Towards Eliciting Weak or Incomplete Preferences in the Lab: A Model-Rich Approach (2111.14431v4)

Abstract: Recovering and distinguishing between the potentially meaningful indifference and/or indecisiveness parts of a decision maker's preferences from their observable choice behaviour is important for testing theory and conducting welfare analysis. This paper contributes a way forward in two ways. First, it reports on a new experimental design with a forced and a non-forced general choice treatment where the 282 subjects in the sample were allowed to choose multiple gift-card bundles from each menu they saw. Second, it analyses the collected data with a new non-parametric combinatorial-optimization method that allows to compare the goodness-of-fit of utility maximization to that of models of undominated or dominant choice with incomplete preferences, in each case with or without indifferences. Most subjects in the sample are well-approximated by some indifference-permitting instance of those three models. Furthermore, the model-specific preference recovery typically elicits a unique preference relation with a non-trivial indifference part and, where relevant, a distinct indecisiveness part. The two kinds of distinctions between indifference and indecisiveness uncovered herein are theory-guided and documented empirically for the first time. Our findings suggest that accounting for the decision makers' possible indifference or inability to compare some alternatives can increase the descriptive accuracy of theoretical models and their empirical tests thereof, while it can also aid attempts to recover people's stable but possibly weak or incomplete preferences.