Costs Per Rider and Willingness to Build

At the Transit Costs Project, we study the costs of urban rail lines per kilometer. This, with our usual controls, is a rough indicator of how efficient a city’s infrastructure construction program is. However, cost-effectiveness is different from efficiency, and is better measured not by just the cost but also the benefits, which are measured by ridership. Thus, cost-effectiveness is best measured by the cost of each project per rider. So what does it mean to compare different cities by their costs per rider, rather than per kilometer?

The answer is, “not much,” at least not as far as efficiency or good decisionmaking. In fact, while some projects persistently have costs per rider that are too high to be cost-effective (for example, it’s endemic to airport extensions), some cities have costs per rider that are too low, such as Berlin. The issue here is that if municipal or state decisionmaking is healthy, it will build cost-effective lines; if a line under discussion has a very high benefit-cost ratio, or equivalently a very low cost per rider, it means it should have been built long ago, and it speaks poorly to the local decisionmaking that it hasn’t been built already, as is unfortunately the case in Berlin.

Cost per rider and cost-effectiveness

As always, let’s start with Second Avenue Subway Phase 1, the project that launched my interest in construction costs and the Transit Costs Project writ large. The projected daily ridership is 202,000; the line came close to it in 2019, reaching about 163,000 (see data here; I’m taking boardings for the three new stations, adding the growth in ridership at Lexington/63rd, and multiplying by two to account for disembarkings), and would have likely reached projection by now but for the pandemic. The cost, in 2023 dollars, was $6.23 billion, or about $31,000 per projected weekday trip.

For another anchoring number, let’s use the Stockholm Metro’s entire construction from the 1940s to the 1970s. This is useful because in addition to costs per rider, there is a published ex post benefit-cost ratio, thanks to the work of Maria Börjesson. The cost of that system, in 2023 PPP dollars, was $3.7 billion, with a ridership of 1,265,900 per workday in 2019, or $2,900/rider, while the benefit-cost ratio is 6 without taking agglomeration into account, or 8.5 with. This does not mean that the value of a rider is only about $24,900; this was the value in the economic situation of postwar Sweden, a country that was substantially poorer than the Sweden of today. In 2023 PPP terms, Sweden’s GDP per capita in 1965, about midway through the opening of the original system, was about $19,400; today, Sweden is 3.5 times richer, and the US is 17% richer than Sweden.

The benefits of urban travel are largely proportional to GDP per capita. The economic benefits of agglomeration are proportional to the value of the economy writ large, and so are the benefits of reduced travel time, which in benefit-cost analyses are usually taken to be proportional to the average hourly wage. Conversely, the ability of a government to spend money on desired outcomes is proportional to its tax base, i.e. the size of its economy. All of this is approximate, but all we’re doing is first-order approximate analysis anyway, and so correcting for GDP per capita is valuable.

As it is, the difference between American and Swedish GDP per capita boils down entirely to working hours. Swedes work shorter hours than Americans, as they, like nearly everyone else in Europe, get much longer vacation time; as of 2022, Sweden’s GDP per hour worked was, by a rounding error, slightly higher than the US’s. However, at the same time, the daily ridership numbers for Sweden are specifically drawn from winter workday ridership, to avoid reporting figures from when ridership is lower during the summer vacation season, and the same is true for daily ridership counts in France. If we give Sweden and France credit by looking at ridership when people aren’t on vacation, we must compare the cost per rider with GDP per capita and not GDP per hour.

The upshot is that countries should be building metros up to a maximum capital cost per rider that’s about as large as their GDP per capita. The $24,900 computation for 1960s’ Sweden is ex post, and usually the ex ante benefit-cost ratio must be at least 1.2 or 1.3 for the project to go ahead due to risk. For light rail the allowable cost should be lower, and for bus projects it should be lower still, due to the higher lifetime operating costs; but for metros and metro-like urban rail projects (such as largely grade-separated light rail, or commuter rail tunnels like Citybanan), this analysis should hold. Second Avenue Subway Phase 1, which opened in a United States with a GDP of $73,300/capita in 2023 prices, is thus very cost-effective; Phase 2, budgeted at something like $70,000/rider in today’s prices, is marginal but makes it (in 2022, US GDP per capita in 2023 prices was $80,300).

Some more costs per rider

Our database of construction costs per km is largely complete, but we don’t have much ridership data. Worse, the costs per rider we do have have some biases. We have better information for the US and Canada than elsewhere, and for Europe than the rest of the non-North American world. Costs are also likeliest to be reported for megaprojects with notable delays or cost overruns and thus an incomplete database will be biased upward; large, international cities have better reporting in English than the rest and this introduces another upward bias in incomplete data (these are typically wealthy and therefore capable of affording a high cost per rider).

With that in mind, here are some (again, incomplete) examples:

  • Crossrail cost $24 billion in 2023 PPP terms; ridership in late 2023 was 4.3 million a week, which is usually about 700,000/weekday, said to be above projections, with long-term projections rising to a million/weekday; on current ridership it’s $34,000/rider, and on future projections it’s $24,000.
  • Paris Métro Line 14’s original section, opened by 2007, cost $2.44 billion in 2023 PPP terms; ridership in 2019, before more recent extensions and before corona, was 92 million, so around 320,000 per workday, which is $7,600/rider.
  • Grand Paris Express was projected in 2021 to cost (in 2023 PPP dollars) around $65 billion, with a ridership projection of 2 million/day from 2022, and 2.3-2.4 million/day from 2012, for around $32,000/rider, or $28,000 if the older ridership projection is used.
  • Milan Metro Line 5 cost $2.2 billion in 2023 PPP terms to date, and was said to carry 180,000 daily passengers in 2019, for $12,000/rider.
  • Milan Metro Line 4 cost $2.63 billion, and was expected to carry 235,000 riders per day when complete, which would be $11,000/rider; it will be completed this year, but ridership so far is for only the half of the line that has opened so far ans is also still somewhat suffering from corona.
  • The U-Bahn extensions in Berlin that are currently in development hell include U8 to Märkisches Viertel projected at 13,160€/rider (around $21,000 in 2023 PPP terms), a U6 extension to the former Tegel Airport projected at 27,200€/rider (around $43,000), and a U7 extension to the current BER airport projected at 23,000€/rider (around $36,000), among others brought up by the new coalition.

