from the ceiling of the Park Avenue tunnel to the top of rail (ToR) is 15’ 1”. Today, the tallest train that runs in it measures 14’ 6” above ToR. That is also just about the maximum train height the North River Tunnels, which have had overhead wire since the 1930s, permit. That means the total space from the top of the train to the structure for an overhead current carrier is 7” or 178 mm. In the UK, for standard 25 kV AC electrification, special permission allows 150 mm of clearance above the contact wire to structure, and the minimum clearance is 200 mm from the contact wire to the train envelope, for a total of 350 mm. However, per a presentation made by Network Rail engineers, new technology can potentially reduce both the required the wire-to-structure and train-to-wire gap: 

• Surge arresters can reduce wire-to-structure clearances to 80 mm or even lower if combined with a contact wire cover and insulated paint.

• Combining surge arresters with shaped metal plates and insulated pantographs can reduce the clearance between the wire and the train roof to as low as 60 mm.

Combined, these technologies could potentially allow electrification with as little as 140 mm or 5.5“ between the train roof and the structure. Furthermore, it is my understanding that the Park Avenue tunnel has ballasted track. Direct fixation track could allow the rail to sit several inches lower in the tunnel, which per an old newspaper clipping is 16’ 8” from the floor to the ceiling. Lowering the track in a tunnel has ample precedent, and I do not know about lowering the tunnel floor itself.

To my eye, as a curious outsider, it seems the technology to equip the Park Avenue tunnel with overhead wire is at least mostly there. However, for the 63rd Street tunnel, it may still be the case that it is cheaper to stick third rail shoes on cars rather than fit the tunnel—the tightest regional rail tunnel in New York—with overhead equipment. In that scenario, the business case for converting Park Avenue to overhead electrification is reduced, since New York would need at least some cars to use both third rail and wire anyway. It must be said that, to my knowledge, the aforementioned technology has not been widely applied. The closer the clearances are, the higher the risks of a tripping event. Installation of any overhead system will take months of disruption.

Having mentioned several caveats, the answer to whether the ex-New York Central network can be converted to overhead electrification is probably yes. Switches in the Grand Central station throat used to have overhead contact strips powered with DC to prevent locomotives from losing power, and most of the platform tracks seem to have enough room for overhead equipment. Compared to Grand Central, whatever track lowering or platform height adjustments are required in the Bronx, Westchester and upstate—where the clearance from structures to ToR is supposedly at least 16’—should be easy.

While New York can use flexible rolling stock in the medium term, modern non-metro rail systems use OLE powered with alternating current for good reasons. It permits higher speeds and requires fewer substations and less power than third rail or wire powered with direct current. OLE is almost certainly safer for passengers and workers than third rail. For example, in the recent Katonah Harlem Line crash, the third rail pierced a train car, igniting a fire. I do not know of any OLE failures that have done similar damage to the inside of a train car. Besides the wire already on the New Haven Line, the Hudson Line feeds into Penn Station. At Amtrak-level electrification costs, which are some of the highest in the world, the entire Harlem Line to Wassaic, Hudson Line to Albany, and Empire Connection would cost around $2 billion to wire. Electric trains would significantly cut travel times for outer Harlem and Hudson riders and intercity riders to Albany and beyond—even with an engine change. Plus cars that only have to use overhead wire likely cost at least somewhat less to acquire and maintain than dual-power stock like the M8 or the Class 700.

In the short to medium term, conversion of the ex-Pennsylvania electrification from the original frequency of 25 Hz to the modern North American standard of 60 Hz would cut out substantial amounts of custom equipment and cut in half the weight of transformer that trains have to carry. Unlike an increase of voltage from 12 kV to 25 kV, this frequency change should not require increasing clearances or replacing OLE components. The change would allow the M8—whose transformer is large enough for 60 Hz but not 25 Hz—to run to New Jersey and beyond without using third rail west of the 60 Hz/25 Hz transition in Queens. As a second-tier project, standardized electrification of New York regional rail is likely to bring some benefit, and OLE in the Park Avenue and 63rd Street tunnels is at least worth a fresh look.

Clinging to Old Ways: Spirals and Cant Deficiency

Northeastern railroad managers continue to make their jobs and riders’ experience harder by clinging to old track design standards. Many curves around the Northeastern rail system can support higher speeds than trains are allowed through them.

The major limitation on speed through a curve stems from centripetal acceleration. Centripetal acceleration redirects an object’s path off a straight line along a curve; passengers on a train feel it as a sensation of being pulled off one’s feet toward the center of a circle. To partially relieve this sensation, tracks are canted or banked so that the reaction force from the track both resists the train’s weight and provides some centripetal force. The centripetal acceleration still felt by the passenger, i. e. that not provided by the track, is expressed as the amount of additional bank needed for the track to supply all the centripetal force. The term for this is unbalanced bank or cant deficiency. Some residual centripetal acceleration acting on the passenger and vehicle is perfectly safe.

Centripetal acceleration a obeys the following formula, where F is force, m is mass, v is velocity (speed), and r is radius.

