Website CV Paper
Fields
Behavioral & Experimental, Theory
Job market paper
A Measure of Bayesian Updating
We introduce a novel measure to quantify deviations from Bayesian updating that accounts for the informativeness of evidence. It is defined as the minimum adjustment to a (potentially unobserved) prior required to rationalize observed beliefs as Bayesian. The measure has a simple closed-form expression and admits an alternative interpretation as the maximum expected monetary payoff that can be extracted from non-Bayesian behavior. It responds intuitively to parameters in well-known models of non-Bayesian updating and provides a cardinal tool for comparing belief-updating behavior across individuals, treatments, and populations. Applying the measure to a pooled sample of 34,569 belief reports from six experimental studies, we uncover novel patterns in updating behavior, including that individuals behave closer to Bayes' rule when they overreact than when they underreact to information, and that deviations from Bayesian updating are larger following information that confirms prior beliefs.