I Gave a Talk About Construction Costs

Two years ago, I gave a talk at NYU about regional rail, and as promised, uploaded slides the next day for discussion. Yesterday I gave another such talk, about construction costs.

But here there are two things to upload: the slides, and the data table. I’ve been intermittently adding cities to a spreadsheet of various urban rapid transit lines and their construction costs, and by now there is a total of 207 distinct items, ranging from 1 km extensions to 3-figure packages like 200-km GPX and 160-km Delhi Metro phases. The total length of the lines in this database right now is 3610 km, of which 2090 are underground. These are almost exclusively new lines – most of them aren’t even open, and most of the rest opened this decade, so be cautious since much of the cost estimation is ex ante and a number of the soon-to-open line on the list have had serious cost overruns.

I hope people make use of this dataset and the preliminary analysis contained in the slides, and I ask that people look at both, since the slides do have some interpretive notes about confounding variables. One note that I did not include in the slides and explained verbally is what source means in the table: media means I’m drawing costs from popular media, trade means trade media like Railway Gazette, plan means official plans (either ex ante or ex post), wiki means Wikipedia (as always, a reliable source for line length and station count, never cost), measured means I measured line length on Google Earth lacking any alternative. One item, Crossrail, has its tunnel cost coming from a freedom of information request submitted by an alert reader who I will credit upon request; the headline budget is somewhat higher as it includes surface improvements, a common confounder for regional rail projects (the RER E extension, for example, splits its budget about 50/50 between the tunnel and above-ground works).

More detailed analysis is forthcoming, either here or in print.


  1. Shaul Picker

    It was a great talk last night. Great job. I will look to add some of the sources you found for construction costs to Wikipedia when I have more time.

  2. Adam

    Prediction, Los Angeles selectively adopts cut and cover for some future projects but their design-build perverted cut and cover winds up taking both longer than their normal (station palace plus bored tunnels) methods and is equal to or more expensive per km than said methods.

    And Scope creep results in LA doing cut and cover subway lines to a minimum depth of forty meters.

  3. Eric

    Very impressive work!

    Two things I think would be useful additions to the table:
    1) A “figure of merit” for the equivalent number of subway-miles. Add the number of subway-miles to 1/2 the number of elevated-miles and 1/5 the number of at-grade-miles (or whatever other fractions you think are appropriate), and show the resulting number. This would allow for an apples-to-apples comparison of the cost per mile between lines that have any proportion of subway, elevated, and at grade.
    2) The length of station platforms in meters. Some lines have 50 meter platforms while others have 200 meter platforms. Presumably longer platforms/stations are much more expensive (if not quite in proportion to the length), and since stations are a large overall cost, they make the whole line significantly more expensive. But frequently this expense is justified by the greatly increased capacity of the line.

    • yuuka

      Another thing worth adding is the size of a train car – since that dictates (or is dictated by) alignment which factors into land acquisition/eminent domain costs, as well as tunnelling costs.

      Things like tolerance of curved platforms could also have a factor, but this is just too detailed to be able to do simple side by side comparisons.

    • Alon Levy

      Ad 1, the problem is that the subway-to-elevated ratio isn’t quite the same everywhere. In particular, it’s very high in New York, maybe 5. But yes, I could try imputing this for most cities.

      Ad 2, yes, it’s a very good idea. But I will caution that station depth matters as well, which is how the cut-and-cover digs in Paris aren’t thaaaaaaat much smaller than the mined SAS stations.

