Abstract
This paper derives and implements a robust test of the short-run neutrality of anticipated policy that is valid when the information set used to form expectations is incorrectly specified and the effect of unanticipated policy on deviations of output from the natural rate varies over time. The test is based on the observation that if unanticipated policy affects output with a maximum lag of
n quarters, then any information prior to
n should be uninformative about output movements. The test is applied to quarterly data for the United States and the paper concludes that policy is not neutral.