Long memory in electricity prices

Abstract

This dissertation explores long-term memory in Australian electricity prices. The issue of anti-persistence in the logarithmic first differences of electricity price complicates direct application of risk management tools commonly used in finance such as Modern Portfolio Theory (MPT). Methods of measuring long-term memory effects are detailed and evidence of anti-persistence is given. Time series which exhibit long term memory effects are well described by the Stable Paretian family of distributions. Therefore, we examine how electricity prices are described by Stable Paretian distributions and what affect generation mix may have on this. Electricity prices from Australia's National Electricity Market (NEM) are examined using the rescaled range analysis as a measure of long memory and an the iterative type estimator for the parameters of the stable distribution. Logarithmic first differences of NEM prices, properly deseasonalized, exhibit anti-persistence as measured by the rescaled-range statistic and are well described by the stable distribution. While the parameters of the stable distribution of electricity prices do seem to vary with time there is no indication that this is due to changes in the electricity generation mix. Rather, an increase in natural gas prices seems to have a significant effect on the stable distribution parameters. Finally, this dissertation presents a preliminary investigation into applying MPT to the electricity sector in the presence of assets whose returns follow a Stable Paretian distribution.