Category: Shoddy Studies

How to Waste Money on Public Transportation

This is the fourth in a series of five (not four) posts about the poor state of political transit advocacy in the United States, following posts about the Green Line Extension in metro Boston, free public transport proposals, and federal aid to operations, to be followed by a post about how to do better instead.

I think very highly of Yonah Freemark. His academic and popular work on public transport and urbanism ranges from good to excellent, and a lot of my early thinking (and early writing!) on regional rail and high-speed rail owes a debt to him.

But I think he’s wrong in his proposal for a Green New Deal for transportation. This is a proposal by the Climate and Community Project (not the Urban Institute as I said in previous posts – sorry) to decarbonize transport in the United States, through fleet electrification and investments in public transport. Yonah is one of several authors; I identify him with the public transit-related parts of the report, but I want to make it clear that it’s the report I’m criticizing, regardless of who wrote what.

The fundamental problem of the CCP report is what I’ve been building up to in the last three posts in this series: it tries to please everybody by throwing money everywhere and making conflicting promises. The Green Line Extension was built this way under Deval Patrick, and costs ballooned, and what passed for discipline under Charlie Baker just reinforced the same long-term loss of state capacity that led to the cost explosion.

For example, here’s its take on fleet electrification:

In other words, there is a compelling and immediate need to decarbonize this fleet within a decade. And that’s feasible: buses are replaced every 10 to 15 years on average, and commuter rail trains about every 25 years; currently, commuter trains in the United States are on average 22 years old. Publicly owned vehicles would be replaced with the electric equivalent; for privately owned contracted vehicles (the case for many school buses), and requirements for electrification would be written into contracts and tax credits given to assist the transition of buses from fossil fuels to electric. The commissioning of thousands of new transit vehicles would produce new, good-paying union jobs in manufacturing. The shift to electric transit vehicles would affect maintenance requirements, and the Department of Transportation must ensure the mechanic and operator workforce is fully prepared for the electric transition through workforce retraining assistance. This may require retraining, such as encouraging mechanics to retrain as electric vehicle charging installers.


Electrifying existing diesel railways would require overhead catenary electrical wires to be useful for electrified trains (though the trains themselves actually cost less than diesel vehicles). The cost of railway electrification infrastructure alone is between roughly $1 and $5 million per mile. There are roughly 6,600 miles of non-electrified commuter rail in the United States, plus roughly 20,800 miles of non-electrified Amtrak service (with some overlap between the two). Amtrak’s routes are mostly owned by freight rail companies, but we suggest joint electrification that includes both passenger trains and freight trains, using this program for Amtrak and another we lay out below for the freight lines. To electrify the national passenger rail network of existing lines would cost between $27 and $137 billion. In addition, new trains would have to be purchased to run on these electrified lines.

I cite this pair of paragraphs because of something they show about the study: it is not uniformly bad. The second paragraph is a decent idea (though $1m/mile is very cheap), and trying to workshop how to wire the national freight network is not necessarily a bad idea, even if the report doesn’t go into enough detail about what the business barrier to electrification is for the private carriers.

But the first quoted paragraph is awful. Here’s the key thing: “The commissioning of thousands of new transit vehicles would produce new, good-paying union jobs in manufacturing” is a giant waste of money. Bus vendors outside North America consistently produce equipment for much less than the protected North American market; the Boris Bus, at £350,000 per unit (around $500,000), is both cheaper than American buses and locally considered expensive, a prime example of Boris Johnson’s poor performance as mayor of London.

The passenger rail industry does not exist in the United States, and attempts by American governments to coerce it to build factories domestically in order to create well-paying jobs have resulted in ballooning costs. The premium for recent American rolling stock orders, behind bespoke regulations, protectionism, informal state-level protectionism, and agency heads that know less than recently-graduated interns who make one quarter of what they do (less, if those interns are European), looks like 50% over European equivalents. Nor does this do much job creation, except perhaps for sitework consultants: the premium for some recent orders has been $1 million per $20/hour 4-to-6-year job created. Those are not objectively good jobs – the wages are not much higher than present-day retail, food service, and delivery jobs – but backward-looking politicians consider them inherently moral, and the report coddles them instead of looking forward.

Then, the report has the following recommendations for how to spend money on improving public transportation:

End the use of federal infrastructure funding for new highway infrastructure, except for focused opportunities that improve equity. Provide immediate funds for a quick-start infrastructure program for walking and cycling. Vastly expand support for transit and metropolitan network planning.

Appropriate $250 billion over 10 years, or $25 billion annually, in federal funding bill to support transit operations funding throughout the United States.

Increase federal support for transit and intercity rail capital projects to $400 billion over 10 years, or $40 billion annually, providing funds for new lines, maintenance of existing infrastructure, and upgrades designed for equitable accessibility.

Require metropolitan planning organization voting systems to be proportional to resident population. Mandate adjustments to local zoning policy to enable more dense, affordable housing near transit in exchange for federal aid. Implement regional commuter benefits throughout the nation.

This, I’m sorry, is a bad program. The $40 billion/year capital investment is not bad, but the proposal explicitly includes maintenance, making it vulnerable to the state of good repair scam, in which agencies demand escalating amounts of money for infrastructure with nothing to show for it. The $25 billion/year operating aid is likely to be a waste as well.

Transit agencies can invest money prudently, but the report says nothing about how to do it, instead proposing to zero out highway funding (which is a good way to save money, but is less relevant to mode shift than American transit advocates think it is). The one concrete suggestion for what to do with the money is “One goal, for example, would be for all residents to have access to a bus or train with a short wait within at most a 15-minute walk at all times of the day.” This is a standard I can get behind in a dense place like New York; nearly everywhere else, it means overfunding coverage routes in low-density areas, often middle-class white flight suburbs, ahead of workhorse urban routes. Writing years ago about New Haven, Sandy Johnston noted that a bus reform there would cannibalize the circuitous suburban bus branches to add service on the core routes through the city and Hamden. The CCP report would do the opposite, boosting frequencies where they are least useful.

Finally, the MPO rules seem weak. I get what Yonah (and perhaps the other authors) wants to do here: he wants to incentivize more housing production near mass transit nodes. But MPO voting weights are not especially relevant. What is relevant is using state power to disempower local communities, which are dominated by NIMBYs even in places where the residents vote YIMBY at the state level, such as San Francisco. The report talks about banning single-family zoning (okay, but duplexes are not TOD), but that’s it. Then it suggests extracting developer profits through mandatory inclusionary housing, which acts as a tax on TOD and reduces housing production. The authors of the study are left-wing, but do not propose public housing, only taxes on TOD to subsidize some local housing; Yonah knows this is not how social housing works in Paris, but he still proposes this for the United States.

The theme of lack of willingness to prioritize flow throughout these recommendations. There is no discussion of how to prioritize good investments, how to increase efficiency (the report points out operating costs for all US transit combined are $50 billion/year; this is 2.5 times the German level, for similar ridership, not per capita), how to make sure that progress does not get extracted by programs for groups thought inherently moral.

The Interborough Study

I was excited about the idea of Interborough Express (IBX) as announced by New York Governor Kathy Hochul, and then last week her office released a preliminary report about the alternatives for it, and I got less excited. But it’s not that the study is bad, or that Hochul is bad. Rather, the study is a by the numbers alternatives analysis, shorter than the usual in a good way; its shortcomings are the shortcomings of all American planning.

The main rub is that the report looks at various options for the IBX route, broken down by mode. There’s a commuter rail option, which bakes in the usual bad assumption about commuter rail operations, including heavier trains (lighter trains are legal on US tracks as of 2018) and longer dwell times that are explained as a product of the heavier trains (dwell times have nothing to do with train mass). That’s par for the course – as we saw yesterday, everything that touches mainline rail in North America becomes stupid even in an otherwise understandable report.

But even excluding commuter rail, the study classifies the options by mode, focusing on bus rapid transit and light rail (and no subway, for some reason). It compares those two options and commuter rail on various measures like expected ridership and trip times. This is normal for American alternatives analyses for new corridors like IBX: they look at different modes as the main decision point.

This is also extraordinarily bad governance. There are some fundamental questions that are treated as afterthoughts, either not studied at all or mentioned briefly as 1-2 sentences:

  • How far north should the line go? The IBX plan is to only go from Jackson Heights to the south, in contrast with older Triboro proposals going into the Bronx.
  • What should the stop spacing be? The stops can be widely spaced, as in the current proposal, which stops mainly at intersection points with other lines, or more closely spaced, like an ordinary subway line.
  • Under a light rail option, should the line be elevated where the trench is too narrow or at-grade?
  • Should freight service be retained? What are the benefits of retaining freight rail service on the Bay Ridge Branch and what are the incremental costs of keeping it versus taking over the right-of-way?
  • How large should the stations be?
  • How frequent should the trains be? If freight service is retained, what frequencies are compatible with running freight on the same tracks for part or all of the line?

