Meteorological Forecasts and the Pricing of Temperature Futures

Abstract

Futures contracts on the weather may seemodd, yet weather risk can be a serious matter. Weather is plainly not a storable commodity, so the cost-of-carry pricing model does not apply. Instead, one expects a temperature futures contract to be priced close to the consensus expectation based on current information for the temperature variable at contract maturity. But everyone knows how difficult it is to predict the weather, even though some patterns are well understood (e.g., it is colder in the winter than in the summer, on average). In this article, the authors build and estimate temperature models for three major cities that have traded weather futures contracts. The models include seasonal variation, positive serial correlation over short periods, a possible longterm warming trend, and random fluctuation. To the model based on historically observed temperatures, they add forwardlooking weather forecasts from an online weather service. These forecast-augmented models are more accurate in matching market prices for weather futures, and they are even closer to futures prices than would be “perfect” forecasts constructed from realized temperature data