Combining Sources of Preference Data

Abstract

Mixtures of revealed preference and stated preference data are recognised by transportation researchers as offering a richness of behavioural input to travel choice modelling that is often absent from the isolated use of each data source. Accumulating evidence from various literatures, especially in marketing, psychology and transportation, provides support for the desirability of combining sources of preference data as a way of transferring increasing power of understanding of travel behaviour from the econometrics of a model to the underlying data inputs. Together with advances in the specification and estimation of discrete choice models, we are beginning to see that the simpler choice models such as multinomial logit (MNL) deliver an amazing amount of behavioural power providing that the underlying data specification is given a statistically rigorous treatment. There still remains however a growing role for more general choice models as a way of establishing the relevance of the simpler MNL model. This paper reviews the major contributions to the literature on combining sources of preference data and suggests new directions for fruitful research.