Willingness to build and Berlin’s problem

The above numbers should not be taken to mean that Italy and Germany are more cost-effective about metro construction. To the contrary, they indicate that they are leaving value on the table by not building.

Germany has a GDP per capita of $66,000 in PPP terms. The benefit-cost ratios of the U-Bahn extensions in Berlin that are being discussed but not actually brought forward and built are very high, and the U8 extension to Märkisches Viertel is scratching a ratio of 4. It is irresponsible that there isn’t a massive construction plan not just under discussion but in design and soon physical construction. This must include not just those lines but also others to be added until the benefit-cost ratio of the marginal line falls to slightly more than the minimum that deserves funding.

Italy has the same problem. No matter how pessimistic one’s view of the future of the economy of Italy is, the metro lines it is currently building, at least in Milan, are so cost-effective that there have to be more of them. In the 2010s, Italy was in a budget crunch and ended up using public-private partnerships to keep debt off-books, at very high implied interest rate based on the rate of payment to the concessionaire, about 8%. Today, fundamentally, even a stagnant economy with a GDP per capita of $55,000, which may finally overtake its 2007 peak this year, should be building more if the current extensive margin of Milan Metro construction is not even in the teens of thousands of dollars. Perhaps the benefit-cost ratios for Sweden rely not just on Sweden’s GDP per capita of the 1960s but on that of today, after 60 years of growth; but then even absent any of this growth, 55/3.5 > 12.

Paris showcases a healthier situation: the success of M14 endeared the region to driverless metros with longer interstations than the older metro lines but shorter ones than the RER, and with further growth in the region, the state decided to build the 200 km Grand Paris Express to improve circumferential rail service and add more fast radial connections from the suburbs to the city not along the RER. France’s GDP per capita is around $58,000 in PPP dollars; some elements of GPE may be marginal, but the project writ large has a solid benefit-cost-ratio starting with a 2.

Spain is like France in that when it can build, as it could before the Great Recession and as it can now after having recovered, it does. This is the right way to do it: low costs per km should translate to massive subway expansion. If you’re Madrid, you can build lines that would be completely ridiculous anywhere else in terms of land use, like MetroSur, because it’s cheap.

A Paris or Berlin cannot be so profligate as Madrid. Building a line like MetroSur here, which in the Spanish context looks silly but does pass cost-effectiveness muster, would be an economic albatross. But medium-cost cities can still cover most of the built-up area with subway and commuter rail lines; Paris is doing so but Berlin is not.

The problem for Berlin is not quite austerity, which afflicted Southern Europe last decade, producing negative economic growth. In the last coalition, it was the two most left-wing parties, the Greens and Die Linke, which opposed U-Bahn construction. In their view, trams are preferable; they complain about the high construction costs of subways, but don’t meaningfully engage with why they’re still necessary for fast urban transport beyond the range of the handful of inner neighborhoods where people vote for the Greens.

That said, in the current coalition, there’s no real political left. SPD is in it, but under an atypically conservative leader in Franziska Giffey, and the more left-wing faction in the party, including the youth league Jusos, is sidelined. Both parties in the coalition, CDU and SPD, are committed to building those U-Bahn extensions. But then nothing is happening; Giffey is not competent enough to do it, and CDU is too wrapped up in anti-environmental populism to do something that pisses off some environmentalists but makes others happy.

This way, the low costs per rider projected for Berlin U-Bahn expansion are not a testimony to cost-effectiveness. They’re a testimony to political unwillingness to build, for largely petty reasons. It is wasteful to build a subway line with a very high projected cost per rider relative to the country’s wealth, but it is equally wasteful not to build one with a low projected cost per rider, and the latter is what is happening in Berlin, unfortunately.

Public Transit Subsidies and Efficiency

A few days ago, Streetsblog covered a new study about the impact of transit subsidies on efficiency. Transportation economics research is skeptical of operating subsidies to public transit, arguing that it incentivizes waste. In contrast, the new study argues, the opposite is the case: systems with more subsidies are more efficient. In reality, this is a bad study – trivially and uninterestingly wrong. The real question of interest is not whether it is saying something – it isn’t – but why it is getting any positive press among transit advocates. The context here is important: non-technical advocates, having given up on building good transit, instead just want more subsidies for operations, as the costs rise (bus service being labor-intensive), with scant interest in outcomes like ridership.

What’s in the study?

The study, by Funk, Higgins, and Newmark, is a regression among the top 15 American metro areas, using data from 2016-9. Among those regions, higher subsidies to public transit correlate with higher efficiency. Efficiency is measured as average vehicle occupancy, regardless of whether the vehicle is a train or bus (but trains are calculated on a per-car basis, following NTD norms). Subsidies are measured per capita.

By Funk-Higgins-Newmark, New York has by far the most transit subsidies in the United States: the region subsidized public transit operations across all agencies by $444.50/capita in 2019 dollars. In contrast, regions that have rounding-error ridership were stingier: Dallas averaged $124.20/capita, and Phoenix $97.20. New York has much higher efficiency measured by vehicle occupancy, and secondary transit cities like Boston and San Francisco have lower efficiency than New York and higher than the no-transit Sunbelt cities.

