Independent Spike Models: Estimation and Validation

Abstract:
We apply a class of Markov switching models (independent spike models) to six European electricity markets and two European gas markets. This paper extends the current framework by introducing Gamma distributed spikes, which improves the fit for most energy markets. The models are quite complex. The robustness of the estimates is therefore evaluated using three different estimation strategies: direct maximization of the likelihood function, the Expectation-Maximization algorithm, and Markov Chain Monte Carlo (MCMC). The seasonal variation is corrected for by using the month-ahead forward price as a predictor. The models provide good empirical results for most markets.