Identification of Singular and Noisy Structural VAR Models : The Collapsing-Ica Approach

When the number of variables is larger than the number of structural shocks driving the economy, the associated structural VAR system is said to be singular. We propose an identification method for singular structural VAR models contaminated by noise that combines a collapsing procedure with the Independent Component Analysis. We discuss the consistency of the proposed combined procedure and examine its finite sample properties with Monte Carlo simulations. The empirical application of the proposed scheme on U.S. data identifies the low dimensional system of structural shocks driving the U.S. economy. Interestingly, despite the flexibility of the identification method proposed, our impulse response functions display no evidence of price-puzzle effect