F = ma = mv²/r

Mass drops out, giving

a = v²/r

Railroads limit speed over curves based on the radius, the cant, and the cant deficiency, so that centripetal acceleration on the car and passenger remains under safe limits. Railroad documents express cant as the distance by which the outer rail is elevated relative to the inner rail. On Metro-North’s and other US companies’ track charts, the cant in inches is located below a line showing the location of curves as bulges. The number above the line is the curvature in degrees per 100 feet of chord. The cant deficiency is similarly expressed as the additional superelevation required for the track to supply the entire centripetal force. The sum of the cant and cant deficiency divided by the track gauge is the sine of the equivalent bank angle; centripetal acceleration a is proportional to its tangent.

a = g tan(sin^-1((E_A + E_U)/G))
g = gravitational acceleration
E_A = cant deficiency
E_U = cant deficiency
G = track gauge
E_A, E_U, and G must be in the same unit.

In Europe, railroads generally allow 150 mm (6”) of cant and another 150 mm (6”) of cant deficiency on non-tilting trains. Tilting trains typically permit a cant deficiency up to 225 mm (9”). The Northeast Corridor contains curves with up to 6” cant, but cant deficiency is limited to 3”. I have a theory as to why.

The length of a spiral is a secondary factor in allowable speeds over a curve. Spirals are the transitions from curves to straight track. They smooth the change in degree of curvature, tilt, and centripetal acceleration. It is not acceleration that causes discomfort; it is the change in acceleration, called jerk. Trains traveling over a transition from a straight section to a curve have to keep the centripetal jerk within set limits. The standard American Railway Engineering and Maintenance-of-Way Association (AREMA) formula, the de facto US standard, is the following:

L = 1.63 E_u V; L is the spiral length in feet, E_u is the cant deficiency, in inches, and V is the speed in mi/h.

In exceptional circumstances, the formula L = 1.22 E_u V is used.

The latter formula, and presumably the former as well, incorporates an arbitrary factor of 2 based on vague comfort standards. There is substantial English language literature substantiating the notion that the AREMA formulas are too conservative.

In Europe, formulas specify maximum jerk as the change in elevation per second; the standard depends on the country but typically lands around 50 — 55 mm/s change in cant or cant deficiency. Tilting trains are typically allowed 70 — 80 mm/s. Essentially, the standard AREMA formula allows 23 mm/s, less than half of what non-tilting trains are allowed in most of Europe; the latter one allows 31 mm/s. In other words, AREMA standards generally require a spiral of double the length as Europe does to handle a given speed. On closely spaced curves, it is likely that AREMA’s overly conservative jerk allowance depresses speeds.

For example, take the reverse curves at Port Chester, New York, on the New Haven Line. On the basis of their radii, they permit 75 mi/h at 12” equivalent cant (sum of cant and cant deficiency), yet trains are restricted to 45 mi/h. At that speed and at Metro-North’s current limit of 3” cant deficiency, the standard AREMA formula (23 mm/s) dictates a spiral of 280 feet is needed to transition into the tightest curve. In contrast, at the speed the curve actually permits at 6” cant deficiency, it prescribes almost a 750 foot spiral and roughly double that for the length needed to transition between the two halves of the reverse curve.

The European standard of 50 mm/s prescribes a 350 foot spiral, a minor increase in length, for 75 mi/h running at 12” equivalent cant. Alternatively, the currently prescribed spiral should support up to 60 mi/h at that much equivalent cant. That is, at a minimum, a 33% speed increase at one of the biggest speed restrictions on the line, which is worth ten seconds over the roughly half-mile slow zone. The braking and acceleration that is avoided saves yet more time. The slowest spots along the line offer many speedup opportunities that merely require adoption of modern operating standards. Since improved operations come a lot cheaper than new right-of-way, realization of these easy speedups behooves riders and the railroad.

NJ Transit Capacity Numbers

In 2014, NJ Transit published a Commuter Rail Fleet Strategy in which the company details its reasoning for favoring Bombardier Multilevels over single-level cars. I skimmed it when it came out and recently went over it again. In brief, NJ Transit itself shows that replacing single-level trains with Multilevels increases the seat count by a much smaller margin than many have been led to believe.

The analysis finds that a 12-car Multilevel electric multiple unit (EMU) set, the power units for which were ordered recently, would provide just a 12 % increase in seats over 12 single-level EMU cars. This increase is less than the 20% that I assumed in my previous post, and there is ample reason to believe that this is not enough to make up for the increased travel time. Recently, I had the luck of catching a rush-hour express from New York Penn to Little Silver. I took train 3255, which was run with a locomotive and Multilevels. With a dramatically reduced passenger load, the trip took 1:10 to reach Little Silver from Manhattan, whereas you would be lucky to make it in under 1:20 with normal ridership. With NJ Transit rail service largely back to pre-pandemic levels, this dramatic change in runtime, a cut of around 15%, likely stems from the lack of passengers turning over at each station. I am fully aware that this is one anecdote, but it is consistent with the outcome of introducing bilevels on Paris’ RER A: a capacity reduction from 30 to 24 trains per hour on its central trunk.

