Estimating the Payoffs of Temperature-based Weather Derivatives

Temperature-based weather derivatives are written on an index which is normally defined to be a nonlinear function of average daily temperatures. Recent empirical work has demonstrated the usefulness of simple time-series models of temperature for estimating the payoffs to these instruments. This paper argues that a more direct and parsimonious approach is to model the time-series behaviour of the index itself, provided a sufficiently rich supply of historical data is available. A data set comprising average daily temperature spanning over a hundred years for four Australian cities is assembled. The data is then used to compare the actual payoffs of temperature-based European call options with the expected payoffs computed from historical temperature records and two time-series approaches. It is concluded that expected payoffs computed directly from historical records perform poorly by comparison with the expected payoffs generated by means of competing time-series models. It is also found that modeling the relevant temperature index directly is superior to modeling average daily temperatures.