Some Calculations of Lifetime Tax Incidence

Abstract

This paper reports a set of lifetime tax incidence calculations using a life cycle simulation model for Canada due to Davies (1979a, 1982). A repeatedly stated qualification to annual calculations in the empirical tax incidence literature is that it would be more satisfactory to make calculations on a lifetime basis. Even though it is acknowledged that lifetime tax incidence could well differ from annual, it is widely believed that data and other difficulties make such calculations next to impossible. Indeed, the widespread acceptance of the data problems of lifetime calculations seems also to have inhibited speculation about how lifetime tax incidence might differ from annual. As a result, redistributive tax policy judgments continue to be based on annual incidence calculations in spite of the reservations many have about their usefulness. Our paper is intended to reorient discussion towards lifetime tax incidence by providing some initial null hypotheses about the shape of lifetime tax profiles. Our main finding is that under the standard competitive assumptions common in the incidence literature, lifetime and annual incidence calculations both produce mild progression in tax rates across household deciles (ignoring the bottom decile in the annual calculation). While the income tax is less progressive in lifetime than in annual calculations, other taxes are for the most part less regressive. Also, lifetime incidence calculations are much more robust to alternative shifting assumptions than annual calculations. In the lifetime context, key distributions such as earnings, transfer payments, and consumption are less heavily concentrated in particular percentiles of the population than is true in annual data. As a result, changing the allocative series for any particular tax does not have the large effect on incidence results found in annual calculations.' Each component of the tax system is allocated to households grouped by lifetime income using particular distributive series following a procedure similar to that employed in annual incidence calculations (for example, Richard Musgrave et al., 1974; Joseph Pechman and Benjamin Okner, 1974; Edgar Browning and William Johnson, 1979; W. Irwin Gillespie, 1980). In the process we are able to compare lifetime and annual incidence calculations using the same data set. In both lifetime and annual calculations, we allocate five groups of taxes among households using distributive series which come partly from the 1971 Statistics Canada Survey of Consumer Finances (SCF) and partly from our life cycle simulation model. The SCF data are used to construct synthetic longitudinal lifetime profiles of earnings and transfer payments for a sample of 500 households. The latter are assigned inheritances by simulating patterns of mortality and bequest. These data are then used in the life cycle model to generate lifetime consumption profiles and bequests. The earnings, transfer, and inheritance data, plus the model output provide the distributive series on which alternative incidence calculations are based. While the incidence calculations presented in this paper use Canadian data, results would likely be similar for the United States