Spot-forward price relationships in the restructured electricity markets

The dissertation illustrates that two common manifestations of the restructured electricity industry in many regions, (1) the absence of long-dated forward markets; and (2) a bias of the short-dated forward prices in forecasting the subsequent spot price, are implied by the temporal and distributional properties of electricity spot prices themselves. With simulation data deduced from an idealized electricity market model, it shows that the constellations of forward prices and the expected volatility of spot price stay around their long-run seasonal means until near-delivery, implying that only a small number of forward markets suffice as an allocative signal and hedging instrument. The predictability of price relationships between electricity and a related commodity allows long-dated futures contracts of one commodity to be duplicated by those of the other. Analyses of the California electricity and the NYMEX natural gas futures price data provides empirical evidence that the spark-spread can be predicted reasonably well.;Previous studies econometrically examining relationships between the spot and short-dated forward price have often understated complications resulting from the distributional and temporal properties of electricity price series. The Monte Carlo simulation reveals that although the coefficient estimates and standard errors of the three regression models are consistent, the null hypothesis of unbiased forward price and risk neutrality are over-rejected by a considerable extent, even with a sample size far greater than for typical data about electricity prices.;Models of electricity price dynamics suggested for derivative pricing are subject to similar econometric issues, as well as to the bias in approximating the true price dynamics by a reduced-form equation. The Monte Carlo simulation reveals that three common specifications allowing non-constant volatility, a level-effect, GARCH, and a stochastic jump, approximate the true price dynamics less accurately than a constant variance specification yet are preferred by a likelihood criterion. A semi-structural model specifying electricity price as a non-linear function of the load and fuel price outperforms the reduced-form models by goodness of fit criterion, yet leads to an incorrect statistical inference about a risk premium. All these results suggest previous studies on the spot-forward price relationships in electricity markets need to be reexamined.