I gave my webinar talk about the Stockholm case and uploaded the video here. I don’t want to repeat either the case or my presentation thereof, but rather just point to one thing I said during the Q&A, about what counts as a megaproject. At the time I thought it was just an extemporaneous answer, but Sandy Johnston highlit it in his livetweeting, and I think it has some deeper meaning.
The issue at hand is that the definition of what a megaproject is is relative to local capabilities and practices. Building 5 km of subway tunnel is a megaproject if you’re an American city or a small European capital, but not if you’re a large European or Asian city. What I mean by this definition is that the usual properties of megaprojects are relative to local capabilities in the following ways:
- Megaprojects are hotly debated politically at the highest level – Crossrail and High Speed 2 were in the manifestos of both Labour and the Conservatives, and Grand Paris Express evolved with direct government involvement. In smaller cities, projects of similar levels of political importance are as one might expect smaller, like Citybanan (which is 6 km) and Nya Tunnelbanan (which is 19 km); in turn in Paris, extensions of the Métro totaling 19 km happen gradually without such political involvement.
- Megaprojects are institutionally new. Grand Paris Express not only was decided by the government, as an expansion of Métro service almost as long as the preexisting system, but also stretched project management capacity to the point of collapse, setting up the cost overrun; thus, the current project is being built using institutionally novel techniques including a single-purpose delivery vehicle with some design-build aspect.
- Megaprojects have a large, noticeable impact on the city or region if built; this can be an economic impact as with transport projects, but also a cultural impact, as with the Sydney Opera House, whose factor of 15 cost overrun is a case study in Bent Flyvbjerg’s oeuvre.
In a way, this means megaprojects are defined by cost. A 3 km expansion of the T-bana is not a megaproject, let alone a 3 km expansion of the Istanbul Metro, but a 3 km expansion of the New York City Subway is, because it’s a full order of magnitude costlier. A lower-cost city or country is one that builds more, simply because more projects are cost-effective, and thus it has more projects that are below the threshold of what counts as a megaproject and instead are routine extensions.
The Transit Costs Project hasn’t consciously made any comparison of megaprojects with technically similar non-megaproject transit expansion. The Istanbul case comes closest with its focus on Marmaray and smaller metro projects, but Istanbul Metro expansion writ large should be viewed as a megaproject (it’s certainly planned and politicized as such), and Marmaray is genuinely more technically difficult than just about any other urban rail project. A vulgar quantitative comparison across our database is probably infeasible – there are too few examples by definition, and, moreover, because megaprojects are in practice defined by cost, they’re likelier to be more expensive, even if their specific features do not raise costs.
That said, I do believe that megaprojects are likely to be costlier than equivalent non-megaproject extensions. Stockholm is not a good example for this because it isn’t doing incremental urban rail expansion, only megaprojects. But Paris is a good example, and the per-km cost rose dramatically in the wake of Grand Paris Express. In Barcelona, L9 is a very expensive megaproject; part of its mega- status comes from its worse-than-factor-of-3 cost overrun, but it’s also a large extension of the metro and its construction technique, a large-diameter tunnel boring machine, was new.
Berlin is more complex. We’ll need to wait to see which of the U-Bahn extensions under discussion are built, but those are liminally mega-, sharing some features of megaprojects (namely, political debate, consisting of modal warfare between U-Bahn and streetcar expansion) but not others (they’re not institutionally new and nobody claims they’re transformational for the city). That said, I tilt toward not viewing them as a megaproject, because the debate over them is more general modal warfare, in the same way I don’t think a project subject to debate over spending versus austerity or road versus public transport investment is a megaproject. The key political attribute is not that there’s any political debate, but rather that the political debate introduces politicization of technical decisions over alignment and construction methods; the modal warfare in urban Germany between streetcars and rapid transit is a proxy for much broader fight between consumption- and quality of life-oriented urbanism on the one hand (favoring streetcars and bike lanes) and production- and job access-oriented urbanism on the other (favoring rapid transit and also motorway construction).
You can read it here. It evolved a lot during writing, partly because of the rising costs in the Nordic countries, partly because of the tension between the forward-looking rhetoric of what the report calls the globalized system and what interviewees with more practical involvement have said, partly because of the voluminous literature on models of capital construction and maintenance that only look at Northern Europe or the UK.
In a similar manner to the webinar about the Italian and Turkish cases, there is going to be a webinar about this one. The date is the 20th of September, 17:00 Central European Summer Time (UTC+2); here is the Zoom registration page. It will take the format of a short presentation, around half an hour, to be followed by a Q&A of indefinite duration, and I will try not to be mostly negative – even with the cost overruns, Nya Tunnelbanan is noticeably cheaper per km than the average 2020s subway, and there are a lot of commendable aspects of the Nordic model of infrastructure construction including at least one (labor efficiency) that is superior to the otherwise-cheaper Southern European models.
Institutional knowledge at agencies that build infrastructure shapes up to be an important factor behind how well they handle projects. Good agencies build up a knowledge base over time that lets them see what works and what doesn’t, and this way they’re capable of making in-house planning decisions, and even when they use consultants, they make sure to learn what the consultants have taught them and implement those lessons in the future. In our Italian, Turkish, and (soon to be released) Swedish cases, the agencies have all built up this knowledge over decades.
