Stephen Smith’s recent post excoriating high US transit costs left me with a weird feeling that took me a while to figure out exactly. The feeling is primarily about the attitude, but the most telling quote about it is the following attack on East Side Access:

East Side Access, the most expensive project, is overpriced by about 1,000%. (Compared to Spain, the world leader in low-cost subway construction, the project is on the order of 10,000% too expensive.) And even in San Francisco, the Central Subway project, which will cost $500 million per kilometer, is – and I’m being generous here – about three-quarters waste.

It is completely true that East Side Access cost a hundred times more per kilometer as two recent commuter rail tunnels in Spain, but it doesn’t really capture the size of the construction cost problem. The range of costs worldwide is quite high; the large majority of projects are significantly more expensive than the single cheapest example. One could just as well criticize Paris’s plan to extend the RER E from its Saint-Lazare terminus to La Défense for costing €1.58-2.18 billion for 8 km of tunnel (see PDF-pp. 59 and 79); although the per-kilometer cost is average for the complexity of the project, and the per-rider cost is also low, it’s still much higher than the two cheapest Spanish projects, by a factor of about 5.

When I write about cost control, beyond just collecting information that isn’t otherwise available in one place, I keep two things in mind:

1. Am I comparing the project in question to the average, including some above-average-cost projects, or to one cheap outlier?

2. Is the project reasonably cost-effective at its present cost, independently of the fact that it could be done for cheaper?

For American subway projects, the answer to question 1 is unambiguously yes. Although there has to be a most expensive line, the projects in the US are persistently more expensive than outside the Anglosphere, and by a large factor, close to but not quite a full order of magnitude. I have less data for light rail and above-ground rapid transit projects, but the non-US numbers I do have are a fraction of some US projects, and American projects that seem more affordable are often very minimalistic, merely upgrading existing tracks to urban rail standards instead of doing construction on city streets. That said, the difference is not 10,000%; to get even 1,000% we need to start looking for the more expensive US projects or the cheaper European projects.

But the answer to question 2 is not always no. As construction costs decline, cities and countries start building more marginal lines, so that the construction cost per rider of urban transit, or the profitability of intercity rail, is not very low. Conversely, some very expensive projects are also so well-patronized that good transit advocates should not oppose them even as they push for cost savings in the future.

For example, Madrid’s MetroSur, built for about $1.7 billion in today’s money, or $45 million per km, gets only 140,000–170,000 riders per day, for a total of around $10,000 per rider. This is fine, but not very low, since the very low construction costs are matched with low ridership per kilometer, more comparable to a tramway than to a subway; most Parisian projects are considerably cheaper per rider, even though Paris builds on-street light rail for the same cost Madrid builds tunnels. In contrast, Second Avenue Subway is about $25,000 per projected rider, high by non-US standards but not obscenely so; I know of no cheaper project in the US under construction right now, including some with quite reasonable per-km costs. New York’s high construction costs mean that the only projects that can pass muster are ones that would set records for cost-effectiveness at normal costs and are still okay at elevated local costs.

The advantage of looking at low-cost outliers like Spain or Calgary is not that American projects are so much more expensive. It’s that we can look at what they do right and imitate some of their practices, in the hopes of getting some of the cost reduction. But it’s important to remember that they’re outliers, and the goal should be to have average costs, not a fraction of the average. To a good approximation, a subway in a dense city will cost $250 million per kilometer – and judging by the low density of the area around MetroSur and conversely the cost escalations on Barcelona’s L9, that’s true even in Spain. It’s possible to do better, but not so much better that it’s worth scuttling lines over.

The lines that are cost-ineffective in the US tend to be the kind that would be bad even at normal cost. In Europe, few of these are built. Those lines – BART’s Livermore extension, Los Angeles’ Foothills Extension, and New York’s 7 extension are favorite punching bags of local transit activists, even relatively political ones – are not necessarily the most expensive, but they’re the most cost-ineffective. The lines that would have been successful at normal costs are not even being proposed: high costs are making them unworkable, and they usually lack value as developer-oriented transit to make players push for them regardless.

The value of international comparisons then is not really for single items or for precise estimates. It’s a first-order estimate inherently. It’s useful as a reality check on certain claims: that it’s unsafe to have a single operator and no conductors on a train with a thousand passengers, that urban transit cannot run on a predictable schedule, that deep-level construction is always preferably to shallow construction. But this is useful exactly because the counters to such claims are frequently universal that claims of special circumstances are less credible – for example, nearly all subway systems in the world run with one employee.

