IMF Working Papers

This paper proposes a “quasi-agnostic” sign restriction procedure to identify structural shocks in frequentist structural vector autoregression (SVAR) models. It argues that low acceptance rates, inherent to agnostic sign restriction procedures, are not necessarily an indication of model misspecification. They can be low because agnostic procedures fail to exploit the ex ante priors on the sign of responses of macro variables to structural shocks.