Proxy VAR Models in a Data-Rich Environment

Discussion Papers 1831, 43 S.

Martin Bruns

2019

get_appDownload (PDF  1.19 MB)

Abstract

Structural VAR models require two ingredients: (i) Informational sufficiency, and (ii) a valid identification strategy. These conditions are unlikely to be met by small-scale recursively identified VAR models. I propose a Bayesian Proxy Factor-Augmented VAR (BP-FAVAR) to combine a large information set with an identification scheme based on an external instrument. In an application to monetary policy shocks I find that augmenting a standard small-scale Proxy VAR by factors from a large set of financial variables changes the model dynamics and delivers price responses which are more in line with economic theory. A second application shows that an exogenous increase in uncertainty affects disaggregated investment series more negatively than consumption series.

Martin Johannes Bruns

Wissenschaftlicher Mitarbeiter in der Abteilung Konjunkturpolitik



JEL-Classification: C38;E60
Keywords: Dynamic factor models, external instruments, monetary policy, uncertainty shocks