A0808
Title: Expectations vs data news on nowcasting US GDP
Authors: Ana Galvao - Bloomberg Economics (United Kingdom) [presenting]
Abstract: Nowcasting models typically rely on the predictive content of a set of indicators to anticipate the first estimate of key economic variables such as GDP growth and employment. The accuracy of these models is frequently not significantly better than consensus expectations surveyed just before the release. The predictive content of monthly indicators is compared to their consensus expectations to nowcast US GDP growth. Optimal weights are estimated for both types of predictors using a Bayesian machine learning method applied to a mixed-data sampling regression (MIDAS). Empirical results support the usefulness of surveyed expectations in forecasting GDP growth in real time.
