CMStatistics 2016: Start Registration
View Submission - CFE
A0931
Title: Structural analysis using factor augmented VARs and three-pass regression filters Authors:  Anindya Banerjee - University of Birmingham (United Kingdom) [presenting]
Massimiliano Marcellino - Bocconi University (Italy)
Igor Masten - University of Ljubljana (Slovenia)
Abstract: A new approach is proposed to sign-restriction-based identification of structural shocks in FAVAR models. Using a new method to estimate factors called the three-pass regression filter (3PRF), sign restrictions are imposed on the responses of industrial production (IP) and consumer price index (CPI) in the first twelve months following a monetary policy (MP) shock. Each of the $N$ variables in our U.S. dataset is then used individually as a proxy for the 3PRF factors and for each of these proxies up to seven factors are extracted and used in a BBE-type FAVAR model. From the models satisfying the sign restrictions we attempt to obtain the one that is most representative. We observe that the impulse responses of non-sign-restricted variables are in accord with economic intuition. Moreover, the adjustment of prices seems to be much faster (and without permanent effect on the price level) than in BBE or similar studies. A further step of our analysis compares the 3PRF responses with impulse responses based upon one to seven factors estimated from standard principal component analysis (PCA). The responses of IP are quite similar under both approaches but price puzzles emerge when looking at the responses of CPI using PCA. Looking through 3PRF lenses we argue that adding more information which may be irrelevant leads to biased responses and erroneous identification of shocks.