Designing efficient stated choice experiments in the presence of reference alternatives

This paper examines various design strategies that might be employed to construct statistically more efficient stated choice designs in the presence of a reference alternative in a choice set. Using data collected in Sydney in 2004 in the context of trading time and cost attributes associated with alternative tolled and non-tolled routes to drive a car to work, we contrast D-efficient designs (based on a number of ways of pivoting attribute levels around a reference alternative) with the more traditional orthogonal designs and conclude that D-efficiency design strategies produce significantly improved results, in a statistical sense of relative efficiency, than the more traditional orthogonal design. Furthermore, the increased use of computer aided personal survey instruments and internet-based surveys enables researchers to structure the experiments around the very specific experiences of each sampled respondent, adding relevance and comprehendability to the attribute levels being assessed in contrast to other averaging methods to construct reference alternatives.