Evolution of subjective hurricane risk perceptions: A Bayesian approach

How do decision makers weight private and official information sources which are correlated and differ in accuracy and bias? This paper studies how traders update subjective risk perceptions after receiving expert opinions, using a unique data set from a prediction market, the Hurricane Futures Market (HFM). We derive a theoretical Bayesian framework which predicts how traders update the probability of a hurricane making landfall in a certain range of coastline, after receiving correlated track forecast information from official and unofficial sources. Our results suggest that traders behave in a way not inconsistent with Bayesian updating but this behavior is based on the perceived quality of the information received. Official information sources are discounted when a perception of bias and credible alternatives exist.

► We empirically analyze subjective hurricane risk perceptions using prediction markets.
► Correlated official and unofficial hurricane track forecasts serve as information sources.
► Our results suggest traders used weights consistent with Bayesian updating for two information sources. Traders discounted a third source, which was the least accurate, but provided information relatively uncorrelated with other sources.
► Official information sources are discounted when a perception of bias and credible alternatives exist.
► Traders predicted more accurately than track forecasts, except for hurricanes close to landfall, perhaps due to a favorite-longshot bias.