Denmark provides an interesting test case for this, because Copenhagen opened its metro in 2002 (Helsinki: 1982; Oslo: 1966, Stockholm: 1950), and so it’s possible to compare it with the other Nordic capitals. The construction costs in Copenhagen are notably higher: the City Circle Line (built 2009-19) cost 25,300 DKK for 15.5 km, which in 2022 PPP dollars is around $280 million/km, and the soon-to-open M4 extension to Sydhavn is 9,100 DKK for 4.5 km, or $330 million/km; in contrast, we have the following costs for the other Nordic capitals:
|City||Line||Length||Years||Cost||Cost/km (2022 PPP)|
|Oslo||Løren||1.6||13-16||1.33b NOK||$110 million|
26.4b NOK, ’21
|Stockholm||Nya Tunnelbanan||19||20-30||32b SEK, ’16||$235 million|
|Helsinki||West Metro phase 1||13.5||09-17||1.171b€||$145 million|
|Helsinki||West Metro phase 2||7||14-23||1.159b€||$275 million|
All of these costs are higher than you may have seen in past posts – this is mostly an inflation artifact (and in particular, you should mentally increment all costs by 25% if you remember them in mid-2010s dollars). But it’s notable that in both Oslo and Helsinki, real costs are sharply up; the Fornebu Line is more complex than the Løren Line, but much of its complexity is an engineering choice to deep-mine the stations.
In Stockholm there’s no similar comparison, but Citybanan cost, also in 2022 PPP dollars, $365 million/km, and a factor of 1.5 is an unusually low premium for city center regional rail carrying 250 meter trains over regular metro trains; the RER premium in Paris looks like a factor of 2, and the Munich S-Bahn tunnel was budgeted at a factor of 2 premium over a current U-Bahn extension and has since announced a factor of 2 overrun over that, for which it has been widely mocked in the German press. It’s plausible that when the regional rail premium is netted out properly, Stockholm has in fact seen a large real increase in costs, which matches the history of Nya Tunnelbanan’s cost overrun, from 23 to 32 billion kronor.
Denmark is seeing a real cost increase as well, but a much smaller one. In effect, what’s happening is that Copenhagen started building its metro in the 1990s at higher cost than Nordic norms, and in the generation since then, costs in the other Nordic countries have converged to Danish costs.
So what’s going on?
Some hints can be found in the details of the most recent Danish extension, M4 to Sydhavn. The soft cost multiplier over hard costs is higher than one would find elsewhere, and the contingency is 30% at the contract award, an unusually high figure; 20% is more typical, or even less at contract award (but more during earlier planning). Moreover, the entire project was awarded as a single design-build contract to a joint venture of Vinci and Hochtief, with hard costs worth 460M€.
The entire Nordic world is trying to transition to that style of contracting. This is inspired by British and Dutch models of privatization, which the state, academic, and consultant studies I’ve read while writing the Stockholm report view positively. The procurement strategy for Trafikverket in Sweden calls for transitioning to a so-called “pure client” model for the next big rail investment, Gothenburg’s West Link, like Citybanan not included on the above table as it is a regional rail through-running tunnel. The emerging model in the Nordic countries, which I call globalized in the report since it aims at international competitiveness attracting global contracting firms, can be compared with the traditional model as follows:
|Itemized contracts||Fixed price contracts|
|Smaller contracts (hundreds of millions of kronor)||Larger contracts (billions of kronor)|
|Product procurement (“how to build”)||Functional procurement (“what to build”)|
|Public client risk||Private contractor risk|
The Nordic project I’m most familiar with, Nya Tunnelbanan, does not use the globalized system; it uses elements of both the globalized and the traditional systems, but the trend is to be more globalized. Moreover, the Fornebu Line uses design-bid-build; its problem is partly that the private risk allocation encourages defensive design. If the builder strictly follows the design, all liability is on the designer, otherwise it’s on the builder; thus, the builder strictly follows the design, and because geotechnical surprises are inevitable during tunneling, the designer is overly cautious and tries to anticipate every potential problem rather than seeing what is actually necessary while the tunnel is dug. The traditional system has problems, especially when the risk allocation is improper like this. What’s more, the preference for larger contracts over smaller ones comes from ongoing industry consolidation – there just aren’t enough domestic contractors anymore, and pan-European ones, let alone global ones, are not going to enter an unfamiliar market for a $100 million contract. Unfortunately, the move to privatization of risk under the pure client model does not improve things, and is associated with higher costs.
I am less familiar with the Copenhagen Metro than with the Stockholm Metro, but from reading both how the expansion is done and what Eno is saying about its model (it did a case there but not in Stockholm), Denmark was an early adopter of the globalized system. Eno even pointed out that it uses design-build to showcase that low-construction cost cities use it successfully.
So the Denmark effect is real – this does appear to be a matter of experience. Having never built a metro before – the last urban rail tunnel in Denmark, the S-tog, opened in 1934 – Copenhagen never had the institutional knowledge of how to use the traditional system, so it opted for (elements of) the globalized system, which was not how the other Nordic countries did things but was what British consultants recommended. Note that this does not mean higher costs (that is, around global average, rather than far less) were inevitable in Denmark – it could have adopted the traditional system by leaning on intra-Nordic connections, which are extensive. But perhaps in the 1990s, and certainly in the 2000s, even the other Nordic countries started to come to believe in greater privatization of risk.