  4. Witold Bartnik

    Hi Alon!
    Writing from Warsaw I would like to suggest some corrections to our rows in your table:
    1) The numbers for the central part of the second line in your table (row 23) include rolling stock. The construction contract was for 4.17 billion PLN (https://pl.wikipedia.org/wiki/Linia_M2_metra_w_Warszawie#Odcinek_centralny). Please note this was a design-and-build.
    2) The second phase, the so called “3+3” includes 6 stations (row 24) and was a build-only contract (with two separate contractors for east and west extension). The numbers in PLN are (together for both) 2 214 741 074 PLN = ~2.21 billion PLN. This is significantly less than the 750 million EUR because that number also includes modernisation of the Kabaty depot and 13 new trains. The eastern phase has indeed been completed in 2019, the western will open April 4th, 2020. Length is 3.4 + 3.2= 6.6 km.
    3) The next phases which are all under construction now are the following:
    3a) Last 4 stations in the west and a new depot for the cost of 1 622 951 790 = ~1.62 billion PLN. The length of the tunnels here is ca. 3.9km.
    3b) 2 stations connecting the 3a) segment and the existing fragment for the cost of 959 249 940 = ~0.96 billion PLN. Length is 2km
    3c) Last 3 stations in the east for the cost of 1 397 895 000 – ~1.4 billion PLN- this is the one that is mentioned in your table, but with 4 stations where there should be three. Length is 3.9km.
    You can see it all here: https://www.metro.waw.pl/ii-linia-metra-109
    The red line is the central fragment from 1), green and dark blue are the 3+3 extensions from 2), light blue is the depot and last fragment from 3a), orange is the 3b) segment and yellow is 3c). All the costs are sourced at Polish wikipedia on 2nd line which itself uses the official metro site as the source (www.metro.waw.pl).

    Best regards from your avid reader!

      • Witold Bartnik

        No problem. Keep up the good work! I will keep a watchful eye on those Polish numbers 🙂

  5. yuuka

    I’m not sure I fully agree with the part about simplifying station designs. The assumption seems to be whatever worked in the 1910s will work now, which clearly isn’t the case.

    It would be nice if consultants didn’t exist and everything could be done in house, but try getting Jarrett Walker to work for TriMet, WMATA, and MARTA all at the same time, for one. There are only so many people in the world who know how to build subways. “Let the designers choose their own materials, methods, etc.—give them a technical score and make that 50% of the weight of the bid.” Isn’t this basically design-build?

    I think there’s a case of the new headhouse at 72/Broadway on the 1 2 3 in NY being a far more dangerous replacement for the original street entrances, since there was now a need to cross the road. The alternative would be building them in plazas, but 1) you’d need to find the space and 2) you’d need to tunnel under buildings, which could be trading one issue for another. At the sort of shallow depths manageable with only lifts and stairs, you also can’t do TBM as well so you’re talking more cut and cover and more land acquisition.

    Also, I don’t believe you’d be able to put a 200m long station beneath a 4-lane road in 18 months, even with third world labour costs meaning you can throw tens of thousands of warm bodies at a subway project. You’re thinking a minimum of three years, provided no incidents happen. Don’t forget the disruption which would result from shutting down the neighbourhood for that long, which would probably result in nasty things happening at the ballot box.

    Even the Chinese, with no ballot boxes and a lot more ways to quell dissent, don’t do that. It’s just not worth making the transit agency the enemy of half the city, and extremely counterproductive too. (speaking of which, Chinese Wikipedia has quite a few info on the costs of the Shanghai Metro, I’d translate if I have the time or just use google)

    • Henry Miller

      Expertise can be grown, and probably should be. The consultants got their expertise somehow, you can do the same.

      I’d start by hiring smart people with useful degrees and fluency in Spanish or Turkish and sending them to Barcelona, Madrid, and Turkey (Istanbul, or are their other cities) to learn for a year. I think the transit departments in each city would agree to help train them (though legal issues might get in the way – this is why I don’t suggest hiring people away). When they come home, compare notes and start designing a small subway. Small is key, a couple km to learn from while containing how high cost overruns could get, it wouldn’t be the most valuable route (probably some outer suburb where I can put in a train maintenance shop to an inner suburb – but I’d ensure it goes from areas zoned for density), once that is running have them build an extension towards downtown, then start building more and larger useful lines.