Except that all they’re proving is that cities with higher transit ridership have both higher vehicle occupancy (partly due to higher crowding, partly due to real efficiency, because they run more trains and not just buses) and higher subsidies per capita. To switch to raw 2019 NTD data just because it’s easier to explain: the MTA, across all sub-agencies recorded in the NTD’s top 50 database linked above, spent $12.381 billion on operations in 2019 and earned just $6.359 billion from fares, which works out to somewhat more than $400 per capita in the MTA service region – but those $6 billion of subsidies covered something like 30 billion p-km and 4.6 billion unlinked trips, working out to $1.3 per trip. Even taking into account that unlinked trips are fake news (they double-count people who transfer between subway lines, even within the system), the subsidy per ride is on the on the order of $2/trip. Dallas, in contrast, spent $568 million on operations, got $66 million in fare revenue, and carried 69 million unlinked trips and 706 million p-km. Far from stingy, Dallas spent $7.28 on subsidizing each unlinked bus or rail trip. Houston likewise spent $5.65/trip and Phoenix (buses only) $4.33.

At no point does the study try to establish that if Dallas quadruples its transit subsidies, it will be able to raise its transit ridership by the factor of 67 required to match New York per capita. A snapshot regression of 15 metro areas has no chance of having enough statistical power to establish anything on the margin. The literature review is negative on transit subsidies, finding that they are associated with higher labor spending (for example, through reluctance to innovate in labor-saving technology) and with spreading service-hours too thinly at too low a frequency. To the contrary, all available research suggests that an increase in Dallas’s transit subsidy to match New York will just get wasted; ridership will increase, but so far below the increase in subsidy that the per trip subsidy will skyrocket even more.

In other words, the higher subsidy in New York per capita just showcases that a larger share of people in the New York region ride public transit, and therefore even a relatively low per-rider subsidy can amount to a hefty subsidy per resident. This is not at all surprising; by the same token, polities with larger numbers of schoolchildren spend a larger proportion of their economy on education, which says nothing about how well-funded the schools are (for example, Israel has a rather high education spending as a percentage of GDP, but low spending per student). Thus the study is not just wrong but also boringly so.

So why is it getting any positive press?

American advocacy and subsidies

For years, American transit advocates have been looking into the idea of federal subsidies to transit operations. Traditionally, these are not allowed, except in very small cities with even less relevant public transit than the systems of Dallas or Phoenix; instead, federal funding only goes to capital construction. In the Streetsblog piece, Kea Wilson writes about this connection to advocacy for federal subsidies; I’ve seen it from talking to advocates, who kept complaining, essentially, that federal subsidies require some clear piece of infrastructure to be built and that’s too constraining.

As corona hit, these efforts grew dramatically, since in an emergency it was valuable to inject cash into the systems to prevent long-term closures; then, as ridership hasn’t quite returned (it looks like 70% of pre-pandemic levels as of late 2023; in Berlin, with large reductions in fares, it was 95%), advocates started agitating for more subsidies, without much of a clear goal, certainly nothing related to outcomes. Instead, the demand is to have buses and trains run just so that cities can say that they’re running them. When such advocates say what they’d like to do with the money, it’s, in my experience, always about inputs, never outputs; one said that light rail trains should have a second crew member working as a train attendant just to make passengers feel safer.

In contrast, when I see advocates demand subsidies for concrete ridership-related improvements, it’s not really connected to any demand for federal subsidies. I don’t know what Riders Alliance thinks about federal subsidies to operations, but in its Six-Minute Service campaign, which in its opinion requires some additional state subsidies (and in mine and in that of ETA pays for itself on the subway), it talks about how it’s such a small increase in subsidy relative to existing ones, and doesn’t directly ask for federal subsidies. When it’s for something this specific, advocacy organizations feel confident asking for money directly.

So the belief that subsidies to operations help efficiency, based on a boringly wrong regression, is there in support of something different – in support of wanting subsidies with no clear goal, other than “fund transit.” It’s not really about frequent service, the most valid use of additional operating funds. It’s certainly not about capital extensions that reduce future operating costs, which are favored under the current regime. It’s about spending for its own sake.

Security Theater on Metros

In comments, Sarapen asked me about security on urban rail. It’s common in developing Asia to require people to go through metal detectors to get to the platforms; I’ve seen this in Bangkok, she mentions this in the context of Manila, and it’s also the case on Indian metros and Chinese ones. Seung Y. Lee, a BART digital media worker and indispensable commentator on American and Asian metro history who sadly doesn’t blog enough, has an excellent post about it, talking about the use of security as a tool of social control, for example in Hong Kong.

But Sarapen is asking about the need for security to deter terrorism and violence, which do exist in the Philippines.

For this, I naturally went to the country facing terrorism and violence that I know most intimately. I haven’t visited in more than 11 years, and so I’ve never taken the light rail in either Tel Aviv or Jerusalem (which was open on my last visit but I didn’t visit the city then). So I went ahead and asked on my Discord server, and got this:

The person covered up by the two shorter white rectangles also posted a picture of a platform in Tel Aviv, with a vault for bomb disposal:

Israel Railways does have security theater – one has to open one’s bags in front of a security guard and go through a metal detector, and this being Israel, there is extensive racial profililng. But the light rail, including the underground portions in Tel Aviv, do not. There’s a lot of visible security presence, including cameras, security personnel, and K-9 units, but no metal detectors. This is in a country that, in the 1990s and early 2000s, underwent a wave of suicide bombings on buses, and still didn’t put metal detectors on them, because it’s infeasible to install such technology on surface transit, and too expensive and frankly not too useful on metros (a security line is an attractive bombing target).

This shouldn’t be too surprising, in a way. Western Europe did not install security theater on trains in response to far-left terrorism in the 1970s and 80s. Israel’s way of building and running public transportation is intended as a pan-European medley, using consultants who have done work in Europe, and with media discourse that looks up to Europe’s urban transportation systems (on other matters Israel looks up to the United States, but Israelis understand American public transportation is not good). If nobody in Europe (or at least Western Europe) does something, it’s unlikely Israel will do it, not on civilian public infrastructure.