The tragedy is that we can see, in real time, how good institutional knowledge is forgotten. Nya Tunnelbanan is, by itself, a pretty straightforward case of cost overrun. But in the context of parallel trends in Helsinki and Oslo, and perhaps an imputation of how much more complex Citybanan was, the situation is different. Real costs increased over time – this was not a mere matter of cost underestimation. Moreover, they increased during a time of ongoing, successful construction of metro projects – the lines that have opened all have healthy ridership, encouraging plans to build even more. And yet, the real problems with the traditional system have led to the adoption of what appears to be a worse procurement system, supported every step of the way by the same agencies that used to compete for world records for low-cost construction.
New York State just announced that per the result of a legal settlement, it is committing to make 95% of the subway accessible… by 2055. Every decade, 80-90 stations will be made accessible, out of 472. Area advocates for disability rights are elated; in addition to those cited in the press release or in the New York Times article covering the news, Effective Transit Alliance colleague Jessica Murray speaks of it as a great win and notes that, “The courts are the only true enforcement mechanism of the Americans with Disabilities Act.” But to me, it’s an example not of the success of the use of the courts for civil rights purposes, in what is called adversarial legalism, but rather its failure. The timeline is a travesty and the system of setting the government against itself with the courts as the ultimate arbiter must be viewed as a dead-end and replaced with stronger administration.
The starting point for what is wrong is that 2055 is, frankly, a disgrace. By the standards of most other old urban metro systems, it is a generation behind. In Berlin, where the U-Bahn opened in 1902, two years before the New York City Subway did, there has been media criticism of BVG for missing its 2022 deadline for full accessibility; 80% of the system is accessible, and BVG says that it will reach 100% in 2024. Madrid is slower, planning only for 82% by 2028, with full accessibility possible in the 2030s. Barcelona is 93% accessible and is in the process of retrofitting its remaining stations. Milan has onerous restrictions such that only one wheelchair user may board each train, but the majority of stations have elevators, and 76% have elevators or stairlifts. In Tokyo, Toei is entirely accessible, and so is nearly the entirety of Tokyo Metro. Even London is 40% accessible, somewhat ahead of New York. Only Paris stands as a less accessible major world metro system.
The primary reason for this is costs. The current program to make 81 stations accessible by 2025 is $5.2 billion. This is $64 million per station, and nearly all are single-line stations requiring three elevators, one between the street and the outside of fare control and one from just inside fare control to each of two side platforms. Berlin usually only requires one elevator as it has island platforms and no fare barriers, but sometimes it needs two at stations with side platforms, and the costs look like 1.5-2 million € per elevator. Madrid the cost per elevator is slightly higher, 3.2 million €. New York, in contrast, spends $20 million, so that a single station in New York is comparable in scope to the entirety of the remainder of the Berlin U-Bahn.
And this is what adversarial legalism can’t fix. The courts can compel the MTA to install elevators, but have no way of ensuring the MTA do so efficiently. They can look at capital plans and decree that a certain proportion be spent on accessibility; seeing $50 billion five-year capital plans, they can say, okay, you need to spend 5-10% of that on subway accessibility. But if the MTA says that a station costs $64 million to retrofit and therefore there is no room in the budget to do it by 2030, the courts have to defer.
This, in turn, is a severe misjudgment of what the purpose of civil rights legislation is. Civil rights laws giving individuals and classes the right to sue the government already presuppose that the government may be racist, sexist, or ableist. This is why they confer individual and group rights to sue under Title VI (racial equality in transportation and other facilities), Title IX (gender equality in education), and the ADA. If the intention was to defer to the judgment of government agencies, no such laws would be necessary.
And yet, the nature of adversarial legalism is that on factual details, courts are forced to defer to government agencies. If the MTA says it costs $64 million to retrofit a station, the courts do not have the power to dismiss managers and hire people who can do it for $10 million. If the MTA says it has friction with utilities, the courts cannot compel the utilities to stop being secretive and share the map of underground infrastructure in the city or to stop being obstructive and start cooperating with the MTA’s contractors when they need to do street work to root an elevator. Judges are competent in legal analysis and incompetent in planning or engineering, and this is the result.
Worse, the adversarial process encourages obstructive behavior. The response to any request from the public or the media soon becomes “make me”; former Capital Construction head and current MTA head Janno Lieber said “file a Freedom of Information request” to a journalist who asked what 400 questions federal regulators asked regarding congestion pricing. Nothing goes forward this way, unless accessibility in 33 years counts, and it shouldn’t.
The Italy case, done by Marco Chitti, is up on the website. I encourage people to read the entire report on how Italy has set things up in the last 20-30 years so as to have one of the lowest-cost urban rail infrastructure programs in the world. The Turkey case, by Elif Ensari, will be up shortly.
This is leading to a webinar, to be done tomorrow at 16:00 my time, 10:00 New York time, in which Marco and Elif will present their cases to the general public. I encourage people to register; you’ll be able to ask us questions and we’ll answer in chat or on video. But if you can’t make it, it will be recorded.
