We estimate the dynamic effects of government spending shocks, using time-varying volatility in US data modeled through a Markov switching process. We find that the average government spending multiplier is significantly and persistently above one, driven by a crowding-in of private consumption and non-residential investment. We rationalize the results empirically through a contemporaneously countercyclical response of government spending and an efficient weighting of observations inversely to their error variance. We then show that the multiplier is significantly smaller when volatility is high, consistent with theories predicting reduced effectiveness of fiscal interventions in uncertain times.
Keywords: Fiscal policy, Government spending multiplier, Uncertainty, Structural vector autoregressions, Heteroskedasticity