      Of course the above plan means I first need some assurance of funding that won’t dry up. Ridership of that first line won’t be good until the system is much more complete: I need some assurance I will keep building for the next 20 years because that it what it takes to get a useful system in place in a reasonable budget. If I can’t get that the first line will go over budget (It will!), but I won’t have any lessons learned to get costs down for the next line. Of course I will also need some sort of metric of when to abandon my hopes – it may be there is some factor meaning costs can never be reasonable in the US. (ie I may discover that the real problem is some obscure report for a powerful special interest that nobody has even mentioned previously and so I’ll never be able to make stations at a reasonable price)

      If you are in a large city you should do the above. However make sure you are developing depth in your planning organization. If you get costs down other cities will look to poach your planners to do subways in your city. The more you get costs down the more your people will be in demand elsewhere.

      • Henry Miller

        A 20 year plan is also a good idea anyway. Planners who look over the city will start saying “in 10 years I’m going to build a station about there, so lets buy that lot now while it is on sale. There is even opportunity to tell the water department while they are digging up the street to fix a failed main that the transit department is going to divert resources to put in 50 feet of subway tunnel in that area while the street is torn up anyway (thus putting 10% of the costs on the water department)

        • yuuka

          A 20 year plan only works so far as it’s actually followed through.

          SAS has been on the agenda since 1920.

      • yuuka

        >If you get costs down other cities will look to poach your planners to do subways in your city. The more you get costs down the more your people will be in demand elsewhere.

        At that point you might as well start your own consultancy, so you’ve lived long enough to become the villain? (see: LT International)

        I don’t think it’s simply “planners”, you’re also talking engineering staff and project management to actually build the damn thing. Good to have in house, but thinking you can tie your people down by the “noble” term of “public service” ignores the fact that free movement of labour is a thing, and that also doesn’t necessary work out when the private sector is paying just that much more, perhaps with opportunities to work on more “exciting” projects elsewhere thrown in too.

        You simply won’t be getting your 50% savings if that’s the case. Arguably it would probably pan out better if this was done at the federal level or something, like how you have JRTT in Japan (https://en.wikipedia.org/wiki/Japan_Railway_Construction,_Transport_and_Technology_Agency) that manages construction of some rail lines. But given the competition for limited central government resources, you have JR Central running off to build the Chuo Shinkansen on their own, and that’s been a shitshow so far.

        • Alon Levy

          Having centralized state services for this would work well too, and I suspect this may be a better solution in the US specifically just because the challenges in so many different cities are similar (though New York’s are somewhat different).

          The way to get project managers is not to talk about the nobility of public service, of course, but to increase the prestige of working in that capacity for the public sector. This means mostly paying competitively; the figures I was told in Boston show a large private-sector premium, but it’s not one that’s unseemly for the public sector to offer, $140,000 vs. $106,000. In fact, one job category, resident engineer (just below project manager), is paid competitively, $104,000 at the MBTA (yes, just $2,000 less than the manager). This is the result of something called collision, in which union payscales are such that the supervisor gets paid less than a senior line worker and thus promoting senior line workers to management is not possible: there was a collision between construction inspectors and resident engineers, so the MBTA bumped up the resident engineer pay and now there’s a near-collision with project managers.

          Non-monetary prestige is more nebulous. Agencies could just hire higher-class people, but then they’d lose the existing talent (there’s a lot of talent in the US public sector that is there because they’re overlooked in the private sector, e.g. because they went to the wrong uni). A better way to do it is to stop politically demagoguing against the civil service; most public transit agencies are not Metro-North, most public-sector workers are not senior management, and generally the worst people in New York are the political appointees, not the professionals.

    • Alon Levy

      A couple points:

      1. No, 50% technical score is not design-build. There are separate contracts for design and construction, evaluated on a combo of a technical score, how fast you can complete the contract, and cost.

      2. Madrid’s trains are 110-115 meters long, New York’s are 165-180 on the new lines built, it can’t be such a huge difference.

