Modeling the Dynamics of Temperature with a View to Weather Derivatives

Abstract:     
The accurate specification of the process that the temperature follows over time is a prerequisite for the pricing of temperature derivatives. To this end, a horse race of alternative specifications of the dynamics of temperature is conducted by evaluating their out-of-sample forecasting performance under different evaluation metrics and forecast horizons. An extensive dataset of the daily average temperature measured at different locations in Europe and U.S. is employed. We find that a developed principal components model and a combination forecasts model perform best in U.S. and Europe, respectively. Point forecasts for popular temperature indices are formed, as well. The results have implications for the pricing and trading of the fast growing class of temperature derivatives, as well as for forecasting temperature.

Number of Pages in PDF File: 43

Keywords: Point forecasts, Predictability, Principal Components, Temperature, Temperature indices

JEL Classification: C53, G10, G12, G13

working papers series