Subway investments can include expansion of the map of lines, for example Second Avenue Subway; proposals for such extensions are affectionately called crayon, a term from London Reconnections that hopped the Pond. But they can also include improvements that are not visible as lines on a map, and yet are visible to passengers in the form of better service: faster, more reliable, more accessible, and more frequent.
Yesterday I asked on Twitter what subway investments people think New York should get, and people mostly gave their crayons. Most people gave the same list of core lines – Second Avenue Subway Phase 2, an extension of the 2 and 5 on Nostrand, an extension of the 4 on Utica, an extension of the N and W to LaGuardia, the ongoing Interborough Express proposal, and an extension of Second Avenue Subway along 125th – but beyond that there’s wide divergence and a lot of people argue over the merits of various extensions. But then an anonymous account that began last year and has 21 followers and yet has proven extremely fluent in the New York transit advocacy conversation, named N_LaGuardia, asked a more interesting question: what non-crayon systemic investments do people think the subway needs?
On the latter question, there seems to be wide agreement among area technical advocates, and as far as I can tell the main advocacy organizations agree on most points. To the extent people gave differing answers in N_LaGuardia’s thread, it was about not thinking of everything at once, or running into the Twitter character limit.
It is unfortunate that many of these features requiring capital construction run into the usual New York problem of excessive construction costs. The same institutional mechanisms that make the region incapable of building much additional extension of the system also frustrate systemwide upgrades to station infrastructure and signaling.
New York has one of the world’s least accessible major metro systems, alongside London and (even worse) Paris. In contrast, Berlin, of similar age, is two-thirds accessible and planned to reach 100% soon, and the same is true of Madrid; Seoul is newer but was not built accessible and retrofits are nearly complete, with the few remaining gaps generating much outrage by people with disabilities.
Unfortunately, like most other forms of capital construction in New York, accessibility retrofits are unusually costly. The elevator retrofits from the last capital plan were $40 million per station, and the next batch is in theory $50 million, with the public-facing estimates saying $70 million with contingency; the range in the European cities with extensive accessibility (that is, not London or Paris) is entirely single-digit million. Nonetheless, this is understood to be a priority in New York and must be accelerated to improve the quality of universal design in the system.
Platform screen doors
The issue of platform screen doors (PSDs) or platform edge doors (PEDs) became salient earlier this year due to a much-publicized homicide by pushing a passenger onto a train, and the MTA eventually agreed to pilot PSDs at three stations. The benefits of PSDs are numerous, including,
- Safety – there are tens of accident and suicide deaths every year from falling onto tracks, in addition to the aforementioned homicide.
- Greater accessibility – people with balance problems have less to worry about from falling onto the track.
- Capacity – PSDs take up platform space but they permit passengers to stand right next to them, and the overall effect is to reduce platform overcrowding at busy times.
- Air cooling – at subway stations with full-height PSDs (which are rare in retrofits but I’m told exist in Seoul), it’s easier to install air conditioning for summer cooling.
The main difficulty is that PSDs require trains to stop at precise locations, to within about a meter, which requires signaling improvements (see below). Moreover, in New York, trains do not yet have consistent door placement, and the lettered lines even have different numbers of doors sometimes (4 per car but the cars can be 60′ or 75′ long) – and the heavily interlined system is such that it’s hard to segregate lines into captive fleets.
But the biggest difficulty, as with accessibility, is again the costs. In the wake of public agitation for PSDs earlier this year, the MTA released as 2019 study saying only 128 stations could be retrofitted with PSDs, at a cost of $7 billion each, or $55 million per station; in Paris, PSDs are installed on Métro lines as they are being automated, at a cost of (per Wikipedia) 4M€ per station of about half the platform length as in New York.
New York relies on ancient signaling for the subway. This leads to multiple problems: maintenance is difficult as the international suppliers no longer make the required spare parts; the signals are designed around the performance specs of generations-old trains and reduce capacity on more modern trains; the signals are confusing to drivers and therefore trains run slower than they can.
To modernize them, New York is going straight to the most advanced system available: CBTC, or communications-based train control, also known as moving-block signaling. This is already done on the L and 7 trains and is under installation on other lines, which are not isolated from the rest of the system. CBTC permits much higher peak capacity in London; in New York, unfortunately, this effect has been weaker because of other constraints, including weak electrical substation capacity and bumper tracks at the terminals of both the L and the 7.
Moreover, in New York, the L train’s performance was derated when CBTC was installed, to reduce brake wear. The effect of such computer control should be the opposite, as computers drive more precisely than humans: in Paris, the automation of Line 1 led to a speed increase of 15-20%, and CBTC even without automation has the same precision level as full automation.
As before, costs form a major barrier. I can’t give the most recent analogs, because such projects tend to bundle a lot of extras, such as new trainsets and PSDs in Paris. In Nuremberg, the first city in the world to permanently convert a preexisting metro system to driverless operations, the cost of just the driverless system is said to have been 110M€ in the late 2000s, for what I believe is 13 km of U2 (U3 was built with driverless operations in mind, and then U2, from which it branches, was converted). It is said that automating U1 should cost 100M€ for 19.5 km, but this project is not happening due to stiff competition for federal funds and therefore its real cost is uncertain. In contrast, Reinvent Albany quotes $636 million for the 7 train in New York, of which $202 million must be excluded as rolling stock conversion; the Flushing Line is 16 km long, so this is still $27 million/km and not the $7-12 million/km of Nuremberg.
