Yesterday at the colloquium we discussed Peter Diamond's paper, co-authored with Emmanuel Saez and published recently in the Journal of Economic Perspectives, entitled The Case for a Progressive Tax: From Basic Research to Policy Recommendations.
The paper makes 3 recommendations: (1) apply marginal tax rates that are both graduated & higher than present law to very high earnings (e.g., perhaps 70% for the top 1%, which kicks in at about $400,000, (2) for low-earners, subsidize earnings but then apply high marginal tax rates in the phase-out rate, (3) what the paper calls “capital income” should be taxed.
Herewith some thoughts about Topics 1 and 3. (It is already a bit on the long side without including Topic 2, on which in any case I had less to say.)
1. Optimal tax rates for high-earners
1. The fact that this paper talks about higher rates at the top, perhaps on the order of 70 percent, is a valuable public service. Whether or not one agrees in the end (or thinks it politically feasible even if one agrees), it expands the range of “permissible” debate, and this is a good thing both intellectually and politically. Diamond and Saez are also to be commended for caring about real world policy debates, as opposed to wanting to play with complicated mathematical models for their own sake (which can be fun, but is less public-spirited).
2. The paper argues that the tax policy aim with respect to people at the top should be almost pure revenue maximization, without regard to the marginal disutility that a high tax rate imposes on them (via both taxes paid and deadweight loss). More specifically, it shows that assuming a very low social weight for these marginal utility losses, rather than a zero weight, wouldn’t change the conclusions very much.
Two grounds are offered for this low social weight: low Ui, and low wi. The former refers to the affected taxpayer’s marginal utility. For example – to draw on someone well above the merely $400,000 level – Warren Buffett probably would experience next to no change in his daily personal circumstances if his tax bill went up by $10 million – it’s not as if he’d have to cut back on anything that he likes doing. By contrast, low wi refers to the social weight that one places on Buffett’s, as opposed to someone else’s disutility. Some versions of welfarism – but not utilitarianism – give less weight to the welfare of people at the top than people at the bottom.
In terms of the robustness of the paper’s conclusion, it makes a big difference which of these two factors one is relying on. For example, it is, not obvious that Ui is effectively zero until one is well above the $400,000 level. Someone at the level very likely can think of things to do with an extra dollar that would be relevant to personal welfare. By contrast, low wi in the social planner’s mind for everyone in the top 1% takes care of the problem all by itself. So a utilitarian, who treats wi as the same for everyone, cannot as easily agree that pure revenue maximization is the proper aim until one gets closer to the Warren Buffett level.
I personally find utilitarianism more persuasive than the versions of welfarism that use egalitarian weighting, although I also view declining marginal utility as a very real and significant phenomenon. (I am aware, of course, that many others view it more skeptically than I do.) Utilitarianism necessarily follows if one favors (as I do) the Harsanyi framework, in which one looks behind the veil and seeks to maximize expected utility under the hypothetical assumption that one is equally likely to be any of the people in the society. In consequence, I may need to go somewhat higher than $400,000 of annual income before I am ready to sign on to setting marginal tax rates based purely or mainly on revenue maximization.
3. Today’s extreme high-end wealth concentration really re quires us to think about the possibility of resulting externalities. The paper doesn’t discuss externalities, and understandably so as they are very hard to evaluate, much less measure precisely. However, they may be very important. Indeed, significant negative externalities from extreme high-end wealth concentration could reasonably lead a utilitarian to support marginal tax rates at the top that were actually above the revenue-maximizing level (just as a pollution tax may be set above the revenue-maximizing point, given that its aim is to price the harm appropriately).
Negative externalities to high-end wealth concentration could relate to the effects on relative status and social power, and in particular on political economy concerns. A society with an exceptionally wealthy elite that has, in effect, taken a rocket ship away from everyone else may be especially prone to rent-seeking and destructive insiderism. It gives us the astonishing politics of a Great Recession in which Washington seems not to care about persistently high unemployment. These types of concerns might conceivably call for a more than revenue-maximizing rate at the very top, insofar as the behavioral response was to earn less rather than simply do more tax planning.
In fairness, I should note that there are also claims of positive externalities from high-end wealth concentration. Suppose, for example, that exceptionally wealthy consumers fund the development of new technologies that at first are prohibitively costly for everyone else, but over time become cheaper and result in greater consumer surplus being enjoyed by the bottom 99 percent (be it from something like HDTV or advances in healthcare).