A better study must focus on these questions. Some of them, moreover, must be decided early: urban planning depends on whether the line goes into the Bronx or not; and industrial planning depends on what is done with freight service along the corridor.

Those questions, moreover, are more difficult than the modal question. A BRT option on a rail corridor without closely parallel arterial roads should be dismissed with the same ease that the study dismisses options not studied, and then the question of what kind of rail service to run is much less important than the scope of the project.

But American planning is obsessed with comparing public transit by mode rather than by corridor, scope, or any other aspect. Canadian planning has the same misfeature – the studies for the Broadway SkyTrain extension looked at various BRT and light rail options throughout, even though it was clear the answer was going to be SkyTrain, and omitted more fundamental questions regarding the cost-construction disruption tradeoff or even the scope of the project (the original studies from 2012 did not look at truncating to Arbutus, an option that had been talked about before and that would eventually happen due to cost overruns).

So overall, the IBX study is bad. But it is interestingly bad. Andrew Cuomo was a despicable governor who belongs in prison for his crimes. Less criminal and yet similarly loathsome people exist in American public transit. And yet, Hochul and her office are not like that, at all. This is not a sandbag, or a corrupt deal. It’s utterly ordinary in its failure; with all the unique failures of the Cuomo era stripped, what is left is standard American practice, written more clearly than is usual, and it just isn’t up to par as an analysis.

Hochul has been moving on this project very quickly, and good transit advocates should laud this. It should not take long to publish a report comparing alternatives on more fundamental questions than mode, such as scope, the role of freight, and the extent of civil infrastructure to be used. The costs and benefits of IBX heavily depend on the decisions made on such matters; they should not be brushed aside.

New York Publishes a Bad Benchmarking Report

I’ve grown to intensely dislike benchmarking reports. It’s not that the idea of benchmarking bad. It’s that they omit crucial information – namely, the name of the system that one is compared with. The indicators always have a wide variety of values, and not being able to match them with systems makes it impossible to do sanity-checks, such as noticing if systems with high costs per car-km are consistently ones that run shorter trains. This way, those anonymized reports turn into tools of obfuscation and excusemongering.

The MTA in New York recently published such a report, including both US-wide and international benchmarking for the subway as well as commuter rail. The US benchmarking is with comparable American systems – exactly the ones I’d compare, with the systems listed by name as NTD data is wisely not anonymized. The international benchmarking for the subway is with CoMET, which includes most of the larger global systems as well as a handful of smaller ones, like Vancouver; for commuter rail, it’s with ISBeRG, which has an odd list of systems, omitting the RER (which is counted in CoMET), all of Japan except JR East, and any S-Bahn, skipping down to Australian systems, Cape Town, and Barcelona.

That, by itself, makes much of the international benchmarking worthless. The standard metric for operating costs is per car-km. This is covered in pp. 8-9, showing that New York has fairly average costs excluding maintenance, but the second highest maintenance costs. But here’s the problem: I’m seeing a comparison to an undifferentiated mass of other systems. One of them is an outlier in maintenance costs, even ahead of New York, but I do not know which it is, which means that I cannot look at it and see what it does wrong – perhaps it has an unusually old fleet, perhaps it is small and lacks scale, perhaps it is domestically viewed as scandal-ridden.

Far more useful is to look at complete data by name. For example, JICA has complete operating cost data for Japanese metro systems. Its tables are complete enough that we can see, for example, that overall operating costs are around $5/car-km for all systems, regardless of scale; so scale should not be too important, or perhaps Tokyo’s wealth exactly cancels out the scale effect. There are, on table 2.37 on PDF-p. 117, headcounts for most systems from which we can impute labor efficiency directly, using train-km data on PDF-p. 254; Yokohama gets 1,072 train-hours a year per driver at 35 km/h (the rough average speed I get from Hyperdia).

And here’s the thing: without the ability to fill in missing data like average speed, or to look at things the report didn’t emphasize, the report is not useful to me, or to other independent researchers. It’s a statement of excuses for New York’s elevated operating and maintenance cost, with officious proclamations and intimidating numbers.

For example, here’s the excuse for high maintenance costs:

High maintenance costs for NYCT are largely attributable to 24-hour service. Most COMET peer agencies shut down every night, allowing for four hours of continuous daily maintenance. In comparison, NYCT subway’s 24-hour service requires maintenance to occur within 20-minute windows between late night trains, reducing work efficiencies. Additionally, maintenance costs for NYCT have risen recently to support the improvements as part of the Subway Action Plan, which have led to a significant improvement to on-time performance year over year since inception.

Okay, so here we’re seeing what starts like a reasonable explanation – New York doesn’t have regular nighttime maintenance windows. But the other American systems studied do and they’d be above global average too; Boston has regular nighttime work windows but still can’t consign all track maintenance to them, and has almost the same maintenance cost per car-km as New York. Moreover, track maintenance costs per car-km should feature extensive scale effects – only at freight rail loads is the marginal track wear caused by each additional car significant – and New York runs long trains.

Then there is the Subway Action Plan line, which is a pure excuse. Other systems do preventive maintenance too, thank you very much. New York is not unusually reliable by global standards, and the benchmarking report doesn’t investigate questions like mean distance between failures or some measure of the presence of slow restrictions – and because it is anonymized, independent researchers can’t use what it does have and get answers from other sources.

The study has a section on labor costs, showing New York’s are much higher than those of some peer cities. Thankfully, that part is not anonymized, which means I can look at the cities with overall labor costs that are comparable to New York’s, like London, and ignore the rest; New York’s construction labor costs are higher than London’s by a factor of about 2, despite roughly even regionwide average wages. Unfortunately, a key attribute is missing: labor efficiency. The JICA study does better, by listing precise headcounts; but here the information is not given, which means that drawing any conclusion that is not within the purview of MTA’s endless cold war on its unions is not possible. As it happens, I know that New York is overstaffed, but only from other sources, never anonymized.

It’s worse with commuter rail. First of all, at the level of benchmarking, the study’s list of comparisons is so incomplete and so skewed (three Australian systems, again) that nothing it shows can be relevant. And second, commuter rail in North America comes with its own internal backward-looking culture of insularity and incompetence.

The report even kneecaps itself by saying,

While it is true that benchmarking provides useful insights, it is also important to acknowledge that significant differences exist among the railroads that pose challenges for drawing apples-to-apples conclusions, particularly when it comes to comparisons with international peers. Differing local economies, prevailing wages and collective bargaining agreement provisions can have dramatic impacts on respective labor costs. Government mandates, including safety regulations, vary widely, and each railroad exists in a unique operating environment, often with different service schedules, geographic layouts and protocols. Together these factors have also have a significant impact on relative cost structures.

To translate from bureaucratic to plain English, what they’re saying is that American (and Canadian) practices for commuter rail are uniquely bad, but controlling for them, everything is fine. The report then lists the following excuses, all of which are wrong:

• Hours of Operation: LIRR provides 24 hours of service 7 days per week, and MNR provides 20-22 hours of service 7 days a week

• Ungated System: Neither LIRR nor MNR operate gated systems, therefore they require onboard fare validation/collection

• Branch Service: Both LIRR and MNR run service to and from a central business district (New York City) and do not have ability to offer through-running service

• Electrification: Both LIRR and MNR operate over both electrified and non-electrified territory, thereby requiring both electric and diesel fleets

It’s impressive how much fraud – or, more likely, wanton indifference and incuriosity – can fit into just four bullet points. Metro-North’s hours of service are long, but so are those of the JR East commuter lines; the Yamanote Line runs 20 hours a day, which means the nighttime maintenance window is shorter. Ungated systems use proof-of-payment ticketing throughout Europe – I don’t know if Rodalies de Catalunya runs driver-only trains, but the partly-gated RER and the ungated S-Bahns in the German-speaking world do. Through-running is a nice efficiency but not all systems have it, and in particular Melbourne has a one-way loop system akin to that of the Chicago L instead of through-running. Finally, electrification on the LIRR and Metro-North is extensive and while their diesel tails are very expensive, they also sometimes exist in Europe, including in London on a line that’s partly shared with the Underground, though I don’t know if they do in the report’s comparison cases.

