FaMIDAS: A Mixed Frequency Factor Model with MIDAS structure

In this paper a dynamic factor model with mixed frequency is proposed (FaMIDAS), where the past observations of high frequency indicators are used following the MIDAS approach. This structure is able to represent with richer dynamics the information content of the economic indicators and produces smoothed factors and forecasts. In addition, it is particularly suited for real time forecast as it reduces the problem of the unbalanced data set and of the revisions in preliminary data. In the empirical application we specify and estimate a FaMIDAS to forecast Italian quarterly GDP. The short-term forecasting performance is evaluated against other mixed frequency models in a pseudo-real time experiment, also allowing for pooled forecast from factor models.