Fundamentally, Multilevels and other bilevels suffer from bottlenecks at the stairwells and doors that do not beset well-designed single level cars. Interior renovations to replace seats with standee space will not help a bilevel put through more passengers; they will just add more dwell time. Worse still, in part because they are designed to the FRA’s old crashworthiness rule, Multilevels add considerable deceleration and acceleration time per stop versus a modern EMU. Cars designed to the new rule, which would be much lighter, should be able to make the New York-Little Silver run in under an hour calling at every intermediate station.

The document makes a dubious claim that Multilevels are cheaper than single level EMUs. It posits that a single-level EMU would cost $5 million per car, or $15 million for a triplet, yet that a Multilevel triplet consisting of two unpowered cars sandwiching a power unit would cost $12.5 million. In contrast, the latest approved contract for 113 Multilevel cars, of which 58 are power units, averages $6 million per car. Its ten-year plan proposes acquisition of 720 Multilevel cars at an average of $4.5 million apiece. In contrast, the evidence suggests single-level EMUs would be far cheaper than $5 million apiece. Siemens recently supplied to London’s Thameslink system 66 -foot Class 700s for $2 million each; an 85-foot North American car should thus cost $2.5 million. A Buy America premium of 25%, which is the threshold beyond which the law allows waivers, brings the cost to $3.2 million apiece.

Don’t take my word that single-level trains are the way to go; NJ Transit’s own data will tell you itself. The company should read and review its own analysis, listen to its own crews, who rightly dislike bilevel trains, and open its eyes to what works in London and what has failed in Paris.

Northeast Corridor State of good repair

Far too much discussion of Amtrak focuses on the factors it does not control. The press coverage of Amtrak tends to focus heavily on the funding it receives, often comparing its paltry funding to the outlays for highways and air infrastructure. As one example, take this piece just published in the Seattle Times. I am not attacking any one author; most Amtrak coverage does this: the right-wing press assails the relatively little funding that it receives, and the left-wing press generally advocates for more. The heavy focus on funding, which depends on Congress, leads the average reader to conclude that more money will solve most of the company’s problems. On the contrary, closer inspection of the recent history of American passenger rail, such as the Northeast Corridor state-of-good-repair backlog, shows that the sector suffers from more ills than just a shortage of input.

Unlike the rest of the country, Amtrak owns most of the Boston-Washington rail spine known as the Northeast Corridor (NEC). In 2009, the Amtrak Office of Inspector General (OIG) issued a report comparing Amtrak’s Northeastern infrastructure upkeep program with those of European rail operators and offering numerous strategies for improvement. Traffic density on the NEC, which is some of the highest in the US, matches that seen on average Western European rail lines—just under 40 train-miles per main-track mile per day. At the time of the report, the Amtrak Engineering Department estimated the steady-state maintenance cost of the NEC plus several Amtrak-owned branch lines, whose length totals 1,558 track miles, at $330 million—$365 million/year in 2015 US dollars.

Amtrak’s monthly reports and other documents show that it has spent well over that figure in recent years. The following table compiles yearly Amtrak Engineering expenditures on infrastructure since 2002.

Year	Expenses (Millions of Constant 2015 USD)
2002	211.2	link
2003	378.1	link
2004	496.9	link
2005	505.8	link
2006	389.0	See above
2007	540.6	link
2008	482.3	link
2009	491.9	link
2010	631.4	link
2011	507.6	link
2012	429.4	link
2013	491.3	link
2014	650.5	link
2015	637.8	link
2016	590.0	link
2017	660.5	link
2018	692.6	link

Moreover, the Union Internationale des Chemins de Fer (UIC) recently released a report about state-of-good-repair expenditures in eight European countries. On average, in 2015, the analysis found the rail networks in the report spent $185,000 (2015 USD) per main-track mile; traffic levels were comparable to those of the NEC. That works to $230 million/year for the 1,243 main-track miles of the entire Northeast Corridor, or $288 million/year for the 1,558 Amtrak-maintained main-track miles considered in the OIG report or master plan.

As shown on the graph in this post, in every year since 2003, Amtrak Engineering has gotten well over its own steady-state estimate. That means that it should have knocked a large chunk off its 2009 backlog.

Annual Amtrak Engineering State-of-Good-Repair Expenditures (2015 USD)

In 2010, the NEC Infrastructure Master Plan estimated the state of good repair backlog at $8.7 billion—including $3 billion on severely outdated New Haven Line infrastructure owned by the state of Connecticut—not Amtrak. So what happened since then? Today, per various reports, it stands at…$33 billion. Amtrak itself estimates its NEC infrastructure is worth some $25 billion. The increase can’t all be coming from the Connecticut-owned portion of the line; it has had its catenary and track renewed, and estimated replacement costs for its movable bridges total around $5 billion. This figure is obscene by the standards of the rest of the world but still insufficient to explain a $20 billion backlog increase.

My own conjecture is that the $33 billion estimate reflects a combination of Amtrak trying to overshoot its needs and incompetent management of resources. Even had Amtrak not implemented the strategies recommended by the OIG report, which should have cut Amtrak’s annual infrastructure maintenance costs by $50 — 150 million (2009 USD), the amount it has spent since then, while not enough to totally eliminate its backlog, should have materially decreased it.

