Fitting semiparametric Markov regime-switching models to electricity spot prices

M. Eichler, D.D.T. Türk*

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review


Recently regime-switching models have become the standard tool for modeling electricity prices. These models capture the main properties of electricity spot prices well but estimation of the model parameters requires computer intensive methods. Moreover, the distribution of the price spikes must be fully specified although the high volatility of the spikes makes it difficult to check such distributional assumptions. Consequently, there are a number of competing proposals for the distribution in the spike regime. As an alternative, we propose a semiparametric Markov regime-switching model that leaves the distribution under the spike regime unspecified. We show that the model parameters can be estimated by employing robust statistical techniques. This presents an alternative to the existing estimation methods that are based on computer intensive numerical maximization of the likelihood function. The model in combination with the estimation framework is easier to estimate, needs less computation time and distributional assumptions. To show its advantages we compare the proposed model with a well-established Markov regime-switching model in a simulation study. Furthermore, we apply the model to logprices for the Australian electricity market. The results are in accordance with the results from the simulation study, indicating that the proposed model might be advantageous whenever the distribution of the spike process is not sufficiently known. The results are thus encouraging and suggest the use of our approach when modeling electricity prices and pricing derivatives. (C) 2012 Elsevier B.V. All rights reserved.
Original languageEnglish
Pages (from-to)614-624
JournalEnergy Economics
Publication statusPublished - 1 Jan 2013


Dive into the research topics of 'Fitting semiparametric Markov regime-switching models to electricity spot prices'. Together they form a unique fingerprint.

Cite this