The report does not question any of the usual assumptions of American mainline rail: that it must run unusually heavy vehicles, that it run with ticket-punching conductors, etc.

For a much more useful benchmarking, without anonymization, let’s look at German S-Bahns briefly. There is a list of the five largest systems – Berlin, Munich, Hamburg, Frankfurt, Stuttgart – with ridership and headcounts; some more detail about Berlin can be found here. Those five systems total 6,200 employees; the LIRR has 7,671 and Metro-North 6,773. With 2,875 employees, the Berlin S-Bahn has more train-hours than the LIRR, Metro-North, and New Jersey Transit combined; about as many car-km pro-rated to car length as the LIRR times 1.5; and more ridership than all American commuter rail systems combined. The LIRR in other words has more workers than the largest five German S-Bahns combined while the Berlin S-Bahn has more riders than all American commuter rail systems combined.

The excuses in the report highlight some of the reasons why – the US sticks to ticket-punching and buys high-maintenance trains compliant with obsolete regulations – but omits many more, including poor maintenance practices and inefficient scheduling of both trains and crew. But those are not justifications; they are a list of core practices of North American commuter rail that need to be eliminated, and if the workers and managers cannot part with them, then they should be laid off immediately.

Platform Edge Doors

In New York, a well-publicized homicide by pushing the victim onto the subway tracks created a conversation about platform edge doors, or PEDs; A Train of Thought even mentions this New York context, with photos from Singapore.

In Paris, the ongoing automation of the system involves installing PEDs. This is for a combination of safety and precision. For safety, unattended trains do not have drivers who would notice if a passenger fell onto the track. For precision, the same technology that lets trains run with a high level of automation, which includes driverless operations but not just, can also let the train arrive with meter-scale precision so that PEDs are viable. This means that we have a ready comparison for how much PEDs should cost.

The cost of M4 PEDs is 106 M€ for 29 stations, or 3.7M€ per station. The platforms are 90 meters long; New York’s are mostly twice as long, but some (on the 1-6) are only 70% longer. So, pro-rated to Parisian length, this should be around $10 million per station with two platform faces. Based on Vanshnook’s track map, there are 204 pairs of platform faces on the IRT, 187 on the IND (including the entire Culver Line), and 165 on the BMT (including Second Avenue Subway). So this should be about $5.5 billion, systemwide.

Here is what the MTA thinks it should cost. It projects $55 million per station – but the study is notable in looking for excuses not to do it. Instead of talking about PEDs, it talks about how they are infeasible, categorizing stations by what the excuse is. At the largest group, it is accessibility; PEDs improve accessibility, but such a big station project voids the grandfather clause in the Americans with Disabilities Act that permits New York to keep its system inaccessible (Berlin, of similar age, is approaching 100% accessible), and therefore the MTA does not do major station upgrades until it can extort ADA funding for them.

Then there is the excuse of pre-cast platforms. These are supposed to be structurally incapable of hosting PEDs; in reality, PEDs are present on a variety of platforms, including legacy ones that are similar to those of New York, for example in Paris. (Singapore was the first full-size heavy rail system to have PEDs – in fact it has full-height platform screen doors, or PSDs, at the underground stations – but there are later retrofits in Singapore, Paris, Shanghai, and other cities.)

The trains in New York do not have consistent door placement. The study surprisingly does not mention that as a major impediment, only a minor one – but at any rate, there are vertical doors for such situations.

So there is a solution to subway falls and suicides; it improves accessibility because of accidental falls, and full-height PSDs also reduce air cooling costs at stations. Unfortunately, for a combination of extreme construction costs and an agency that doesn’t really want to build things with its $50 billion capital plans, it will not happen while the agency and its political leaders remain as they are.

The Invention of Bad Railroad Timetables

The rail advocate Shaul Picker has uploaded a fascinating potpourri of studies regarding commuter rail operations. Among them, two deserve highlight, because they cover the invention of bad timetable practices in New York, and, unfortunately, not only think those practices are good, but also view their goodness as self-evident. They are both by Donald Eisele, who was working for the New York Central and implemented this system on the lines that are now Metro-North, first introducing the concept to the literature in 1968, and then in 1978 asserting, on flimsy evidence, that it worked. Having implemented it in 1964 based on a similar implementation a few years earlier in the Bay Area, Eisele must be viewed as one of the people most responsible for the poor quality of American mainline service, and his idea of zone theory or zonal operations must be discarded in favor of the S-Bahn takt.

Zone theory

Eisele’s starting point is that commuter rail service should be exclusively about connecting the suburbs with city center. He contrasts his approach with urban transit, which is about service from everywhere to everywhere; trips short of Manhattan were 20% of single-trip ticket revenue for New York Central suburban operations and 5% of multi-ride pass revenue, and the railroad wanted to eliminate this traffic and focus on suburb-to-city commuters. From this inauspicious starting point, he implemented a timetable in which suburban stations are grouped into zones of a few contiguous stations each, typically 2-4 stations. At rush hour, a train only stops within one zone, and then expresses to city center, which in the original case means Grand Central.

The idea behind zone theory is that, since all that matters is a rapid connection to city center, trains should make as few stops as possible. Instead of trying to run frequently, it’s sufficient to run every 20 or 30 minutes, and then once a train fills with seats it should run express. This is accompanied by a view that longer-haul commuters are more important because they pay higher fares, and therefore their trips should not be slowed by the addition of stops closer in.

It’s important to note that what zone theory replaced was not an S-Bahn-style schedule in which all trains make all stops, and if there’s more demand in the inner area than the outer area then some trains should short-turn at a major station in the middle. American railroads had accumulated a cruft of timetables; Eisele goes over how haphazard the traditional schedules were, with short but irregular rush-hour intervals as some trains skipped some stations, never in any systematic way.

The first paper goes over various implementation details. For example, ideally a major station should be the innermost station within its zone, to guarantee passengers there a nonstop trip to city center. Moreover, considerable attention goes to fare collection: fares are realigned away from a purely distance-based system to one in which all stations in a zone have the same fare to city center, simplifying the conductors’ job. The followup paper speaks of the success of this realignment in reducing fare collection mistakes.

The failure of zone theory

We can see today that zone theory is a complete failure. Trains do not meaningfully serve anyone except 9-to-5 suburban commuters to the city, a class that is steadily shrinking due to job sprawl and a change in middle-class working hours. Ridership is horrendous: all three New York commuter railroads combined have less ridership than the Munich S-Bahn, a single-trunk, seven-branch system in a metropolitan area of 3 million. Metro-North would brag about having an 80% market share among rush hour commuters from its suburban shed to Manhattan, but that only amounts to about 90 million annual riders. In contrast, the modal split of rail at major suburban job centers, even ones that are adjacent to the train station like White Plains and Stamford, is single-digit percent – and Metro-North is the least bad of New York’s three railroads in this category.

Even on the original idea of providing fast service from the suburbs to city center, zone theory is a failure. The timetables are not robust to small disturbances, and once the line gets busy enough, the schedules have to be padded considerably. I do not have precise present-day speed zones for Metro-North, but I do have them for the LIRR courtesy of Patrick O’Hara, and LIRR Main Line service is padded 30% over the technical travel time of present-day equipment on present-day tracks. A textbook I have recently read about scheduling best practices cites a range of different padding factors, all single-digit percent; Switzerland uses 7%, on a complex, interlined network where reliability matters above all other concerns. With 30% padding, the LIRR’s nonstop trains between Ronkonkoma and Penn Station, a distance of 80 km, take about as long as local trains would with 7% padding.

Eisele is right in the papers when he complains about the institutional inertia leading to haphazard schedules. But his solution was destructive, especially in contrast with contemporary advances in scheduling in Europe, which implemented the all-day clockface schedule, starting with Spoorslag ’70 and then the Munich S-Bahn takt in 1972.

Zone theory and reliability

The first paper claims as self-evident that zonal timetables are reliable. The argument offered is that if there is a short delay, it only affects trains within that zone, and thus only affects the stations within the zone and does not propagate further. There is no attempt at modeling this, just claims based on common sense – and transport is a field where intuition often fails and scientific analysis is required.

The problem is that zone theory does not actually make trains in different zones independent of one another. The second paper has a sample timetable on PDF-p. 4 for the evening rush hour, and this can also be reversed for the morning. In the morning, trains from outer zones arrive in city center just after trains from inner zones; in the afternoon, trains serving outer zone stations depart city center first, always with a gap of just a few minutes between successive trains. In the morning, a delay in a suburban zone means that the trains in the zones behind it are delayed as well, because otherwise they would clash and arrive city center at literally the same minute, which is impossible.

