How to Distinguish Motivated Reasoning from Bayesian Updating

Abstract

Can we use the way that people respond to information as evidence that partisan bias or directional motives influence political beliefs? It depends. Using one natural formalization of motivated reasoning as wanting to believe certain things (“once-motivated reasoning”), this is not possible. Anyone exhibiting this kind of motivated reasoning has a “Fully Bayesian Equivalent” with a different prior, who has identical posterior beliefs upon observing any signal. This result clarifies what we can and cannot learn from several prominent research designs and identifies a set of results inconsistent with both Bayesian updating and once-motivated reasoning. An expanded version of the model where subjects sometimes completely reject signals that lead to less pleasant beliefs (“twice-motivated reasoning”) can explain these anomalies. The models clarify which empirical tests can provide evidence for different kinds of motivated beliefs and can be incorporated into decision- and game-theoretic models.