4. The paper suggests that the proper response to high-income taxpayers’ greater tax planning ability is simply to broaden the base, expand anti-tax planning rules, and increase enforcement efforts. All those are perfectly good suggestions, and the paper is right to observe that the availability of tax planning is somewhat endogenous to the design choices we make. But I would think that tax planning flexibility is almost bound to be greater for high-earners, especially (though not uniquely) under a realization-based income tax. After all, they are bound to have greater flexibility, liquidity, and risk tolerance, along with access to better advice, than everyone else. This probably needs to be kept in mind when evaluating what is likely to be the revenue-maximizing rate at the high end.
2. Taxing capital income
1. The paper devotes a lot of time to arguing that “capital income” should be taxed. I found this discussion baffling, because I do not regard “capital income” as a coherent category. OK, we can talk about the usual things (a la the Jake Brooks paper in Week 3) such as returns to waiting, risk, embedded labor income or inframarginal returns, etc. But in practice it seems to mean “income of people who happen to be at least incidentally using some capital.”
I gather that what the paper has in mind is categories that tend to be classified as “capital income” under current income tax law. For example, income earned by or through corporations, along with dividends and capital gains. But only a fool would argue that, say, support for a consumption tax approach means that we should exempt those items under our existing system. That would mean in practice that, say, Steve Jobs would have gotten classified as a very poor individual who only earned $1 a year. (I gather that the paper is expressly meant to respond to fools who are making such arguments, however.)
When people discuss whether we should tax “capital income,” what they typically mean is the risk-free return to waiting, which a well-designed income tax reaches and a well-designed consumption tax generally exempts. Again, the paper appears to be responding to the fact that idiots write Wall Street Journal op-eds in which they argue for exempting capital gains, corporate income, and dividends under the existing system. But I wasn't convinced that an article in the Journal of Economic Perspectives needs to respond to Wall Street journal op-ed writers who are aiming at the broad public rather than at JEP readers.
2. Once we are talking about “capital income” under the existing system, it becomes a bit questionable whether we can really have a 70 percent rate, as the other part of the paper argues. The revenue-maximizing rate for capital gains, under the existing system with its realization requirement and tax-free step-up in basis at death, is probably more on the order of 30 percent. Likewise, taxing dividends and corporate income generally at 70 percent might be more than a bit questionable. To be sure, the paper does say that “capital income” doesn’t necessarily have to be taxed at the same rate as everything else, just not at zero given (among other things) the difficulty of telling the two types of income apart. But this certainly was not the best-developed part of the paper.
3. The paper criticizes the well-known (in the public economics literature) Chamley-Judd and Atkinson-Stiglitz-based lines of argument, to the effect that the return to waiting should be taxed at zero. The Chamley-Judd literature shows that a positive tax on waiting leads to exploding tax rates on greatly deferred consumption, which becomes a major problem if you have infinite-lived consumers (such as multigenerational households) optimizing over unlimited time. The paper rightly argues that relevant time perspectives are likely to be a lot shorter. The Atkinson-Stiglitz literature shows that, if it is most efficient to tax alternative commodities the same, then one can fruitfully think of present consumption and future consumption as involving alternative commodities. Taxing future consumption at a higher rate is inefficient, at least at this particular choice margin, because it distorts the inter-temporal choice.
The paper says: Ah, but reliance on Atkinson-Stiglitz to support exempting the return to waiting requires that consumers be homogeneous with respect to saving. Another way of putting the point is that, although the inefficiency at the margin is clear, there may be other reasons for taxing the return to waiting, e.g., in response to heterogeneity.
4. Before turning to those other issues, a general background point is worth making. Optimal taxation is about how best to respond to an information problem (since we lack information about ability and, underlying that, marginal utility). So of course information about people’s saving, or the relative timing of earning and consumption, is potentially relevant. A consumption tax that treats this as irrelevant is failing to use potentially pertinent information. Of course, this by itself doesn’t tell us if/when saving should increase or reduce one’s lifetime tax liability in present value terms.
4. The paper notes a couple of good arguments that potentially make it plausible to tax saving. In particular:
(a) saving may be a “tag” that is correlated with ability. Even as between two individuals with the same observed earnings, the saver may have other attributes (e.g., foresight, planning depth, and self-control) that translate to being better-off in ways that cannot be directly observed.
(b) the income effect of saving means that one can better afford to reduce one’s labor supply – in effect creating a negative revenue externality from saving that undermines use of the tax system to provide insurance against earnings volatility.
5. Getting back to the first point in this section, no good argument for a positive tax rate on the return to waiting arises from the difficulty in distinguishing between labor income and capital income. No good consumption tax model requires so distinguishing. For example, if you use expensing as under the Blueprints cash flow consumption tax that David Bradford and the Treasury Tax Policy staff developed in the 1980s, the issue disappears altogether. Same point if one uses the X-tax. If anything, administrative arguments weigh heavily against continued use of the income tax, especially if it relies on realization.