The dramatic increase in Amtrak’s claimed backlog mirrors the increase in price for expansions such as the Amtrak Gateway Program. The total estimate for the project, including the unnecessary Penn Station South, grew from around $13 billion in 2011 to $30 billion today. Despite Amtrak drumming up ample doomsday rhetoric about the necessity of Gateway, no one appears prepared to commit that much money to it. And they shouldn’t be; the $25 billion Crossrail project in London will carry 500,000 daily riders, whereas Gateway is to add around 200,000. Crossrail was also legitimately tough to construct, including eight stations along 13 miles of tunnel through central London. Compare that to Gateway’s 2.7 miles of tunnel—mostly underwater or too deep to conflict with much of anything—and single unneeded station expansion.

Ultimately, people that see no problem with completing 1/9 as much tunnel per dollar as even Crossrail managed have no place at the helm of Amtrak. The company has pursued a maximalist strategy over the last few years that has even driven the Democratic-governed Northeast to its breaking point. It is time for it to stop making excuses and improve the yields from the significant outlays it already receives.

The Importance of Good Single-Level Trains

NJ Transit just released its capital project draft, and while some good projects made it into there, like increased circulation space and added stairwells at New York Penn Station and high-level platforms at several stations, the company is proposing to waste money on subpar trains that if purchased will reduce capacity.

Contained within the plan is an ask of $3.52 billion to purchase 720 Bombardier Multilevel cars and 25 dual-mode locomotives, presumably ALP45DMs, which run about $9 million apiece. Therefore, $3.30 billion is estimated for the 720 cars, giving $4.5 million apiece. Depending on configuration each Multilevel contains 127 — 142 seats; figure perhaps 180 with standees. In contrast, BR Class 700 electric multiple unit (EMU) cars used on the Thameslink system carrying around 145 people each—many as standees, to be fair—cost about $2 million apiece. Each one of those cars measures 66’ long by 9’2” wide; a North American-sized counterpart that measures 85’ long by 10’4” wide should cost around $2.5 million. The cost per passenger is not the only place that Multilevels lag EMUs.

In procuring Multilevels, NJ Transit is treating the problem of trans-Hudson capacity as a 3-D problem when it is really a 4-D one. Moving lots of passengers requires trains that accelerate and brake nimbly and allow rapid boarding and alighting. The Class 700 power-to-weight ratio is about 11 kW/metric ton. An ALP46 straight electric furnishing 5,600 kW plus 9 Multilevels weighs about 700 metric tons, giving 8 kW/t. Worse still, an ALP45DM plus 8 Multilevels, the maximum such engines haul, weighs 627 metric tons. The ALP45DM puts out 4,400 kW, for a power-to-weight ratio of 7 kW/t. Plus, the Class 700 can accelerate at 2 mi/h/sec vs half that for an ALP46/ML set. That all works out to a difference of about 45 seconds in the time it takes to decelerate from 100 mi/h to a stop and get back to 100 mi/h or about 30 seconds if the speed is 80 mi/h. Once one factors in the slower boarding of Multilevels vs. single levels, a Multilevel train adds around a minute to the time needed for each stop. That plus additional slowdown incurred from reduced acceleration and braking around speed restrictions adds 15—20 minutes to the time an EMU needs for a 40-to-50-mile regional rail journey.

Take the New York-Long Branch journey of about 50 miles. It is timetabled at 1:40 today. A few days ago, I timed train 7265 composed of eight Multilevels hauled by an ALP46 from Penn Station to Little Silver, the second to last stop, at 1:22 making every intermediate stop but North Elizabeth. It is reasonable to expect that tightening up dispatching would permit a reliable 1:30 Long Branch-New York run with Multilevels. In contrast, Alon Levy estimates a standard European EMU would make that trip in just over 1:00—making every intermediate stop.

Costs to a transit agency depend on the length of trips. Over a 1:30 trip from New York City to Long Branch, an eight-car Multilevel-ALP train delivers twelve car-hours of service. Let’s be conservative and say an eight-car EMU shaves 1/4 that travel time; the trip would then take nine car-hours of service to furnish. A Bombardier Multilevel has 127 — 142 seats and should be expected to accommodate around 180 total passengers. Let’s say a 85’ EMU with 100 — 110 seats (like NJ Transit’s current single-level fleet) holds 140 total passengers. Although the bilevel accommodates 22% more riders, it requires 33% more time, and therefore operating cost, to carry them, canceling out any advantage in capacity per car.

Unlike NJ Transit’s current fleet, trains like the Thameslink EMUs that use both third rail and overhead wire can through-run via Penn Station between New Jersey and Long Island. Today, many NJT and LIRR trains still turn back at Penn Station; Amtrak finds that converting those to through trains would boost each track's capacity from 3 — 4 trains per hour to 5. The report notes that passenger flow is the limiting factor, but ignores the stairwells that are being built under the Moynihan Station project. This omission is understandable since the report predates the Moynihan project, but it almost certainly causes the analysis to underestimate capacity. Crucially, a single-level car has no stairwells where passengers bottleneck and thus will turn passengers over faster than a bilevel. For the same reason, squeezing more passengers into Multilevels by replacing seats with handholds is unlikely to help since the bottlenecks at the train stairwells would remain.