      3. Crossing the street is fine in Berlin – the medians are in roads that aren’t so wide, so pedestrians only have to cross 1-2 lanes. It could work well on Upper Thomson Road, which isn’t any wider than Brunnen Strasse, where U8 runs.

      4. Ad “even the Chinese,” China is not a low-construction cost country, and evidently Continental European democracies build stations this way and it’s fine (often with side platforms rather than island platforms, which mostly just drives up elevator construction costs).

      5. Ooh, can you send some links re Shanghai? I poked around a bit and saw some cost items but without references I could click links on, even in Chinese.

      • yuuka

        1. I believe part of “how fast you can complete the contract” also includes “do you have any better way to build it?” – which apparently LU did with some success for the Bank upgrade. https://www.londonreconnections.com/2013/bank-station-part-3-revised-upgrade-proposals/

        2. The LTA loves to trumpet that the longest station structures on the Downtown Line are Fort Canning and Upper Changi, both over 200m long. And for a 70m long train! (presumably the interlockings integrated into the station structure, with station ancillaries above, play a part too) It can, if you put it in absolute percentages.

        3. Not apparently in Asia – they like to have pedestrians avoid having to cross the street, which is understandable given that the scale here is, to be honest, quite different from Berlin. See the underground arcades of Tokyo and Taipei. (RIP the one in Beijing)

        4. The last heavy rail stations I know of being built this way are in Taipei on the Green Line. Copenhagen is using plazas, and they get away with it because their trains are so tiny, allowing for much smaller stations and more flexible routing of the tunnels.

        5. https://zh.wikipedia.org/wiki/上海地铁#项目投资金额一览

        I punch my calculator and it tells me they average below 100 million USD. Would like to check the references but it appears the Great Firewall doesn’t like my internet connection, though the one or two I pulled look legit even if it takes some searching.

      • Marco

        Hi Alon,

        I’ve been following your work on construction costs and transportation for a while now, since I got stuck by the very high construction costs in public works when I moved to Canada 5 years ago.
        Coming from Italy, I can tell you that some of your points resounds pretty well with me.
        In your table you have some italian examples and the most expensive one, Rome C line, has been precisely blamed for using a so called “general contractor”, similar to a design-build, awarding to a private consortium both the design and the construction based on a very preliminary project and cost estimates. Costs has since incresed because of “unexpected” archeological and geological findings. The same formula has been used since the 1990s for major infrastructures, like the M.O.S.E. system in Venice lagoon and some of older section of the High Speed network. This typology of contract as been blamed by the anticorruption authority has one of the most plagued by corruption and/or mismanagment.

        The general contractor has been emended recently and made less attractive (but not completely repealed) by the new “codice degli appalti”, the comprehensive regulatory framework for public contracts. It has since moved back to a preference for
        aseparation between design and build contract and the use of both technical and costs scores is now the norm.

        Another thing the struck me here is the fact that there are no “reference prices” for construction costs. detailed cost estimations in both private and public sectors in Italy relay on detailed tables of “reference prices” compiled by the provincial chambers of commerce (prices, especially of labor, vary quite a lot from north to south). This prices are not mandatory bat are the official baseline for public works awarding. The only time I was able to see detailed pricing here in Montreal, on some modification proposed for the REM project, I found that earthworks were 7 times (!) more exepnesive by cubic meter than prices in northern Italy, where I used to work. Mining and removal of hard rock 12 times higher, more than 900 CAD/cubic meter compared to 40-50 (!!).

        By the way, I was guessing if the prices in your table are fully comparable. For example some includes rolling stock, others not, and/or financing costs. The Milan price you quoted for M4 includes rolling stock purchase, but excludes TVA or financing costs for example. The estimates I’ve seen for blue line extension in Montreal don’t, but they add the financing costs (weird practice from my perspective…), that in Italy is never included in the price tag.
        here you have a breakdown of costs for the blue line in Montreal : http://www.stm.info/sites/default/files/affairespubliques/Communiques/ap_ligne_bleue_briefing_23_avril_final.pdf
        and here you have the cost breckdown for Milan, for example. Rolling stock (materiale rotabile) is priced at 249 mil and it’s counted in the construction costs:

        Anyway, you are doing a great and necessary work! There is indeed a problem in how public works are done across north america. If needed,I can help with italian and french documents search 😉

        • Alon Levy

          I’m trying to not include financing costs, just construction. And I’ll update the Montreal and Milan numbers accordingly, thanks.

