This paper tests a version of Barro's tax smoothing hypothesis using Australian data for the period 1964/65 to 1993/95, The model assumes intertemporal optimization by a government seeking to minimize the distortionary effects of tax collection. The model predicts that the budget surplus is stationary, even if government expenditure and tax collections are nonstationary. In addition, the surplus should be a linear function of expected future changes to government expenditure. The results indicate that Australian fiscal policy has been too volatile to be consistent with optimal tax smoothing.