Identification of Structural VAR Models Via Independent Component Analysis: A Performance Evaluation Study

Independent Component Analysis (ICA) is a statistical method that transforms set of random variables in least dependent linear combinations. Under the assumption observed data are mixtures non-Gaussian and independent processes, ICA able to recover underlying components, but scale order indeterminacy. Its application structural vector autoregressive (SVAR) models allows researcher impact shocks on series from estimated residuals. We analyze different estimators, recently proposed within field SVAR identification, compare their performance recovering coefficients. Moreover, after suggesting an algorithm solve indeterminacy problem, we assess size distortions estimators hypothesis testing. conduct our analysis by focusing distributional scenarios get gradually close Gaussian case, which case where methods fail components. In terms properties find no evidence outperforms all others. finally present empirical illustration using US identify effects government spending tax cuts economic activity, thus providing example techniques can be used for