Megaprojects and Risk

Bent Flyvbjerg, Nils Bruzelius and Werner Rothengatter assembled data on large infrastructure projects and found that cost overruns and demand shortfalls are systematic rather than exceptional. Their database covers hundreds of transport projects across twenty countries and several decades.

The empirical findings are specific. Cost overruns occur in roughly nine out of ten projects. For rail projects the average overrun is around forty-five per cent in constant prices; for bridges and tunnels around thirty-four per cent; for roads around twenty per cent. Demand forecasts are equally unreliable and biased in a particular direction: rail passenger forecasts overstate actual traffic by an average of more than half, while road traffic forecasts are less biased but highly variable. The errors have not improved over seventy years, which the authors treat as evidence that they are not forecasting mistakes.

Their explanation is that the errors are produced deliberately. Project promoters need approval, approval depends on favourable cost-benefit numbers, and the people producing those numbers are engaged by the promoters. Flyvbjerg calls this strategic misrepresentation and distinguishes it from optimism bias, which is unconscious. Their proposed remedies are institutional: require promoters to put private capital at risk, separate the forecasting function from the promoting function, use reference class forecasting based on outcomes of comparable completed projects, and impose accountability for forecasts after the fact. The reference class forecasting recommendation was subsequently adopted by the United Kingdom Treasury. The book is the foundation of a large subsequent literature, including Flyvbjerg's own How Big Things Get Done.