Weather Derivatives – What are they, where are they traded and how are they priced?
Recently, global financial markets have witnessed a rapid escalation in the popularity and the trading of weather derivatives. These derivatives are predominantly temperature-based but can include derivatives written against other weather phenomena such as precipitation, wind, frost and snowfall.
The weather derivative market in the U.S. has grown from around US$500 million in 1998 to over US$ 12 billion in 2003, and is still growing. Such growth has been attributed to the deregulation of energy markets and the significant effect that weather risk has on the volatility of revenue generated by a large number of companies. Significantly, these companies include electricity generators and retailers. Indeed, the weather derivative market in the U.S. was stimulated by the mild winters of 1997 and 1998, the consequence of the ‘El-Nino’ phenomenon. The associated potential for corporate earnings decline signalled the need for a hedge against abnormal weather behaviour.
What are weather derivatives and where and how are they traded?
Importantly, weather derivative contracts differ from insurance contracts. Unlike insurance, with a weather derivative there is no need to file a claim or to prove damages. Further, there is virtually no moral hazard involved except in extraordinary cases such as the party who was long in precipitation attempting to seed clouds to induce rainfall. Also, a weather derivative can be used to hedge against abnormally good weather conditions in one location that can adversely impact on product prices in another. An excellent wine vintage in one region can lead to premium bottle prices over wine from other regions. This is but one of many agricultural commodity examples of this type of hedge.
The most widely traded derivative contract written against weather is a temperature-based derivative. In the U.S., prior to the introduction of trading in weather contracts on the Chicago Mercantile Exchange (CME) in 1997, all such derivatives were traded on the Over-The-Counter (OTC) market. The CME introduced both futures and options contracts on temperature-based derivatives for ten major U.S. cities; namely, Atlanta, Chicago, Cincinnati, Dallas, Des Moines, Las Vegas, New York, Philadelphia, Portland and Tucson.
A temperature derivative allows purchasers to protect themselves against temperature being above or below an agreed threshold for a prescribed future time period. A popular U.S. example of this type of derivative is found in the electricity generation and supply industry. It is well understood that electricity load requirements are highly correlated with temperature. Generators and retailers can use temperature-based weather derivatives to hedge themselves against exposure to highly volatile electricity pool prices that can result from the occurrence of a period of unanticipated high or low daily temperatures.
For temperature-based contracts, the spot price level is represented by either the monthly-accumulated Heating Degree Day (AccHDD) or the monthly-accumulated Cooling Degree Day (AccCDD) index. A degree day is a measure of the disparity between a day’s average temperature and a specified threshold. The daily average temperature is defined to be the arithmetic average between the daily maximum and minimum temperature between 12.01 a.m. and midnight. Daily HDD and CDD measure the coldness and warmness, respectively, of the daily temperature compared to a threshold level of 65° F or 18° C. The choice of 65° F as an industry standard for the threshold level is due to the belief that, for every degree above and below this level, increasing amounts of energy are required for cooling and heating, respectively.
Seasonality in Daily Average Temperature for Atlanta and Chicago
For the most part, HDD and CDD contracts are written against the accumulation of HDD (AccHDD) or CDD (AccCDD) for each day over a defined future period which is typically a calendar month. Contracts longer than a month usually comprise a series of monthly contracts. Similar to other futures contracts, the AccHDD or AccCDD index futures are legally binding agreements to purchase or sell the value of the HDD or CDD index for a specified price and date. Consistent with other traded derivative contracts, the CME Clearing House performs the service of novation. In order to protect itself against assumed risk, the CME Clearing House imposes a daily marking-to-market. The full value of the contract is not transferred but rather, a final marking-to-market on the expiry date is based on the level of the cumulative value of the HDD and CDD, accumulated for each month. This method is known as cash settlement.
How are temperature-based derivatives priced?
