The Design of Stated Choice Experiments: The State of Practice and Future Challenges

Since the work of Louviere and Woodworth (1983) and Louviere and Hensher (1983), stated choice (SC) methods have become the dominant data paradigm in the study of behavioural responses of individuals and households as well as other organizations, in fields as diverse as marketing, transport and environmental and health economics, to name but a few. In SC experiments, it is usual for sampled respondents to be asked to choose from amongst a number of labelled or unlabelled alternatives defined on a number of attribute dimensions, each in turn described by pre-specified levels drawn from some underlying experimental design. The choice task is then repeated a number of times, up to the total number of choice sets being offered over the experiment. Several experimental design strategies are available to the practitioner, however, within the transport literature, it appears that the most common form of experimental design used are orthogonal fractional factorial designs. In this paper we review the properties of such designs, and demonstrate that these properties are unlikely to be retained through to the estimation process. We also discuss an alternative design construction strategy, used to construct statistically optimal designs