Identifying News Shocks from SVARs

Résumé

This paper investigates the reliability of SVARs to identify the dynamic effects of news shocks. We show analytically that the dynamics implied by SVARs, using both long–run and short–run restrictions, are biased. However, the bias vanishes as long as news shocks account for most of the variability of the endogenous variable and the economy exhibits strong forward–looking behavior. Our simulation experiments confirm these findings and further suggest that the number of lags is a key ingredient for the success of the VAR setup. Furthermore, a simple correlation diagnostic test shows that news shocks identified using both restrictions are found to exhibit a correlation close to unity, provided that news shocks drive an overwhelming part of aggregate fluctuations.

Mots-clés

News shocks; SVARs; Identification; Diagnostic Test; Non–fundamentalness;

Codes JEL

C32: Time-Series Models • Dynamic Quantile Regressions • Dynamic Treatment Effect Models • Diffusion Processes
C52: Model Evaluation, Validation, and Selection
E32: Business Fluctuations • Cycles