The empirical literature using vector autoregressive models to assess the effects of fiscal policy shocks strongly disagrees on even the qualitative response of key macroeconomic variables to government spending and tax shocks. We provide new evidence for the U.S. over the period 1955-2006. We show that, controlling for differences in specification of the reduced-form model, all identification approaches used in the literature yield qualitatively and quantitatively very similar results as regards government spending shocks. In response to such shocks real GDP, real private consumption and the real wage all significantly increase following a hump-shaped pattern, while private employment does not react. In contrast, we find strongly diverging results as regards the effects of tax shocks, with the estimated effects ranging from non-distortionary to strongly distortionary. The di¤erences in results can to a large extent be traced back to differences in the size of automatic stabilizers estimated or calibrated for alternative identification approaches. These differences also translate into uncertainty about the effects of policy experiments typically considered in theoretical models.