A1736
Title: A mixed-frequency factor model for nowcasting French GDP
Authors: Marie Bessec - University Paris Dauphine (France)
Julien Andre - Banque de France (France) [presenting]
Abstract: A novel nowcasting model is introduced for quarterly real GDP growth in France, developed at the Banque de France. The model relies on the mixed-frequency three-pass regression filter (MF-3PRF) and aims to forecast the initial release of French GDP growth. Three distinct models designed to nowcast French GDP growth are presented during each month of the quarter. Through Monte Carlo simulations and analysis of French data, it is demonstrated that the accuracy of the forecasts can be significantly enhanced by employing an appropriate temporal aggregation scheme for the monthly indicators in the first step of the method. With a pseudo-real-time assessment, it is found that the new model performs well when compared to simple benchmarks and existing tools at the Banque de France, particularly during the first two months of the quarter. Furthermore, by extending the formulae for the contributions of the predictors to the mixed-frequency case, the contributions of various groups of variables are analyzed on the demand and supply side of French growth. It is found that all groups of variables have exerted a negative influence on French growth since the outbreak of the COVID-19 pandemic in 2020.
