Transit Costs Website

Go here to see the our construction costs website. The static dataset is here, but I encourage people to go to the site, which has some interesting mapping – in particular, because the coverage is close to comprehensive, it is easier to see where many subways are being built (China!) and where they are not.

There are still gaps in coverage, plus some numbers that I am not perfectly certain about because the projects are still under construction. Please email us if you have corrections or additional data, whether it’s current or historic. For example, I wish I had complete historical data for Paris, Berlin, and Tokyo – in all three cities I have current data, and in the first two I also have early 20th century costs, but I don’t know what the postwar costs were, or the 1930s costs in Berlin. (In London and New York I have better though still imperfect historical costs, they’re just not integrated into the site yet.)

And please thank everyone who has worked on this. The lines in the database that I added are not even a plurality of the database – the Chinese data comes from Yinan Yao, the Arab data comes from Anan Maalouf, we’re adding massive amounts of current and historic Korean data due to Abdirashid Dahir, Marco Chitti has added some Italian data, Eric has been invaluable in checking some of the Spanish-language numbers, and the Turkish data comes from Elif Ensari, who also built the website and is responsible for the data visualization and mapping.


  1. Jacob Manaker

    I hate to be “that guy”, but there’s an encoding error on your website. The pop-up box for Karlsruhe displays “Kombilösung” instead of “Kombilösung”. (In case the same problem affects here: instead ‘o’-with-umlaut, I get ‘A’-with-tilde, followed by paragraph conclusion.)

    Otherwise, it looks great!

      • jcranmer

        From what I can tell, is serving the data as mojibake’d UTF-8, which suggests to me (not knowing your precise server setup) that the data file is UTF-8 and the server is configured to send everything as UTF-8, but there’s some read-the-data-file method that is using the default of Windows-1252(/ISO-8859-1) instead of UTF-8.

    • jcranmer

      The ¶ is actually called the pilcrow. Why it and the ampersand (&) are the only typographical characters with weird names, I have no idea.

      • Herbert

        I know the etymology of “ampersand”

        The “&” symbol is derived from Latin “et” (meaning and) and was often placed at the end of alphabets. To avoid confusion, symbols were often called “per se […]” (meaning in and of itself) so it would be “per se and”. Apparently during rote memorization of the alphabet it would often end with the phrase “and per se and” – you can see how that became slurred to what it is today…

  2. grendelkhan

    Apologies if this is redundant, but I didn’t see the U4 in Berlin on the map.

    The Electric Railway Journal had an article on the Schoeneberg-Berlin Underground Railway, originally opened on December 1, 1910, and now the U4, in the August 5, 1911 issue. Page 230:

    Three-fourths of the total construction cost, namely, 7,919,200 marks ($1,979,800), was required for the building of the tunnel structure.

    I think this means that the total cost was 7.9 million marks, right?

  3. Mikel

    It looks great! Congrats on all this (ongoing) work.

    The scatter plot is a very interesting addition. Since we usually talk in terms of cost per kilometer, may I suggest switching the axes so that cost is the ordinate and length is the abscissa?

    As an aside, if you end up choosing Madrid as a case for study, the upcoming Line 11 extension to Conde de Casal (projected at 69M euros/km after a new station was added in the community review stage) might be worth keeping an eye on because:
    -Metro Madrid has suffered from egregious mismanagement in the last decade, and also hasn’t done any major construction, so it remains to be seen if the engineering know-how built over the 90’s and 00’s remains in place.
    -Unlike many recent projects, it’s an inwards extension, crossing under the river and several existing Metro and Cercanías tunnels, so it doesn’t use cut-and-cover and is more comparable to e.g. the central sections of Lines 7 and 9.

  4. Gok (@Gok)

    I still don’t understand the criteria for inclusion. Why exclude the already-completed Phase 1 and 2 of the BART Silicon Valley extension, but include the non-yet-started phase 3? For that matter, what about the BART SFO extension? In Seattle, there’s only 5km of a 35km system, not even including the 60km currently under construction.

