Pricing of weather derivatives

Values of weather derivatives depend on weather outcomes, such as temperature or precipitation. Academics have differed on how to value this type of financial instrument, since weather is not tradeable and the No-Arbitrage Pricing Theory cannot be applied. Cao and Wei (2004) propose a valuation model and I test its predicting accuracy by comparing simulated futures prices to market prices, for cumulative Heating/Cooling Degree Day futures for New York City for contracts offered by the Chicago Mercantile Exchange. The simulation of weather futures prices requires assumptions of values for the risk aversion parameter. Following Cao and Wei, the values -2, -10 and -40 are used. The simulation requires values for the speed of mean reversion of aggregate dividends. Following Cao and Wei, the values of 0.8, 0.9 and 0.99 are used. Due to the lack of a sufficiently long time series data to determine the daily correlation between temperature and aggregate dividends, Cao and Wei assume that the aggregate dividends depend on either the contemporaneous temperature or the 30 lagged temperatures. There are consequently 18 simulation settings. Results indicate that Cao and Wei's (2004) model is useful in predicting weather derivative prices, especially when the risk aversion parameter is -10. Forecast accuracy is very sensitive to the risk aversion parameter, followed by the number of temperature lags that aggregate dividends depend on. The speed of mean reversion of aggregate dividends is not found to be a crucial paramete