Modelling and forecasting temperature based weather derivatives

The temperature-based Weather Derivative market in Australia is largely illiquid. One of the barriers to enhancing liquidity and reducing premiums in this market is the uncertainty surrounding the pricing of these derivatives. With the aim of reducing the uncertainty in mind, this study reviews both time-series and stochastic approaches to modelling prices, and settles on time-series models to forecast the Sydney accumulated cooling degree day (CDD) and heating degree day (HDD) index levels. Two daily and one intraday models are proposed based on a Fourier Transformation of temperature, as well as a wavelet reconstructed Fourier Transformation. All models are compared to a current weather index pricing model and a naïve benchmark model. The results suggest that, overall, the HDD index forecast is superior to the CDD index forecast and that the proposed models show forecast improvement over the current model and benchmark for the CDD index.