Statistically Identified SVAR Model With Potentially Skewed and Fat-Tailed Errors
64 Pages Posted: 20 Sep 2021 Last revised: 9 Mar 2023
Date Written: March 8, 2023
Abstract
We introduce a structural vector autoregressive (SVAR) model with each of the mutually independent errors following a skewed generalized t-distribution that is more flexible than a Student's t-distribution typically considered. Hence, the danger of distributional misspecification is diminished and, more importantly, identification is substantially strengthened. Due to non-Gaussianity, the model is statistically identified, so that the plausibility of any economic identifying restrictions can be formally assessed. In an empirical application to U.S. monetary policy, the data lend support to narrative sign restrictions in identifying the monetary policy shock. In contrast to some of the previous literature, we find a strong negative response of real activity to contractionary monetary policy after a few months' delay.
Keywords: structural vector autoregression, non-Gaussian time series, narrative sign restrictions, monetary policy
JEL Classification: C32, C51, C54, E52
Suggested Citation: Suggested Citation