A regime switching long memory model for electricity prices

In this paper we develop a regime switching model which can generate long memory (fractional integration) in each of the regime states. This property is relevant, e.g., for the deregulated market for electricity power in the Nordic countries, which is characterized by electricity spot prices with a high degree of long memory. It occurs that in some time periods bilateral prices are identical whereas in other periods the prices differ. If the price series are fractionally integrated, then in the former regimes, an extreme form of fractional cointegration amongst prices will exist. The latter regimes occur when a capacity congestion exists across regions and multiple price areas will result. We define a Markov switching fractional integration model from which the fractional orders of integration in separate states can be estimated using maximum likelihood techniques. The model is adapted to data for the Nordic electricity spot market, and we find that regime switching and long memory are empirically relevant to co-exist. In particular, we find that the price behaviour for single markets can be very different depending upon the presence or absence of bottlenecks in electricity transmission. Using Monte Carlo forecasting we find that the regime switching model appears to be especially attractive in forecasting relative prices.