Bent Flyvbjerg is an academic at Oxford who studied a large number of megaprojects costing 1Bn or more.

His conclusion was that 9/10 projects of this size run over budget, 9/10 run over time and 9/10 have benefits that are oversold.

Rail projects are particularly bad. The average traffic on them was 61% of that forecast and some are as low as 25%.

His homepage: http://www.sbs.ox.ac.uk/community/people/bent-flyvbjerg

An interview with him : http://www.econtalk.org/archives/2015/05/bent_flyvbjerg.html

One of his books that is very readable: http://www.amazon.com/Megaprojects-Risk-An-Anatomy-Ambition/...

This isn't to say that such projects should not be built, just that those selected should be the ones with the best realistic cost benefit, not those where the backers have made up the numbers most creatively.