Forecasting issues in stated preference survey research.

The family of market research survey techniques termed stated preference methods (e.g conjoint analysis) has been used quite often in transportation planning over the past decade. While these techniques have been primarily used to gather marketing information, such as willingness to pay for service improvements, an important recent trend is the use of stated preferences in forecasting. Such methods are now becoming seen as a complement to the more traditional revealed preference survey methods in cases where the latter cannot provide the full information necessary. Examples of such cases are the introduction of a qualitatively new transport mode or route (e.g. the Channel Tunnel between France and England), the impact of qualitatively new transport policies (e.g. electronic road pricing), or the reactions to large changes in travel conditions which cannot be observed at present (e.g. changes in trip departure times due to increasing peak hour traffic congestion). The need to accurately predict travel behaviour across the entire population places special requirements on the design, analysis and application of stated preference survey experiments. Many of these requirements are analogous to those faced in revealed preference studies. There remain important issues specific to stated preference, however, because forecasting requires that a stated preference survey be built upon a hypothetical choice context which will yield realistic and unbiased statements of behaviour. As these issues are addressed, stated preference techniques appear to be evolving into an integrated part of the transport forecasting methodology. The purpose of this paper is to give an overview of this evolution, in terms of the study design, model estimation and forecasting issues involved. The discussion includes a treatment of the various errors and biases which may arise in stated preference as opposed to revealed preference surveys, along with reference to practical examples based on recent projects in the Netherlands, Australia, France and the UK.