Abstract:

An n-variable structural vector auto-regression (SVAR) can be identified (up to shock order) from the evolution of the residual covariance across time if the structural shocks exhibit heteroskedasticity (Rigobon (2003), Sentana & Fiorentini (2001)). However, those moments are only available under specific parametric assumptions on the variance process. I propose a new identification method that identifies the SVAR up to shock orderings using unconditional moments implied by an arbitrary stochastic process for the variances. These moments are available without the parametric assumptions of existing approaches. I offer intuitive criteria to select among shock orderings; this selection does not impact inference asymptotically. The identification scheme performs well in simulation. I apply it to the debate on fiscal multipliers. As opposed to being too low, as argued by Mertens & Ravn (2014), the tax multipliers of Blanchard & Perotti (2002), are somewhat high. Narrative tax measures may be invalid instruments, explaining the discrepancy.

Notes: