Common Components Structural VARs

Structural VAR models (SVAR) produce results that can vary dramatically with the choice of variables, because information is deficient. We argue that if the variables of interest belong to a High-Dimensional Factor Model and are replaced in the SVAR by their common components, the information issue finds a solution, provided that the number of common components is larger than the number of structural shocks, so that the SVAR is dynamically singular. This is the Common Components Structural VAR (CC-SVAR). Our main contribution is that we prove consistency of our CC-SVAR estimates, which is far from trivial as our estimated SVAR tends to dynamic singularity. We apply our procedure to monetary policy shocks, finding that, with the CC-SVAR, results are robust to the choice of variables and well-known puzzles disappear.