Worthless Canadian Initiative

Canada just announced a few days ago that it is capping the number of international student visas; the Times Higher Education and BBC both point out that the main argument used in favor of the cap is that there’s a housing shortage in Canada. Indeed, the way immigration politics plays out in Canada is such that the cap is hard to justify by other means: traditionally, the system there prioritized high-skill workers, to the point that there has been conservative criticism of the Trudeau cabinet for greatly expanding low-skill (namely, refugee) migration; capping student visas is not how one responds to such criticism.

The issue is that Canada builds a fair amount of housing, but not enough for population growth; the solution is to build more – in a fast-growing country like Canada, the finance sector expects housing demand to grow and therefore will readily build more if it is allowed to.

Vancouver deserves credit for the quality of its transit-oriented development and to a large extent also for the amount of absolute development it permits (about 10 units per 1,000 residents annually); but its ability to build is much greater than that, precisely because rapid immigration means that more housing is profitable, even at higher interest rates. The population growth coming from immigration sends a signal to the market, invest in long-term tangible goods like housing. Thus, Vancouver deserves less credit for its permissiveness of development – large swaths of the city are zoned for single-family housing with granny flats allowed, including in-demand West Side neighborhoods with good access to UBC and Downtown jobs by current buses and future SkyTrain.

The rub is that restricting student immigration is probably the worst possible way to deal with a housing shortage. Students live at high levels of crowding, and the marginal students, who the visa cap is excluding, live at higher levels of crowding than the rest because they tend to be at poorer universities and from poorer backgrounds. The reduction in present-day demand is limited. In Vancouver, an empty nester couple with 250 square meters of single-family housing in Shaughnessy is consuming far more housing, and sitting on far more land that could be redeveloped at high density, than four immigrants sharing a two-bedroom apartment in East Vancouver.

In contrast, the reduction in future demand is substantial, because those students then graduate and get work, and many of them get high-skill, high-wage jobs (the Canadian university graduate premium is declining but still large; the American one is larger, but the US is also a higher-inequality society in general); having fewer students, even fewer marginal students who might take jobs below their skill level, is still a reduction in both future population and future productivity. What this means is that capital owners deciding where to allocate assets are less likely to be financing construction.

The limiting factor on housing production is to a large extent NIMBYism, and there, in theory, immigration restrictions are neutral. (In practice, they can come out of a sense of national greatness developmental conservatism that wants to build a lot but restrict who can come in, or out of anti-developmental NIMBYism that feels empowered to build less as fewer people are coming; this situation is the latter.) However, it’s not entirely NIMBYism – private developmental still has to be profitable, and judging by the discourse I’m seeing on Canadian high-rise housing construction costs in Toronto and Vancouver, it’s not entirely a matter of permits. Even in an environment with extensive NIMBYism like the single-family areas of Vancouver and Toronto, costs and future profits matter.

The Need for Ample Zoning Capacity

An article by Vishaan Chakrabarti in last month’s New York Times about how to make room for a million more people in New York reminded me of something that YIMBY blogs a decade ago were talking about, regarding zoning capacity. Chakrabarti has an outline of where it’s possible to add housing under the constraints that it must be within walking distance of the subway, commuter rail, or Staten Island Railway, and that it must be of similar height to the preexisting character of the neighborhood. With these constraints, it’s possible to find empty lots, parking lots, disused industrial sites, and (in near-Midtown Manhattan) office buildings for conversion, allowing adding about half a million dwellings in the city. It’s a good exercise – and it’s a great explanation for why those constraints, together, make it impossible to add a meaningful quantity of housing. Transit-oriented development successes go far beyond these constraints and build to much higher density than is typical in their local areas, which can be mid-rise (in Europe) or high-rise (in Canada and Asia).

The issue with the proposal is that in practice, not all developable sites get developed. The reasons for this can include any of the following:

  • The parcel owner can’t secure capital because of market conditions or because of the owner’s particular situation.
  • The parcel is underdeveloped but not empty, and the owner chooses not to redevelop, for a personal or other reason.
  • The area is not in demand, as is likely the case near the Staten Island Railway or commuter rail stations in Eastern Queens, or in much of the Bronx.
  • The area is so auto-oriented, even if it is technically near a station, that prospective buyers (and banks) demand parking, reducing density.

New York has, on 2020 census numbers, 8.8 million people; 1 million additional New Yorkers is 11% more people. Cities that permit a lot of housing have an envelope for much more than 11% extra population. In New York’s history, it was computed in 1956 that under the then-current 1916 zoning code, the city’s zoned capacity was 55 million people, but under the proposed code that would be enacted in 1961, the zoned capacity would fall to 12 million. In Los Angeles, The Homeowner Revolution makes the point (on PDF-page 19) that in 1960, zoned capacity was 10 million in the city proper, four times the city’s population, but by 1990 it fell to 3.9 million, 11% more than the city’s population.

Technically, the same extra zoned capacity that Chakrabarti finds for New York has existed in Los Angeles for a generation. In practice, for all the above reasons why development never reaches 100% of capacity even in expensive areas, Los Angeles builds very little housing, and rents are very high, perhaps comparable to those of New York even as wages are much lower.

What this means is that the way forward for any transit-oriented development plan has to get out of the mentality that the buildings need to be of similar size to the existing character of the neighborhood. This constraint is too strict, and not at all observed in any successful example that I know of.

To Americans, the most accessible example of transit-oriented development success is Vancouver; some sections of the Washington suburbs (especially Arlington) qualify as well, but the extent there is much less than in Canada, and consequently ridership on Washington Metro has lagged that of its Canadian counterparts. In Vancouver, the rule that Chakrabarti imposes that the preexisting parcels must be empty or nonresidential is largely observed – as far as I can tell, the city has not upzoned low-density residential areas near SkyTrain, and even in Burnaby, the bulk of redevelopment has been in nonresidential areas.