This isn’t purely an artifact of short headways between running trains. Subway systems routinely have to deal with this issue. The key is that on a subway system, trains do not have much of their own identity; if a train is delayed, the next train can perfectly substitute for it, and cascading delays just mean that trains run slightly slower and (because the equipment pool is fixed) are slightly more crowded. The principle that individual suburban stations should only be served every 20-30 minutes means no such substitution is possible. S-Bahn trains are not as interchangeable as subway trains, which is why they cannot run as frequently, but they still manage to run every 2-3 minutes with 7% padding, even if they can’t reach the limit values of a train very roughly 1.5 minutes achieved by some big city subways.

Eisele did not think this through and therefore made an assertion based on intuition that failed: reliability did not improve, and with long-term deterioration of speed and lack of reduction in operating expenses, the express timetables at this point are slower than an all-stops S-Bahn would be.

Consultants and Railroaders Turn New Haven Line Investment Into Shelf Art

The state of Connecticut announced that a new report concerning investment in the New Haven Line is out. The report is damning to most involved, chief of all the Connecticut Department of Transportation for having such poor maintenance practices and high construction costs, and secondarily consultant AECOM for not finding more efficient construction methods and operating patterns, even though many readily exist in Europe.

What started out as an ambitious 30-30-30 proposal to reduce the New York-New Haven trip time to an hour, which is feasible without construction outside the right-of-way, turned into an $8-10 billion proposal to reduce trip times from today’s 2 hours by 25 minutes by 2035. This is shelf art: the costs are high enough and the benefits low enough that it’s unlikely the report will lead to any actionable improvement, and will thus adorn the shelves of CTDOT, AECOM, and the governor’s office. It goes without saying that people should be losing their jobs over this, especially CTDOT managers, who have a track record of ignorance and incuriosity. Instead of a consultant-driven process with few in-house planners, who aren’t even good at their jobs, CTDOT should staff up in-house, hiring people with a track record of success, which does not exist in the United States and thus requires reaching out to European, Japanese, and Korean agencies.

Maintenance costs and the state of good repair racket

I have a video I uploaded just before the report came out, explaining why the state of good repair (SOGR) concept has, since the late 1990s, been a racket permitting agencies to spend vast sums of money with nothing to show for it. The report inadvertently confirms this. The New Haven Line is four-track, but since the late 1990s it has never had all four tracks in service at the same time, as maintenance is done during the daytime with flagging rules slowing down the trains. Despite decades of work, the backlog does not shrink, and the slow zones are never removed, only replaced (see PDF-p. 7 of the report). The report in fact states (PDF-p. 8),

To accommodate regular maintenance as well as state-of-good-repair and normal replacement improvements, much of the four-track NHL typically operates with only three tracks.

Moreover, on PDF-p. 26, the overall renewal costs are stated as $700-900 million a year in the 2017-21 period. This includes rolling stock replacement, but the share of that is small, as it only includes 66 new M8 cars, a less than second-order item. It also includes track upgrades for CTRail, a program to run trains up to Hartford and Springfield, but those tracks preexist and renewal costs there are not too high. In effect, CTDOT is spending around $700 million annually on a system that, within the state, includes 385 single-track-km for Metro-North service and another 288 single-track-km on lines owned by Amtrak.

This is an insane renewal cost. In Germany, the Hanover-Würzburg NBS cost 640 million euros to do 30-year track renewal on, over a segment of 532 single-track-km – and the line is overall about 30% in tunnel. This includes new rails, concrete ties, and switches. The entire work is a 4-year project done in a few tranches of a few months each to limit the slowdowns, which are around 40 minutes, punctuated by periods of full service. In other words, CTDOT is likely spending more annually per track-km on a never-ending renewal program than DB is on a one-time program to be done once per generation.

A competent CTDOT would self-abnegate and become German (or Japanese, Spanish, French, Italian, etc.). It could for a few hundred million dollars renew the entirety of the New Haven Line and its branches, with track geometry machines setting the tracks to be fully superelevated and setting the ballast grade so as to improve drainage. With turnout replacement, all speed limits not coming from right-of-way geometry could be lifted, with the possible exception of some light limits on the movable bridges. With a rebuild of the Grand Central ladder tracks and turnouts for perhaps $250,000 per switch (see e.g. Neustadt switches), trains could do New York-New Haven in about 1:03 making Amtrak stops and 1:27 making all present-day local stops from Stamford east.

Infrastructure-schedule integration

The incompetence of CTDOT and its consultants is not limited to capital planning. Operations are lacking as well. The best industry practice, coming from Switzerland, is to integrate the timetable with infrastructure and rolling stock planning. This is not done in this case.

On the contrary: the report recommends buying expensive dual-mode diesel locomotives for through-service from the unelectrified branches instead of electrifying them, which could be done for maybe $150 million (the Danbury Branch was once electrified and still has masts, but no wires). The lifecycle costs of electric trains are half those of diesel trains, and this is especially important when there is a long electrified trunk line with branches coming out of it. Dual-mode locomotives are a pantomime of low electrification operating costs, since they have high acquisition costs and poor performance even in electric mode as they are not multiple-units. Without electrification, the best long-term recommendation is to shut down service on these two branches, in light of high maintenance and operating costs.

The choice of coaches is equally bad. The report looks at bilevels, which are a bad idea in general, but then adds to the badness by proposing expensive catenary modifications (PDF-p. 35). In fact, bilevel European trains exist that clear the lowest bridge, such as the KISS, and those are legal on American tracks now, even if Metro-North is unaware.

The schedule pattern is erratic as well. Penn Station Access will soon permit service to both Grand Central and Penn Station. And yet, there is no attempt to have a clean schedule to both. There is no thought given to timed transfers at New Rochelle, connecting local and express trains going east with trains to Grand Central and Penn Station going west, in whichever cross-platform pattern is preferred.

The express patterns proposed are especially bad. The proposal for through-running to Philadelphia and Harrisburg (“NYX”) is neat, but it’s so poorly integrated with everything else it might as well not exist. Schedules are quoted in trains per day, for the NYX option and the GCX one to Grand Central, and in neither case do they run as frequently as hourly (PDF-p. 26). There is no specific schedule to the minute that the interested passenger may look at, nor any attempt at an off-peak clockface pattern.

Throw it in the trash

The desired rail investment plan for Connecticut, setting aside high-speed rail, is full electrification, plus track renewal to permit the elimination of non-geometric speed limits. It should cost around $1 billion one-time; the movable bridge replacements should be postponed as they are nice to have but not necessary, their proposed budgets are excessive, and some of their engineering depends on whether high-speed rail is built. The works on the New Haven Line are doable in a year or not much more – the four-year timeline on Hanover-Würzburg is intended to space out the flagging delays, but the existing New Haven Line is already on a permanent flagging delay. The trains should be entirely EMUs, initially the existing and under-order M8 fleet, and eventually new lightweight single-level trains. The schedule should have very few patterns, similar to today’s off-peak local and express trains with some of one (or both) pattern diverting to Penn Station; the express commuter trains should take around 1:30 and intercity trains perhaps 1:05. This is a straightforward project.

Instead, AECOM produced a proposal that costs 10 times as much, takes 10 times as long, and produces half the time savings. Throw it in the trash. It is bad, and the retired and working agency executives who are responsible for all of the underlying operating and capital assumptions should be dismissed for incompetence. The people who worked on the report and their sources who misinformed them should be ashamed for producing such a shoddy plan. Even mid-level planners in much of Europe could design a far better project, leaving the most experienced and senior engineers for truly difficult projects such as high-speed rail.

Randal O’Toole Gets High-Speed Rail Wrong

Now that there’s decent chance of US investment in rail, Randal O’Toole is resurrecting his takes from the early Obama era, warning that high-speed rail is a multi-trillion dollar money sink. It’s not a good analysis, and in particular it gets the reality of European and Asian high-speed rail systems wrong. It displays lack of familiarity with rail practice and rail politics, to the point that most nontrivial assertions about rail in Europe and Asia are incorrect.

More broadly, the way O’Toole gets rail investment here wrong comes from making unexamined American assumptions and substituting them for a European or Japanese reality regarding rail as well as rail politics. If the US can’t do it, he thinks other countries can’t. Unfortunately, he’s even unfamiliar with recent work done on American costs, when he compares the Interstate system positively with recent high-speed rail lines.