Finally, bad regional trains negatively affect intercity operations. To keep trip times in check with sluggish equipment, trains run zone local services that involve crossovers between express and local tracks. These moves gum up express tracks that intercity trains would otherwise have to themselves. Conversely, the combination of nimble EMUs and good dispatching will allow NJ Transit to convert express trains to locals and still cut travel times. Fully local trains would be able to stay entirely on the outer tracks on multi-track segments like the Northeast Corridor, which should improve reliability by cutting merges. Plus, additional stops increase the utility of a train, especially for inner-ring suburbs that tend to have large nonwhite populations and which today’s highly zoned service currently underserves.

NJ Transit should spend the savings from buying single-level third rail/wire EMUs versus Multilevels on electrification of its entire system. Amtrak electrified the northern Northeast Corridor for $6 million per route mile in today’s dollars. Although that is at the high end of the scale worldwide, America seems to pay a smaller premium for electrification than it does for other types of rail work, another reason to heavily favor it. Including the Port Jervis Line, there are around 200 route miles on the NJ Transit network left to wire, which should cost around $1.2 billion. In contrast to the $4.5 million per car it is prepared to pay for Multilevels, single-level 85’ EMUs should cost around $2.5 million each. Even after tacking on a 25% premium for Buy America requirements, the threshold beyond which the law permits waivers, the change to Thameslink-style EMUs should save $1 billion. Whatever track can’t be electrified for that amount—the Newark Division should take priority over the Hoboken Division—can be served with Multilevel-ALP trains in the medium term until more money becomes available. At least ALP45s can draw from the wire that’s already in place, which should hasten today’s Hoboken Division run times at least somewhat.

NJ Transit put some good things into its capital plan ask, but some harmful projects also passed through. First to be chopped should be the Multilevels, which reflect a woefully poor understanding of what makes a high-capacity rail system. The current bleak financial picture presents an opportunity to revise the plan for the better. I hope that continued engagement raises awareness of what the bad projects are and why they are bad so that they are ultimately scrapped.

The Advantages of Connecting Penn Station to Grand Central

For anyone interested in really good rail service for greater New York, the Access to the Region’s Core (ARC) Major Investment Study Summary Report is required reading. It describes three major alternatives for increasing trans-Hudson suburban (and intercity) train capacity evaluated under ARC. Unlike the ultimate design that Governor Christie cancelled, all three alternatives as of this 2003 report linked the new tunnels the southern (lower numbered) tracks of New York Penn Station, a feature that the Amtrak Gateway Project retains. The three alternatives are the following, differing on what would happen to the newly added trains after they got to Penn Station:

• Alternative G, where trains would continue in a new tunnel from Penn Station to Grand Central’s lower level;

• Alternative P, where trains would either turn around at Penn Station or at tail tracks just east of it; and

• Alternative S, where trains would continue to Queens in new tunnels.

Notice that Alternative G has the best net operating costs, the greatest increase in passenger revenue, and the greatest diversion in passengers from other modes. Notice also that it has the greatest increase in trans-Hudson passengers and the smallest increase in Penn Station-bound passengers. Herein lies the key: by affording Manhattan-bound riders two center city stations to choose from, it limits outflow at any one, limiting dwells and increasing capacity. This is why 20th century urban rail tunnels have several stations in them. Finally, we notice that Alternative G provides the smallest increase in trans-Hudson trains. It is largely on this basis that the report recommends it not proceed.

The basis for the discrepancy between high utility (diverted trips) and low additional train capacity is, I contend, explained in the following passage on page 19:

“The projected Metro-North 2020 service plan was used as a basis for future train movements into and out of Grand Central Terminal. This service plan included 11 trains in the AM peak hour arriving at the lower level of Grand Central Terminal. The combined Post-Secaucus Service Plan/Metro-North operating plan was the foundation for simulating train movements between the Secaucus Transfer Station, Penn Station, and Grand Central Terminal.

Detailed physical and operating characteristics of NJ TRANSIT, Amtrak, and Metro-North were used to simulate the functional, fully integrated rail network to be in place by 2020. The capacity analysis determined the maximum number of trains capable of operating in the AM peak hour in the key segments, Secaucus to Penn Station and Penn Station to Grand Central Terminal. The results yielded a conceptual service plan that indicated an increase of 13 additional NJ TRANSIT inbound trains (Secaucus to Penn Station) over the Post-Secaucus Service Plan during the AM peak hour, for an inbound total of 36 trains (34 NJ TRANSIT, two Amtrak) occupying 38 slots. Additional service beyond the 13 would make the entire operation
unreliable.

During the AM peak hour, 20 NJ TRANSIT trains would travel from Penn Station to Grand Central Terminal. Of these, 13 would turn in Grand Central Terminal and head back south to Penn Station and points west. To allow NJ TRANSIT trains to enter Grand Central Terminal, the conceptual service plan requires that nine of the 11 Metro-North trains arriving at the lower level of Grand Central Terminal in the AM peak hour would continue through the new tunnels to Penn Station and thence to either the Twelfth Avenue Yard or the Boonton Yard for storage.
The remaining two Metro-North trains and seven NJ TRANSIT trains would head north from the lower level of Grand Central Terminal in revenue and non-revenue service.

