Feature-based attributes and the roles of consumers' perception bias and inference in choice

This paper considers decision contexts wherein consumers make choices among alternatives that contain a manifest feature-based attribute: i.e., a discrete, salient and important attribute that describes a dichotomous quality, such as “genetically modified”, “organic”, or “locally grown”. We propose a choice model that can explicitly account for a) perception bias with respect to such an attribute when its information is present, and b) inference formation if this attribute information is missing for some alternatives. The impact of different information presentation formats on consumers' perception bias and inference formation is then examined by applying theories from social psychology. Our model outperforms standard Random Utility models that omit explicit representation of these phenomena. Consistent with theories considered, we find significant evidence of perception bias and inference in the choice data. Our results also provide insights on how consumers may infer the quality of a missing attribute in different competitive framing contexts. Finally, our welfare estimates show that consumers may benefit simply from the information improvement regarding government labeling policies.