The maintenance regime in New York involves heavy slowdowns and capacity restrictions. Trains run 24/7 without any breaks for regular maintenance. Instead, maintenance is done one track at a time during off-peak periods, with flagging rules that slow down trains on adjacent tracks and have gotten more onerous over the last 10-20 years; only recently have planners begun to use temporary barriers to reduce the burden of flagging.
The result of this system is threefold. First, track maintenance productivity is extremely low – the train on an adjacent track slows down as it passes but the work stops as it passes as well. Second, speeds are unreliable off-peak and the timetable is in perpetual firefighting mode. And third, parts of the system are claimed to be incapable of running more than about 16 trains per hour off-peak, which means that if there is any branching, the branches are limited to 8, which is not enough frequency on a major urban metro system.
It takes a small amount of capital spending to increase efficiency of maintenance, through procuring more advanced machinery, installing barriers between tracks, and installing crossovers at appropriate locations. But it takes a large degree of operations and management reform to get there, which is necessary for reducing the high operating costs of the subway.
New York has the most complicated interlining of any global metro network. Only four lines – the 1, 6, 7, and L – run by themselves without any track sharing with other lines. The 2, 3, 4, and 5 share tracks with one another. Then the lettered trains other than the L all share tracks on various segments, without any further segregation. Only some commuter rail networks are more complex than this – and even Tokyo has greater degree of segregation between different trunk lines, despite extensive through-service to commuter rail. The New York way guarantees more direct service on more origin-destination pairs, but at low frequency and with poor speed and reliability.
London, the second most interlined system, has long wanted to reduce interlining to increase capacity. The Northern line traditionally had just one southern segment reverse-branching to two central trunks, combining and splitting into two northern branches. When CBTC opened, the busier of the central trunks got 26 peak trains per hour; the more recent Battersea extension removed the interlining to the south, permitting boosting capacity up to 32 tph, and full deinterlining to the north would boost it to 36 tph, as on the most captive Underground lines.
In New York, it is desirable to remove all reverse-branching. At DeKalb Avenue in Downtown Brooklyn, the interlocking switches the four express (bridge) tracks from an arrangement of the B and D on one track pair and the N and Q on the other to the B and Q on one track pair and the D and N on the other; the process is so complex that every train is delayed two minutes just from the operation of the switches. Everywhere within the system, interlining creates too much dependency between the different trains, so that delays on one line propagate to the others, reducing reliability, speed, and capacity.
Some of the problem is, as usual, about high costs. Rogers Avenue Junction controls the branching of the 2, 3, 4, and 5 trains in Brooklyn, transitioning from the 2 and 3 sharing one track pair and the 4 and 5 sharing another to the 3 and 4 running on dedicated tracks and the 2 and 5 sharing tracks. For a brief segment, the 2, 3 and 5 trains all share tracks. This devastates capacity on both trunk lines, which rank first and third citywide in peak crowding as of the eve of the opening of Second Avenue Subway. There are already internal designs for rebuilding the junction to avoid this problem – at a cost of $300 million.
But some of the problem is also about operating paradigms. New York must move away from the scheduling ideas of the 1920s and 30s and understand that independently-operated lines with dedicated fleets and timetables, with passengers making transfers as appropriate, are more robust and overall better for most riders. DeKalb can be deinterlined with no capital spending at all, and so can Columbus Circle. It’s Rogers and Queens Plaza where spending is ideal (but even then, not strictly required if some operational compromises are made), and the 142nd Street Junction in Harlem where an extensive rebuild is obligatory in order to permit splitting the 2 from the 5 in the Bronx permanently.
Staffing levels in New York are very high. Trains have conductors and not just drivers; this is not globally unheard of (Toronto and some lines in Tokyo still have conductors) but it’s rare. With good enough signaling, a retrofit even for full automation is possible, as in Nuremberg, Paris, and Singapore. Maintenance work is likewise unproductive, not because people don’t work hard, but because they work inefficiently.
Improving this situation involves changes on both sides of the ledger – staffing and service. Conductors have to be cut for efficiency and not all of them can be absorbed by other roles, and the same is true of some station facilities and maintenance functions. In contrast, the low productivity of drivers in New York – they spend around 550 hours a year driving a revenue train whereas Berlin’s drivers, who get 6 weeks of annual paid vacation, scratch 900 – is the result of poor off-peak frequency, and must be resolved through increases in off-peak service that increase efficiency without layoffs.
Ultimate goal: six-minute service
I wrote two years ago about what it would take to ensure every public transit service in New York runs every six minutes off-peak, calling it a six-minute city.
Riders Alliance argues for the same goal, with the hashtag #6minuteservice; I do not know if they were basing this on what I’d written or if it’s convergent evolution. But it’s a good design goal for timetabling, with implications for labor efficiency, maintenance efficiency, the schedule paradigm, and the bus system.