        • Nathanael

          Personally I think the absolute key is a record of blacklisting chiselling and corrupt bad-actor contractors. Firing ONE contractor and blacklisting them brought costs down massively on the Green Line extension in Boston. Just making ONE contractor an example. ONE blacklisting chased out the corrupt contractors and got good contractors to bid.

  6. seb

    What are your thoughts on the construction of the Grand Paris Express, it seems quite expensive, as there average interstation distance is 2km and with only 100m and 50m trains. The stations are constructed in a cut and cover manner with the same system for nearly all the stations, but so many stations have a depth of 30 to 40 meters which seems unnecessary, especially in suburbia. In general, Societe de Grand Paris manages the project, which was specifically created for it, so there was no in house knowledge.

    • Alon Levy

      Yeah, the lack of prior project management experience is troubling, esp. since there is clearly a wealth of such experience within the region (for Eole, Météor, 2000s extensions, etc.), and my understanding is that this is why costs are running over the original projection.

      Do you know if all (or most) stations are built at such great depth, or only ones in difficult areas, like La Défense, Saint-Denis, Saint-Cloud, etc.?

      Mostly my criticism of the line so far isn’t how it’s being built, but where. M18 is pretty useless and M17 is hardly better; instead, they should have committed money, IIRC around 1 billion euros, to quadruple-track the RER B+D tunnel.

        • yuuka

          If I recall correctly, 20m or something is the minimum viable depth for tunnels to be constructed with TBM, won’t be surprised if there are stations in Copenhagen of similar depth or even deeper.

          Anything above that and you’d have to cut and cover the entire route, since you wouldn’t have much soil cover to shield the TBM from whatever on the street.

  7. Michael Schaeffer

    Look forward to further analysis. Perhaps Nostrand Avenue extension can be built cut and cover, but with side platforms.

  8. Witold Bartnik

    Do you have some sort of a spam mechanism? I posted a comment on Warsaw numbers but it didn’t show.

  9. Pingback: I Gave a Talk About Transportation in Connecticut | Pedestrian Observations
  10. Tom M

    Have you considered the recent (last ~15 years) projects in Perth, Western Australia? Two of note are the Mandurah rail extension (notably the short tunnelling section in the central city area), and the more recent extension from Forrestfield to the Airport.

    Forrestfield-Airport Link key stats:
    – length 8.5 km dual-track 100% (tbc) TBM tunnel
    – Three new stations
    – Design, construct and 10-year maintenance contract, all with the same consortium (often seen in Australian public sector infrastructure projects)
    – Construction start 2016, opening 2020
    – Total cost of AUD 1.86 billion (2018 $) – note this includes the 10-year maintenance costs as well
    Reference: https://www.forrestfieldairportlink.wa.gov.au/

    Mandurah Rail Extension (all costs 2003 $)
    – Opened 2007
    – Total length 72 km, 11 new stations
    – Work split into packages, some all design and construction, some construction only packages
    – Package A: 70 km of twin above-ground rail from the city centre to Mandurah, AUD 310M
    – Package B: three stations, AUD 32M
    – Package C: three stations, AUD 38M
    – Package D: three stations, AUD 32M
    – Package E: bridge upgrades as part of the mainline going south for AUD99M
    – Package F: 600m of cut-and-cover and 700m of twin bored tunnel, two stations in the city centre, AUD324M, design + construct + 10-year maintenance contract
    Reference: Archive media releases at https://www.pta.wa.gov.au/news/media-statements/

  11. Pingback: I’m Giving a Talk About Construction Costs Tomorrow | Pedestrian Observations

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.