An empirical comparison of alternate regime-switching models for electricity spot prices

a r t i c l e i n f o Keywords: Electricity spot price Spikes Markov regime-switching Heteroscedasticity Inverse leverage effect One of the most profound features of electricity spot prices are the price spikes. Markov regime-switching (MRS) models seem to be a natural candidate for modeling this spiky behavior. However, in the studies published so far, the goodness-of-fit of the proposed models has not been a major focus. While most of the models were elegant, their fit to empirical data has either been not examined thoroughly or the signs of a bad fit ignored. With this paper we want to fill the gap. We calibrate and test a range of MRS models in an attempt to find parsimonious specifications that not only address the main characteristics of electricity prices but are statistically sound as well. We find that the best structure is that of an independent spike 3-regime model with time-varying transition probabilities, heteroscedastic diffusion-type base regime dynamics and shifted spike regime distributions. Not only does it allow for a seasonal spike intensity throughout the year and consecutive spikes or price drops, which is consistent with market observations, but also exhibits the 'inverse leverage effect' reported in the literature for spot electricity prices. The valuation of electricity contracts is not a trivial task. If the model is too complex the computational burden will prevent its on-line use in trading departments. On the other hand, if the price process chosen is inappropriate to capture the main characteristics of electricity prices, the results from the model are unlikely to be reliable. The uniqueness of electricity as a commodity prevents us from simply using models developed for the financial or other commodity markets. Electricity cannot be stored economically and requires immediate delivery, while end-user demand shows high variability and strong weather and business cycle dependence. Effects like power plant outages or transmission grid (un)reliability add complexity and randomness. The resulting spot price series exhibit strong seasonality on the annual, weekly and daily level, as well as, mean reversion, very high volatility and abrupt, short-lived and generally unanticipated extreme price changes known as spikes or jumps. Despite numerous attempts (for reviews see e.g. Benth and Benth (2008), Bunn (2004) and Kaminski (2004)), the need for realistic models of price dynamics capturing the unique characteristics of electricity and adequate derivatives pricing techniques still has not been fully satisfied. It is the aim of …