Yesterday at the NYU Tax Policy Colloquium, David Gamage presented the above paper. The following discussion is adapted from my notes for the session. Part 1 is in this blog post, and I’ll put Part 2 in the next one.
The paper responds to the “double distortion” literature in legal tax scholarship which draws the conclusion that distributional objectives should be pursued entirely through a progressive consumption tax (leaving aside the use of transfers to address the lower end). Under this view, there should be no other distributional instruments in the tax system (e.g., no income tax, inheritance tax, or luxury taxes), and no use of legal rules to address distributional objectives.
The paper, by contrast, argues for using lots of smaller separate systems to address distribution. For example, it suggests that we might want to use of all a labor income tax, a VAT, a capital income tax, a wealth tax, and legal rules to address distribution.
We’ll discuss (1) the double distortion literature, and (2) the paper’s argument for using multiple tax instruments to address distribution. [But this post only includes Part 1.]
1. THE DOUBLE DISTORTION LITERATURE
This is a legal literature based on certain economics literature. Economists may be bemused by how this legal literature treats this economics literature.
The double distortion literature’s central point is simply to rebut a particular fallacy. Suppose we are discussing the choice between an “ideal consumption tax” and an “ideal income tax.” The consumption tax just distorts one decisional margin: labor supply. The income tax distorts two margins: labor supply and saving.
So far, so good. But the fallacy involves positing that, by also distorting savings decisions, the income tax inherently does less to distort labor supply decisions. To illustrate, suppose we are making a revenue-neutral choice between a 25 percent consumption tax and a 20 percent income tax. (The income tax raises the same revenue with a lower rate because its base is “broader.”) Seemingly, labor supply is less discouraged by the income tax because the tax rate is lower.
This analysis fails, however, if one accepts that a worker who may save is funding present consumption plus future consumption through his current labor supply. Suppose that I will spend some of my labor income this year and some in the future. Whereas the 25 percent consumption tax imposes a 25 percent excise tax on all consumer goods that I might choose, the income tax, by reason of its hitting intermediate saving, in effect imposes a higher excise tax on future consumption goods than current ones. Suppose the effect of the income tax on my deferred consumption is the equivalent of hitting it with a 30 percent excise tax. We don’t inherently reduce labor supply distortions by converting the uniform 25 percent excise tax on consumption into the equivalent of a 20 percent excise tax on some consumption and a 30 percent excise tax on other consumption.
To a first approximation, therefore, the income tax merely layers the savings distortion on top of the labor supply distortion, rather than imposing it partly in lieu of that distortion. This means that taxing saving would require some motivation other than the mistaken idea that it inherently permits one to reduce labor supply distortions.
The famous Atkinson-Stiglitz (1976) article on which the double distortion literature draws also rules out (by express hypothesis) one further possible argument for the income tax. It assumes separability between labor and all consumer goods, so that how much you work in the current period does not affect (among other things) your preferences between current and future consumption. If separability did not hold, and saving for future consumption turned out to be a leisure complement while higher current consumption was a leisure substitute, then one might actually offset some of the labor supply distortion by taxing future consumption at a relatively high rate (as the income tax does, although there’s no reason to think it would supply the optimal rate differentiation). But even if we question separability, for all we know the relationship between current and future consumption on the one hand and labor supply on the other might lie in the opposite direction. Separability is arguably a reasonable presumption from ignorance, at least until we have some question not to question it but to think that it errs in a particular discernible direction.
What about other possible motivations for taxing saving? They simply aren’t in the Atkinson-Stiglitz model. And there is no reason why they should be. A given model can only reasonably do so much. But possible motivations for taxing saving might include at least the following:
(1) Suppose that saving is a tag of high ability. Then we might want to tax it for the same reason that one taxes earnings (and might decide to tax, say, height) in an optimal tax model.
(2) Suppose we believe that people make many consumption decisions on a per-period basis, rather than a lifetime basis. Then current resources (i.e., savings) may have distinctive current-period distributional relevance.
(3) Savings offer a cushion to hide one’s ability by working less. They therefore undermine the use of an earnings tax to provide ability insurance, supporting an argument that they should be taxed.
(4) Suppose that, especially after having read Piketty’s Capital in the Twenty-First Century, we believe that saving plus inheritance leads to accelerating inequality that has serious adverse consequences. Then one could view saving (or at least inheritance) as having significant negative externalities, like pollution. This provides a reason to tax saving (or at least inheritance) if we believe that these adverse consequences outweigh any positive externalities from saving.