Deriving a fair price for a weather derivative is a contentious issue. Disagreement on the development of an accepted pricing model for these instruments translates into higher volatility and a larger bid-ask spread which, in turn, reduces the liquidity. Despite the recent increases in trading volume, the lack of a universally accepted pricing model is perceived by some to be the crucial issue restricting further growth of the weather derivative market.
Why can’t we use existing derivative pricing models?
Weather is a non-tradable underlying asset. By definition, the absence of a tradable underlying asset violates the no-arbitrage and market completeness assumptions. These are core assumptions on which the standard derivative pricing models are based. As a consequence, the direct application of the standard models for pricing weather derivatives is inappropriate. As previously noted, for temperature-based contracts, the spot price level is represented by the AccHDD or the AccCDD index that reverts to zero at he beginning of each month. This is in contrast to other financial indices whose compounded value of the spot price at the time a contract is purchased is represented by the expected value of the spot price at maturity. Accordingly, in pricing AccHDD or AccCDD derivative contracts, the ability to accurately derive the expected value of a particular month’s AccHDD or AccCDD index is crucial. Temperature forecasting models are used to provide interval forecasts of temperature based on simulation and, as a result, expected values of the above-mentioned temperature-based weather indices.
Researchers have confronted the challenge of developing temperature forecasting models which can be integrated into the options pricing framework while, at the same time, providing accurate estimates. These models can be categorised into two distinct groups. The first group relies on the standard Stochastic Brownian Motion framework, while the second focuses on the use of time series analysis to develop temperature forecasting models. In a recent study, Oetomo and Stevenson (Working Paper, Discipline of Finance, The University of Sydney, 2003) reviewed six different forecasting models proposed to price temperature-based weather derivatives. Three of these models were from each of the above two categories. The data set used in their study consisted of the daily maximum, minimum and average temperature, along with the monthly-accumulated HDD and CDD indices for the period from the 1st January, 1979, to the 31st December, 2002, for all contracts and for all ten cities listed on the CME. From the perspective of correct model specification it is important to recognise the seasonal nature of temperature data. Daily average temperature exhibits a full seasonal cycle of one calendar year and this behaviour is consistent throughout the full observation period from 1979 to 2002. The seasonality implicit in the data and its tendency to follow a sine-wave pattern is evident in the figure left that graphs the three year period between the 1st January, 1984, and the 31st December, 1986, for the cities of Atlanta and Chicago.
In order to compare the ability of the six models to estimate temperature and, therefore, the AccHDD and AccCDD indices, temperature data from January, 1979 to December, 1997 was used to estimate the parameters of the models that generated interval forecasts from January, 1998, through to the end of December, 2002. This latter (forecast) period covers the time from the commencement of trading on the CME. After comparing the six models, Oetomo and Stevenson concluded that no one model was able to consistently outperform the others. It appeared more difficult to forecast the AccCDD index, while the most appropriate forecasting model varied between cities and months, as well as across the estimation and forecasting time periods. The accuracy of forecasts beyond thirty days (or one month) was found to be unreliable. After controlling for seasonality, forecast quality for both classes of models was increased but no improvement in forecast accuracy resulted after accounting for long-term trend.
To fairly evaluate and compare the forecasting models, Oetomo and Stevenson also compared the present value of the estimated price generated by each model against the traded market price of a contract. Based on this analysis, the results provided an insight into the ability of the models to price the settlement level relative to that of the market. The analysis was conducted using data on 212 CDD futures contracts and 158 HDD futures
contracts traded on the CME between the 1st January, 1998, and the 31st December, 2002. While the models were able to price the AccHDD futures contracts more accurately than the market, this was not the case for the AccCDD contracts.
Weather derivatives, and particularly temperature-based derivatives, are likely to follow the global trend and experience rapid growth in Australia. At present they are only thinly traded Over-The-Counter. However, if the U.S. experience is to be repeated, then full exchange trading in Australia seems inevitable. The only drawback to the rapid escalation of the use of this type of derivative contract seems to be the lack of a general, functional and accurate pricing model.
