This paper investigates the direction of causal relationship between taxes and expenditure
in South Africa, using quarterly data for the period 1960:1-2006:2, and annual data for
1960 to 2005. For both frequencies, gross domestic product and government debt are
included in the VAR system as control variables. For quarterly data, the Johansen’s
(1991, 1995) methodology suggest two cointegrating equations among the four variables.
Our findings support the fiscal synchronisation hypothesis, since Granger causality tests
in a Vector Error Correction framework suggests bi-directional causality between taxes
and expenditure for the period under study. In contrast to the VECM for quarterly data,
the VECM for annual data disprove any possibility of Granger causality between taxes
and expenditure. The apparent ambiguity is indication of the fact that causality, among
other factors, depends on the frequency of data.