The other problem with trying to rely on case studies of cheap outliers is that the reasons some places have higher construction costs than others may not be the same as those that the builders think. For example, the list of factors Calgary cites as reasons for its low construction costs include its standardized equipment, proof-of-payment system with high discounts for season passes, and a minimum of tunnels and viaducts. Those are fairly normal on American light rail lines as well; they distinguish the C-Train more from more expensive (and vendor-limited) Canadian subway systems.

In reality, the differences are subtler, involving contracting practices, and the health of the local political system. It’s of course not easy to think of Spain, Turkey, and Italy as leaders of good government and of Germany and the Netherlands as Continental Europe’s high-cost leaders, but government on the agency level works differently from on the national level. The US scores very poorly on this measure, with a transportation-industrial complex that sees transit revival as a grand national project, one that like all the previous ones is about image and not about prudence.

The importance of this more political and institutional view is that it’s not enough to just say construction costs should be lower. Insofar as reducing costs is a matter of increasing efficiency, it is essentially a form of economic growth; economic growth happens in spurts in individual industries, and rapid cost controls are possible, but not instantaneous ones. There’s a multi-century average of economic growth, of a little less than 2% per capita in developed countries, and thinking that those efficiency measures will average out to much more is unwise.

Moreover, the way rapid efficiency measures are usually implemented is not one that causes efficiency. I think this is what concerned me the most about Stephen’s article: it’s the implication that all US transit needs is an outsider like me giving it an honest look. I think what I do is interesting, but without very deep insider knowledge, and the cooperation of the trained workforce, it’s not going to lead to much. If I were given a detailed cost breakdown of subway operation in New York and Tokyo, I’d probably be able to see a large number of potential savings in New York. Maybe four out of five would work if I knew what I was doing and were careful enough; one out of five would instead lead to a disaster. It wouldn’t be possible to know in advance which one it would be; it would often not be possible to even know after the fact what caused the problem and what could remain reformed. The best a reformist like that could do is do everything quickly and move on before the edifice collapses; even then, eventually scandal would catch up, as it did to Chainsaw Al.

The process of reform from outside tends to fail for precisely this reason. The outside reformer has no use for insiders – he scorns them, and they return the favor. The Atlantic identifies Newt Gingrich with this mentality, but there are better, less national examples. In Israel, it’s identified with waste in the military and in business: leaders make themselves indispensable by constantly reorganizing everything to make themselves look important. Second, in the US, Bloomberg and allied reformists have a similar mentality, of running the city like the businesses they are used to. As a result, Bloomberg is unable to achieve anything that required the cooperation of people who are not his subordinates; this was made painfully obvious by the failure of congestion pricing.

The alternative to this process is much more painful, but more reliable, in both cases because it’s by design slower. It requires multiple levels of government to come together, inject money into new capital construction, and then add service in such a way that workers lost to efficiency improvements can be reassigned one-to-one to new service. For example, if Amtrak builds high-speed rail in the Northeast well, it will need to hire thousands of new trackworkers and other employees, and could potentially make an agreement to take redundant commuter rail employees in exchange for running faster and more frequently on commuter rail-owned tracks. Such agreements are necessarily complex, requiring the consent of multiple agencies and unions, but are the only way to secure insider support and knowledge for reform.

Recall that in Japan’s great shedding of mainline rail workforce immediately before and after JNR privatization, Japan was undergoing an economic boom, and the government made an effort to find the laid off employees private-sector work. Since the US is not in that position, it needs to find another way, and a spurt of growth in off-peak mainline service could partially do it; in combination with some FRA reforms, it could allow much better service for the same operating cost, using ridership gains to reduce state subsidies. Even then, it wouldn’t be enough everywhere, not with an agency as big as New York City Transit, and as in Japan the entire process could take decades of attrition. In construction, it’s technically simpler to reform work rules, and it is possible that new projects could be authorized more or less simultaneously, so that the cost reductions would be directed to more service rather than fewer construction jobs.

But what would not work is to decree that costs must be lower, and cancel all projects that don’t meet those goals. Cancellation threats on marginal projects could work; on the other hand, the worst projects are typically those with the most backing by power brokers, and since they’re justified by reasons other than cost-effectiveness, a more hawkish line on cost effectiveness would not reduce their support. Actual cancellations of unfixably bad projects could also work. But more than a Chainsaw Al style of management is needed here. What I write about comparative construction costs may be the beginning, but is not more than that, certainly not the end.