Working Paper: CEPR ID: DP3098
Authors: Carlo A. Favero; Massimiliano Marcellino
Abstract: Nowadays a considerable amount of information on the behaviour of the economy is readily available, in the form of large datasets of macroeconomic variables. Central bankers can be expected to base their decisions on this very large information set. Yet the academic profession has shown a clear preference for using small models to highlight stylized facts and to implement policy simulation exercises. Omitted information is then a potentially relevant problem. Recent time-series techniques for the analysis of large datasets have shown how vast an amount of information can be captured by few factors. In this paper we combine factors extracted from large datasets with more traditional small-scale models to analyse monetary policy in Europe. In particular, we model hundreds of macroeconomic variables with a dynamic factor model, and summarize their informational content with a few estimated factors. These factors are then used as instruments in the estimation of forward-looking Taylor rules, and as additional regressors in structural VARs. The latter are then used to evaluate the effects of unexpected and systematic monetary policy.
Keywords: Dynamic Factors; Monetary Policy; Small Models
JEL Codes: E50; E52; E58
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| large datasets (C55) | monetary policy outcomes (E52) |
| dynamic factor models (C22) | monetary policy outcomes (E52) |
| factors extracted from datasets (C38) | estimation of monetary policy effects (E60) |
| factors extracted from datasets (C38) | accurate estimation of effects on inflation and output gaps (E31) |
| use of dynamic factor models (C22) | better policy simulations (C54) |