AMIDST THE CHAOS that surrounds the budget-making process in most countries, it is often hard to believe that there is any rationale at all to the time pattern of budget deficits and taxes. Yet, under the plausible assumption that the distortionary effects of tax policies become more severe the higher the tax rate, economists can derive testable implications about the dynamic path of an optimal fiscal policy. In a seminal paper, Barro (1979) proposed a simple theory of tax smoothing: the government should spread the burden of raising taxes across time periods in order to minimize their burden.(1) This has important implications for the pattern of budget deficits: when faced by a temporary increase in expenditure, for example, the government should issue debt in order to spread the increase in taxes over a longer time horizon and to minimize the welfare costs of high tax rates. In analogy to consumption theory--where the representative agent seeks to smooth consumption over time (Hall 1978)--the tax-smoothing model of government finance predicts that taxes should follow a random walk (Barro 1981). Changes in the tax rate should be unpredictable because otherwise the government could not have been optimizing before.
There are at least two reasons, however, why it may be worth going beyond the random-walk tests of tax-smoothing models. First, it is often difficult to reject the null hypothesis of a random walk for many economic time series in finite samples. Thus such tests may have very low power. In fact, while Barro (1981) reports that he is unable to reject the random-walk model for tax rates, Sahasakul (1986) is able to reject it. Second, even if tax rates really do follow a random walk this does not necessarily imply that governments smooth taxes: tax smoothing is only one of many possible explanations for tax-rate changes being unpredictable.(2)
In this paper we exploit the close analogy between a consumer trying to smooth consumption and a government trying to minimize tax distortions in order to test a very tight set of restrictions on the joint time series process for budget deficits, government revenues, and expenditures that are implied by the assumption of tax smoothing. Our strategy is to focus on the optimal path of the budget surplus rather than on tax rates themselves. The assumption of tax smoothing implies that the budget surplus should equal the expected present discounted value of the change in government expenditure. If the government expects an increase in its expenditure at some point in the future, for example, it should raise taxes immediately in order to minimize the tax rate it will have to adopt when expenditure actually rises. Likewise, a purely temporary increase in expenditure--such as a war--should be (largely) financed by means of a budget deficit so that the cost of the expenditure can be spread over time, thereby minimizing the distortionary effects of the taxes. The behavior of the optimal budget surplus is thus very similar to the optimal pattern of savings, as analyzed by Campbell (1987), or the optimal behavior of current account surpluses (Ghosh 1995).
The prediction that the budget surplus should equal the present discounted value of expected changes in government expenditure allows us to construct a time series for the optimal budget surplus and to compare it to the actual surplus. Under the null hypothesis that governments seek to smooth taxes these two series should differ at most by sampling error. An additional, somewhat weaker, implication-stressed by Campbell (1987)--is that the budget surplus should Granger-cause changes in government expenditure. This will be true whenever the government has more information about the future path of its expenditure--such as news of political or other events--than that contained in past values of the expenditure series. Under the null hypothesis that the budget surplus equals the expected discounted value of changes in government expenditure, where the expectation is conditional on the government's entire information set, the budget surplus should embody this additional information and should thus Granger-cause changes in expenditure.
Aside from its focus on the budget surplus rather than the tax rate, our methods differ from previous analyses of tax smoothing in an important respect: we distinguish between what we term tax smoothing and tax tilting.(3) Tax smoothing refers to intertemporally …