Random consideration and choice: A case study of “default” options

Abstract

A growing number of stochastic choice models include a "default" option for situations where the decision maker selects none of the feasible alternatives. While this is a welcome development, these models also present an empirical challenge-since the situations where the decision-maker chooses nothing may be difficult to observe. Taking Manzini and Mariotti's (2014) independent random consideration model as a point of departure, I investigate what can be learned about models of choice with default when the no-choice behavior of the decision-maker is unobservable.

Inferring Cognitive Heterogeneity From Aggregate Choices

“… Horan () considered an unobserved default outcome in the context of a dataset with choices from multiple menus, showing that the identification properties of the independent random consideration model in Manzini and Mariotti () remain largely intact. …”

Reference Dependence and Random Attention

“…In the random choice literature, Manzini and Mariotti [2014], Brady and Rehbeck [2016], and Aguiar [2017] are the most closely related and have been discussed at length already. Horan [2019] extends the analysis in Manzini and Mariotti [2014] by assuming that the no-choice behavior of the DM is unobservable. Recently, Cattaneo et al [2020] study a general random attention model which also features an unstructured attention process.…”

The Decision-Conflict and Multicriteria Logit

“…The latter axiom implies that the choice probability of the outside option decreases as more alternatives are added to the menu, whereas (1) and ( 2) allow for it to either increase or decrease through the channel that was discussed earlier. It should be emphasized, however, that unlike the decision-conflict and multicriteria logit where choice of the outside option is a vital element of the underlying decision process, the role of that option in Manzini and Mariotti (2014) and Brady and Rehbeck (2016) is secondary and, as shown in Horan (2019), the general features of these models are unaffected when this option is removed from the analysis. The decision-conflict and multicriteria logit models are also distinct from the multinomial logit model with costly information sampling (and without an outside option) that was studied in Matějka and McKay (2015) (see also Caplin, Dean, and Leahy, 2019) within the rational-inattention framework, and which was extended to a dynamic setting in Steiner, Stewart, and Matějka (2017).…”