Circulation improvements would be needed at both Penn Station and Grand Central Terminal, the latter having certain landmark preservation considerations, to accommodate incremental pedestrian flows. ”

The key is that use of the Metro-North 2020 service plan implies inefficient practices. Today, Metro-North uses a 3 + 1 strategy on the Park Avenue approach to Grand Central. Under this strategy, three out of four tracks take inbound trains in the morning and outbound trains in the evening. Part of the bottlenecking that makes 3 + 1 operation look attractive stems from the low speed limit in the Grand Central throat. Although most of the switches in the approach are good for 15 mi/h or higher, the railroad slows all trains to 10 mi/h for the last mile to the bumpers. Trimming the time trains spend at the platform from today’s 20+ minutes to a more typical dwell of 10 minutes or less should also simplify matters. Alternative G itself would convert many reversals to through movements, thus helping this problem in its own right. The only remaining, and somewhat legitimate, reason for 3 + 1 operation is to keep Hudson Line trains to the westernmost Park Avenue approach track. A flyover at Melrose would easily fix this; a similar flyover proposed for Newark is estimated at $300 million. This cost, while high compared to the Hitchin flyover along the UK’s East Coast Main Line, amounts to rounding error in what Governor Cuomo seems prepared to spend on Penn South.

Essentially, the Alternative G service plan would export to Grand Central the bad practices used at Penn Station today. During the morning peak hour, thirteen of the twenty trains from New Jersey through Penn Station to Grand Central would turn back at Grand Central. These turnarounds would necessarily introduce delays since opposing trains would have to cross paths. It is way simpler to extend these trains to turn around at outlying terminals like Croton-Harmon, North White Plains, and Stamford. Just eleven Metro-North trains out of a total of 52 would arrive at Grand Central’s lower level from the north; nine of those would continue through Penn Station to New Jersey yards. Running just eleven Metro-North trains each morning peak to Penn Station dramatically underutilizes the connection; two-track lines routinely accommodate two-way train flows of 25 — 30 trains each way.

The key to using Grand Central’s potential is to use tighter turn times, a new flyover, and the Alternative G tunnel itself to switch from 3 + 1 to 2 + 2 operations. This way, two tracks, rather than one, would accommodate Westchester-bound trains in the morning, letting in much more than 7 trains per hour from New Jersey. These New Jersey-Westchester trains would bolster reverse peak service and increase capacity upstream at Penn Station. The same would be true in reverse for trains arriving from Westchester. The Alternative G tunnel would enable two parallel through operations at Penn Station: trains linking the new Hudson tunnels with Grand Central and trains linking New Jersey with Queens using the existing Hudson and East River tunnels.

In the same vein, connecting the highest-numbered tracks at Penn Station via a short tunnel to the Empire Line would create a third through running operation. Trains would be able to use the northern two East River tubes relatively independently from those using the southern ones to link to New Jersey.

The insularity in the New York railroading world oozes through the language in the report. The analysis clearly ignores major pieces of modern train operations. The choice made to turn trains back as soon as possible after reaching Manhattan almost certainly stems from a desire to have each railroad’s trains handled by its own crews. This approach is ridiculous; modern railroads perform crew changes constantly, and systemwide electrification (to eliminate slow, heavy, finicky diesel dual-modes) would allow all the local railroad entities to harmonize their trains. On a Tokyo area train, I personally witnessed at least one driver change through the front window; had the window not been there, I would have had no clue since the station stop took no longer than any other on the journey. I see little reason why similar handoffs cannot be done here in New York.

For those with open minds, the contrast between Alternative G and Penn South shows the folly of using new construction to cement bad practices. Paris, Milan, Tokyo, and many other cities have overcome the challenges of institutional turf to secure for their inhabitants much higher quality rail service than New Yorkers enjoy. Ultimately, we can do the tough, unsexy work of untangling institutional dysfunction too. Will we?

Penn Station South Is Unnecessary

Several prominent voices have claimed that trans-Hudson rail riders need an enlarged Penn Station to go along with new Hudson tunnels. Having used the station many times, it is clear why most travelers would support the $8 billion Penn South project, which would condemn an entire Midtown city block to expand Penn Station with eight station tracks that would be linked to New Jersey but not to Queens. However, a closer look reveals that everything Penn South would do can instead be done in the station’s existing footprint with improved operations and passenger circulation.

