October 2020

IZA DP No. 13821: Reverse Bayesianism: Revising Beliefs in Light of Unforeseen Events

Bayesian Updating is the dominant theory of learning in economics. The theory is silent about how individuals react to events that were previously unforeseeable or unforeseen. Recent theoretical literature has put forth axiomatic frameworks to analyze the unknown. In particular, we test if subjects update their beliefs in a way that is consistent "reverse Bayesian", which ensures that the old information is used correctly after an unforeseen event materializes. We find that participants do not systematically deviate from reverse Bayesianism, but they do not seem to expect an unknown event when this is reasonably unforeseeable, in two pre-registered experiments that entail unforeseen events. We argue that participants deviate less from the reverse Bayesian updating than from the usual Bayesian updating. We provide further evidence on the moderators of belief updating.