The redevelopment in Vancouver proper looks like this:

And here is Metrotown, Burnaby:

The surface parking may look obnoxious to urbanists, but the area has more jobs, retail, and housing than the parking lots can admit, and the modal split is very high.

European transit-oriented development is squatter – the buildings are generally less tall, and they’re spread over a larger contiguous area, so that beyond the resolution of a few blocks the density is high as well. But it, too, often grows well beyond tradition. For example, here is Bercy, a redevelopment of the steam-era railyards at Gare de Lyon, no longer necessary with modern rail technology:

In the future, a New York that wants to make more room will need to do what Vancouver and Paris did. There is no other way.

Quick Note: Different Anti-Growth Green Advocacies

Jerusalem Demsas has been on a roll in the last two years, and her reporting on housing advocacy in Minneapolis (gift link) is a great example of how to combine original reporting with analysis coming from understanding of the issue at hand. In short, she talks to pro- and anti-development people in the area, both of which groups identify with environmentalism and environmental advocacy, and hears out their concerns. She has a long quote by Jake Anbinder, who wrote his thesis on postwar American left-NIMBYism and its origins, which are a lot more good-faith than mid-2010s YIMBYs assumed; he points out how they were reacting to postwar growth by embracing what today would be called degrowth ideology.

I bring this up because Germany is full of anti-growth left-NIMBYism, with similar ideology to what she describes from her reporting in Minneapolis, but it has different transportation politics, in ways that matter. The positioning of German left-NIMBYs is not pro-car; it has pro-car outcomes, but superficially they generally support transportation alternatives, and in some cases they do in substance as well.

In the US, the left-NIMBYs are drivers. Jerusalem cites them opposing bike lanes, complaining that bike lanes are only for young childless white gentrifiers, and saying that soon electric cars will solve all of the problems of decarbonizing transportation anyway. I saw some of this myself while advocating for rail improvements in certain quarters in New England: people who are every stereotype of traditional environmental left-NIMBYism were asking us about parking at train stations and were indifferent to any operating improvements, because they don’t even visit the city enough to think about train frequency and speed.

In Germany, many are drivers, especially outside the cities, but they don’t have pro-car politics. The Berlin Greens, a thoroughly NIMBY party, are best known in the city for supporting removal of parking and moving lanes to make room for bike lanes. This is not unique to Berlin or even to Germany – the same New Left urban mayors who do little to build more housing implement extensive road diets and dedicated lanes for buses, streetcars, and bikes.

These same European left-NIMBYs are not at all pro-public transportation in general. They generally oppose high-speed rail: the French greens, EELV, oppose the construction of new high-speed lines and call for reducing the speed on existing ones to 200 km/h, on the grounds that higher speeds require higher electricity consumption. In Germany, they usually also oppose the construction of new subway and S-Bahn tunnels. Their reasons include the embedded carbon emissions of tunneling, a belief that the public transport belongs on the street (where it also takes room away from cars) and not away from the street, and the undesirability (to them) of improving job access in city center in preference to the rest of the city. However, they usually consistently support more traditional forms of rail, especially the streetcar and improvements to regional rail outside major cities. For example, the NIMBYs in Munich who unsuccessfully fought the second S-Bahn trunk line, whence the expression Organisation vor Elektronic vor Beton (the Swiss original omits Organisation and also is very much “before” and not “instead of”), called for improvements in frequency on lines going around city center, in preference to more capacity toward city center.

I’m not sure why this difference works like this. I suspect it’s that American boomer middle-class environmental NIMBYism is rooted in people who suburbanized in the postwar era or grew up in postwar suburbs, and find the idea of driving natural. The same ideology in Europe centers urban neighborhood-scale activism more, perhaps because European cities retained the middle class much better than their American ones, perhaps because mass motorization came to Europe slightly later. It also centers small towns and cities, connected to one another by regional rail; the underlying quality of public transportation here is that environmentalists who can afford better do rely on it even when it’s not very good, which hourly regional trains are not, whereas in the United States it’s so far gone that public transportation ridership comprises New Yorkers, commuters bound for downtown jobs in various secondary cities, and paupers.

Eurostar Security Theater and French Station Size

Jon Worth has been doing a lot of good work lately pouring cold water on various press releases of new rail service in Europe. Yesterday he wrote a long post, reacting to some German rail discourse about the possibility of Eurostar service between London and Germany; he explained the difficulties of connecting Eurostar to new cities, discussing track and station capacity, signaling, and rolling stock.

Jon, whose background is in EU politics, wastes no time in identifying the ultimate problem: the UK demands passport controls, and this demand is unlikely to be waived in the near future due to concerns over Brexit and the need to have visible border control theater. In turn, the passport control and the accompanying security theater (not strictly required, but the UK insists for Channel Tunnel security) mean that boarding trains is a slow process since platforms must be kept sterile; thus, a Eurostar station requires dedicated platforms, and if it has significant rail traffic then it requires many of them, with low throughput per track. This particularly impacts the prospects of Eurostar service to Germany, because it would go via Belgium and Cologne, which has far from enough platforms for this operation.

What I’d like to add to this analysis is that Eurostar made a choice to engage in such controlled operations in the 1990s. The politics of Brexit can explain why there’s no reform that is acceptable to the British political system now; it cannot explain why this was chosen in the 1990s. The norm in Europe before Schengen was that border control officers would perform on-board checks while the train traveled between the last station in the origin country and the first station in the destination country; long nonstop trains between Paris and London or even Lille and London are ideal for such a system. Britain insists on the current system of border control before boarding because this way it can deny entry to people who otherwise would enjoy non-refoulement protections – but in the 2000s the politics in Britain was not significantly more anti-immigration than in, for example, Germany, or France.

