Attribute non-attendance in wine choice: Contrasts between stated and inferred approaches

A growing number of studies employing discrete choice experiments provides evidence that respondents ignore some attributes while making a choice, thus violating preference structure assumptions. Using a discrete choice experiment, we explore consumers preferences for wine. A greater understanding of the choice process accounting for attribute non-attendance may contribute to the definition of more accurate policy recommendations and market strategies. The result is checked in three different databases. This paper adds to the literature by contrasting the stated and inferred modelling approaches accounting for attribute non-attendance. The main contributions of the paper are (1) to compare the respondents stated non-attendance with probabilities inferred by the model, and explores the concordance between both methods; (2) examine the differences in model fit measures and willingness to pay values, adding as reference a standard model that assumes all attributes are attended. Results provide evidence of the presence of attribute non-attendance in all markets, the inferred approach reaching the best model fit. Few respondents behave in a fully compensatory manner, but a high proportion considers the price attribute, which is an important tip for marketers and wine firms. In addition, it is a key factor for the robust estimation of willingness to pay by attribute. Moreover, respondents overstate their levels of attendance and we found little concordance between both methods, in all datasets. The willingness to pay measures are lower in the inferred approach, while the estimates from the mixed logit model are found to be more congruent with those from the standard model.