This paper develops a new measure of US fiscal policy shocks that intends to avoid the anticipation problem affecting conventional measures, being also arguably free from endogeneity. The shocks are intended to capture changes to the component of anticipated fiscal policy that is exogenous to economic developments. Key economic variables such as output and interest rates respond quickly and significantly to a realization of the estimated shock and, in the first part of the sample, 1969-1988, in a way consistent with the Keynesian prior. In contrast, over the period 1989-2008 the effects are at odds with that prior, with fiscal loosening producing contractionary impacts.