"The antidote for bad cost-benefit analysis is good cost-benefit analysis." This is from the usual introductory lecture on CBA to public policy students. The remark obscures the analytical problems that go with CBA (what are "existence" values?, which discount rates?, etc.) and the bigger problem of inevitable politicization.
Bent Flyvbjerg recently visited USC and reported on some of his work on the topic. Ex poste, he finds the "megaproject disaster gene": vastly underpredicted costs and vastly overestimated benefits. Bent notes that this is an international problem and cites recently published work on the Chunnel (by R. Anguera, Transportation Research A40, 2006) that its NPV and IRR are both in the bizarre-negative area ($-17.8 billion; -14.5%, respectively).
On a very different note, Frank Ackerman, Lisa Heinzerling and Rachel Massey ("Applying Cost-Benefit to Past Decisions: Was Environmental Protection Ever a Good Idea?" http://ssrn.com/abstract=576161) cite three recent episodes where the "right" policy choices were made in spite of (not because of) CBA.
I imagine that one can easily find many more where CBA (done right) would have had the opposite effect. But Flyvbjerg's work shows that CBA (done any which way) by itself is never a barrier; ex ante, it can be all smoke and mirrors and awful mega-projects go forward.
I read this morning that "... in 1987, President Reagan vetoed a spending bill because it contained 121 earmarks. The number of earmarks has skyrocketed over the past decade, from 4,126 in 1994 to 15,268 in 2005 ..." (WSJ op-ed by Sen. Tom Coburn, p. A18).
Who or what will stop them?