Working Paper: CEPR ID: DP15926
Authors: Juan Antolin-Diaz; Thomas Drechsel; Ivan Petrella
Abstract: A key question for households, firms, and policy makers is: how is the economy doing now? We develop a Bayesian dynamic factor model and compute daily estimates of US GDP growth. Our framework gives prominence to features of modern business cycles absent in linear Gaussian models, including movements in long-run growth, time-varying uncertainty, and fat tails. We also incorporate newly available high-frequency data on consumer behavior. The model beats benchmark econometric models and survey expectations at predicting GDP growth over two decades, and advances our understanding of macroeconomic data during the recession of spring 2020.
Keywords: nowcasting; daily economic index; dynamic factor models; real-time data; bayesian methods; fat tails
JEL Codes: E32; E23; O47; C32; E01
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Bayesian Dynamic Factor Model (DFM) (C22) | US GDP growth (O49) |
| high-frequency data on consumer behavior (D12) | US GDP growth (O49) |
| US GDP growth (O49) | economic activity (E20) |
| economic activity (E20) | aggregate economic activity (E10) |
| economic crises (COVID-19 pandemic) (G01) | US GDP growth (O49) |
| Bayesian Dynamic Factor Model (DFM) (C22) | understanding of economic dynamics (E32) |