Comment

In this paper, Debortoli, Galı́, and Gambetti offer compelling empirical evidence that extraordinary actions taken by the Federal Reserve were able to shield the macroeconomy from many of the policy constraints associated with the zero lower bound (ZLB) on nominal interest rates. As Debortoli et al. argue, if these extraordinary actions had been ineffective, the United States would have witnessed a change in the volatility of macro aggregates and a change in their response to specific nonfinancial shocks. Yet volatility and impulse responses remained largely unchanged during the ZLB period. There is no one more qualified than the paper’s first discussant to discuss the ZLB, the Fed’s actions, and their effects on the macroeconomy. With this in mind, I will offer no comments of substance about this excellent paper, beyond the observation that I am in agreement with Debortoli et al.’s overall empirical conclusions. Instead, I will focus my comments on a methodological issue: statistical inference in sign-restricted structural vector autoregressions (SVARs), which is one of the methods used in Debortoli et al.’s paper. Sign-restricted SVARs are an increasingly popular method for estimating dynamic causal effects in macroeconomics. Many researchers use a variant of Uhlig’s (2005) Bayes method for imposing these sign restrictions and conducting inference. Thismethod has both strengths and weaknesses. The strengths are widely recognized by macroeconomists but the weaknesses far less so. This discussion explains and highlights these weaknesses. I make two initial comments. First, Debortoli et al. use a sophisticated time-varying SVAR identified by both long-run equality restrictions and shorter-run sign restrictions. To keep things simple, I will focus on a time-invariant SVAR. Second, there is nothing original in my comments beyond a few numerical calculations. Sign-restricted SVARs are a special