Boston's GLX: a case study in cost disease

2023 October 01

I read with interest Transit Costs Project’s report on the expense of the overbudget and decades-late Boston Green Line Extension, hoping for insight into how it came to be so expensive. This is part of a larger phenomenon sometimes called cost disease, to wit, why does everything (health care, infrastructure, college, etc) cost 10 times what it cost 50 years ago? In New York, for example, building new subway lines costs 20 times as much today (per kilometer, inflation adjusted) than in 1900. This is not purely a temporal phenomenon: subway construction in New York costs similarly much more than in Seoul today.

The latter is shown, for example, in figure 1 of the above report:

The US is the 6th most expensive country to build “rapid-rail transit infrastructure” in, but the situation is worse when accounting for the proportion of track that is underground: in the top 5 countries, at least 80% of new track is tunneled (which of course is much more challenging than surface track), but only 37% in the US.

From the TCP executive summary we find a breakdown of approximate sources of costs based on a case study in NYC:

…from which we can see that there is not one reason the NYC subway is expensive, but rather every reason.

Moving to Boston, table 1 (excerpted below) has a comparison of the most recent major subway expansions:

Project km Tunnel % Cost, million 2021 USD Cost / km
GLX 2012 - 2021 7.6 0 2289 301.2
RLX 1978 - 1985 5.1 100% 1642 321.9
OLX 1966 - 1977 8.6 14% 1156 134.4

(Note that the listed dates for the GLX excludes the planning process that started considerably before 2012.) The green line extension cost as much as the red line extension did 4 decades earlier, despite being entirely above ground, and more than twice as much as the orange line, which at the time had been heavily criticized for delays and going over budget by a factor of two.

However, after reading the TCP report on the GLX, I am unfortunately not left with a clear understanding of what causes the high costs and delays in the abstract, as opposed to the numerous concrete factors that have been identified in specific for any particular project. In synthesizing the GLX report with other retrospective analyses, I have written my own highly speculative list of factors that contribute to successful infrastructure projects.

  1. Leadership. A common feature I have observed in successful projects is having a specific person who is in charge of the project and passionate about delivering it to completion. To some extent this may be an illusion caused by retrospective stories inevitably favoring a great man interpretation, but I think there exists a contribution in practice.

  2. Authority. American governance splits authority across many different actors: the three branches of federal government; considerable executive power being retained by state or even municipal governments; and many different government agencies at each level with competing or overlapping mandates. I dimly recollect reading in The Great Thirst: Californians and Water – A History by Norris Hundley Jr1the only book I’ve read that had 200 pages of endnotes and bibliography that California has 500 state agencies concerning regulating water use; a brief taste of that can be seen also in Governing the Commons by Elinor Ostrom, which has a case study on California water. Such split authority, especially exacerbated by today’s extreme partisanship, helps protect us from would-be autocrats (even with Republican control over all three federal branches, Republicans were impotent to do little more than tear down the administrative apparatus) but hinders productive work.

    Any major infrastructure project will have winners and losers, and a central factor in a project’s success is being led by an agency (or individual!) with the authority to make such decisions unilaterally. Each American infrastructure project has to survive a gauntlet of lawsuits from private interests or the government itself, and hold up against environmental or other regulations that are sensible in general but may become ridiculous in certain contexts; without a central voice saying which conflicting interests are unimportant and which actually matter, the project will expand to encompass all purposes. The GLX project was repeatedly hobbled by interagency conflict and by a desire (or political need?) to appease everyone – for example growing to include the Somerville community path, a walking/biking path at a cost of $30 million per kilometer! Split ownership of projects also causes delays due to needing to seek approval from multiple parties.

    Of course, there are downsides to the extreme opposite: China with its high levels of central authority can churn out massive networks of high-speed rail242000 km since 2008. The US has 80 km of HSR. The California HSR project has been in formal planning since 1996, construction began in 2016, and it is expected to have 192 km of single track in service by 2033, at a cost of $19B., but how safe are they for construction workers and passengers?3…though even the worst rail is far safer than car travel, and Chinese HSR may be the safest HSR globally. Accurate numbers do not exist but it seems Chinese HSR may have 20x the number of total fatalities as American HSR, almost all from people struck by trains. Excessive or superfluous regulation can cripple an industry (cough nuclear power plants) but neither do you want to ignore all regulation as some of it is critical for safety (cough also nuclear power plants). Other nations strike a balance in central authority somewhere between the US and China that, at least to me, seems better than both.

