A0272
Title: Inference based on time-varying SVARs identified with sign and zero restrictions
Authors: Juan Rubio-Ramirez - Emory University (United States)
Daniel Waggoner - Federal Reserve Bank of Atlanta and Emory University (United States)
Jonas Arias - Federal Reserve Bank of Philadelphia (United States) [presenting]
Minchul Shin - Federal Reserve Bank of Philadelphia (United States)
Abstract: An approach for Bayesian inference is proposed in time-varying structural vector autoregressions identified with sign and zero restrictions. The linchpin of the approach is a class of rotation-invariant time-varying SVARs in which for any given sequence of structural parameters belonging to the class it is possible to find another sequence that has the same posterior density for any realization of the data. We develop two algorithms. The first applies to the case in which the identification strategy involves only sign restrictions. The second applies to the case in which the identification strategy involves sign and zero restrictions.
