Comments on John B. Shoven and John Whalley's, "Irving Fisher's spendings tax in retrospect"

American Journal of Economics and Sociology, The, Jan, 2005 by Alan J. Auerbach

It is useful to start my review of the Shoven-Whalley paper with a quote from Irving Fisher himself:

   All students of the problem are fully agreed on at least one
   point--namely, that existing income taxes must be radically
   reformed. But some of the reformers would include all
   capital-increases, while others would exclude them all. There are
   arguments on both sides, but both agree that either system would be
   an improvement over the present system or lack of system.
   (Fisher 1939)

This could well be the statement of a serious observer today, rather than of Irving Fisher six decades ago. Fisher favored a tax on "net paid income"--income distributed and consumed--because he thought it fairer and not subject to the income measurement problems of the other pure alternative, based on the Schanz-Haig-Simons accretion concept of income.

These are points that have continued to loom large in the more recent debate over consumption taxes. Indeed, issues of fairness and administration are probably receiving more attention from economists now than they did just a few years ago, when our most commonly studied economy consisted of a representative agent residing in a very simple world. It is useful to start with the developments in this world, largely related to economic efficiency, and then reconsider the various complications, some of which occurred to Fisher and others of which did not.

I

Efficiency

IN A WORLD WITH EVEN THREE GOODS, two of which are subject to excise taxes, Corlett and Hague (1953-1954) told us that we can't know for sure which good should be taxed more heavily. This result notwithstanding, a consensus seems to have arisen that a consumption tax-equivalent to equal tax rates on current and future consumption--is superior to an income tax, which implies a higher tax rate on future consumption (e.g., Feldstein 1978). There is some sense to this--why should future consumption be a much greater complement to leisure?--but there is no general theorem here.

On the other hand, adding more periods to life leads to a much stronger result: that capital income taxes should converge to zero, in other words, that consumption should face uniform tax rates in the long run (Chamley 1986). This result is often misunderstood, but the easiest intuition is to imagine the contrary, with a steady-state capital income tax bounded away from zero. This implies implicit taxes on future consumption that rise without bound as time goes on. Given that the welfare cost of a distortion rises in nonlinear fashion with respect to the size of the distortion, one can see that eventually this ever-increasing consumption distortion is likely to be suboptimal. If implicit taxes on consumption cease rising over time, this implies that capital income taxes cannot be positive in the long run. But a long-run tax rate of zero in an infinite horizon model doesn't translate into a low or zero tax rate in a finite horizon model, and most of us don't behave as if we expect to live forever.

A more serious problem with this result is that the economy doesn't consist of a single individual living for two periods or an infinite number of them. Tax systems don't change on a generation-specific basis, and the enactment of a consumption tax has different generational consequences than enactment of a wage tax, which, with constant returns to scale in production, would be equivalent in a single-agent model. This is not simply an issue of distribution, for the same differences in burden have efficiency consequences. Failing to tax existing assets by adopting a tax on wages instead of a tax on consumption not only shifts the burden from the old to the young, but also reduces the efficiency of the tax reform, a difference that Larry Kotlikoff and I found, in our simulation model, to be large enough to turn an efficiency gain into an efficiency loss (Auerbach and Kotlikoff 1987). My more recent simulations of realistic tax proposals find that we would lose essentially all the potential efficiency gains from adopting a consumption tax if we maintain the existing level of progressivity and protect asset values in transition (Auerbach 1996).

II

Distribution

THIS LAST FINDING brings out a key issue regarding the consumption tax--whether it is inherently less progressive than an income tax. The previous discussion reminds us that it is necessary to take a perspective greater than one year; one can't compare the old and young using current income as a measure of ability to pay. If we do use income, even measured over a period of several years, recent work has emphasized that we should look at risk-adjusted returns in measuring capital income, a point that goes back to Tobin's work on taxation and risk taking (Gentry and Hubbard 1997).

But others ask why we should use income, rather than consumption, as a measure of ability to pay. If we do use consumption, then clearly we can make the tax progressive simply by having a personal consumption tax with a progressive rate structure. In standard models, this is likely to be a better approach than that adopted under European value-added taxes, which achieve progressivity via tax rate variations on luxuries and necessities. Variations in indirect taxes offer limited scope for redistribution at the cost of considerable distortion of consumer choice. However, this result is less clear if we expand models to incorporate conspicuous consumption (Bagwell and Bernheim 1996).