The mean-reverting nature and seasonal patterns of electricity prices throughout the world provide evidence that pricing techniques which rely on random walk assumptions and storage costs may be inappropriate for pricing electricity derivatives. Instead, economic intuition and examination of empirical data suggest that forces of supply and demand produce some predictable spot price behavior which nevertheless remains consistent with no arbitrage theories.; Through formal empirical comparison of four diffusion models and their jump-diffusion counterparts, we show that "geometric" Ornstein-Uhlenbeck mean-reversion (with jumps) is generally superior at characterizing the dynamic behavior of electricity spot prices than Brownian motion, Ornstein-Uhlenbeck mean-reversion, and geometric Brownian motion (without and with jumps). Specifically, we conduct twelve log-likelihood comparisons of these models: three each for data series from Norway, the United Kingdom, California, and Victoria, Australia, and find that this model is the best performer in nine of the races and is a strong performer in the remaining three.; We then derive the dynamics of electricity derivative prices implied by this "winning" model of spot price dynamics. By examining the implied dynamic behavior of the expected prices, we avoid estimation of storage costs and convenience yield in deriving a general derivative pricing formula. For the diffusion model, the derivation is based upon replication arguments, while in the jump-diffusion model, we assume the existence of a state-price deflator which prices both the jump and diffusion risks systematically. The derived solutions for forward, call, and put prices reflect this dynamic behavior and suggest a methodology for determining derivative prices in non-storable commodity markets. Examination of related literature shows consistency with existing theories and empirical studies concerning expected risk premium, the theory of storage, convenience yield, and seasonality.
展开▼