At track level, the railroads prioritize independent, stub-ended New Jersey and New York suburban train operations at the expense of capacity. Although most platform tracks connect to both New Jersey and Queens, many trains reverse out, causing conflicts that decrease throughput. Trains that don’t reverse discharge all passengers and run to yards, a procedure called “drop and go”. These outdated practices enable New York and New Jersey trains to use independent crew pools and power sources—third rail in New York and overhead wire in New Jersey. By contrast, other cities’ suburban rail networks feature through running, where trains carry passengers from a suburb through the city center to another suburb without reversing. Among them is London, whose network, like ours, features both overhead wire and third rail. It uses dual-power trains similar to our M8s. Replacing the NJ Transit and LIRR fleets—totaling around 2,200 cars—with dual-power units should cost $4—5 billion and yield large benefits. Today’s service, which runs frequently during peak times and sparsely otherwise, requires many crews to split their workday—half in the morning and half in the evening. In contrast, a through running operation would keep trains and staff in continuous service, which would cost less per service hour. Moreover, the new cars would brake, accelerate, and load passengers more quickly than many current ones, especially the NJ Transit Multilevels. Most express trains could be converted to locals and still cut travel time. These additional stops combined with one-seat service from New Jersey to Long Island and Connecticut would induce new ridership, especially from inner suburbs that are currently underserved.

A 2014 Amtrak study confirms that through running is feasible at Penn Station. It finds that each track could handle 5 trains per hour with through running, noting that passenger flow to and from the trains is the limiting factor. The fifteen through tracks alone, seven per direction, thus afford 35 trains each way per hour, today’s peak-hour volume from Queens. Use of tracks 1 – 4 to turn trans-Hudson trains and using tracks 20 – 21 to “drop and go” from Queens contributes additional capacity. The analysis ignores the platform stairwells being built under the Moynihan project, meaning it likely understates the potential of through running.

Upstairs, fragmented management of the concourse levels causes inefficient flow. Although Amtrak owns the whole station, it only manages the western portion of the concourse, renting the northern part to the LIRR and the southeastern portion to NJ Transit. Signage and design cues encourage each company’s passengers to only board trains through “their” area of the station. For example, prominent overhead signs just inside the 32nd Street and 7th Avenue entrance tell NJ Transit passengers to immediately turn left. They give no indication that further access to the same tracks lies straight ahead or that many NJ Transit trains board to the right, where ticketing and information for them is nonexistent. Instead, the departure information for all trains should be displayed all around the station. Moreover, all signage needs to be unified and use a clear naming convention for the different station areas. For example, each row of platform gates could be assigned a letter. Finally, all suburban ticket machines should sell fares to all destinations served by suburban trains.

On top of unified management and information systems, Penn Station still requires some construction. Today, very little of the station’s upper or lower concourse floors is open to passengers, and unlike the original design, little light reaches the platforms. Retail and office space needs to be converted to circulation area and use as much glass as possible, as the refurbished West End Concourse already does.

Small changes in the time a train spends at the platform, the dwell, have big effects. Amtrak assumes that through running allows each track to process a train every 12 minutes, presumably using the worst pre-Moynihan platform clearance times. Fortunately, by next year, the Moynihan project will have added enough stairwells to cut the worst clearance times by a minute or so. Those stairwells will also speed access for boarding passengers; the combination of faster egress and ingress between concourse and train should cut the time per train to 10 minutes from today’s values of 15 — 18 minutes, allowing 42 trains per hour each way on the through tracks. Continuing to turn a limited number of trains back to New Jersey on tracks 1 — 4 should give a total station capacity around double today’s peak trans-Hudson volume of 25 trains per hour.

A 2008 analysis showed that while 1,600 passengers could clear a train via one of the northern platforms in 2 – 3 minutes, it took around 6 minutes at the southern platforms, which had fewer stairwells. Today, Penn Station gets 34,000 arriving passengers in the peak hour from Queens plus 26,000 from New Jersey. East Side Access is to divert around half of the Queens passengers to Grand Central, and the Gateway Program would roughly double inbound passengers from New Jersey, giving a total of around 70,000 inbound passengers during the peak hour. The Moynihan project will allow up to 800 passengers to enter and exit each of the eleven platforms per minute, which would let detraining passengers clear most trains within 3 – 4 minutes and allow ample time for new boardings. Other projects to further speed passenger flow, such as completion of the “Central Concourse”, which is now in design, should be prioritized. The cost of it plus enlargement of nearby hallways is estimated at $340 million. There is little reason to worry that such projects depend on Penn South; NYC Transit and Amtrak are undertaking circulation work at Grand Central and at Washington Union Station, respectively, with little to no property acquisition. Note that the cost of circulation improvements and new trains fits well within the $8 billion that Governor Cuomo is apparently prepared to spend on Penn South.

Summary of Penn Station circulation parameters expected after completion of the Moynihan Train Hall project.

Penn Station circulation capacity and clearance times at each platform with (B) and without (NB) stairwells being delivered under the Moynihan Train Hall project.

Longer term, New York should further modify Penn Station’s track level to speed boarding and minimize instances where opposing trains cross paths. Removal of station tracks to widen platforms and stairwells would allow ample two-way passenger flow, cutting train dwells from several minutes to 1 — 2 minutes. After new Hudson tunnels are dug, terminal tracks at Penn Station (tracks 1 – 4) and the lower level of Grand Central (tracks 105 — 112) should be connected. This project would eliminate turnarounds, provide even more one-seat rides, and spread out passengers between two stations rather than one, increasing efficiency. Judging by the Second Avenue Subway and East Side Access, $2 – 3 billion should afford this connection and the associated station work. Similarly, tracks 20 – 21, which currently lead to the West Side Yard, should be connected to the Empire Line to take LIRR trains to the Upper West Side and the Hudson Valley. This connection requires a much shorter tunnel than that needed to connect to Grand Central.

