I first saw this book in the hands of a student in my policy analysis class several years ago, and I regret that I waited so long to read it.
Now I have, and I think that many of you will want to read it as well.
The book, authored by Flyvbjerg, Bruzelius and Rothengatter (FBR), is subtitled "an anatomy of ambition" but its subtitle should be "an examination of why big transportation projects go wrong, how to make them right, and why these ideas apply to all large infrastructure projects," but the marketing department probably favored snappy over accurate.
The main point that FBR make, in clear and painful detail, is that megaprojects (projects that have big financial, economic, social and/or environmental impacts) usually fail because their proponents and constructors do not bear the risk of failure when estimating the cost of constructing the project or demand for its services once completed. The risk is instead carried by taxpayers who end up paying more than expected to fund a project that's less useful than promised.
If that sounds familiar (it reminds me of Cadillac Desert, desalination in San Diego, the Colorado Aqueduct, Aridzona's CAP, the Delta "Conveyance", Coyote's critiques of light rail, and chapter 8 of my book), then you will STILL gain something by reading this book due to its clear analysis, useful case studies and forthright declaration of how to prevent failures in the future.
Here are the notes I made while reading:
- FBR make the important point that "dialogue" cannot overcome power. It's therefore necessary to change rules to improve accountability (they recommend that private money fund at least one-third of a project's cost, with no government guarantee of repayment in the event of poor performance).
- Many projects are "born in sin" - with unrealistic forecasts of costs and revenues. They should not be built, but developers often get paid their full fee when they are. That's no way to bring discipline to this arena.
- A huge share of environmental impact reports are carried out when the project is already going forward, do not contain adequate baseline data (to make it possible to understand actual impacts) and are very rarely updated after project completion (this is a common problem; development aid projects are not often subject to performance reviews).
- It's not unusual to ignore new costs that will result from reducing environmental impacts -- costs that may push the benefit:cost ratio below 1.0!
- Many simulations of futures follow a EGAP protocol (Everything Goes According to Plan) when it would be much more useful and enlightening to consider Worst Case or MLD (Most Likely Development) scenarios.
- Such myopia is expected when the developer does the simulation for a government that is the proponent of a project that it's ALSO supposed to regulate and oversee.
- To improve accountability, FBR propose accountability,* performance (not technology) specifications, clearly defined regulations and risk-bearing, and involving (private) risk capital.
- If private capital is not willing to invest without guarantees,THEN MAYBE THE PROJECT SUCKS.
- The book's full of checklists and flowcharts. It could be used as a cookbook for analyzing megaprojects. I wonder if anyone at the Bureau of Reclamation has a copy?
- FBR think that private or state companies can build megaprojects, but they need to do so as risk-carrying, stand-alone companies that are supervised by the government.
- Never underestimate the potential manipulation of rent-seeking politicians and/or special interests (developers, construction companies, transport companies, et al.)
* Dan Ariely (review coming!) has found that transparency can result in MORE inaccuracy, as when the scrutinized estimator increases the fudge factor (e.g., +20 percent to cost) in the knowledge that it will be bargained down, but then makes more when it's not bargained down enough (e.g., -10 percent to cost). FBR's transparency is less vulnerable to this problem as they propose posting all documents in public AND allowing peer review, but this transparency will only really work if there's some SERIOUS exploration of numbers and estimates.