Purpose: The purpose of this paper is to contibute toward understanding the behaviour of government debt by examining two implications of tax smoothing hypothesis for the case of Indonesia. Key literature/ theoretical perspective: A tax smoothing results when the government smooths the tax rate in order to minimize the implied distortionary costs of taxation over time for a given path of government spending. Minimization of the total cost of taxation implies that planned tax rate should be set constant across all future time periods. Temporary changes in government spending and output will result in government’s budget deficit or surplus, but there will be no change in the tax rate. Meanwhile, public debt levels are permitted to vary over time to allow smooth tax rates. Key literature on this topic including Barro (1979, 1981), Kingston and Layton (1983), Sahasakul (1986), Kingston and Layton (1986), Huang and Lin (1993), Ghosh (1995), Cashin, Ul Haque and Olekalns (2002), Fisher and kingston (2004, 2005). Originality: There are only a few studies of tax smoothing for developing countries. The originality of this paper is to examine the time series for Indonesia to ascertain whether tax smoothing behaviour is consistent with the fiscal policies of Indonesia. Methodology: In a stochastic environment, the equivalent condition for optimal tax smoothing is that the tax rate would follow a random walk and that changes in the tax rate are unpredictable. First, we examine the random walk implication for the tax rate by testing the null hypothesis of a unit root in the tax rate. For this purpose, we use two methods of unit root test, namely the Augmented Dickey-Fuller (ADF) test and the Phillip-Perron (PP) test. Second, we test whether changes in tax rate are predictable. For this purpose, we perform autoregression for changes in the tax rate variable and vector autoregression (VAR) on a vector of variables including changes in the tax rate, changes in the government expenditure ratio and growth of real GDP. Findings: From the unit root tests, we find that the tax rate series is not stationary, indicating that the tax rate series follows a random walk. Next, we also find that changes in the tax rate cannot be predicted either by its own lagged or by lagged values of changes in government expenditure ratio and real output growth. The acceptance of both random walk behaviour and unpredictability of the tax rate confirms tax smoothing by the government. Research limitation: There are actually some other methods that can be used to examine tax smoothing hypothesis. This research only focus on the random walk behaviour of the tax rate and the predictability of changes in the tax rate. Future research can be extended by combining different approaches. Implication: This paper finds evidence of tax smoothing. This implies that the government keeps the tax rate relatively smooth and most likely use the public borrowing as the preferred response to shocks to government expenditure and output.