Effectively, these connections would result in Penn Station hosting the following three trunk lines, listed from south to north:

1. New Jersey — Westchester County via new “Gateway” trans-Hudson tunnel and new tunnel to Grand Central.

2. New Jersey — Long Island/Connecticut via existing Hudson and East River tunnels 1 and 2.

3. Upstate New York — Long Island via new Empire Line tunnel and existing East River tunnels 3 and 4.

Since New York gets the least transit construction per dollar in the world, it needs to choose projects very carefully. The MTA budget has increased markedly over the last decade, and Amtrak’s finances have (justifiably) stabilized, without a corresponding improvement at Penn Station. This shows that a new approach that challenges institutional inertia and territoriality is needed to maximize the reach of dollars. Political leaders seem prepared to spend billions of dollars on Penn Station. If that cash is channeled into a modern through operation rather than into accommodation of the outdated terminal strategy, an enviable station experience and rail network is within the reach of New York.

Portal Bridge funding is a pyrrhic victory

A few days ago, the New York press was abuzz about a report that Trump has agreed to let the Portal Bridge North project obtain federal construction funds. This project would replace the drawbridge that carries the Northeast Corridor over the Hackensack River between Newark, NJ, and New York City with one that is high enough to not need to open. Make no mistake, Trump has been playing political games with infrastructure. Replacement of the existing drawbridge with a fixed bridge is a worthy project. But securing federal funding would be one battle won in a war New York is losing with itself.

As has been extensively chronicled, no urban area on earth is getting less transit construction per dollar than greater New York. Even though the available examples of rail bridge project costs are fewer in number and less well compiled than metro and rail tunnel projects, the data points still suggest a substantial New York premium. The asking price of $1.8 billion for 2.3 miles of construction is double the per mile cost of replacing the Crum Creek viaduct, which carries the SEPTA Media-Elwyn Line over a deep gorge. Worse still, the Portal Bridge North cost includes approach spans that are comparatively easy to construct, whereas the Crum Creek project limits are entirely within the canyon. Overseas, the Erasmusbrug in Rotterdam cost around $550 million per mile in today’s dollars. It carries an urban boulevard with four road lanes and a tram line and has a movable span. The Po River bridge along the Milan-Bologna HSR line cost $70 million per mile, including approaches. The Portal Bridge North cost of nearly $800 million per mile would, nearly everywhere else in the world, suffice for an under-river tunnel.

In contrast to the Hackensack crossing, the new tunnel proposed to span the Hudson remains mostly unfunded. Like Portal North, the asking price reaches highs seen nowhere else in the world to date. The tunnel proponents are asking $9.5 billion for 2.7 miles of tunnel from North Bergen to Penn Station. By contrast, the tunnels and systems for the Second Avenue Subway Phase I cost around $500 million/mile. Tunnels and systems on London’s Crossrail project, which spans the Thames twice, cost $400 million/mile. Both of those projects suffered significant schedule and cost overruns. The Transbay Tube connecting San Francisco and Oakland cost about 40 % more than the corresponding segments entirely under land, whereas the Hudson tunnels are to cost sevenfold more per mile than the Second Avenue ones.

Politically, if New York and New Jersey want to feed the narrative of Trump swooping in to save the day, taking the money for Portal is the perfect way to do it. He will be able to say he saved the incompetent New York/New Jersey bureaucracy from itself, and in that regard he will largely be right. Excusing away poor yield per dollar by pointing out that the Northeast states are “donor states” is also not politically tenable. While the Northeast certainly receives less in federal benefits than the tax money it sends to Washington, the end result is in large part exactly the downward redistribution of wealth to the relatively poor interior states that the dominant Northeastern political ideology—and the one offered up to “resist” Trump—claims to, and should, espouse.

Instead of relying on a couple of tweets on which our mercurial Commander-in-Chief can renege on a dime, the New York infrastructure sector needs to work on increasing its yield per dollar. The available evidence suggests a substantial premium that cannot be explained by geography or the political complication of a densely populated, high-cost union stronghold. Moreover, California has coaxed substantial sums out of the Trump Administration for transit construction. It strains credulity to deny that their relatively high yield per dollar did not factor into their relative success. The tunnels on the Los Angeles Purple Line are to cost $500 million/mile, and the overall cost is just over $1 billion/mile. Compare that to the $3.5 billion/mile asking price for the Second Avenue Subway second phase, the tunnels for which are partially in place.

The good news is that the project management just has to improve yield per dollar to the low end of that experienced elsewhere to make the Hudson tunnel and the larger Gateway Program work. Just achieving the productivity realized under Second Avenue would build the tunnel for $1.4 billion. Clearly, something about Amtrak’s project management methods increases costs far beyond what even the famously inefficient MTA incurs. It’s time to figure out what that is and do the hard work of fixing it.