It is fortunate that the agenda of systemwide improvements does not exhibit significant tradeoffs in investment. Other parts of the transit agenda do not need to suffer to implement those improvements. On the contrary, they tend to interact positively: accessibility and PSDs can be combined (and federal law is written in such a way that PSDs void the grandfather clause permitting the subway to keep most of its stations inaccessible), faster and more reliable trains can be run more frequently off-peak, better service means higher ridership and therefore higher demand for extensions. Only the issue of labor exhibits a clear set of losers from the changes, and those can be compensated in a one-time deal.
Moreover, the budget for such an agenda is reasonable, if New York can keep its construction costs under control. At the per-elevator costs of Berlin or Madrid, New York could make its entire network wheelchair-accessible for around $3.5-4 billion. Parisian PSDs, pro-rated to the greater size of New York trains, would be around $10 million a station, or $5 billion systemwide. Full automation at German costs would be maybe $6 billion with triple- and quad-track lines pro-rated. The entire slate of changes required for full deinterlining, including a pocket track for the 3 train at 135th Street, a rebuild of the 36th Street station in Queens, and a connection between Queensboro Plaza and Queens Plaza, should be measured in the hundreds of millions, not billions.
The overall program still goes into double-digit billions; it requires a big push. But this big push is worth two to three years’ worth of current New York City Transit capital spending. A New York that can do this can also add 50-100 km to its subway network and vice versa, all while holding down operating costs to typical first-world levels. For the most part, the planners already know what needs to be done; the hard part is getting construction costs to reasonable levels so that they can do it on the current budget.
The Transit Costs Project adjusts all construction costs for purchasing power parities. This means that, for example, a Chinese subway is converted into dollars not at the exchange rate of $1 = 6.7¥, but at the PPP rate of $1 = 4.2¥; this means that present-day Chinese subways look 1.5 times more expensive in our analysis than in analyses that use exchange rate values, and projects from 10 years ago look twice as expensive. I believe our choice is correct, and would like to explain why, since it has gotten some criticism from serious people, who’s prefer exchange rates.
I started this comparing mature developed countries. The US and Europe have largely separate markets for construction, and so American work is almost entirely done in dollars and European work in euros (or pounds, or kronor, etc.). Japan is likewise very local and so is China. In that case, local costs matter far more than international ones.
But what’s interesting is that even in countries that use imported technology and international consultants and contractors and have low wages, costs are almost entirely local. I wrote about this last year, referencing an article out of India about the small cost impact of indigenization and an interview I made with a Philippine planner who told me 90% of the value of civil works is local. Rolling stock is internationally traded, but we exclude it from our cost estimates whenever possible.
The impact of currency changes
Using PPPs, if a country undergoes a bout of inflation, this should be reflected in changes in construction costs. This is intentional. The example given to me in the critique linked in the lede is that if Bangladeshi food prices rise, then this makes the PPP exchange rate look less favorable (a taka in Bangladesh can then buy less relative to a dollar in the US). But that’s fine – if Bangladeshi food prices rise then this forces Dhaka to pay higher wages to MRT construction workers, so overall it’s just domestic inflation. It’s no different from how, today, we’re seeing nominal construction cost growth in the United States and Europe because of high inflation.
At least the inflation today is moderate by any developing-country standard. Core inflation in the United States is 6%; in Germany it’s 3%. This may introduce third-order errors into the database as we deflate costs to the midpoint of construction. In contrast, 50-60% annual inflation is sustained over years in some middle-income countries like Iran, and then the choice of year for prices has significant impact, to the point that Iranian costs have a significant error bar. But that’s regardless of whether one adjusts for PPP or not, since usually inflation leads to deteriorating terms of trade.
In contrast, if prices are compared in exchange rate terms, then international fluctuations create fictitious changes in construction costs. When China permitted the renminbi to appreciate in the mid-2000s, this would have looked like an increase in costs of about 20% – but the costs of local inputs did not change, so in reality there was no increase in costs. The euro:dollar rate peaked around 1€ = $1.58 in 2008, before tumbling to 1€ = $1.28 in the financial crisis – but nothing material happened that would reduce European construction costs by 19% relative to American ones; right now it’s trading at 1€ = $1.05, but this again does not mean that construction in Europe is suddenly a third cheaper compared with in the US relative to 15 years ago.
Unusual currency values
Some patterns are systemic – richer countries have stronger currencies relative to PPP value than poor countries. But others are not, and it’s important to control for them. A currency can be weak due to the risk of war or disaster; the Taiwanese dollar is unusually weak for how rich Taiwan is, and this should not mean that Taiwanese construction costs are half what they really are. Or it can be strong or weak based on long-term investment proposition: investors will bid up the value of a currency in a country they expect to profit in in the long term, perhaps due to population growth coming from high birthrates or immigration, and this does not mean that today, it builds infrastructure more expensively.
In any of those cases, the unusual value of the currency really reflects capital availability. Capital for investment in Australia is plentiful, but this by itself does not raise its construction costs; capital for investment in Taiwan is scarce, but this certainly does not make it a cheap place to build infrastructure.
In some peripheral countries with unstable currencies, costs are quoted in foreign currency – dollars or euros. Some Turkish contracts are so quoted, and this is also common in Latin America and sometimes Southeast Asia. But ultimately, the vast majority of the contract’s value is paid out in the local currency, not just labor but also locally-made materials like concrete. This creates a weird-looking statistical artifact in which we convert dollars or euros to local currency in exchange rate terms and then back in PPP terms.