  3. In-house capacity. The MBTA and other transportation agencies have increasingly been handing off work to private contractors that was once done in-house. It is not obvious to me a priori that private contractors would be worse than in-house employees, but observationally this seems to contribute to cost overruns, and the TCP repeatedly cites contracting and lack of in-house expertise to run projects (or, in the case of the MTBA, to even oversee contractors) as directly contributing to costs.

  4. Institutional experience. There has been a general decline in investments in public infrastructure in the US since (roughly) the 70s, and consequentially a loss of expertise as leaders and managers for major projects in the past retired without passing on their skills. This experience decay occurs both in public sectors and in the larger private construction industry. People are not interchangeable, and it is very hard to train someone on how to lead a megaproject; and hiring managers from outside means they will lack agency-specific experience and inter-personal knowledge (see: people are not interchangeable).

  5. Economies of scale. Countries that deliver cheap infrastructure do so by building many copies of the same thing, with minor variations. The main advantage here is not saving actual physical construction costs but in reducing overhead, such as design, management, and regulatory compliance. Most importantly, it reduces uncertainty and therefore interagency friction: if you’re just doing the same thing as last time, everyone already knows what they need to do and how much it is going to cost, and doesn’t have to budget as much money (and other resources!) for potential overruns. Furthermore, you can successively iterate improvements on the same design as you build experience at all levels.

  6. Speed. Going over budget goes hand-in-hand with missing project deadlines. Whenever I hear about major infrastructure projects being completed under budget it always seems to be lightning-fast, too. Being fast means less time for scope creep, changing leadership, changing laws/regulations/policies, and cost sunk into cancellations (all of which happened to the GLX, for example). Also, being fast means less money spent on paying for management and design (and re-design…), which otherwise have unlimited potential to grow, whereas craft labor is usually somewhat tethered to the actual finished project. I would be very curious to see a formal analysis correlating projects going over time versus over budget.

  7. not English speaking. For whatever reason, it is English-speaking countries that are consistently running infrastructure costs into the stratosphere. This has not always been so uniformly the case, and is surely due to some cultural reasons which now appear to be spreading into non-English speaking countries too. TCP writes (executive report page 16):

Across the English-speaking countries other than the United States, there are some striking similarities. The institutional problems we have heard about from experts in Britain, Canada, and Australia are largely the same, and it appears that Canada, Australia, Singapore, Hong Kong, and New Zealand are imitating British practices. Moreover, the cost histories of this region are similar. Where New York began displaying a large construction-costs premium over the rest of the world in the 1930s, London only did between the 1970s and the 1990s, shortly followed by Hong Kong, Canada, and Singapore.

Imitation of British practices is also seen elsewhere: as detailed in the Stockholm case study, some of the British aspects of privatization and devolution of expertise to private consultancies are making their way to Scandinavia. We cannot definitively connect such Anglicization with higher costs, as the process is in its infancy, but the Anglicization process correlates with cost increases across countries.

  1. Political will? Successful infrastructure projects are highly visible, and their value is self-evident to the public, leading to more appetite for future investments. However, despite obvious failures and broad public criticism of repeated cost overruns, public demand for increased investment in the US remains high, making this factor somewhat uncertain. It is unclear to me whether this public demand translates into sufficient political will, as politicians may be reluctant to tie themselves to projects that are certain to go over budget, and are aware of how fickle voters are4eg, each part of Obamacare is individually popular but “Obamacare” as a whole is not.

Of course, this speculative list is just my idle thoughts on the subject… if you want a more carefully thought out and informed take, with specific examples and concrete numbers, you should read the Transit Costs Project’s reports!

Follow RSS/Atom feed for updates.