Ex-post evaluation of traffic forecasts
Panteia has extensive expertise in traffic models and forecasts. But to what extent do traffic forecasts, made for major infrastructure projects, actually come true? Every forecast comes with a degree of uncertainty: various factors can cause actual outcomes to deviate from the forecast. Nevertheless, it is essential to make accurate forecasts so that infrastructure managers can make reproducible and well-founded decisions for a future-proof design in large infrastructure projects. Because we place great importance on the quality of our models and forecasts, we have taken the extra step of conducting an ex-post evaluation.
Ex-post evaluations at the completion of projects are not common in the Netherlands, even though our literature review found repeated calls for such evaluations. As a result, there is hardly any available data in this field. The data we did find were ex-post evaluations of forecasts made using the National Model System (LMS). These showed that mobility forecasts often closely align with actual outcomes. However, these are national forecasts, not focused on specific road sections or routes.
Expanding our literature review to international sources, we found a database containing 210 projects across 14 countries. Analysis of this data revealed that traffic forecasts generally deviate significantly from actual outcomes. In 50% of road projects, actual traffic volumes differed from forecasts by as much as 20%. This deviation is not systematic: overestimations and underestimations balance each other out.
For our case study, we selected five cases where we compared forecasts and actual outcomes for one or more relevant road sections. The actual outcomes deviated between a minimum of 2% and a maximum of 29% from the forecasts, with deviations exceeding 10% for six of the nine analysed road sections. There were both underestimations and overestimations, although there was an average overestimation of around 4%. These initial results from our limited case study align with findings based on international data.
A broader case study is needed to draw stronger conclusions. In the coming period, we aim to expand the research by including more cases. Students with an interest in transport and logistics who find this an interesting internship project are warmly invited to contact researchers Servé Hermens (s.hermens@panteia.nl) or Jan Kiel (j.kiel@panteia.nl).
Finally, we invite you to reflect on the following statement:
"To make (traffic) forecasts accessible for evaluation studies, they need to be systematically stored in a database."
