Hourly Electricity Prices

Day-ahead hourly electricity prices cannot be seen as either a time-series or a mean reverting process over days and hours since the agents on the international day-ahead power markets that set these prices valuate the delivery of power for each hour on the next day at the same moment in time. We propose a panel data set-up that exploits the fact that the agents information set updates from one day to the next but remains constant for each individual hour. We perform model selection tests on a fixed-effects model that combines the time-dimension for the inter-day price movements and a cross-section for the intra-day price movements, hence matches the above described market microstructure, and a time-effects model that only exploits the time-dimension of the panel. We document empirical evidence that the fixed-effects model is most capable of describing the stylized facts of the panel. We also report hourly-varying mean reversion behavior, which supports our observation that each hour should be treated differently.