Identifying noise shocks

We study identifying restrictions that allow news and noise shocks to be recovered empirically within a Bayesian structural VARMA framework. In population, the identification scheme we consider exactly recovers news and noise shocks. Monte Carlo evidence further demonstrates its excellent performance, as it recovers the key features of the postulated data-generation process–the real-business cycle model of Barsky and Sims (2011) augmented with noise shocks about future total factor productivity-with great precision. We provide several empirical applications of our identification scheme. Evidence uniformly support the conclusion that noise shocks play a minor role in macroeconomic fluctuations.