Preferences with Costly Bayesian Learning

Abstract

In this paper, we study a general model of information acquisition: costly Bayesian learning. Using a menu choice framework, we provide an axiomatic characterization of the model, identify its parameters (a utility function, an increasing transformation, a second-order prior belief, and an information cost function), and behaviorally compare the costs. Our results show that the rational inattention model, which has found various applications in the literature, is a special case of the costly Bayesian learning model. We identify several behavioral conditions each of which can be used to test if the decision maker is rationally inattentive or is of a more general type Bayesian learner including those who exhibit aversion to uncertainty. We argue that our decision makers can have flexible attitudes towards the timing of resolution of uncertainty.