" 'How (In)accurate Are Demand Forecasts in Public Works Projects?' examines 183 road and 27 rail projects in 14 countries on five continents, including in the U.S. It found that rail passenger forecasts were overestimated by 106 percent on average, resulting in an actual ridership of less than half of that predicted. For 9 of 10 rail projects, estimates were high, according to the study. One project in eight overshot estimates by more than 400 percent. The study also noted that forecasts are no more accurate now than they were 10, 20, and 30 years ago.
...'My team and I were shocked by our findings. For professional forecasters to be wrong by more than 100 percent and to make no improvement for decades is unheard of. Simple forecasting errors cannot explain this,' Flyvbjerg said. 'By routinely overestimating benefits and underestimating costs, promoters make their projects look good on paper, which helps get them approved and built. The only ones to pay are the taxpayers.' "
The study also covers road investments. For road projects studied, half of all traffic forecasts were found to be wrong by more than 20 percent. This is better than for rail forecasts. But for roads, too, there has been no improvement in the accuracy of forecasts for 30 years, despite millions of dollars spent on improving forecasting models.
From the journal article abstract: This article presents results from the firststatistically significant study of trafficforecasts in transportation infrastructureprojects. The sample used is the largestof its kind, covering 210 projects in 14nations worth U.S.$59 billion. The studyshows with very high statistical signifi-cance that forecasters generally do a poorjob of estimating the demand for transportationinfrastructure projects. For 9out of 10 rail projects, passenger forecastsare overestimated; the average overestimationis 106%. For half of all road projects,the difference between actual and forecastedtraffic is more than ±20%. Theresult is substantial financial risks, whichare typically ignored or downplayed byplanners and decision makers to the detrimentof social and economic welfare.Our data also show that forecasts have notbecome more accurate over the 30-yearperiod studied, despite claims to the contraryby forecasters. The causes of inaccuracyin forecasts are different for rail androad projects, with political causes playinga larger role for rail than for road. Thecure is transparency, accountability, andnew forecasting methods. The challengeis to change the governance structuresfor forecasting and project development.Our article shows how planners may helpachieve this.
[Editor's note: The link below is to a 750KB PDF document.]
Thanks to Chris Steins