Rank-order conjoint experiments: efficiency and design

In a rank-order conjoint experiment, the respondent is asked to rank a number of alternatives instead of choosing the preferred one, as is the standard procedure in conjoint choice experiments. In this paper, we study the efficiency of those experi-ments and propose a D-optimality criterion for rank-order conjoint experiments to find designs yielding the most precise parameter estimators. For that purpose, an expression of the Fisher information matrix for the rank-ordered multinomial logit model is derived which clearly shows how much additional information is provided by each extra ranking step made by the respondent. A simulation study shows that Bayesian D-optimal ranking designs are slightly better than Bayesian D-optimal choice designs and (near-)orthogonal designs and perform considerably better than other commonly used designs in marketing in terms of estimation and prediction accuracy. Finally, it is shown that improvements of about 50 % to 60 % in estimation and prediction accuracy can be obtained by ranking a second alternative. If the re-spondent ranks a third alternative, a further improvement of 30 % in estimation and prediction accuracy is obtained