Identification, Vector Autoregression, and Block Recursion
Federal Reserve Bank of Atlanta
Working Paper 96-8
In the applications of identified VAR models, finite-sample properties are not obvious to obtain when identifying restrictions are imposed on some lagged relationships. As a result, researchers have either left lagged relationships unrestricted even though some restrictions clearly make economic sense or failed to provide correct inference of the estimates. We extend the Bayesian methodology in the existing literature to these cases and develop the blockwise Monte Carlo methods. We show how to implement these methods to obtain the estimation and inference.
JEL classification: C11, C15, C32, C50
Key words: Contemporaneous recursive blocks; identifying restrictions; likelihood; finite samples, posterior; blockwise Monte Carlo methods