Rather, the issue is that Britain insisted on some nebulous notion of separateness, and this interacted poorly with train station design in France compared with in Germany. Parisian train stations are huge, and have a large number of terminating tracks. Dedicating a few terminal tracks to sterile operations is possible at Gare du Nord, and would be possible at other Parisian terminals like Gare de Lyon if they pointed in the direction of a place that demanded them. SNCF has conceived of its operations, especially internationally, as airline-like, and this contributed to complacency about how the train stations are being treated like airports.

Germany developed different (and better) ways of conceiving of train operations. More to the point, Germany doesn’t really have Paris’s terminals with their surplus of tracks, except for Frankfurt and Munich. Cologne, the easiest place to get to London from, doesn’t have enough tracks for sterile operations. This is fine, because German domestic trains do not imitate airlines, even where there is room (instead, the surplus of tracks is used for timed connections between regional trains); this also cascades to international trains connecting to Germany, whether from countries that have more punctual rail networks like Switzerland or from countries that work by a completely different paradigm like Belgium or France.

And now Eurostar politically froze a system that was only workable at low throughput, at a handful of stations with more room for sterile operations than is typical. The system is still below its ridership projections from before opening; it was supposed to be part of a broader international rail network, but that never materialized, because of the burden of security theater, the high fares, and the indifference of Belgium to extending high-speed rail so that it would be useful for international travelers (the average speeds between Brussels-Midi and the German border are within the upper end of the range for upgraded classical lines, even though HSL 2 and 3 are new high-speed lines).

And now, with the knowledge of the 2010s, it’s clear that any future expansion of Eurostar requires forgoing the airline-like paradigm that led SNCF to stagnation in the same decade. This clashes with British political theater now, but there’s no other way forward.

And this even affects domestic British rail planning. London planners are fixated on Paris as their main comparison. This way, they are certain trains must turn slowly at city terminals, requiring additional tracks at Euston and other stations that are or until recently were part of High Speed 2, at a total cost of several billion pounds. In Germany and the Netherlands (at Utrecht) trains can move faster, down to turns of seven to eight minutes on German regional trains and four to five minutes on intercity trains pinching at terminal stations like Frankfurt. But planners in large cities look down on smaller cities; it’s no different from how planners in New York assume that because New York is bigger than Stockholm, Second Avenue Subway’s stations have higher ridership than the stations of Citybanan (in fact, Citybanan’s two stations, located in city center, are significantly busier).

This way, a particular feature of historic Parisian stations – they have a lot of tracks – got turned into something that every city’s train station is assumed to have. It means Eurostar can’t operate into other stations, because there is no surplus of platforms allowing segregating service to the UK away from all other traffic; it also means that planners in the UK that are trying to engineer stations assume British stations must be overbuilt to Parisian specs.

Small Metros Aren’t Lean, They’re Underbuilt

Reece Martin does very good vlogs on public transportation, and has begun text-blogging more regularly, which I appreciate greatly. But a post of his from six days ago, talking about lean metros, misses a key aspect of short metro systems. He compares old legacy systems like Paris or New York or Berlin’s with newer ones, like Hong Kong’s, and points out that the newer ones are rather short relative to city size, saying that it’s a leaner, lower-cost way of doing things. But in fact, the reason we see such short metros relative to city size is not efficiency, but underbuilding, leading to overcrowding.

What’s a lean metro?

Reece divides leanness into two kinds. The first is the ability of some cities to build driverless metros with very short trains at very high frequencies, to save money on station construction. He gives the examples of Copenhagen and Vancouver. This is particularly common in Italy: Milan Metro Lines 4 and 5 have 50 meter long trains, the Turin Metro has 55 meter long trains, the Brescia Metro has 40 meter long trains. With this setup and with the generally low construction costs of Italy, even Brescia can afford a metro, in a city proper of 200,000 with a built-up urban area of 673,000 (and rising) as of 2011.

The second kind of leanness is just building fewer lines. He talks about Toronto’s system, with two main trunk lines, one branch line with a transfer to a main line, and a total of 70 km of length. He also brings up Hong Kong, which has, counting both proper metro lines and the two inherited commuter lines, around 212 route-km, with very high ridership. This can be supported through transit-oriented development, for which Hong Kong is famous. It can also come from strong bus-rail connections as in Toronto: a blog post from last decade that I can no longer find points out that York Mills has what looks like 14,000 weekday boardings on pre-corona numbers, despite low-density land use immediately surrounding it, because of the strong connecting buses on the Toronto grid, favorably comparing it with American metros like Washington’s.

Small station metros

I don’t want to criticize the Italian trend too much, but I do want to separate it from the other systems Reece calls lean. The issue with the style of construction used in Brescia is that it’s really good, if your city is the size of Brescia. Small-station, partly cut-and-cover driverless metros should be in the toolkit of metro areas of about a million people, in order to save money. Other tools should be heavily relying on legacy commuter lines (as in Zurich) and using trams if they’re available (as in Bratislava or Brno), and likely combining all three solutions when feasible (in fact Zurich has a large tramway network in addition to the S-Bahn).

In a larger city, such light metro lines are only useful in a very restricted set of circumstances. Singapore has short trains on the Circle Line – but the Circle Line has not been cheap to build, and its last section has been extraordinarily expensive. On a radial line, it’s a nonstarter. A large city needs the very high throughput of a driverless metro but also larger trains; those can be medium-length trains, like the 90 meter trains of Paris Métro Line 1, or longer, like the 138 meter long trains of the now-driverless legacy lines in Singapore, or the 200 meter long platforms of the Shinbundang Line in Seoul. If the line is too short, the city may find it needs to build another just for relief, as the area that was once thought peripheral develops.

