Identifying Shocks via Time-Varying Volatility

Abstract I propose to identify an SVAR, up shock ordering, using the autocovariance structure of squared innovations implied by arbitrary stochastic process for variances. These higher moments are available without parametric assumptions on variance process. In contrast, previous approaches exploiting heteroskedasticity rely path innovation covariances, which can only be recovered from data under specific The conditions identification testable. compare scheme existing in simulations and provide guidance estimation inference. use methodology estimate fiscal multipliers peaking at 0.86 tax cuts 0.75 government spending. find that shocks explain more variation output longer horizons. empirical implications my estimates consistent with theory narrative record than those based some leading approaches.