This past Tuesday at the colloquium, we discussed the above paper by Manoj Viswanathan. If I had to give a two-sentence summary / overview / elevator pitch for the paper, it would be: The notion of tax progressivity relates to things that are important, although it doesn’t frame them in the best possible way. Given, however, that people are going to be discussing progressivity in any event, they ought to be clearer about how they are defining it, what underlying assumptions this presupposes, and how all that relates to the conclusions that they draw.
This is a very useful and needed project design, and one that the paper is ably carrying out. It brought to mind a similarly motivated paper that I once wrote concerning tax expenditures. The author might consider, however, pushing further by more aggressively defining and advocating “best practices” in the progressivity measurement business. That is, while a key feature of the paper is its demonstrating that there is no single “right answer” regarding how to define and measure progressivity, it could (a) say “these are some plausible approaches to particular design choices, no one of which is indisputably the best, and (b) call out blatant misuses of the concept.
I’ll comment in 5 parts: a background section that I will include here, and then a response in turn to each of the paper’s sections 1-4, which I will put in a follow-up (i.e., Part 2) blogpost.
Background – Tax progressivity is a measure or concept that has to do with distributional issues pertaining to inequality. To evaluate such issues, one needs to apply vertical measures of people’s relative economic positions. For tax progressivity itself, this involves in particular applying what the paper very usefully calls a progressivity base. That is, to evaluate the progressivity of, say, the federal income or a state property tax, we look at the tax burdens borne by people at different vertical levels of one’s measurement framework (i.e., the progressivity base), which need not be the same as the particular system’s choice of tax base.
Often, progressivity analyses are engaged in a comparative statics exercise – that is, comparing State of the World A to State of the World B. These might differ due to the passage of time (e.g., the tax system has become either more or less progressive, reflecting legal and/or economic changes in the interim). Or the question might be how particular legislation would affect it or has done so.
Such comparative exercises work best with a fully specified counterfactual. This is especially a problem if we ask how progressive the tax system “is.” (Compared, at least implicitly, to what? No tax system? No government? A uniform head tax or flat tax to pay for everything?) But even when we are looking at, say, current law vs Proposed Set of Tax Changes XYZ, we really need a budget-neutral, or perhaps even revenue-neutral, counterfactual in order to avoid being potentially very misleading.
For example, consider the large tax cuts enacted in 2001, 2003, and 2017. They were designed to offer tax cuts at pretty much every income level. But these tax cuts were generally small at the bottom and large at the top. Given that these tax cuts would need, pretty much as a matter of basic arithmetic, to be funded eventually in some way (e.g., through higher taxes or lower spending than would have applied in their absence), do we really need to get into some of the measurement games that enliven (to put it kindly) the debate? Or is it pretty clear that people at the bottom, or their kids or grandkids, were pretty definitely losing overall, while those higher up were very likely winning?
In a budget-neutral and indeed revenue-neutral comparison, typically rising average tax rates (ATRs) as one vertically ascends are what you need for the system to be “progressive” within common nomenclature. The ATR is a fraction in which the numerator is something like tax liability or tax burden, and the denominator is the underlying distributional measure (i.e., the progressivity base).
In a non-revenue-neutral comparison, however, a focus on ATRs may prove highly misleading. Consider, for example, the fact that most of the US’s peer countries have less progressive tax systems, but more progressive fiscal systems, than we do. Suppose that we became more like them, by reason of adopting a VAT and using the revenues to better fund healthcare, childcare, education, etcetera. Our tax system would now look less progressive, focusing on ATRs, but our overall fiscal system would probably now be far more progressive than it had previously been. Or at least, to put it differently, after-tax-and-spending distribution would now be less unequal than it had been before the change.
By contrast to ATRs, marginal tax rates (MTRs) are merely a technical tax design feature that would lack the distributional significance of ATRs. Thus, suppose that (in the spirit of a Mirrlees OIT model) we applied 100% MTRs at all income levels, so that all national economic production was nationalized and then paid out in uniform demogrants. This might be a very bad system, but it would NOT fail to be duly “progressive” because its MTR structure was flat!
Still, under various fairly common design features, rising MTRs may be necessary to create rising ATRs.
As a sidenote, when we speak of “progressive consumption taxes” (such as here), we typically are referring purely to the technical MTR structure. But this reflects that consumption taxes, such as those remitted by sellers, typically have a completely flat MTR structure (at least as to consumers, if not as to consumer goods). So the word “progressive” here calls attention to a distinctive design choice.
In sum, when used with proper care, progressivity assessments that are based on how ATRs change as one vertically ascends the progressivity base. But one has to do this carefully (e.g., using budget-neutral comparisons and looking at both taxes and spending), in order to address the risk of being badly misled.