Estimation of Consumer Preferences via Ordinal Decision-Theoretic Entropy

We introduce the first non-parametric approach to the problem of inferring a model of a consumer’s preferences on the basis of available evidence. The evidence can take a variety of different forms, including survey and transaction data. We use an axiomatic approach to infer a unique utility model that complies with the given data and is maximally non-committal with regard to missing information. Moreover, our method does not rely on a priori assumptions about the functional form of the utilities. We develop a consumer model, and demonstrate the applicability of our method to a variety of preference estimation problems using synthetic data. In our examples we infer different consumer utilities over bundles of goods or product characteristics. We compare our method with Conjoint Analysis, and show that it can compete favorably even in areas which are ideally suitable for the latter. Furthermore, we show that our approach can be useful in important marketing domains where Conjoint Analysis cannot generally be applied, including the case of inferior goods and that of perfect complements.