claim or prove damages. Second, there is little moral hazard. Third, unlike insurance, weather derivatives
allow one to hedge against comparatively good weather in other locations, which may be bad for local
business (e.g., a bumper crop of California oranges may lower the prices received by Florida growers).
Weather forecasting is crucial to both the demand and supply sides of the weather derivatives
market. Consider first the demand side: any firm exposed to weather risk either on the output (revenue)
side or the input (cost) side is a candidate for productive use of weather derivatives. This includes obvious
players such as energy companies, utilities and insurance companies, and less obvious players such as ski
resorts, grain millers, cities facing snow-removal costs, consumers who want fixed heating and air
conditioning bills, and firms seeking to avoid financial writedowns due to weather-driven poor
performance. The mere fact that such agents face weather fluctuations, however, does not ensure a large
hedging demand, because even very large weather fluctuations would create little weather risk if they were
highly predictable. Weather risk, then, is about the un predictable component of weather fluctuations -
“weather surprises,” or “weather noise.” To assess the potential for hedging against weather surprises, and
to formulate the appropriate hedging strategies, one needs to determine how much weather noise exists for
weather derivatives to eliminate, and that requires a weather model. What does weather noise look like
over space and time? What are its conditional and unconditional distributions? Answering such questions
requires statistical weather modeling and forecasting, the topic of this paper.
Now consider the supply side - sellers of weather derivatives who want to price them, arbitrageurs
who want to exploit situations of apparent mispricing, etc. How should weather derivatives be priced? It
seems clear that standard approaches to arbitrage-free pricing (e.g., Black-Scholes, 1973) are inapplicable
in weather derivative contexts. In particular, there is in general no way to construct a portfolio of financial
assets that replicates the payoff of a weather derivative. Hence the only way to price options reliably is by
using forecasts of the underlying weather variable, in conjunction with a utility function, as argued for
example by Davis (2001). This again raises the crucial issue of how to construct good weather forecasts
(not only point forecasts, but also, and crucially, complete density forecasts), potentially at horizons much
-2-