Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Crop insurance programs rely on conditional predictive distributions of loss random variables (e.g., yield, revenue, loss costs, etc.) to determine probabilities and magnitudes of loss. The loss variables may be related to stochastic variables that are not known at the time the policy is priced. Such is the case for weather; weather is stochastic, realizations are not known when the crop insurance policy is sold, and there is often additional historical information on weather relative to the loss variable itself. We provide a Bayesian methodology for incorporating historical weather information in crop insurance rating. We apply the method in empirical applications to county‐level U.S. corn yields and loss cost ratios in the Midwest. The historical weather‐conditioned distributions differ from those based on shorter samples. In the yield distribution setting, additional temporal weather information leads to economic gains relative to other rating approaches; the magnitude of these gains increases with the amount of historical weather information included in the analysis.