Author(s): Roberto Perotti
With fiscal foresight, the shocks identified by standard Vector Autoregression (SVAR) techniques can be non-fundamental for the variables of interest. In an important paper, Ramey (2011) uses direct measures of the private sector's forecast revisions of defense or federal spending to estimate the effects of government spending shocks in a VAR, obtaining the 'expectations - augmented' VAR, or EVAR. The response of GDP to these shocks is smaller than 1, and consumption and the real wage fall: this is consistent with the neoclassical model, but the opposite of recent results from SVARs. In this paper, I make three points. First, EVARs and SVARs give virtually the same results. Ramey reaches the opposite conclusion because she never estimates the two specifications on the same sample and with the same government spending variable. Second, the evidence from EVARs is not robust. It is enough to dummy out just two quarters during WWII (when rationing was introduced) or during the Korean War (when new Fed regulation di couraging the purchase of durables was introduced) for the negative effects of defense spending shocks to disappear. Third, the forecast revision of federal spending from the Survey of Professional Forecasters has high explanatory power for government spending, but for the 'wrong' reason: the predictive power of expected government spending growth is extremely low, so that the forecast error is effectively actual spending growth less noise.
Keywords: Government Spending, Vector Autoregressions, Fiscal Multiplier
JEL codes: E62, H30, H60