In this paper we analyze whether the effect of fiscal policy differs across the business cycle. To tackle this question, we use a regime-switching error-correction framework, where nonlinearities are only modeled in the short-run and have no impact on the long-run equilibrium. Regime specific shocks to government revenue and government purchases are identified using sign restrictions. Linear combinations of the impulse responses of these basic shocks are used to construct a deficit-spending shock and a deficit-financed tax-cut shock. We find that active spending policies have a stronger impact in recession, with multipliers exceeding unity, and should be preferred to deficit-financed tax-cuts.