(5) Suppose we believe that income taxes will generally be more progressive in practice than consumption taxes, for reasons of political economy. If one wants the tax system to be more progressive within the relevant range, this may motivate supporting income taxation even if it is otherwise inferior.
Now, all these may be either good arguments or bad ones. The point of interest in relation to the “double distortion” literature is simply that it provides absolutely no ground for evaluating them – nor should it; that isn’t its “job,” which is simply to address the fallacy described above. Lawyers who think that Atkinson-Stiglitz 1976 is relevant to any of these questions are mistaken – as I rather suspect Atkinson and Stiglitz themselves would agree.
(Random digression, reflecting that Atkinson and Stiglitz can speak for themselves on this point, if they are so minded: Recently I mentioned to a class that the author of a particular article might disagree with how his argument was being interpreted. This reminded me of the scene in Woody Allen’s Annie Hall where Woody refutes a loudmouth on a movie line who is purporting to explicate Marshall McLuhan, by bringing out McLuhan himself, who happens to be in the lobby. I discovered that almost none of the students had seen Annie Hall. A canonical cultural reference 20 or more years ago – as, say, The Godfather still is – Annie Hall apparently has lost that status.)
Anyway, back to the double distortion literature. The Kaplow-Shavell stricture against using legal rules to address distribution addresses the same fallacy. To give this point contemporary policy relevance, suppose that we are concerned about the effect of CEO compensation and financial sector returns on high-end inequality. Now, we might have independent regulatory reasons for addressing CEO pay and financial sector returns. For the former, perhaps their pay often vastly exceeds the marginal private value of their production because of agency costs, collusive boards, etc. For the latter, perhaps financial sector returns often vastly exceed marginal social value because they reflect front-running, duping customers through complexity and opacity, etcetera.
If these critiques are correct (and I certainly find them generally plausible), then we have straight efficiency reasons for addressing CEO pay and financial sector returns. But the Kaplow-Shavell double distortion / don’t use legal rules for distribution line of argument would say: We shouldn’t over-correct, and make CEO pay and financial sector returns too low, rather than too high, in response to the distributional effects, even if we strongly dislike those effects. Rather, we should address the distributional issue purely through the tax system (i.e., with a progressive consumption tax) so that we don’t create labor supply distortions PLUS (and by no means instead of) particular distortions in these markets.
Once again, the double distortion argument addresses a particular fallacy, which is that we can inherently avoid labor supply effects by targeting narrower areas of high-wage labor. But it does not address other possible arguments for using legal rules in these two contexts, if (as in the case of the above-noted arguments for taxing savings) plausible rationales of some entirely different kind are offered.
Teaser for my next post, where I will turn to the particular main arguments in the Gamage paper: Suppose that over-addressing CEO pay and financial sector returns through targeted regulation has the advantage of eliminating the use of tax avoidance techniques that would inevitably hamper using the distribution system. In particular, even with a shift from the current income tax to a progressive consumption tax, suppose that we couldn’t eliminate particular “tax gaming” tricks that would allow CEOs and financial players to wipe out the intended tax liability. Then, even if they also separate have bags of tricks to deploy against the regulatory regimes, we might want to use the regulatory system after all to do some of the distributionally-minded work. This would merely involve trading off the reduction in opportunities for tax gaming against the increase in sectoral inefficiency and regulatory gaming. But arguably the optimal extent to which we would use the regulatory system, by reason of this tradeoff, is greater than zero.
Closing comment on the double distortion literature: I’d analogize the usefulness of Atkinson-Stiglitz (“AS”) in the public economics literature to that of Modigliani-Miller (“M-M”) in the finance literature, with a twist. MM shows that the choice between debt and equity is irrelevant to firm value unless there are tax issues, bankruptcy issues, other relevant regulatory regimes that treat the two differently, or agency cost / asymmetric information issues. MM doesn’t show that debt-equity choices ARE irrelevant, given that those issues may exist. Rather, it shows where we would have to look in order for such choices to matter.
Likewise, AS doesn’t show that we should only use a (potentially progressive) consumption tax to address all distribution issues. Rather, it shows us where NOT to look for an argument in favor of using something else. That clears the decks so that the analysis can move on to the questions of real interest that remain.