Date : Jeudi | 2025-05-15 à 12h30
Lieu : Salle des thèses

Lien TEAMS : Cliquez ici pour rejoindre le séminaire du LÉO

Cyril DELL'EVA (Université de Rouen / LERN)

Inflation expectations in the QPM are approximated with the BER Survey of Inflation Expectations. These are forecasted by using an exogenous AR(1) process, constrained to converge to the target at the two year horizon. The AR(1) process imposes strong judgment on the forecasting of inflation and it is unlikely to forecast expectations correctly, outside a small short run window. We substitute the AR(1) process with an adaptive learning process, driven by data, to forecast BER expectations. We show that an adaptive learning process is a realistic approximation of the inflation expectations process and leads to a more accurate inflation forecast.