Government spending multipliers over business cycle

Purpose - The purpose of this paper is to show the asymmetric effects of government spending shocks for South Africa over the period 1960Q1–2014Q2. Design/methodology/approach - A threshold vector autoregressive model that allows parameters to switch according to whether a threshold variable crosses an estimated threshold is employed to address the objective of this paper. The threshold value is determined endogenously using Hansen (1996) test. Generalized impulse responses introduced by Koopet al.(1996) are used to study the effects of government spending shocks on growth depending on their size, sign and timing with respect to the economic cycle. The author also uses a Cholesky decomposition identification scheme in order to identify discretionary government spending shocks in the non-linear model. Findings - The empirical findings support the state-dependent effects of fiscal policy. In particular, the effects of 1 or 2 standard deviations expansionary or contractionary government spending shock on output are very small both on impact and in the long run; and a bit larger in downturns but has only a very limited effect or no effect in times of expansion. This result gives support to the evidence in the recent literature that fiscal policy in developing countries is overwhelmingly procyclical. Originality/value - It adds to the scarce empirical fiscal literature of the South African economy in particular and developing economies in general by allowing non-linearities to estimate the effect of government spending shocks over economic cycle.