High-Speed Rail Profitability: France

I’m currently working on building a database similar to our urban rail costs for high-speed rail. Between this and previous iterations of analyzing the TGV, I’ve been reading a lot of internal French reports about its system. Thankfully, France makes available very good public information about the costs and technical specifications of its system. It helps that I read French, but the gap between what’s available for France and Belgium (see for example line schemas) is vast. This provides crucial background that O’Toole is missing.

The most important thing to understand is that the TGV network is profitable. The Spinetta report on the fiscal losses of SNCF makes it clear, starting on p. 60, that the TGV network is profitable, and recommends favoring its development over the money-losing legacy networks, especially the branch lines. The report even calls for closing weak branch lines with only a few trains a day, which I called the Spinetta Axe at the time, in analogy with the Beeching Axe. Due to public outcry the state rejected the cuts and only implemented the organizational changes promoted by the report.

Moreover, all lines are very profitable excluding the cost of fixed capital. The Spinetta report’s TGV section says that operating costs average €0.06/seat-km, which is around 0.085€/p-km, despite overstaffing of conductors (8 per conventional 400-car TGV) and extensive travel on legacy track at low speed and higher per-km labor costs. Average TGV fare revenue per an ARAFER report from 2016 is 0.10€/p-km – compare p-km on p. 15 and revenue on p. 26. This is typical for Europe – RENFE and DB charge similar fares, and the nominal fares seem to have been flat over the last decade.

What’s dicier is cost of capital. In all other European countries for which I’m aware of the process, all of which are Northern rather than Southern, this is done with benefit-cost analysis with a fixed behind-the-scenes discount rate. France, in my view wisely, rates lines by their financial and social rates of return instead. A 2014 report about the Bordeaux-Toulouse LGV, recently given the go-ahead for 7.5 billion €, warns that the profitability of LGVs decreases as the system is built out: the LGV Sud-Est returned 15% to SNCF’s finances and 30% to French society (including rider consumer surplus), but subsequent lines only returned 4-7% to SNCF’s finances, and Bordeaux-Toulouse is likely to return less, 6% including social benefits per the study and at this point slightly less since the study assumed it would cost slightly less than the current budget.

The general theme in the French discourse on trains is that the TGV network is an obvious success. There absolutely is criticism, which focuses on the following issues:

  • Regional rail, that is not intercity rail, is underdeveloped in France outside Paris. The ridership of TER networks is pitiful in comparison with German-speaking and Nordic metropolitan areas of comparable size. For example, sourced to a dead link, Wikipedia claims 64,300 TER PACA trips per day, comprising the metropolitan areas of Marseille (1.8 million), Nice (1), Toulon (0.6), and Avignon (0.5); in Helsinki (1.5) alone, there are 200,000 daily commuter rail trips. But this isn’t really about high-speed rail, since TER planning and subsidies are devolved to regional governments, and not to SNCF.
  • SNCF has contentious labor relations. In the early 2010s, the unions went on a wave of strikes and got wage concessions that led to the evaporation of SNCF’s 600 million €/year primary surplus. The railway unions in France (“cheminots”) are unpopular, and Macron has been able to pass reforms to SNCF’s governance over their strikes and objections.
  • Future LGVs are not as strong as past ones. Real costs in France are rising, and the network already links Paris with all major secondary cities in airplane-competitive time save Nice. Interprovincial links on the network are weak, despite the construction of the LGV Rhin-Rhône, and nothing like the Deutschlandtakt is on the horizon enabling everywhere-to-everywhere travel.
  • SNCF thinks like an airline and not like a railroad. It separates passengers into different buckets as airlines do, has many executives with airline background (and Spinetta is ex-Air France), thinks passengers do not ride trains for longer than 3 hours even though at 4 hours the modal split with air is still better than 50-50, and has poor integration between the TGV and legacy rail.
  • SNCF still has a lot of accumulated debt from past operating losses, some predating the TGV and the start of regional subsidies for regional rail. It was hoped that TGV profits could cover them, but they can’t. This mirrors the controversy in Japan in the 1980s, where, in the breakup of JNR into the JRs and their privatization, debt from past operating losses was wiped but not debt from Shinkansen construction (see Privatization Best Practices, PDF-p. 106).

However, saying that the existing network is a failure is the domain of cranks and populists. It is unrecognizable from the discussion of transportation investments in France.

What O’Toole says about high-speed rail

O’Toole’s understanding of internal French (or Spanish, or Japanese) issues is weak. This isn’t surprising – Americans to a good approximation never have good insights on the internal issues of any other country, even when it speaks English. The American political sphere, which includes political thinktanks like Cato, is remarkably ignorant globally, and rather incurious. As a result, what he says about the TGV is based on an Americanized understanding. To wit:

Bus-rail competition

The Northeastern United States has a weak rail network: Amtrak averages vintage 1960s speeds and charges 2-4 times the per-km fare of the TGV. As a result, an ecosystem of private intercity buses has developed, starting with unregulated ones like Fung Wah and, as they were shut down, corporate systems like Megabus and Bolt. O’Toole is fond of these buses, with their lower fares and road-like lack of integration between infrastructure and operations.

And thus, he claims, falsely, that European high-speed rail cannibalized profitable buses. This is unrecognizable from within Europe, where intercity buses were underdeveloped until recently. In France, US-style intercity buses are called Macron buses, because the deregulation that brought them into existence passed in the mid-2010s, when Macron was the economy minister. They complement high-speed rail but do not replace it, because trains get me from Paris to the German border in 1:45 and buses don’t.

To be fair, TGV ridership has been stagnant in the last few years. But this stagnation goes back to the financial crisis, and if anything ridership picked up starting 2017 with the opening of the LGV Sud-Europe-Atlantique. So the buses are not even outcompeting the trains – they thrive in the gaps between them, just as historically they did on international routes, where rail fares are considerably higher and ridership lower.

High-speed rail construction costs

O’Toole looks at the most expensive few lines possible:

Britain’s 345‐​mile London–Scotland HS2 high‐​speed rail line was originally projected to cost £32.7 billion (about $123 million per mile) and is currently expected to cost £106 billion ($400 million per mile).

International comparisons of high-speed rail costs exist, and Britain’s costs are by far the worst. For example, a 2013 Australian comparison looking at the prospects for such a system in Australia finds that High-Speed 1/CTRL, the line linking the Channel Tunnel with London, cost A$134 million/km, and the second costliest line in the dataset was thee 94% tunneled Bologna-Florence line, at A$95 million/km.

French costs up until the LGV Bordeaux-Toulouse stood around $25-30 million per km in 2021 dollars, net of tunnels. German costs are similar, but German lines have far heavier tunneling than France, a range of 26-51% in tunnel compared with 0-6% in France. One reason is topography. But another is that Germany prefers mixed-use passenger-freight lines, which forces higher construction costs as freight requires gentler grades and, since superelevation must be lower, wider curves; France, like Japan and China, builds dedicated passenger lines, and, unlike Japan or China, keeps them largely at-grade to reduce costs.

O’Toole says, without more references, that it would cost $3-4 trillion to build a US-wide high-speed rail network. But the official Obama-era crayon, at 20,000 km, would be $500 billion at tunnel-free European costs, or maybe $600 billion with 5% tunneling, mostly in difficult places like California and across the Appalachians.

Freeway costs

O’Toole proposes more freeways, and says that to build the Interstate system today would cost $530 billion so it’s better than high-speed rail. Here is where his lack of knowledge of the most recent literature on infrastructure costs is a serious drag on his analysis: Brooks-Liscow establish that there was a large real increase in Interstate cost throughout the life of the program, so a budget that’s really a mixture of cheaper early-1960s construction and more expensive construction in the 1970s is not applicable today.

The same issue affects rail costs: the LGV Sud-Est cost, in today’s money, around $8 million/km, which cost would never recur. Brooks-Liscow explain this by greater surplus extraction from citizen voice groups, which demanded detours and route compromises raising costs. This appears true not just diachronically within the US but also synchronically across countries: so far, the low-cost subways we have investigated are all in states with bureaucratic rather than adversarial legalism, while medium-cost Germany is more mixed. Politicized demands leading to more tunneling are well-documented within Germany – the Berlin-Munich line was built through a topographically harder alignment in order to serve Erfurt, at Thuringia’s behest.

So no, today costs from the 1960s are not relevant. Today, urban motorway extensions cost double-digit millions of dollars per lane-km, sometimes more. The I-5 improvement project in Los Angeles is $1.9 billion for I-5 South, a distance of 11 km, adding two lanes (one HOV, one mixed traffic) in each direction. It’s possible to go lower than this – in Madrid this budget would buy a longer 6-lane tunnel – but then in Madrid the construction costs of rail are even lower, for both metros and high-speed lines.