Short metros

If a metro system is short, even if its trains are long, it’s not generally a sign of efficient construction in the city. It’s a sign of underbuilding and overcrowding.

Hong Kong has very high crowding levels, even with a system length that, counted properly, is not that unreasonable: the MTR’s total route-length is almost the same as that of the Paris Métro, which has 227 km, and its ridership is, on pre-corona numbers, slightly higher, 1.7 billion a year compared with 1.5 billion. Now, to be clear, Paris has very high ridership for the system’s size. I suspect the reason I’ve never seen overcrowding on the Métro is that the nature of Parisian job concentration is such that the lines that get overcrowded are ones connecting the suburbs with the city, that is, the RER and M13, rather than predominantly intra-city lines like M1. But the situation in Hong Kong is overall less one of leanness and more one of not being to expand as fast as it would like due to extreme construction costs, which are a strong contender for the world’s second worst, after New York’s. (Toronto is an even stronger contender.)

Then there are the developing-world metros that are just far too short for their city size. Hong Kong is a city of 7 million with a little more than 200 km of metro and commuter rail. Cairo is a metro area of 22 million with 100 km of metro. Cairene construction costs are high, but when, depending on how much one trusts dead links from 10 years ago, the city has the world’s highest rail ridership per km, it needs a lot more; that Cairo has 100 km of metro rather than 800 like Shanghai, a similar-size city in a country that, during its peak expansion, was about as rich as Egypt, is not about leanness but about the Egyptian government’s spending priorities.

For a middle-income country that wants to get out of the middle-income trap that Egypt is stuck in (or Brazil, home to the almost as underbuilt São Paulo Metro), China is a decent benchmark. So is Paris – France is rich but also, precisely because it’s rich, rather motorized by any developing country standards, leading to a modal split of about 43% public transit, 43% cars for work trips in Ile-de-France. Using these benchmarks, your city should have on the order of 30 km of metro and new commuter rail per million residents. If you have 4.5 like Cairo, it’s not efficiency, it’s total disinterest in the living standards of the urban population.

Cost-Plus Contracting is Good

Because I write a lot about government waste, I get to interact with people who complain a lot about government waste, which exposes me to a range of views, some more reasonable than others. The less reasonable views themselves range from unhinged rants to things that sound like they could be true, that may even be true for parts of contracting other than public transportation or infrastructure megaprojects. It’s one of the latter views that I wish to talk about here. It’s all too common, in my experience, for people who complain about government waste in the United States to speak ill of cost-plus contracting, which they claim (incorrectly but plausibly) is a form of a waste and even (much less plausibly) fraud, in which contractors get paid regardless of how poor the job is. In truth, the only good way to build infrastructure megaprojects is with itemized contracts and with generous public-sector assumption of risk; this isn’t called cost-plus in the literature but itemization, but is close enough that from the perspective of citizen-activists it should be viewed as a cost-plus variant.

Itemization versus fixed-price contracting

Contracts can, broadly, be either itemized or fixed-price, the latter also known as lump-sum. A fixed-price contract has a single cost line for the entire project; an itemized contract instead prices each section, and certain elements that are agreed to be outside the contractor’s control, like materials, are adjusted automatically if prices on the world market change.

There are also mixed models. For example, in Stockholm, some contracts are fixed-price, some itemized, and others use an intermediate system called fixed price with adjustable quantities, in which some elements are itemized but others are not. Turkish construction contracts itemize about 65% of the costs in detail, covering the civil infrastructure, and then take each of the main sub-items of the other 35% (at the detail level of ventilation, power design, sewage, etc.) as a fixed-price contract.

In Southern Europe and Turkey, the items are public, published transparently by the government, as an anti-corruption measure. These are detailed, and take different costs of living into account, and thus in Italy the prices, published by the state, may differ by region. If the itemized prices are unrealistic, then the contractors are allowed to demand more, but those itemized lists help ground the contract.

This system far outperforms fixed-price contracts. We have standard references on this, namely Ryan (1, 2) and Bolotnyy-Vasserman, finding that itemization for power plants (in Ryan) or road maintenance (in Bolotnyy-Vasserman) leads to reduced overall costs, by about 20%. The reasons are that fixed-price contracting leads to increased friction in change orders, and that, relatedly, fixed-price contracting imposes risk on the private contractor, leading to higher bids for compensation. For this reason, all low-construction cost environments we’ve investigated itemize, to a significant extent. The Nordic countries are moving from itemized to fixed-price contracting under the influence of British practices and the requests of large international firms that are used to doing business in English-speaking countries or in other high-cost countries, and as they do so, their costs are moving from low to medium, in addition to other bad reforms.

So why do people think itemization is bad?

The big multinationals are just used to fixed-price contracts. In New York (where it’s usually not big multinationals), an additional issue is that the contractors prefer to be paid upfront, due to delays in payment under itemization. But more commonly, the contractors just don’t like having to give exact specifications, and are used to environments where they don’t and don’t notice how the costs in those places are much higher than in places that don’t govern in English.

But government contractors are hardly a popular figure among people who complain about government waste. So what gives?

I suspect that this is about the sight of the government assuming risk for something a private contractor does. The sight of a contractor complaining about a problem that the government then has to pay extra for is repulsive to the general public, which assumes that something shady is going on, regardless of whether it is. This leads to political pressure to insulate the public sector from unexpected changes.

The problem is that there is no way to fully insulate the public sector. Once the megaproject has been decided, the political cost for the government of walking away is larger than the economic penalties for the concessionaire. In practice, every attempt to privatize risk will get contractors to bid higher, while still not protecting the public from the risk of large cost overruns. Turkey is somewhat exceptional in this, but it still itemizes, and the way it enforces the maximum 20% cost overrun is by having so many contractors that it’s easy to blacklist malefactors; this is not available in richer countries with fewer, larger firms.

