This paper undertakes both a narrow and wide replication of the constant coefficients vector autoregression (VAR) identified with sign restrictions considered by Peersman (Journal of Applied Econometrics 2005; 20(2): 185-207. His results for the US are robust to an increase in the sample period from 2002:Q2 to 2014:Q2, but the extension to time-varying parameters highlights the importance of testing the robustness of results against time variation. In particular, there are differences across models regarding the role of individual shocks during the 2001 US slowdown.