The discourse on profits

In contrast with the basic picture I outlined for the TGV, French media and researchers often point out threats to rail profitability. This can easily be taken to mean that the TGV is unprofitable, and if one has an American mindset, then it’s especially easy to think this. If SNCF officials say that 20% of TGVs lose money, then surely they must be hiding something and the figure is much higher, right? Likewise, if Spinetta says that the TGV network is profitable but not all trains are, then surely the situation is even worse, right?

But no. This is an Americanized interpretation of the debate. In the US, Amtrak is under constant pressure to show book profits, and its very existence is threatened, often by people who cite O’Toole and other libertarians. Thus, as a survival strategy, Amtrak pretends it is more profitable than it really is.

This has no bearing on the behavior of railroads elsewhere, though. SNCF is not so threatened. The biggest threat from the perspective of SNCF management is union demands for higher wages, and therefore, its incentive is to cry poverty. Nobody in France takes out yardsticks of farebox recovery ratios, and therefore, nobody needs to orient their communications around what would satisfy American libertarians.


Within the European high-speed rail research community, the energy efficiency of high-speed rail is well-understood, and many studies look at real-world examples, for example the metastudy of Hasegawa-Nicholson-Roberts-Schmid. In fact, it’s understood that high-speed rail has lower energy consumption than conventional rail. For example, here is García Álvarez’s paper on the subject. This is counterintuitive, because higher speeds should surely lead to higher energy consumption, as Hasegawa et al demonstrate – but high-speed lines run at a uniform speed of 200 or 250 or 300 or 350 km/h, whereas legacy rail has many cycles of acceleration and deceleration. At speeds of up to about 200 km/h, nearly all electricity consumption is in acceleration and not maintaining constant speed, and even at 300 km/h, a late-model high-speed train consumes only above one third of its maximum power maintaining speed.

Instead of this literature, O’Toole picks out the fact that all else being equal energy consumption rises in speed, which it is not equal. Garcia in fact points out that higher speeds are better for the environment due to better competition with air, in line with environmental consensus that trains are far superior on well-to-wheels emissions to cars and planes. Worse, O’Toole is citing Chester-Horvath’s lifecycle analysis, which is not favorable to California High-Speed Rail’s energy efficiency. The only problem is that this paper’s analysis relies on a unit conversion error between BTUs and kWh, pointed out by Clem Tillier. The paper was eventually corrected, and with the correct figures, high-speed rail looks healthy.

Competition with cars and planes

Where high-speed rail exists, and the distance is within a well-understood range of around 300-800 km, it dominates travel. A 2004 report by Steer Davies Gleave has some profiles of what were then the world’s main networks. For Japan, it includes a graphic from 1998 on PDF-p. 120 of modal splits by distance. In the 500-700 km bucket, a slight majority of trips all over Japan are made by rail; this is because Tokyo-Osaka is within that range, and due to those cities’ size this city pair dominates pairs where rail is weaker, especially inter-island ones. In the 300-500 km bucket more people drive, but the Shinkansen is stronger than this on the Tokyo-Nagoya pair, it’s just that 300-500 includes many more peripheral links with no high-speed rail service. It goes without saying that high-speed rail does not get any ridership where it does not exist.

In France, this was also studied for the LGV PACA. On p. 14, the presentation lists modal splits as of 2009. Paris-Toulon, a city pair where the TGV takes around 4 hours, has an outright majority for the TGV, with 54% of the market, compared with 12% for air and 34% for driving. Paris-Cannes is 34% and Paris-Nice is 30%, both figures on the high side for their 5:00-5:30 train trips. Lyon-Nice, a 3:30 trip with awful frequency thanks to SNCF’s poor interprovincial service, still has a 25% market share for the TGV.

In general, competition with cars is understudied. Competition with planes is much more prominent in the literature, with plenty of reports on air-rail modal splits by train trip length. JR East, Central (PDF-p. 4), and West all report such market shares, omitting road transport. Many European analyses appeared in the 2000s, for example by Steer Davies Gleave again in 2006, but the links have rotted and Eurostat’s link is corrupt.

O’Toole misunderstands this literature. He lumps all air and road links, even on markets where rail is weak, sometimes for geographical factors such as mountains or islands, sometimes for fixable institutional ones like European borders. In fact, at least measured in greenhouse gas emission and not ridership, all air travel growth in Europe since 1990 has been international. International high-speed rail exists in Europe but charges higher fares and the infrastructure for it is often not built, with slowdowns in border zones. This is a good argument for completing the international network in Europe and a terrible one against building any network at all.


Even at the level of basic topography, O’Toole makes elementary errors. He discusses the Tokaido Shinkansen, pointing out its factor-of-2 cost overrun. But its absolute costs were not high, which he characterizes as,

The Tokyo–Osaka high‐​speed rail line supposedly made money, but it was built across fairly flat territory

So, first of all, the “supposedly” bit is painful given how much JR Central prints money. But “fairly flat territory” is equally bad. Japan’s mountainous topography is not an obscure fact. It’s visible from satellite image. Per Japanese Wikipedia, 13% of the route is in tunnel, more than California High-Speed Rail.

The United States can and should do better

The report is on stronger grounds when criticizing specifics of Amtrak and California High-Speed Rail. American rail construction is just bad. However, this is not because rail is bad; it’s because the United States is bad.

And there’s the rub. Americans in politics can’t tell themselves that another country does something better than the US does. If it’s in other countries and the US can’t do it, it must be, as O’Toole calls rail, obsolete. This is especially endemic to libertarians, who are intellectually detached from their European right-liberal counterparts (Dutch VVD, German FDP, etc.) even more than the American center-left is from social democrats here and the right is from the mainline and extreme right here.

So here, faced with not too hard to find evidence that high-speed rail is profitable in Europe and Asia, and in fact intercity rail is profitable here in general (direct subsidies are forbidden by EU law unless the line is classified as regional), unlike in the United States, O’Toole makes up reasons why trains here are unprofitable or unsuccessful. He says things that are not so much wrong as unrecognizable, regarding topography, buses, construction costs, debt, the state of the TGV debate, or greenhouse gas emissions.

O’Toole is aware of our transit costs comparison. I imagine he’s also aware of high-speed rail cost comparisons, which exist in the literature – if he’s not, it’s because he doesn’t want to be so aware. And yet, no matter how loudly the evidence screams “the United States needs to become more like France, Germany, Japan, Spain, etc.,” American libertarians always find excuses why this is bad or unnecessary. And then, when it comes to expanding freeways, suddenly the cost concerns go out the door and they use unrealistically low cost figures.

But figuring out why the US is bad requires way deeper dives. It requires delving into the field and understanding how procurement is done differently, what is wrong with Amtrak, what is wrong with the California High-Speed Rail Authority, how engineering is done in low- and medium-cost countries, various tradeoffs for planning lead time, and so on. It requires turning into the kind of expert that libertarians have spent the last 60 years theorizing why they need not listen to (“public choice”). And it requires a lot of knowledge of internal affairs of successful examples, none of which is in an English-speaking country. So it’s easier to call this obsolete just because incurious Americans can’t do it.

High-Speed Rail and the Pacific Northwest

The Pacific Northwest seems like the perfect region for high-speed rail: its cities form a neat line from Vancouver to Portland and points south, grow at high rates with transit-oriented development, and have sizable employment cores around the train station. And yet, when I generated my high-speed rail maps, I could only include it as a marginal case, and even that inclusion was charitable:

(Full-size image is available here.)

There’s been a lot of criticism over why I’m including Atlanta-Jacksonville but not Vancouver-Seattle-Portland, and I’d like to explain why the model says this.


The population density in the Western United States is very low. What this means in practice is that cities are far apart – the best example is Denver, a large metropolitan area that is 537 km from the nearest million-plus metro area (Albuquerque). A high-speed line can connect two cities, maybe three, but will not form the multi-city trunk that one sees in Germany or Italy, or even Spain or France. Lines can still make sense if they serve enormous cities like Los Angeles, but otherwise there just isn’t much.

This relates to Metcalfe’s law of network effects. In a dense region, the 500-800 km radius around a city will have so many other cities that network effects are obtained as the system grows. Even Florida, which isn’t dense by European standards, has cities placed closely enough that a medium-size system can connect Miami, Orlando, Tampa, and Jacksonville, and then with a 500 km extension reach Atlanta. The I-85 corridor can likewise accrete cities along the way between Washington and Atlanta and get decent ridership.

