On the Predictability of Tax-Rate Changes

Some previous analyses have suggested that the smoothing of tax rates over time would be a desirable guide for public debt management. One implication of this viewpoint is that future changes in tax rates would be unpredictable based on current information. This proposition is tested by examining the behavior of U.S. federal and total government tax (and "non-tax")receipts relative to GNP. The sample for the federal government goes back to1879, while that for total government starts in 1929. Some econometric problems with using time-averaged data are discussed. The main empirical results accord with the theoretical analysis -- in particular, there is first, little indication of drift in the tax rates; second, insignificant relations of tax-rate changes to the own history of changes; and third, little explanatory value for tax-rate changes from a vector of lagged variables, which include the behavior of government spending and real output. If the findings are sustained, they imply that the existing IJ.S. time series data do not isolate periods in which current overall tax rates would be perceived as high or low relative to expected future rates. Accordingly, it may be impossible to use these data to evaluate policies that entail intertemporal manipulation of aggregate tax rates.