Abstract: This paper is concerned with the e ects of weather uncertainty on the electricity
future curve. Following the approach used by Lucia and Schwartz (2002), the behavior of
the underlying spot price is assumed to consist of two components: a totally predictable
deterministic component that accounts for regularities in the evolution of prices and a
stochastic component that accounts for the behavior of residuals from the deterministic
part. The regime-switching with time-varying transition probabilities has been used to
account for the deterministic part of the spot price. The weather uncertainty is modeled
consistently with seasonal forecast probabilities from the CPC (Climate Prediction Center)
outlook. It is assumed that all the temperature paths are known under given seasonal
forecast probabilities. If the temperature process is known, the electricity load can be
predicted very accurately by the time series model using temperature and other explanatory
variables. So it can be assumed that all the load paths can be known if all temperature
paths are given. Furthermore, if temperature and electricity load are known, the spot price
can be predicted by the model specification, since the time-varying transition probabilities
can be calculated by the logistic functions of temperature and load, and since the tworegime
spot prices can be predicted by the time series model with exogenous variables such
as temperature and load. Therefore, it can be assumed that the behavior of the spot price
can be predictable if the temperature process is known. So the electricity future price can
be calculated, under the given seasonal forecast probabilities from the CPC outlook that
reflects the weather uncertainty. The future curve reflects clearly behaviors of the spot
price a ected by the weather patterns. As the summer gets warmer, the high price regime
is more likely to be realized so that the future price may increase.
JEL Classification: G13.
Keywords: electricity future price, incomplete market, regime-switching model with
time-varying transition probabilities, weather generator, seasonal forecast.
