Daniel Hauser, Aalto University

"Representing Heuristics as Misspecified Models"

Abstract

A growing literature in economics considers how to model heuristics that agents use to process information and the resulting biases that emerge in belief updating. In this paper, we link two common approaches used to model inaccurate belief updating: (i) defining an 'updating rule' that specifies a mapping from the true Bayesian posterior to the agent's subjective posterior, and (ii) modeling an agent as a Bayesian with a misspecified model of the signal process. We establish conditions under which an updating rule can be represented as a misspecified model and conditions under which a misspecified model can be represented as an updating rule. This result connects the two approaches and clarifies the implicit assumptions about an agent's learning rule required for each approach.

Contact person: Julia Salmi