    • Alon Levy

      The criteria for inclusion were “could we find reliable sources for costs?”, “is this rapid transit?” (Stadtbahns count, but entirely at-grade light rail doesn’t), and “is this complete or under construction/engineering?”. BART to SFO and to Berryessa missed the first criterion – if you have cost data I’ll be happy to integrate it. Seattle, same thing, I have costs for U-Link and for some of the future lines and I can look again for the under-construction stuff but IIRC none of the rapid transit is under construction?

      There will be updates with more data, e.g. I found LA Red Line costs as the database was going to press. There’s also extra Korean data in the pipeline, in addition to what people are telling me in comments here and on Twitter.

    • Alon Levy

      Try again now? The bandwidth is more limited than we anticipated, but now that the big wave on Twitter has subsided it should be okay again.

  5. yuuka

    Shanghai Line 20 is just an outer orbital and it’s kind of scaring me. Or is it not “outer orbital” enough and thus still has to deal with the challenges of ground conditions and existing lines?

  6. Patrick Jensen

    What is the cutoff criteria for a project to be considered? For example, the Finnish list might include the Western track addition in Pasila (49 M€ / 1.5 km), but it didn’t add or extend any lines, merely simplified operations.

    I’m asking because I might be able to surface expenditures for rapid transit investments in the Helsinki region between 1970-2005, but many of them were built in such a piecemeal fashion that it’s not always clear what constitutes a project.

    • Alon Levy

      Honestly, any extension of metro service counts, there are some sub-1 km extensions on the list, like the U2 extension to Pankow.

  7. Eric2

    Looks very nice! Some thoughts:

    1) Cost per km is not a very useful comparison when some lines are tunneled and some are not. Maybe you could add another metric to adjust for this, e.g. “cost per adjusted tunneled km” where each tunneled km counts as 1km and each nontunneled km counts as 0.3lkm (or whatever you think is the right ratio)?
    2) What if, rather than showing line details in a new window, you showed them at the bottom of the map page (it’s only 3-4 lines of text after all) while zooming in the map page?
    3) Could the map be used to
    4) The “Projects” tab sounds like projects you are doing, not projects transit authorities are constructing. Maybe another title like “Database” would be better? (And “Raw data” or “Sources” for what is now “Data”)
    5) The data table is very cool, technically. Is it best to have a separate color for each country, or else one color per region (Europe, Anglosphere, China, Southeast Asia, etc)? Could you put a slope=1 line on the graph to show if cost really scales with length? Could you add more graphs with cost compared to other variables, for example % tunneled, or year, or number of projects executed in the city/country, or PPP ratio? Could you add the city name to the tooltip above each point?

    • Alon Levy

      Ad 1: I should follow up on this ratio. But note that we include the tunnel % in the table on the website as the only independent variable (in addition to country, city, and project name), because it matters so much.

      And ad 5, we can do that, yeah.

  8. Paul Davidson

    Really minor, but unless I’m being blind, the data from the MIN calculation in cell R553 looksto be wrong. I can’t find any values below $66.25m

  9. grendelkhan

    Some more digging: the initial segment of the PATCO Speedline, according to the chair of the Delaware River Port Authority reporting to Congress in 1977:

    The Lindenwold Line, built in 1965-1968 for a capital cost of $94 million dollars, is 14.5 miles in length connecting downtown Philadelphia with suburban, southern New Jersey.

    (These are almost certainly not adjusted for inflation, as the same number is quoted in this 1996 report. Note also that the line completed construction in 1968, but opened in 1969.)

    Also, according to a consultant named Peter Alouche:

    In São Paulo, the price of metro company CMSP’s line No. 4 (yellow line) was calculated at US$233mn/km

    It might be worth trying to get in touch with him; he’s still being quoted in articles as of this year.

    • Hugh B

      Two things about PATCO
      1. The Western ~3.5 miles were built in the 1930s along with the Ben Franklin Bridge.
      2. The remaining 11 miles were in a preexisting rail ROW, though I assume there were at least a few grade separations built as part of PATCO, as the previous owner was the Pennsylvania-Reading Seashore line. Using 1967 Dollars, this comes out to ~$66 million per mile.

