Yesterday was our first colloquium session of the year. I've been doing this for 15 years now. The institution has come a long way, and we now have 25+ very strong students. The total audience at the PM session was about 50, with the students (who participated actively) being joined by people from several types of academic institutions as well as practice. We managed to hand it off to the audience much earlier than is the case in other scholarly colloquia I have attended, at NYU and elsewhere.
Our rule of not having the authors present their own papers is a huge benefit. In addition to encouraging the audience to come prepared, it saves us from having 20 wasted minutes at the start. The only thing more boring to me than listening to an author present a paper I have already read carefully is being myself that author, and thus being forced to prepare and deliver as well as listen to said repetitive 20-minute talk. The result of the opening talk is always to diffuse subsequent discussion, since the author presumably touches on all of the main themes. By contrast, when we start out by offering comments on particular aspects, one at a time, we may succeed in stimulating a more focused and in-depth discussion. (With the help of the audience - the model we have in mind as lead discussants is Jason Kidd, not Alan Iverson.)
This week's paper was by Lily Batchelder of NYU and Eric Toder of the Urban Institute, presenting the opening chapters of a book (being co-written with Austin Nichols) entitled "$750 Billion Misspent? Getting More from Tax Incentives." My colleague Mihir Desai was the lead commentator. One of the main issues we raised went to the book's focus. The $750 billion figure is from an OMB measure of the annual cost of provisions officially classified as tax expenditures - in practice, a polyglot of items. By dollar amounts, the 5 biggies on the list, in order, are favorable tax provisions for (1) retirement plans, (2) special capital gains and dividend rates, (3) health insurance, (4) home mortgage interest, and (5) charitable contributions.
The authors are interested, not in officially listed tax expenditures as such, but rather in provisions that they call tax incentives, and describe as "spending" through the tax code that is designed to change behavior and thereby address externalities. They agree that this does not, for example, have anything to do with the special capital gains and dividend rates, which are structural rules related to the niceties of income realization and the double corporate tax.
I would describe their interest as pertaining to allocative rules within the income tax, which we think of as a distributional system, in the sense that the rationale for taxing income is to affect distribution (or distribute tax burdens relative to "ability to pay") rather than to have its admitted effect of discouraging work. Rules addressing purported externalities are an example of allocative rules, but perhaps are just a subset. Thus, if one had to rationalize the exclusion of employer-provided health insurance on allocative grounds, one might call it a response (whether or not a good one) to the adverse selection problems that create market failure in healthcare, rather than as mainly aimed at externalities.
The paper also contains much discussion of allocative rules as responding to internalities (e.g., irrationally failing to save enough due to myopia - a problem that could support positive incentives for retirement saving even in the otherwise inter-temporally neutral context of a consumption tax). We argued that, so far as externalities are concerned, one should think of getting the "price" right whether there are behavioral issues or not. Those issues, in turn, may need to be addressed whether or not one needed to fix externalities in order to ensure that people face the "right" prices. But the externality framework may not have that much to do with most of the big tax incentives that interest the authors, even if (at least for purposes of argument) we accept those rules as good policy reflecting an allocative motivation.
A big part of the project, which may perhaps become more prominent in later drafts, is to pursue in detail possible applications of the idea, from Batchelder's earlier work, of replacing tax incentives that take the form of deductions or exclusions with fixed refundable credits. Thus, for example, rather than a charitable or home mortgage interest deduction, the value of which depends on one's marginal tax rate (and on one's being an itemizer with positive taxable income to offset), one would instead have a rule under which everyone who incurs these items gets a refundable credit for, say, 30% of the amount spent. (Refundable means allowed to exceed the net income tax liability otherwise arising.)
This idea has virtues even if the incentive has nothing to do with externalities, and even if it is a really bad idea. Thus, consider home mortgage interest deductions. If two people in different marginal rate brackets (or only one of whom is an itemizer) are considering buying the same house, one might prefer that it go to the person who values it more, as shown by their offer prices. But if the one who otherwise values it less can make more use of the deductions, the "wrong" person in a pre-tax sense may end up buying it. Uniform refundable credits eliminate this problem.
Batchelder and Toder agree that the credits shouldn't always be uniform, as a matter of optimal subsidy design. For example, if research on charitable giving suggests that high earners are more responsive to the subsidy than low earners, the optimal credit for gifts might be a higher percentage for the former group. And if only poorer people are considered to need a nudge in order to save enough or buy enough health insurance (from the standpoint of their own self-interest, if internalities are the key here), then the optimal credit might be higher for them. Of course, changing the credits as income rises also is an input to marginal rate design, since it can effectively raise the marginal rate on earnings, but that merely complexifies rather than defeating the idea. But in many cases the optimal credit percentage (so far as we can determine it) may be the same for everyone, or may vary based on factors other than earnings.
One way of putting the general point is that, for incentives that take the form of rebating to people a percentage of particular favored outlays, there is in principle an optimal marginal reimbursement rate (MRR). But there is generally no reason to think that this will have anything to do with people's marginal tax rate (MTR), which is established via entirely separate considerations. Tax incentives that take the form of a deduction or exclusion automatically match the MRR to the MTR (creating a MRR of zero once net taxable income for the year is gone), often for no good reason other than unrelated political optics.