A Cautionary Note on Designing Discrete Choice Experiments: a Comment on Lusk and Norwood’s “effect of Experiment Design on Choice-based Conjoint Valuation Estimates”

Recent Monte Carlo work on choosing experimental designs for discrete choice experiments seemed to greatly simplify this choice for applied researchers. It suggested that (a) commonly used designs can generate unbiased estimates for indirect utility function specifications with main effects only and main effects plus higher order terms, and (b) random designs are more efficient than main effects designs. We show that these results are very specific to the particular indirect utility specifications studied and do not generalize well. We further show that conclusions drawn concerning random designs are problematic and potentially dangerous for applied researchers.