      • stewartclamen

        The ROW was inherited, but the legacy railroad ran at grade. PATCO has a viaduct (Camden to Westmont), and a trench (Haddonfield). East of Haddonfield it mostly runs at grade.

  10. jonsalmans

    Have you considered including construction costs for privately owned/constructed freight rail? I know there are differences in construction requirements that affect cost between freight and passenger rail, but I have wondered if there are any private sector practices that could be leveraged to improve public sector passenger rail.

    • Alon Levy

      I occasionally see such costs on Railway Gazette, but data is limited. When it comes to North American freight railroads in particular, they don’t really do anything comparable to what urban rail construction involves – they do things like double-tracking lines outside built-up areas, increasing clearance for double-stacked freight in existing tunnels, and installing a signal system optimized for remote lines with way less traffic than suburban rail lines.

  11. df1982

    A correction on Sydney Metro NW: the new line was only 23km and eight stations (15km tunnel, 8km viaduct). In addition to that there was the conversion of an existing 13km/5 stations underground line to metro (which was essentially platform screen doors and new signalling, plus two new platforms at one station). So this would affect your calculations.

    Also I’m not sure where you got the $7.8b figure from. The initial budget was $8.3b, but there were reports that the project came in $1b under budget (not so surprising because the initial budget was very generous with contingencies). Although this amount includes rolling stock and I think maintenance for a set period of time.

    I also have figures for two other lines in Sydney:

    Airport line: 9km (fully tunnelled), five stations, $900m, completed in 2000
    Epping-Chatswood line: 13km (fully tunnelled), five stations, $2.3b, completed in 2008 (this was the line later converted to metro)

    Would have to dig around for sources backing these figures up, but they have been consistently reported as such.

  12. Pingback: Friday’s Headlines for a Long Weekend – Streetsblog USA
  13. Herbert this half hour episode of Germany’s leading regular program on all things railway (tellingly called einsenbahn *Romantik*) has historical footage of the 1934 Berlin S-Bahn construction. Please keep in mind that you’re literally looking at (edited) Nazi propaganda…

  14. Jonathan Stone


    I’d be interested to see the cost per km for New Zealand’s 19702 electrification of the North Island Main Trunk Line, over the volcanic plateau. My recollection is that that electrification was done by Swiss contractors; news at the time said the NZ Govenrment borrowed Swiss francs to finance it.

    IMHO that’d make a great comparison point to the Papakura-to-Pukekohe electrification, once it’s complete. (Contract was signed in August). Head-to-head comparison of German-speaking rail electrification costs, to Anglosphere costs 🙂

    • Alon Levy

      Brightline also imported the train-mounted platform extenders, which the FRA is wisely making a national standard for wheelchair accessibility.

    • Jonathan Stone

      Note the sentence in that article:
      “This is the first time the box-jack system will be used for a train project outside of the Northeast United States.”

      Tentative translation: “This is the first time the box-jack system will be used for a train project in the United States outside of the Northeast.” (LIRR?)

      • adirondacker12800

        Very likely because later in the article it says “Commonly used in Europe for rail construction, the box-jacking method was first used in the U.S. earlier this year on the Long Island Railroad expansion project in New York City. ” Though the grade separation project is likely to be one of the ones in Nassau County.

  15. BindingExport

    Since you included Leipzig City Tunnel, CItybanan and Malmö City Tunnel u should also include Västlänken currently under construction Göteborg 20 billion SEK in 2009 money. 3×4-track underground stations with two island-platforms 200 meters in length as well as 8 km worth of tunneling.

    Trafikverket has every document u need imo for costs including construction methods used on each section on its site.

    Short cost break down

    Detailed planning:—smidigare-pendling-och-effektivare-trafik/Dokument/Jarnvagsplan/

    Project page with lots of documents:—smidigare-pendling-och-effektivare-trafik/

Leave a Reply to Jonathan Stone Cancel reply

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

You are commenting using your 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.