In this paper we propose a three-factor spike model that accounts for different speeds of mean reversion between normal and spiky shocks in the Scandinavian power market. In this model both short and long-run factors are unobservable and are hence estimated as latent variables using the Kalman filter. The proposed model has several advantages. First, it seems to capture in a parsimonious way the most important risks that practitioners face in the market, such as spike risk, short-term risk and long-term risk. Second, it explains the seasonal risk premium observed in the market and improves the fit between theoretical and observed forward prices, particularly for long-dated forward contracts. Finally, closed-form solutions for forward contracts, derived from the model, are consistent with the fact that the correlation between contracts of different maturities is imperfect. The resulting model is very promising, providing a very useful policy analysis and financial engineering tool to market participants for risk management and derivative pricing particularly for long-dated contracts. (C) 2010 Elsevier Ltd. All rights reserved.