Modelling toll roads - where have we gone wrong?

Transport modelling has made a rare intrusion into the attention of the media and the public due to inaccuracies in recent projections done for toll roads in Australia. Generally the modelling of toll roads in Australia has been fairly poor, with many roads attracting only half of their projected volumes and some roads, such as Brisbane's Clem 7, attracting much less. However this is not a uniquely Australian phenomenon, with similar outcomes around the world. A number of papers have been written to explore the reasons for this, and some have come to the conclusion that the most likely cause is biased analysis, either inadvertent or deliberate. Genuine modelling uncertainly, it is argued, would lead to as many underpredictions as over-predictions, and would tend to have reducing errors as models improve. However there are some specific modelling problems that could lead to similar patterns of errors. In this paper we review the approaches used in recent Australian toll road projects, examine their underlying theory, and explore the ways in which the models may lead to over predicted demands. The paper concludes with an examination of alternative approaches to modelling toll roads and their practicality in an Australian context, and some recommendations for improving modelling outcomes. Disclaimer The author of this paper is a practicing transport modeller, and has done peripheral work on some of the projects discussed in this paper. Much of the work done on toll roads is closely covered by confidentiality agreements and is treated as commercial-in-confidence. None of the confidential information that the author has had access to has been used in the preparation of this paper – all of the referenced material is publicly available and, where possible, web links are provided. As discussed in the paper, there is limited detailed information publicly available on the modelling approaches used for toll roads and so this paper focuses heavily on those projects that did release information, particularly Brisbane City Council’s Hale St Link Project (now the Go-Between-Bridge) Business Case and the Hills M2 Upgrade Environmental Assessment. The arguments presented here should not be taken as criticisms of those particular projects, their proponents or their modellers – in fact they should be commended for making the information available. The nature of a paper such as this one is that not every argument can be supported by evidence, and some issues are contentious. It almost goes without saying that the unsupported views expressed in this paper are simply the author’s opinions and expressed in the interest of improving the state of the modelling profession.