Title: Combination of forecasts in dynamic factor models: Application to the Italian power exchange
Authors: Andres M Alonso - Universidad Carlos III de Madrid (Spain)
Guadalupe Bastos - Universidad Carlos III de Madrid (Spain)
Carolina Garcia-Martos - Universidad Politecnica de Madrid (Spain) [presenting]
Abstract: Nowadays, electricity markets are liberalized in the vast majority of countries of our socio-economic context. That is why forecasting electricity prices is a crucial task, both in the short (one day up to a week ahead) and long run (year-ahead). In the last decade, many methodological contributions have been developed in order to cope with empirical features of electricity price series. Particularly, Dynamic Factor Models (DFM) have emerged as a very interesting alternative for reducing forecasting errors both in the short and long run. However, when estimating a DFM, a model for the common factors must be selected. This task can be difficult, and it can occur that there is no a unique valid model for each factor. We propose model averaging as a possible alternative to deal with this issue, expecting that this would give smaller forecasting errors than modelling common factors just by a single model. Numerical results are provided for electricity spot prices in the Italian Power Exchange (IPEX) but the presented methodology can be applied to any other market. Results of combinations of forecasts are shown for different forecasting horizons.