Michael Todd has a disappointing article in Miller McCune arguing that High Speed Rail projects are likely to go over budget. His justification is, well, that they're large capital projects, and large capital projects often go over budget. Yes, they do. But this isn't an argument against high speed rail or any other major capital project. It's an argument for better approaches to cost estimation and control.
Nowhere does Todd justify the implication that high speed rail projects are at greater risk of cost overrun than, say, huge urban freeway tunnels such as Boston's Big Dig, Sydney's Cross-City Tunnel, or the likely tunnel in downtown Seattle's future.
But I'm grateful to Todd for linking to a very useful 2008 article by Ryan Blitstein on the work of Danish economist Bent Flyvbjerg. Read it. It's a good overview of Flyvbjerg's important work on cost overruns and strategies for predicting them. For more, see Flyvbjerg's book Megaprojects and Risk.
Note, by the way, that the problem of cost overruns is conceptually separate from questions about why capital costs, predicted or not, are so high.
Thanks for this. I was just chatting with folks today who said part of the cost escalation for streetcars was that they were being over designed to light rail standards jacking up the price. Over engineering such as specific low grades needed for streetcars (the J Church grade in Dolores Park would never be built today if they got a hold of it)is some of the issue. Part of the cost issue I think in this country is us just not having enough expertise given our long time away from mass production of systems like these.
I haven’t read all your links yet, but I’m of the opinion that most boondoggles are a natural outcome of all political processes (including those in the private sector). Someone can look great by getting something done, and be gone before they have to be held accountable…so they work to get things done. Once a cost is sunk, there is a huge biological urge to keep going and so we do.
What makes some decisions terrible and other decisions merely bad is the degree to which forecasts are multiplied to get a final outcome.
For example…if I am building a McDonalds franchise, my forecasting probably starts at total revenue level…whereas if I am building a restaurant that has never been built before, I have to forecast multiple things: Number of “leads”, conversion rates, average expenditures, average # of return visits in a year, etc.
You see, if I am forecasting just revenue, being optimistic by 10% will only be a 10% setback when reality hits.
But if I am forecasting # of leads optimistically by 10%, and conversion rates optimistically by 10%, and average expenditure per visit by 10%, and # of return visits by 10%, then the final forecast is off by more than 45%. The more things you have multiplying in your forecast, the worse your final outcome can when you are optimistic.
So for example, with CAHSR, it is pretty reasonable to assume that they have underestimated their operating costs, underestimated their capital costs, overestimated ridership and overestimated ticket prices…at least if you go off of historical performances of other rail systems. But the degree to which each forecast is optimistic will result in the difference between a $40B rail system and a $100B rail system.
To add to what Danny said: mature-technology ridership projections have been much better than new-technology projections. The light rail systems built in the US in the early 1980s overestimated ridership, by a lot – I think on average by a factor of two. And conversely, the first TGV line overshot expectations, because SNCF hadn’t figured in induced demand.
After such experiences, the planners learned more, and since then ridership projections have been in line with reality. For every VTA missing ridership estimates and there’s a Charlotte light rail overshooting them. And newer LGVs still overperform expectations, but not by a lot, and I’ve heard rumors that SNCF lowballs ridership estimates slightly so that people will be impressed more when the traffic numbers exceed them.
This is really interesting Jarret and seems very useful, although one question I’d have over Flyvbjerg’s work is how well this approach can be used to inform decision making without accurately costing the benefits of these projects? Engineering is engineering, and so an approach like this where a rail line’s costs are benchmarked against comparable schemes makes sense. But would adopting the same approach to (for example) ridership (where socio-economic, political and cultural influences are likely to play a bigger role) be appropriate? Having an accurate estimation of the costs will allow you to plan better and actually know if something is affordable, but unless you can tell what all the benefits will be then using that as a tool for decision making seems shortsighted.
Flyvbjerg’s technique and especially his anti-politician stance reminded me of an article from the December 13th New Yorker that discussed the diminishing rate of repeatability of studies as years go by after the original publication; it seems that even scientists with little financial stake suffer from “optimism bias.”
I got kind of a “politics bad, engineering good” feeling from the Miller-McCune article, but politics is just part of life. Unfortunately you can’t quantify the benefits of removing an elevated highway that cuts through the middle of a city; at least parts of the Big Dig were necessary at any cost. Similarly, no matter how cost-effective it is, you will rarely convince someone that their house needs to be removed for a project, and if that person donates a lot of money to politicians or if their house was the birthplace of Ghandi that is going to influence the course of the project.
Maybe I’m stating the obvious here but it seems that the biggest benefit to Flyvbjerg’s model was only mentioned briefly in the article: planning for these overruns will help to contain them. So even if they had used this model on the Big Dig, and it had found a 100% chance of a 1000% cost overrun, the project still would have been necessary because of the giant highway cutting off half of Boston from the other half, but they could have spent more on the planning piece (the article mentioned soil borings) and saved on the overruns later.
One reason overruns occur, both in public and in private projects, is the “if we tell the sponsors how much it will really cost, they won’t approve the project” problem. The Portland Aerial Tram (a worthy project, but one that came in over 3x the original budget) is a good example; as Oregon Health and Science University officials were found to have withheld rising cost estimates from the City of Portland, until sufficient costs were sunk. Whether the project would have been greenlighted had the true cost been known up front is unclear.
Certainly, rail projects are no more prone to this sort of fraud than other types of infrastructure, but it does occur.
Slightly off topic, but related. I would be interested to know how much public participation is allowed in large capital project in other countries (like say Australia or New Zealand vs US)? One of the problems with using infrastructure as a stimulus project in the US has been that it takes so long to go from project proposal to build out, I am wondering if that long delayed process is true of most wealthy western countries or is this a distinct feature of US litigation culture?
I suspect that even if transportation consultants don’t directly bill for the cost of meeting with all of the various layers of governments and time spent at various public hearings, that cost is probably capitalized into the project cost as overhead.
Similar to what EngineerScotty said, there’s also the issue of having to go with the lowest bidder, especially prevalent here in the US. The bidding contractors push to make their estimate the lowest just so they can get the job, at which point it’s much easier to get adjustments for whatever “unexpected circumstances” were conveniently unaccounted for. This happens at all levels, from multi-billion dollar infrastructure projects to small residential remodeling. Maybe a second-lowest bid requirement would help.
I wonder if the planners in Copenhagen read their Flyvberg. The Copenhagen metro has had some pretty impressive cost escalations.