Estimating US Monetary Policy Shocks Using a Factor-Augmented Vector Autoregression: An EM Algorithm Approach

Economy-wide e§ects of shocks to the US federal funds rate are esti-mated in a state space model with 120 US macroeconomic and Önancialtime series driven by the dynamics of the federal funds rate and a few dy-namic factors. This state space system is denoted a factor-augmented VAR(FAVAR) by Bernanke et al. (2005). I estimate the FAVAR by the fullyparametric one-step EM algorithm as an alternative to the two-step prin-cipal component method and the one-step Bayesian method in Bernankeet al. (2005). The EM algorithm which is an iterative maximum likelihoodmethod estimates all the parameters and the dynamic factors simultane-ously and allows for classical inference. I demonstrate empirically that thesame impulse responses but better Öt emerge robustly from a low orderFAVAR with eight correlated factors compared to a high order FAVARwith fewer correlated factors, for instance four factors. This empirical re-sult accords with one of the theoretical results from Bai & Ng (2007) inwhich it is shown that the information in complicated factor dynamics maybe substituted by panel information