Managing Traffic Forecast Uncertainty

Abstract

Traffic volume forecasts are fundamental to policymaking; yet, the impacts of forecast errors are rarely considered. Accordingly, traffic forecasts in 39 studies were compared to observed volumes after the forecast year had elapsed, yielding a mean error of 40%. To evaluate the effect of forecast accuracy, five case studies examined how forecast error affected investment decisions. Because errors of 11%, 12%, 20%, 33%, and 34% can alter investment decisions, the respective observed errors of 20%, 38%, 64%, 6%, and 52% meant the forecast error was potentially large enough to alter the investment decision based on each study’s stated criterion. Although striving for more accurate forecasts is laudable, the results from the 39 studies showed that only a quarter of the variation in study accuracy is not attributable to random variation. Accordingly, until highly accurate forecasts become the norm, an interim recommendation for planners is to use empirical decision intervals, as shown herein, indicating the magnitude of error can alter investment decisions.