Cyclical variation in US government spending multipliers

To estimate state‐dependent US output multipliers of government spending, we build a Markov‐switching fiscal VAR assuming that the economy is subject to unobservable regime shifts. We combine variables in the VAR and external qualitative indicators to infer the regimes and find that they match well with the business cycle. To account for the possibility of regime shifts after a shock, we compute generalized impulse responses using a novel recursive method. We find the multipliers smaller than 1 in both regimes, but statistically larger in recessions than in expansions. We reconcile our results with the main findings in the previous studies.