Final NYU Tax Policy Colloquium session for fall 2019

Yesterday at the colloquium, after marking the completion of my 25th year co-running the thing, we discussed Josh Blank’s and Ari Glogower’s Progressive Tax Procedure. This is still an early draft of an ambitious project, hence plenty of opportunities to discuss the way forward. (Not presented when we discuss, as sometimes happens, recently published papers.)
Each of the three words in the title could be interrogated a bit. However, the basic idea is that procedural rules in the federal income tax – for example, concerning statutes of limitation, penalty rules, and standards of care in taking reporting positions – might vary with the income or wealth of the taxpayer. Audit rates are also in the ballpark, although to what extent within scope remains unclear. The clearest contrast, although here I seem to have begun interrogating the third word in the title, lies between procedural and “substantive “ rules – establishing, for example the tax rate and base.

“Progressive” raises numerous definitional issues, but the broader category might be called “means-based.” Suppose you want average or effective or statutory or marginal rates to rise with the taxpayer’s income. Then you favor income tax progressivity as defined or measured one way or another, but the broader point is that you favor a positive relationship between the rate of particular interest to you and the taxpayer’s overall income (which is a measure of the taxpayer’s means).

In that example, we also know how to define a regressive tax system. The rate of chosen interest goes south rather than north, with a perfectly flat tax standing in between them as the benchmark of a means-neutral system so far as these aspects are defined. (Of course, in a flat rate tax system, those with higher income still pay more overall tax, but the rate that one is focusing on, is distinct from overall liability, doesn’t vary with the measure of means.)

“Progressive tax procedure” therefore implies that item one is looking at grow less favorable in some way as the overall measure of the taxpayer’s means increases. Illustrative examples that the paper is at least willing to contemplate might involve, for example, having penalty rates go up as a percentage of the underpaid tax liability, statutes of limitation increase, or standards of taxpayer care to avoid penalties grow more demanding, as the taxpayer’s income (or, say, wealth, if a measure of that was available) increases. 

Having audit rates rise with income would be within the paper’s scope if that qualifies as “procedure,” which remains to be determined by the authors. This helps raise the point that once is talking about means-based tax procedure, without specifying as yet that it might be progressive, one might be motivated, not just by distributional preferences, but also the question of what information is relevant to tax administration. For example, supposed that the IRS’s information audits found that the amount of one’s income (at the start of the audit, or at the end) was informative regarding the likely revenue yield from a given audit. We know, of course, that the IRS must be looking at such things as whether, say, cash businesses or those in particular industries offer greater audit yields, or perhaps returns with large vs. small charitable contributions of a given type. If they find that something relating to the taxpayer’s overall means is also relevant to expected audit yield, one could ask (among other questions) whether using or ignoring this information would be, not only the better approach all things considered, but even the more “neutral” one, if one was attempting to define and apply such a benchmark. But while I suspect that a consistently applied audit yield metric would result in a significant upward shift, along the income scale, in who is audited, it wouldn’t necessarily be “progressive” all the time. E.g., suppose EITC claimants tend to yield greater audit yield than those earning above the phase-out. Or suppose there is more audit yield from the merely rich in the 99.0 to 99.5% percentile, than from those at the very top. Then one’s audit yield strategy wouldn’t be “progressive” at all margins, even when it was means-based.

This distinction can be an important one – looking at “means” because it has relevant informational content wholly apart from one’s distributional policy preferences, vs. because it is itself a topic of interest under one’s distributional preferences.

A further distinction to have in mind here lies between formal and substantive means-based variation in tax procedural rules. You know the old gag: “The law, in its majesty, forbids the rich and poor alike to sleep under bridges.” An opposite version of the same thing is FATCA, requiring information reporting about US taxpayers’ foreign bank accounts. As between full-time U.S. residents, this has progressive impact, at least to a degree, because you have to be at a certain level of wealth and/or income before one starts availing oneself of foreign bank accounts. (But perhaps it tapers down at some point towards the top? And of course for U.S. taxpayers who spend enough time abroad to need local banking outside the country, FATCA looms even if their resources are decidedly modest.) Likewise, if one applies particular penalties above a flat dollar amount of overall tax liability shortfall, or if one disfavors the use of tax advisor opinions as penalty shields, the rule even if formally neutral will have upwards-tilting effects.

In thinking about the various approaches that the paper puts in play, both the Kaplow-Shavell work on restricting distribution policy to the “tax system” and the Kaplow work on the social value of determining income (or whatever) accurately offer important orienting devices. Rules that might be described as implementing progressive tax procedure are contrary to the Kaplow-Shavell approach if they are used to increase the overall progressivity of the tax system – except insofar as by, say, reducing tax avoidance opportunities they affect optimal rates. But if they are using means-based information that is relevant to efficient implementation, the case is different. The point here isn’t to insist on Kaplow-Shavell conformity, as that’s a live issue under debate, but it’s useful for situating and understanding the claims.

And here’s where “accuracy” as discussed by Kaplow and others may enter the analysis. Suppose we used means-based, whether or not progressive, tax procedural rules to change the taxation of rich people in the following way. E.g., suppose that initially half were paying tax at a 40% effective rate and others at a 20% rate, due to tax avoidance opportunities available disproportionately to the latter. Then we used tax procedural rules, such as cutting back on the use of penalty shield tax opinions, or more broadly (whether or not within the term’s scope) by increasing audits of high-income taxpayers. One might think of the shift as being distributionally neutral, in an aggregate group sense, if now all the rich paid 30%, but for multiple reasons this might now be a better system (leaving aside the costs of getting there). Whereas, if we got all of them up to 40%, the system would now apparently be more accurate, but it would also be more progressive – which might be fine, but muddies the waters a bit regarding why we might favor (if we did) the tax procedural changes that brought about this new state of affairs. In Kaplow terms, a key question in the now-all-30% scenario would be measuring the benefit vs. the cost (if positive) of the greater accuracy – we obviously wouldn’t be willing to spend infinite resources in order to measure everyone’s income accurately and assure the uniformly “correct” application of statutory tax rates.

My point here is simply that this helps to demarcate the different issues raised by means-based tax procedure that the paper will be exploring as it develops.