This, we do because the quotation of the contract (in dollars or euros) is not the real value. Rather, it comes out of one of two artifacts. The first is data reporting: we rely on international trade media, and those often quote prices in exchange rate dollars or euros, even if the contract is in local currency (and in all cases where we’ve seen both, they match in exchange rate value).
The second is that an international consultancy may demand actual payment in foreign currency as a hedge against currency depreciation; in that case its rate of profit should be dollar- or euro-denominated. However, this again is a small minority of overall contract value. Moreover, if a country’s institutions can’t produce enough capital stability to do business in their own currency, it’s a problem that should be reflected in global indices; ultimately, if costs are higher in PPP terms as a result, this means that the country really does have greater problem affording infrastructure.
A posteriori justification
The above reasoning is all a priori. When I started comparing costs in the early 2010s, I was comparing developed countries and the euro:dollar rate was in flux in the early financial crisis, so I just went with one long-term PPP rate.
However, a posteriori, there is another positive feature of PPP adjustment: it levels the differences in construction costs by income. There is positive correlation between metro cost per km and the GDP per capita of the country the metro is built in, about 0.22, but it comes entirely out of the fact that poorer countries (especially India) build more elevated and fewer subway lines; correcting for this factor, the correlation vanishes. This is as it should be: PPP is a way of averaging out costs in different countries, first because it levels short-term fluctuations such as between different developed countries, and second because exchange rate value is dominated by internationally tradable goods, which are relatively more expensive in poor countries than non-tradable goods like food and housing.
What this says is that infrastructure should be viewed as an average-tradable good, at least a posteriori: its variation in costs across the world is such that there is no correlation with GDP per capita, whereas food prices display positive correlation even after PPP adjustment, and tradables like smartphones display negative correlation (because they cost largely the same in exchange rate terms).
I’m about to complete the report for the Transit Costs Project about Sweden. For the most part, Sweden is a good comparison case: its construction costs for public transport are fairly low, as are those of the rest of Scandinavia, and the projects being built are sound. And yet, the Nordic countries and higher-cost countries in the rest of Northern Europe, that is Germany and the Netherlands, share a common prejudice against Southern Europe, which in the last decade or so has been the world leader in cost-effective infrastructure. (Turkey is very cheap as well but in many ways resembles Southern Europe, complete with having imported Italian expertise early on.)
This is not usually an overt prejudice. Only one person who I’ve talked to openly discounted the idea that Italy could be good at this, and they are not Nordic. But I’ve been reading a lot of material out of Nordic countries discussing future strategy, and it engages in extensive international comparisons but only within Northern Europe, including high-cost Britain, ignoring Southern Europe. The idea that Italians can be associated with good engineering is too alien to Northern Europeans.
The best way to illustrate it is with a toy model, about the concept of livable cities.
Consider the following list of the world’s most livable cities:
The list, to be clear, is completely made up. These are roughly the cities I would expect to see on such a list from half-remembering Monocle’s actual lists and some of the discourse that they generate: they should be Northern European cities or cities of the peripheral (non-US/UK) Anglosphere, and not too big (Berlin might raise eyebrows). These are the cities that urbanist discourse associates with livability.
The thing is, prejudices like “Northern Europe is just more livable” can tolerate a moderate level of heresy. If I made the above list, but put Taipei at a high place shifting all others down and bumping Vancouver, explaining this on grounds like Taipei’s housing affordability, strong mass transit system, and low corona rates (Taiwan spent most of the last two years as a corona fortress, though it’s cracked this month), it could be believed. In effect, Taipei’s status as a hidden gem could be legitimized by its inclusion on a list alongside expected candidates like Vienna and Stockholm.
But if instead the list opened with Taipei, Kaohsiung, Taichung, and Tainan, it would raise eyebrows. This isn’t even because of any real criteria, though they exist (Taiwan’s secondary cities are motorcycle- and auto-oriented, with weak metro systems). It just makes the list too Taiwanese, which is not what one expects from such a list. Ditto if the secondary Taiwanese cities were bumped for other rich Asian cities like Singapore or Seoul; Singapore is firmly in the one-heresy status – it can make such a list if every other city on the list is as expected – but people have certain prejudices of how it operates and certain words they associate with it, some right and some laughably wrong, and “livable” is not among them.
The implication for infrastructure
A single number is more objective than a multi-factor concept like livability. In the case of infrastructure, this is cost per kilometer for subways, and it’s possible to establish that the lowest-cost places for this are Southern Europe (including Turkey), South Korea, and Switzerland. The Nordic countries used to be as cheap but with last decade’s cost overruns are somewhat more expensive to dig in, though still cheaper than anywhere else in the world; Latin America runs the gamut, but some parts of it, like Chile, are Sweden-cheap.
Per the one-heresy rule, the low costs of Spain are decently acknowledged. Bent Flyvbjerg even summarized the planning style of Madrid as an exemplar of low costs recently – and he normally studies cost overruns and planning failures, not recipes for success. But it goes deeper than just this, in a number of ways.
