In addition, our results that high-income households save a greater fraction of income while working than low-income households imply that a flat rate consumption tax will (on a lifetime basis) be regressive relative to a flat rate income tax. There are two reasons for this. First, even if there were no bequests (so that all households consumed their wealth during their lifetime), households with higher incomes would pay more under an income tax compared to an equal revenue consumption tax, simply because such households can expect to receive a disproportionate share of interest and dividend income.
Second, one needs to take into account that bequests are not zero and may vary with lifetime earnings. A number of studies have calculated the distributional effects of moving from an income-based tax system to one that taxes just consumption. Fullerton and Rogers (1993), Metcalf (1994), and Altig et al. (2001) use data from Menchik and David (1982)to estimate how bequests differ by lifetime income group. The Menchik and David data, however, exhibited a U-shaped relationship between the fraction of resources bequeathed and lifetime income; bequests were the largest share of lifetime resources for the lowest income decile.46 This pattern is consistent with the poor (in the lowest income decile saving, on average, a larger fraction of their lifetime resources than the rich! Not surprisingly, Metcalf (1994) found, using these data, that adjusting for bequests added progressivity to a consumption tax relative to an income tax. In contrast, our results show substantially higher saving rates (and wealth-income ratios) among families with higher lifetime income(see also Mieszkowski and Palumbo 2002). These results, together with the lack of evidence that the elderly high-income households dissave at a higher rate, suggest that, on average, higher-income households bequeath a larger fraction of lifetime earnings than lower-income households. Because bequests are effectively exempt from a consumption tax (at least for the current generation) but are not exempt from an income tax, our evidence suggests further regressivity in a consumption tax relative to an income tax.
Finally, the results have implications for the “choice versus chance” question first raised by Friedman (1953) and more recently by Venti and Wise (1998). Is the considerable variation in accumulated wealth the consequence of choice (preferences and tastes) or chance? Venti and Wise argue that most of the variation in wealth within lifetime income
groups is due to saving decisions, or choice. Our finding that differences in saving behavior across income groups are also an important source of the overall variation in wealth of the U.S. population suggests a diminished role for choice. For example, wealth accumulation because of government or private policies that differentially affect saving (such as asset-based means testing or the availability of 401(k) plans) cannot be readily attributed to tastes or preferences. Nor can a lucky career outcome such as joining Microsoft in 1984, and a subsequently higher rate of saving, be attributed entirely to choice. In sum, much remains to be learned about household saving behavior, especially that of elderly households and that of the very top of the income distribution. Still, we believe that our work has established one fact: The rich do, indeed, save more.