Data and Code for: Signing Out Confounding Shocks in Variance-Maximizing Identification Methods

Recent papers have examined the dominant drivers of business cycles using variance-maximizing techniques for identification. However, identification is poor when shocks other than the target of interest play large roles in driving volatility at the targeted frequency or horizon, leading them to capture a "hybrid" shock. This paper suggests a simple fix that lowers biases in the impulse responses. The fix is to include theoretically informed sign and magnitude restrictions at the identification stage of the vector auto-regression. Applying this to U.S. data we find a broadly equal role for demand and supply shocks in generating business-cycle fluctuations.