Preference Modelling: Conjoint Analysis and Multi-attribute Models

While many multi-variate techniques have been applied frequently in Ireland, academics and practitioners alike have tended to shy away from incorporating conjoint analysis in their research. Conjoint analysis has been used extensively in other countries primarily to estimate consumers’ preferences for products whereas a less complex, easily applied technique, compositional multi-attribute modelling, has been used in Ireland for the same task. This study sets out to compare the powers of estimation or prediction of preference of both techniques. Unresolved areas in preference modelling, namely attribute order bias, learning effects and heterogeneity of preference structures, are addressed in an attempt to clarify the application and interpretation of both techniques under study. The analysis contradicts assumptions generally accepted and previous research work in relation to these areas. However, conjoint analysis is found to predict more accurately than compositional multi-attribute models. Those contemplating the application of either are advised not to always view them as interchangeable.