In the Pacific Northwest, any intercity infrastructure has to live off Vancouver, Seattle, and Portland – that’s it. Spokane is small, orthogonal to the main line, and separated by mountains; Salem and Eugene are small and Salem is technically in the Portland combined statistical area; California’s cities are very far away and separated by mountains that would take a base tunnel to cross at speed. And Seattle is just not that big – the CSA has 5 million people, about the same as Berlin, which has within 530 km every German metropolitan area.

The model

The model thinks that with Vancouver (2.6)-Seattle (5) at 220 km and Seattle-Portland (3.2) at 280, ridership is as follows, in millions of passengers per year in both directions combined:

City S\City NVancouverSeattle

In operating profits in millions of dollars per year, this is,

City S\City NVancouverSeattle

This is $135 million a year. It’s actually more optimistic than the official WSDOT study, which thinks the line can’t make an operating profit at all, due to an error in converting between miles and kilometers. The WSDOT study also thinks the cost of the system is $24-42 billion, which is very high. Nonetheless, a normal cost for Vancouver-Portland HSR is on the order of $15 billion, a bit higher than the norm because of the need for some tunnels and some constrained urban construction through I-5 in Seattle.

It isn’t even close. The financial ROI is 0.9%, which is below the rate of return for government debt in the very long run. Even with social benefits included, the rate is very low, maybe 2.5% – and once social benefits come into play, the value of capital rises because competing government investment priorities have social benefits too so it’s best to use the private-sector cost of capital, which is 4-5%.

This exercise showcases the value of density to intercity rail networks. You don’t need Dutch density, but Western US density is too low – the network effects are too weak except in and around California. It would be mad to build Atlanta-Jacksonville as a high-speed rail segment on its own, but once the Florida network and the I-85 network preexist, justified by their internal ridership and by the Piedmont’s connections to the Northeast, connecting Atlanta and Jacksonville becomes valuable.


The one saving grace of the Pacific Northwest is growth. That’s why it’s even included on the map. Lines in the 1.5-1.8% ROI region are not depicted at all, namely Houston-New Orleans and Dallas-Oklahoma City-Kansas City-St. Louis, both discounted because none of the cities connected has local public transportation or a strong city center. The Pacific Northwest is not discounted, and also benefits from strong growth at all ends.

The gravity model says that ridership is proportional to the 0.8th power of the population of each city connected. To get from 0.9% to 2% requires a factor of 2.2 growth, which requires each city to grow by a factor of 2.2^0.625 = 1.65.

Is such growth plausible? Yes, in the long run. In 2006-16, Metro Vancouver grew 16%; in 2010-9, the core three-county Seattle metro area (not CSA) grew 16% as well, and the core Portland metro area (again, not CSA) grew 12%. At 16% growth per decade, the populations will rise by the required factor in 34 years, so building for the 20-year horizon and then relying on ridership growth in the 2050s and 60s isn’t bad. But then that has a lot of risk embedded in it – the growth of Seattle is focused on two companies in a similar industry, and that of Vancouver is to a large extent the same industry too.

Moreover, the region’s relative YIMBYism can turn into NIMBYism fast. Metro Vancouver’s housing growth is healthy, but the region is fast running out of developable non-residential areas closer in than Surrey, which means it will need to replace single-family housing on the West Side with apartment buildings, which it hasn’t done so far. Growing construction costs are also threatening the ability of both Vancouver and Seattle to feed commuters into their central business districts by rail – Seattle may have built U-Link for costs that exist in Germany, but the Ballard/West Seattle line is $650 million/km and mostly above-ground, and the Broadway subway in Vancouver, while only C$500 million/km, is still on the expensive side by non-Anglo standards. It’s useful to plan around future growth and safeguard the line, but not to build it just on the promise of future growth, not at this stage.

Sorry Eno, the US Really Has a Construction Cost Premium

There’s a study by Eno looking at urban rail construction costs, comparing the US to Europe. When it came out last month I was asked to post about it, and after some Patreon polling in which other posts ranked ahead, here it goes. In short: the study has some interesting analysis of the American cost premium, but suffers from some shortcomings, particularly with the comprehensiveness of the non-American data. Moreover, while most of the analysis in the body of the study is solid, the executive summary-level analysis is incorrect. Streetsblog got a quote from Eno saying there is no US premium, and on a panel at Tri-State a week ago T4A’s Beth Osborne cited the same study to say that the US isn’t so bad by European standards, which is false, and does not follow from the analysis. The reality is that the American cost premium is real and large – larger than Eno thinks, and in particular much larger than the senior managers at Eno who have been feeding these false quotes to the press think.

What’s the study?

Like our research group at Marron, Eno is comparing American urban rail construction costs per kilometer with other projects around the world. Three key differences are notable:

  1. Eno looks at light rail and not just rapid transit. We have included a smattering of projects that are called light rail but are predominantly rapid transit, such as Stadtbahns, the Green Line Extension in Boston, and surface portions of some regional rail lines (e.g. in Turkey), but the vast majority of our database is full rapid transit, mostly underground and not elevated. This means that Eno has a mostly complete database for American urban rail, which is by construction length mostly light rail and not subways, whereas we have gaps in the United States.
  2. Eno only compares the United States with other Western countries, on the grounds that they are the most similar. There is a fair amount of Canada in their database, one Australian line, and a lot of Europe, but no high-income Asia at all. Nor do they look at developing countries, or even upper-middle-income ones like Turkey.
  3. Eno’s database in Europe is incomplete. In particular, it looks by country, including lines in Britain, Spain, Italy, Germany, Austria, the Netherlands, and France, but even there it has coverage gaps, and there is no Switzerland, little Scandinavia (in particular, no ongoing Stockholm subway expansion), and no Eastern Europe.

The analysis is similar to ours, i.e. they look at average costs per km controlling for how much of the line is underground. They include one additional unit of analysis that we don’t, which is station spacing; ex ante one expects closer station spacing to correlate with higher costs, since stations are a significant chunk of the cost and this is especially notable for very expensive projects.

The main finding in the Eno study is that the US has a significant cost premium over Europe and Canada. The key here is figure 5 on takeaway 4. All costs are in millions of PPP dollars per kilometer.

Tunnel proportionMedian US costMedian non-US cost

However, the study lowballs the US premium in two distinct ways: poor regression use, and upward bias of non-US data.

Regression and costs

The quotes saying the US has no cost premium over Europe come from takeaways 2 and 3. Those are regression analyses comparing cost per km to the tunnel proportion (takeaway 3) or at-grade proportion (takeaway 2). There are two separate regression lines for each of the two takeaways, one looking at US projects and one at non-US ones. In both cases, the American regression line is well over the European-and-Canadian line for tunneled projects but the lines intersect roughly when the line goes to 0% underground. This leads to the conclusion that the US has no premium over Europe for light rail projects. Moreover, because the US has outliers in New York, the study concludes that there is no US premium outside New York. Unfortunately, these conclusions are both false.

The reason the regression lines intersect is that regression is a linear technique. The best fit line for the US construction cost per km relative to tunnel proportion has a y-intercept that is similar to the best fit line for Europe. However, visual inspection of the scattergram in takeaway 3 shows that at 0% underground, most US projects are somewhat more expensive than most European projects; this is confirmed in takeaway 4. All this means that the US has an unusually large premium for tunneled projects, driven by the fact that the highest-cost part of the US, New York, builds fully-underground subways and not els or light rail. If instead of Second Avenue Subway and the 7 extension New York had built high-cost els, for example the plans for a PATH extension to Newark Airport, then a regression line would show a large US premium for elevated projects but not so much for tunnels.

I tag this post “good/interesting studies” and not just “shoddy studies” because the inclusion of takeaway 4 makes this clear: there is a US premium for light rail, it’s just smaller than for subways, and then regression analysis can falsely make this premium disappear. This is an error, but an interesting one, and I urge people who use statistics and data science to study the difference between takeaways 2 and 3 and takeaway 4 carefully, to avoid making the same error in their own work.

Upward bias

Eno has a link to its dataset, from which one can see which projects are included. It’s notable that Eno is comprehensive within the United States, but not in Europe. Unfortunately, this introduces a bias into the data, because it’s easier to find information about expensive projects than about cheap ones. Big projects are covered in the media, especially if there are cost overruns to report. There is also a big-city premium because it’s more complicated to build line 14 of a metro system than to build line 1, and this likewise biases incomplete data because it’s easier to find what goes on in Paris than to find what goes on in a sleepy provincial town like Besançon. Yonah Freemark thankfully has good coverage of France and includes low-cost Besançon, but Eno does not – its French light rail database is heavy on Paris and has big gaps in the provinces. French Wikipedia in fact has a list, and all of the listed systems, which are provincial, have lower costs than Paris.

