Polarized Preferences In Homegrown Value Auctions

Incentive compatible auction experiments, often referred to as homegrown value auctions, have become a popular tool for exploring how controversial product attributes and knowledge of these attributes affect consumer willingness to pay. A common observation in these experiments is a prevalence of zero bids and bimodal bid distributions. One possible explanation is that individuals have polarized preferences: find all products with a particular attribute desirable (positive polarization) or undesirable (negative polarization). The purpose of this paper is to explore three questions. Do polarized preferences exist? If they do exist, can they be identified? If they can be identified, does their identification provide useful information? To answer these questions, polarized preferences are theoretically formalized. This theory is used to discuss bidding behavior and how common experimental design features can facilitate or hinder the identification of polarization. The weaknesses of common econometric models used to analyze auction bids are reviewed in the context of polarization and a new model is proposed. Finally, the new model is tested using data from a home grown value auction. The results of this analysis suggest that polarized preferences do exist and that accounting for them can improve estimates of willingness to pay and likelihood that a product is valued at all.