Rational Misspecification: Framework and Applications

This paper proposes a framework for assessing whether misspecified decision-makers would be willing to pay for information that can potentially make them less misspecified. We introduce a prior-free approach, based on “constrained” maximal regret, to derive an upper bound on the subjective assessment of potential gains from acquiring a more accurate model. The constraint stems from the information currently available to the decision-maker. We apply our approach to three prominent models of misspecified beliefs: coarse expectations, causal misperceptions, and sampling equilibria. (JEL D83, D84, D91)