Proxy SVAR Identification of Monetary Policy Shocks - Monte Carlo Evidence and Insights for the Us

In empirical macroeconomics, proxy structural vector autoregressive models (SVARs) have become a prominent path towards detecting monetary policy (MP) shocks. However, in practice, the merits of proxy SVARs depend on the relevance and exogeneity of the instrumental information employed. Our Monte Carlo analysis sheds light on the performance of proxy SVARs under realistic scenarios of low relative signal strength attached to MP shocks and alternative assumptions on instrument accuracy. In an empirical application with US data we argue in favor of the specific informational content of instruments based on the dynamic stochastic general equilibrium model of Smets and Wouters (2007). A joint assessment of the benchmark proxy SVAR and the outcomes of a structural covariance change model implies that from 1973 until 1979 monetary policy contributed on average between 2.2 and 2.4 percent of annual inflation. For the so-called Volcker disinflation starting in 1979Q4, the benchmark structural model shows that the Fed's policy measures effectively reduced the GDP deflator within three years (i.e. by -12.26 pp until 1982Q3). While the empirical analysis largely conditions on a small-dimensional trinity SVAR, the benchmark proxy SVAR shocks remain remarkably robust within a six-dimensional factor-augmented model comprising rich information from Michael McCracken's database (FRED-QD)