- While Madrid most likely has the world’s lowest urban subway costs, the rest of Southern Europe achieves comparable results and so does South Korea. So it’s important to look at shared features of those places and learn, rather than just treat Spain as an odd case out while sticking with Northern European paradigms.
- Like Italy, Spain has not undergone the creeping privatization of state planning so typical in the UK and, through British soft power, other parts of Northern Europe. Design is done by in-house engineers; there’s extensive public-sector innovation, rather than an attempt to activate private-sector innovation in construction.
- Southern European planning isn’t just cheap, but also good. Metro Milano says that M5 carries 176,000 passengers per day, for a cost of 1.35b€ across both phases; in today’s money it’s around $13,000 per rider, which is fairly low and within the Nordic range. Italian driverless metros push the envelope on throughput measured in peak trains per hour, and should be considered at the frontier of the technology alongside Paris. Milan, Barcelona, and Madrid have all been fairly good at installing barrier-free access to stations, roughly on a par with Berlin; Madrid is planning to go 100% accessible by 2028.
- As a corollary of point #3, there are substantial similarities between Southern and Northern Europe. In particular, both were ravaged by austerity after the financial crisis; Northern Europe quickly recovered economically, but in both, infrastructure investment is lagging. In general, if you keep finding $10,000/rider and $15,000/rider subways to build, you should be spending more money on more subway lines. Turkey is the odd one out in that it builds aggressively, but on other infrastructure matters it should be viewed as part of the European umbrella.
- Italian corruption levels in infrastructure are very low, and from a greater distance this also appears true of Spain. Italy’s governance problems are elsewhere – the institutional problems with tax avoidance drag down the private sector, which has too many family-scale businesses that can’t grow and too few large corporations, and not the public sector.
I’m not going to make a list of the cities with the best urban rail networks in the world, even in jest; people might take this list as authoritative in ways they wouldn’t take a list I made up about livability. But in the same way that there are prejudices that militate in favor of associating livability with Northern Europe and the peripheral Anglosphere, there are prejudices that militate in favor of associating good public transport with Northern and Central Europe and the megacities of rich Asia. All of those places indeed have excellent public transportation, but this is equally true of the largest Southern European cities; Istanbul is lagging but it’s implementing two large metro networks, one for Europe and one for Asia, and already has Marmaray connecting them under the Bosporus.
And what’s more, just as Southern Europe has things to learn from Northern Europe, Northern Europe has things to learn from the South. But it doesn’t come naturally to Germans or Nordics. It’s expected that every list of the best places in Europe on every metric should show a north-south gradient, with France anywhere in between. If something shows the opposite, it must in this schema be unimportant, or even fraudulent. Northerners know that Southerners are lazy and corrupt – when they vacation in Alicante they don’t see anyone work outside the hospitality industry, so they come away with the conclusion that there is no high-skill professional work in the entire country.
But at a time when Germany is building necessary green infrastructure at glacial rates and France and Scandinavia have seen real costs go up maybe 50% in 20 years, it’s necessary to look beyond the prejudice. Madrid, Barcelona, Rome, Milan, Istanbul, Lisbon, and most likely also Athens have to be treated as part of the European core when it comes to urban rail infrastructure, with as much to teach Stockholm as the reverse and more to teach Berlin than the reverse.
Prague has one of Europe’s busiest metros, and what looks like the highest per capita rail ridership in not just Europe but also the non-Tokyo world. And yet, expansion is seeing exploding costs.
In our database, the past extensions in Prague are not especially expensive. The most recent expansion to open was that of Line A to Nemocnine Motol, built 2010-5. It cost 20.2 billion CZK for 6.1 km, or 3.3 billion CZK/km; in PPP dollars, this is around $250 million/km. This is just how much things in Czechia would cost. The previous extension was that of Line C to Letňany, built 2004-8; it cost 15 billion CZK/4.6 km, the same as the later Line A extension per km, and in the interim period, Czechia had practically no inflation. Both lines had a feature that should slightly suppress costs: the Line C extension was partly cut-and-cover and partly bored, and the Line A extension, otherwise fully underground, has a daylit terminus built into the side of a hill.
And now Prague is building Line D, at a far higher cost. The current estimate is 73 billion CZK/10.5 km. This is in PPP terms $540 million per km, making it the most expensive metro (not S-Bahn) line I know of in Continental Europe, and only marginally cheaper per km than the Battersea extension of the Northern line in London.
The map provided in the link shows the line not even going all the way to city center. Its northern terminus, Náměstí Míru, connects with Line A, is in the center, but is just outside the historic core where the three current lines meet; from there the line is to go south, intersecting Line C peripherally and Line B not at all. Nor is the line quite fully underground – like the Nemocnine Motol extension, it has minor daylit segments, including a river crossing, a station, and a depot; overall, it looks 90% underground, not 100%.
I do not know what’s going on there. The Czech economy is growing, but there’s no singular boom that should explain why the 2020s are so profoundly different from the 2000s and 10s. On my Twitch stream, a Czech commenter speculated that the contractor ecosystem is breaking, with only 5-6 contractors, all domestic, and reticence to hire foreign, whereas for example in Sweden there’s a steady influx of Turkish and Chinese contractors, and in the private sector Prague’s construction sites are full of immigrants from poorer countries. But then Skanska was one of the lead contractors for the extension to Letňany.
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.