There is also no coverage of German tramways; we don’t have such coverage either, since there are many small projects and they’re in small cities like Bielefeld, but my understanding is that they are not very expensive. Traditionally German rail advocates held the cost of a tramway to be €10 million/km, which is clearly too low for the 2010s, but it should lower the median cost compared to the Paris-heavy, Britain-heavy Eno database.

The Subway is Probably not Why New York is a Disaster Zone

New York is the capital of the coronavirus pandemic, with around 110,000 confirmed cases and 10,000 confirmed deaths citywide, and perhaps the same number across its suburbs. There must be many reasons why this is so; one possibility that people have raised is infection from crowded subways, so far without much evidence. Two days ago, MIT economist Jeffrey Harris wrote a paper claiming that the subways did in fact seed the Covid-19 epidemic in New York, but the paper cites no evidence. Sadly, some people have been citing the paper as a serious argument, which it isn’t; the purpose of this post is to explain what is wrong with the paper.

New York and other subways

In multiple other countries, one cannot see the transit cities in the virus infection rates. In Germany the rates in the largest cities are collectively the same as in the rest of the country. In South Korea, the infection is centered on Daegu; Seoul’s density and high transit usage are compatible with an infection rate of about 700 in a city of 9.5 million, about 1.5 orders of magnitude less per capita than in most Western countries and 2.5 orders of magnitude less than in New York. In Taipei, the MRT remains crowded, with weekday ridership in February and March down by 15-16%. In Italy, car usage is high outside a handful of very large cities like Milan, and Milan’s infection rate isn’t high by the standards of the rest of Lombardy.

However, rest-of-world evidence does not mean that the New York City Subway is safe. The Taipei MRT has mandatory mask usage and very frequent cleaning. German U- and S-Bahn networks are a lot dirtier than anything I’ve seen in Asia, but much cleaner than anything I’ve seen in New York, and also have much less peak crowding than New York. New York uniquely has turnstiles requiring pushing with one’s hands or bodies, and the only other city I know of with such fare barriers is Paris, whose infection rates are far below New York’s but still high by French standards.

So the question is not whether rapid transit systems are inherently unsafe for riders, which they are not. It’s whether New York, with all of its repeated failings killing tens of workers from exposure to the virus, has an unsafe rapid transit system. Nonetheless, the answer appears to be negative: no evidence exists that the subway is leading to higher infection rates, and the paper does not introduce any.

What’s in the paper?

A lot of rhetoric and a lot of lampshade hanging about the lack of natural experiments.

But when it comes to hard evidence, the paper makes two quantitative claims. The first is in figure 3: Manhattan had both the least increase in infections in the 3/13-4/7 period, equivalent to a doubling period of 20 days whereas the other boroughs ranged between 9.5 and 14, and also the largest decrease in subway entries in the 3/2-16 period, 65% whereas the other boroughs ranged between 33% and 56%.

The second is a series of maps showing per capita infection levels by zip code, similar to the one here. The paper also overlays a partial subway map and asserts that the map shows that there is correlation of infection rates along specific subway routes, for example the 7, as people spread the disease along the line.

I will address the second claim first, regarding line-level analysis, and then the first, regarding the borough-level difference-in-differences analysis; neither is even remotely correct.

Can you see the subway on an infection map?

Here is a static version of the infection map by zip code:

This is cases for 1,000 people – note that my post about Germany looks at rates per 10,000 people, so the range in New York is consistently about an order of magnitude worse than in Germany. The map shows high rates in Eastern Queens, the North Bronx, and Staten Island, hardly places with high public transportation ridership. The rates in Manhattan and the inner parts of Brooklyn are on the low side.

There are no ribbons of red matching any subway line – there are clumps and clusters, as in Southern Brooklyn in Orthodox Jewish neighborhoods, and in Central Queens around Corona and East Elmhurst. There is imperfect but noticeable correlation with income – working-class areas have higher infection rates, perhaps because they have higher rates at which people are required to still show up to work, where they can be infected. East Asian neighborhoods have lower rates, like Flushing and environs, or to some extent Sunset Park; Asians are infected at noticeably lower rates than others in New York and perhaps in the rest of the Western world, perhaps because they took news in China more seriously, began practicing social distancing earlier, and wear masks at higher rates. There are many correlates, none of which looks like it has anything to do with using the public transportation network.

What’s more, the paper is not making any quantitative argument why the graph shows correlation with subway usage. It shows the graph with some lines depicted, often misnamed, for example the Queens Boulevard Line is called Sixth Avenue Local, leading to a discussion about higher infection rates on local trains than on express trains where in fact the F runs express in Queens. But it does not engage in any analysis of rates of subway usage or changes therein, or in infection rates. The reader is supposed to eyeball the graph and immediately agree with the author’s conclusion, where there is no reason to do so.

Manhattan confounders

The claim about Manhattan is the only real quantitative claim in the paper. Unlike the zip code analysis, the borough analysis does make some statistical argument: Manhattan had larger reduction in subway usage than the rest of the city and also a slower infection rate. However, this argument relies on an N of 2. Among the other boroughs, there is no such correlation. The argument is then purely about Manhattan vs. the rest of the city. This is incorrect for so many reasons:

  1. Manhattan is the highest-income borough, with many people who can work from home. If they’re not getting infected, it could be from not commuting as much, but just as well from not getting the virus at work as much.
  2. The Manhattan subway stops are often job centers, so the decline in ridership there reflects a citywide decline. A Manhattanite who stops taking the subway is seen as two fewer turnstile entries in Manhattan, whereas a New Yorker from the rest of the city who does the same is likely to be seen as one fewer Outer Borough entry and one fewer Manhattan entry.
  3. Many Manhattanites left the city to shelter elsewhere, as seen in trash collection data.
  4. Manhattan’s per capita subway usage is probably higher than that of the rest of the city counting discretionary trips, so 65% off the usual ridership in Manhattan may still be higher per capita than 56% off in Brooklyn or 47% in Queens. (But this is false on the level of commuting, where Manhattan, the Bronx, and Brooklyn all have 60% mode share.)

Does the paper have any value?


I have heard people on Twitter claim that correlation is not causation. This argument is too generous to the paper, which has not shown any correlation at all, since the only quantitative point it makes has an N of 2 and plenty of confounders.

For comparison, my analysis of metro construction costs has an effective N of about 40, since different subway  projects in the same country tend to have similar costs with few exceptions (such as New York’s extreme-even-for-America costs), and I consider 40 to be low enough that Eric Goldwyn and I must use qualitative methods and delve deep into several case studies before we can confidently draw conclusions. The paper instead draws strong conclusions, even including detailed ones like the point the paper tries to make about local trains being more dangerous than express trains, from an N of 2; it’s irresponsible.

But what about the workers?

A large and growing number of New York City Transit workers have succumbed to the virus. The current count is close to the citywide death toll, but transportation workers are by definition all healthy enough to be working, whereas citywide (and worldwide) the dead are disproportionately old or have comorbidities like heart disease. Echoing the union’s demands for better protection, Andy Byford had unkind words to say about Governor Andrew Cuomo’s appointees in charge of the system, MTA chair Pat Foye and acting NYCT chair Sarah Feinberg.

However, this is not the same as infection among passengers. The dead include workers who are in close proximity to passengers on crowded vehicles, such as bus drivers, but also ones who are not, such as train operators, maintenance workers, and cleaners. Train cleaners have to remove contaminated trash from the platforms and vehicles without any protective equipment; NYCT not only didn’t supply workers with protective equipment, but also prohibited them from wearing masks on the job even if they’d procured them privately. Contamination at work is not the same as contamination during travel.

So, should people avoid public transportation in New York?

Absolutely not.

If the best attempt to provide evidence that riding the subway is a health hazard in a pandemic is this paper, then that by itself is evidence that there is no health hazard. This is true even given New York City Transit’s current level of dirt, though perhaps not given its pre-crisis peak crowding level. Social distancing is reducing overall travel and this is good, not necessarily because travel is hazardous, but mostly because the destination is often a crowded place with plenty of opportunity for person-to-person infection.

In preparation for going back to normal, the current level of cleanliness is not acceptable. The state should make sure people have access to masks, even if they’re ordinary ones rather than N95 ones, and mandate their usage in crowded places including the subway once they are available. It should invest far more in cleaning public spaces, including the subway, to the highest standards seen in the rich countries of Asia. It should certainly do much more to protect the workers, who face more serious hazards than the riders. But it should not discourage people who are traveling from doing so by train.