Monday, january 20th, 2020
Seminar A, at 15 p.m.
Queen Mary University of London
"SVARs Identification through Bounds on the Forecast Error Variance"
This paper provides tools for estimation and inference in Structural Vector Autoregressions (SVARs) that are set-identied through bound restrictions on the Forecast Error Variance Decomposition (FEVD).
The contributions can be summarized as follows. First, the paper shows FEVD bounds correspond to quadratic inequality restrictions on the columns of the rotation matrix transforming reduced-form residuals into structural shocks. These restrictions could be imposed alone or alongside the linear restrictions that are currently considered in the literature on SVARs that are set-identied through equality and/or sign restrictions.
Second, the paper establishes theoretical conditions such that bounds on the FEVD lead to
a reduction in the width of the impulse response identied set relative to only imposing sign restrictions.
Third, this article proposes a robust Bayesian approach to inference, although the insights could also apply to standard Bayesian or frequentist inference. Fourth, the article shows that elicitation of the bounds could be based on DSGE models with alternative parametrizations; the method is extended to incorporate uncertainty about the bounds. Fifth, simulation studies and an empirical application illustrate the potential usefulness of FEVD restrictions for obtaining informative inference in set-identied monetary SVARs, where loose bounds on the FEVD suggest a signicant eect of monetary policy on the short-run real activity.
Keywords: Bounds, Forecast Error Variance, Monetary Policy, Set Identication, Sign Restrictions,
Structural Vector Autoregressions (SVARs). JEL: C32, C53, E10, E52.
|Go to News|