Janno Lieber Lies to New York About Costs and Regulations

After being criticized about the excessive size of subway stations designed on his watch, MTA head Janno Lieber fired back defending the agency’s costs. In a conversation with the Manhattan Institute, he said about us, “They’re not wrong that the stations are where the MTA stations add cost. But they are wrong about how they compare us – the cost per mile is misleading” (see discussion on social media here). Then he blamed labor and the fire code. Blaming labor is a small but real part of the story; this is common among the white-collar managers Eric and I have talked to, and deserves a separate explanation for why this concern is overblown. But the issue of the fire code is fraud, all the way.

I’ve previously seen some journalists and advocates who write about American construction costs talk about fire safety, which is mentioned occasionally as a reason designs cannot be changed. It’s not at all what’s going on, for two separate reasons, each of which, alone, should be grounds to dismiss Lieber and ensure he never works for the state again.

The first reason is that the fire safety regulation in the United States for train stations, NFPA 130, has been exported to a number of other countries, none of which has American costs or the specific American tradition of overbuilding stations. China uses NFPA 130. So does Turkey. Spain uses a modification. We can look at their designs and see that they do not build oversize stations. I’ve seen an environmental impact analysis in Shanghai, with the help of a Chinese student studying this issue who explained the main planning concerns there. I could write an entire blog post about China (not a 10,000-word case report, of course), but suffice is to say, if the train is projected to be 160 m long, the station dig will be that plus a few meters – and Chinese stations have mezzanines as I understand it. Spanish and Turkish stations have little overage as well; building a dig twice as long as the station’s platforms to house back-of-the-house spaces is unique to most (not all) of the Anglosphere, as design consultants copy bad ideas from one another.

Even the claim that NFPA 130 requires full-length mezzanines is suspect. It requires stations to be built so that passengers can evacuate in four minutes in emergency conditions, rising to six minutes counting stragglers (technically, the throughput needs to be enough to evacuate in four minutes, but with latency it can go up to six). The four minute requirement can be satisfied on the lettered lines of the subway in New York with no mezzanines and just an access point at each end of the platform, but it’s close and there’s a case for another access point in the middle; no full-length mezzanine is required either way. If the stations are any shorter, as on the numbered lines or in other North American cities, two escalators and a wide staircase at the end of each platform are more than enough, and yet the extensive overage is common in those smaller systems too (for example, in Vancouver, the Broadway extension is planned with 128 m long digs for 75 m trains, per p. 9 here).

“Fire safety” is used as an excuse by people with neither engineering background nor respect for anything quantitative or technical. Lieber is such a person: his background is in law and he seems incurious about technical issues (and this is also true of his successor at MTA Capital Construction, public policy grad Jamie Torres-Springer).

Perhaps due to this lazy incuriosity, Lieber didn’t notice that the MTA has extensive influence on the text of NFPA 130, bringing us to the second reason his claim is fraudulent. NFPA 130 is not to blame – again, it’s the same code as in a number of low- and medium-cost countries – but Nilo Cobau explains that the NFPA process is such that big agencies have considerable input, since there aren’t many places in the US that build subways. Nolan Hicks pointed out in the same thread, all linked in the lede paragraph, that the MTA has a voting member and two alternates on the board that determines NFPA 130 and hasn’t requested changes – and that Montreal, subject to the same codes, built a station with little overage (he says 160 m digs for 150 m platforms).

The handwaving of a fire code that isn’t even different from that of cheaper places is there for one purpose only: to deflect blame. It was a struggle to get Lieber and other New York leaders to even admit they have high costs, so now they try to make it the fault of anyone but themselves: fire safety regulations, organized labor, what have you.

Labor is a real issue, unlike fire safety, but it’s overblown by managers who look down on line workers and have generally never been line workers. Lieber graduated law school, was hired by USDOT at either junior-appointed or mid-level civil servant role, I can’t tell which, and then did managerial jobs; his successor as head of MTA Construction and Development, Jamie Torres-Springer, graduated public policy. These aren’t people who worked themselves up from doing engineering, architecture, planning, or ethnographic work; add the general hostility American white-collar workers have toward blue-collar workers, and soon people in that milieu come to believe that just because their top 5%er wages are much higher than they could earn anywhere else in the world, the sandhogs also earn much more than they could anywhere else in the world, when in truth New York sandhog and Stockholm miner wages and benefits are very close.

Occasionally the point that it’s not wages but labor productivity seeps in. There, at last, we see a real problem with labor. Eric and I found that about a third of the sandhogs on Second Avenue Subway didn’t really need to be there. Further cuts could be achieved through the use of more labor-efficient techniques, which the MTA is uninterested in implementing. The rest of the American labor premium comes from excessive staffing of white-collar supervisors, including representatives from each utility, which insists that the MTA pay for the privilege of having such representatives tell them what they can and cannot do in lieu of mapping the utilities and sending over the blueprints. All included, labor was around 50% of the cost of Second Avenue Subway, where the norm in Italy, Turkey, and Sweden is around 25% (note how higher-wage Sweden is the same as lower-wage Italy and much lower-wage Turkey); excessive labor costs contributed a factor of 1.5 premium to the project, but the other factor of 6 came from excessive station size, deep mining of stations (which thankfully will not happen at 106th and 116th Street; it will at 125th but that’s unavoidable), lack of system standardization, and a litany of project delivery problems that are generally getting worse with every iteration. Lieber personally takes credit for some of the privatization of planning to design-build consultancies, though to be fair to him, the project delivery problems predate him, he just made things slightly worse.

A New York that wants to build will not have incompetent political appointees in charge. It will instead hire professionals with a track record of success; as no such people exist within the American infrastructure construction milieu, it should use its own size and prestige to find someone from a low-cost city to hire, who will speak English with an accent and know more engineering than American legal hermeneutics. And it will not reward people who defraud the public about the state of regulations just because they’re too lazy to know better.