Application of Rational Inattention in Discrete Choice Modeling

• Applicant: Professor Dr. Thomas Otter
• Subject Area: Management and Marketing
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 496422953

Discrete choice models (DCM) based on random utility theory are widely used in economics and marketing as a tool to learn about the preferences of decision makers (DM) in multi-attribute, multi-alternative settings. This knowledge allows to infer counterfactual outcomes that may pertain to changes in attributes or the choice set. Key assumptions in DCMs concern the information on which DMs base their choice, and clearly, misspecifications likely have a detrimental effect on subsequent counterfactuals. Classical models assume that DMs incorporate all available information into their choice, while DCMs accounting for inattention do so typically via modifications of the random utility index. For instance, variable selection models impose an exogenous probability that an attribute is ignored. When attention is costly, however, these models do not fully incorporate that attention becomes a strategic variable. Then, the specifics of a choice context as per its easily accessible features (e.g., brand) together with expectations of what can be gained from attentional effort will guide its deployment. Recent consumer search models address this by allowing DMs to choose which alternatives to inspect. Yet, the assumptions required by these models for tractability have been questioned by empirical evidence. We propose to develop a DCM based on the recent theory of rational inattention (RI) in order to improve on the shortcomings of extant models. In RI, a rational DM faces cognitive costs of processing information and chooses what and how much to learn about the available alternatives. The RI model generates a rich set of behavioral patterns while being rooted in rational utility optimization. Thus, our objectives are the following. First, we develop a parsimonious method that estimates preferences while accounting for optimal attention allocation. Second, we provide experimental evidence that choice sets with features common in marketing research impact the attention allocation of individuals. We view the translation of RI theory, that has been (mostly) developed in models with one-dimensional uncertainty, into a multi-attribute, multi-alternative setting with respondent heterogeneity, typical for marketing, as the key challenge in our project and its key contribution. In addition, we will develop a methodological extension aimed at identifying cost structures. Empirical identification of cost structures is important in applied work where the question which pieces of information are more or less accessible often lacks an answer a priori. Finally, to illustrate the general purpose of our framework, we will study implications for policy design, in particular, the interplay between the complexity and the effectiveness of regulations faced by heterogeneous respondents. Notably, this is relevant for policy makers who try to attain a specific goal while facing a trade-off between consumer inertia and costs of incentives and transparency.

• DFG Programme: Research Grants