This work discusses potential pitfalls of applying linear regression models for explaining the relationship between spot and futures prices in electricity markets. In particular, the bias coming from the simultaneity problem, the effect of correlated measurement errors and the impact of seasonality on the regression results. Studying a 13-year long (1998-2010) price series of spot and futures prices at Nord Pool and employing regression models with GARCH residuals, we show that the impact of the water reservoir level on the risk premium is positive, which is to be expected, but contradicts the results of Botterud et al. (2010). We also show that after taking into account the seasonality of the water level, the storage cost theory proposed by Botterud et al. (2010) to explain the behavior of convenience yield has only limited support in the data.