Attribute Specific Impacts of Stated Non-Attendance in Choice Experiments

In this paper, we generalise existing approaches to the treatment of stated attribute non-attendance data in discrete choice experiments by allowing attribute specific impacts. We implement this approach by employing an extended hierarchical Bayes logit model specification. To illustrate this approach, we consider data collected to examine Indian consumers' preferences for traditional aromatic rice varieties. Our results regarding stated attribute non-attendance reveal that, our new approach shrinks marginal utilities of non-attenders substantially compared to stated attenders, with significant differences in the shrinkage between some of the attributes. In addition, our results reveal the way in which non-attendance of attributes interact with each other and the impact that this has on the distribution of willingness to pay estimates.