And how the part-political, part-psychological problem can be avoided.
Last week the San Francisco Chronicle reported that the Transbay Transit Center, a massive transportation hub calling itself the "Grand Central Station of the West," will cost at least $300 million more than project officials estimated. One city official characterized the situation as unfortunate but said it wouldn't have a "meaningful impact" on the project. The comment may have been meant as optimism, but it also reflects the fact that enormous cost overruns have become such a normal part of urban mega-projects that they barely even register as a problem.
So how did it get to the point where the only thing we can confidently expect from a big infrastructure project is that it will cost way more than expected?
One thing's for sure: the people who predict the cost of urban mega-projects do a terrible job. Several years ago the University of Oxford scholar Bent Flyvbjerg, who's made a career researching mega-project mismanagement, analyzed 258 transportation infrastructure projects from around the world and found that nine in ten exceeded their cost estimates. The overruns were greater on rail projects than road projects but averaged 28 percent across the board.
What struck Flyvbjerg most about the problem was how very un-random it was. If people were simply very bad at estimating the costs of huge projects, then one might expect some projects to come in under budget and others over. But an under-budget mega-project is about as rare as a dodo riding a unicorn. Instead, wrote Flyvbjerg and some collaborators in 2002, it's more likely that when it comes to mega-projects, public officials engage in "strategic misrepresentation" — aka lying:
The policy implications are clear: legislators, administrators, investors, media representatives, and members of the public who value honest numbers should not trust cost estimates and cost-benefit analyses produced by project promoters and their analysts.
Flyvbjerg's explanation is no doubt true in some cases, but there's also a less sinister reason why people associated with a project might be bad at predicting its costs. From a psychological standpoint, people are saddled with a cognitive bias that causes them to be unjustifiably upbeat (some might say delusional) about the prospects of their own plans. So they do whatever it takes to get them approved — certain that whatever problems have plagued others in the past will be avoided.
Writing in the late 1970s, the psychologists (and would-be Nobel Laureates) Daniel Kahneman and Amos Tversky called this problem the "planning fallacy" [PDF]. At its root is a tendency for people to see their plans from the inside rather than evaluating them as an external referee might. Whereas an outside observer might consider a project and recognize its "distributional" risks — the combined possibility of a delivery delay or a worker strike or plain bad weather, for instance — the person directly involved might be blind to these possibilities, write Kahneman and Tversky:
The prevalent tendency to underweight, or ignore, distributional information is perhaps the major error of intuitive prediction.
Whatever the underlying causes of cost overruns may be, there seems to be one promising means of addressing them: creating a "reference class" of similar projects to serve as a platform for comparing costs. The idea, as explained by Kahneman in 2003 [PDF], is that old outcomes can serve as a barometer for recognizing just how unrealistic a biased new prediction might be — and help adjust it accordingly. Such a strategy controls for both political chicanery and cognitive biases alike.
In 2005, Flyvbjerg put this theory into practice at the request of Dutch officials who were considering a massive rail project called the Zuiderzee Line, reported Ryan Blitstein in a great 2008 profile for Miller-McCune. Flyvbjerg predicted that the officials had underestimated construction costs by $2.5 billion and also miscalculated the chance of an additional cost overrun (they put it at 20 percent; Flyvbjerg, at 65 percent). The project was canceled before bids even went out.
Therein may lie the practical flaw with the "reference class" response to the cost overrun problem. More canceled projects means fewer ribbon-cuttings, fewer consulting gigs, fewer construction jobs, and so on. One more thing that's easy to underestimate is the power of the status quo.