Improving Predictive Power of Conjoint Analysis by Constrained Parameter Estimation

The prediction of first choice preferences by the full-profile method of conjoint analysis can be improved significantly by imposing constraints on parameters based on a priori knowledge of the ordering of part worths for different levels of an attribute. Constrained estimation, however, has little effect on the predictive validity of the tradeoff method because the preference judgments within rows (or columns) of tradeoff tables have largely the same role as the constraints.