A comparison of different methods of modelling long distance highway demand for toll roads and their effect on the revenue forecasts

Abstract

There is a debate raging about how accurate toll road traffic and revenue forecasts are (eg Bain, 2009). The accuracy of these forecasts is critical to the success or failure of a particular scheme and to whether they ever get built. We have undertaken many toll road traffic and revenue modelling exercises with different kinds of model to suit each toll road’s sometimes very different, circumstances. This paper brings together that collective wisdom with the objective of drawing some conclusions about the recommended approaches to take under different circumstances.

Diversion curves – common practice in the US until recently - have at last been superseded. Common modelling practice in some circumstances is to develop a simple elasticity model considering the point-to-point traffic from one end of the proposed tolled road to the other – perhaps having different elasticities for different market segments. But this considers traffic as having one origindestination pair so a better method is to use a trip matrix approach with elasticities or even better with a logit model of route choice. The logit model of route choice can be enhanced to model the longer term changes of destination and trip frequency with a nested set of logit models. Sometimes these models include stated or revealed preference surveys to measure the value of time, willingness to pay, perception of the toll road etc. Sometimes they have only a few market segments and sometimes they split the demand into many. These different methods are compared and contrasted using a toll road model based on a model recently developed for a set of tolled roads in Nigeria.

The paper compares and contrasts the strengths and weaknesses of the different forecasting methods and draws conclusions about their relative accuracy. This is a very important subject – vital to the success or failure of important transport infrastructure - but it does not get sufficient consideration, so this paper attempts to open up the whole subject for debate.

.