Not to return to a periodic gripe of mine - but then again, why not indulge myself by engaging in it after all - but it is sometimes frustrating to see people writing about "tax expenditures" without any notion of whether this is a coherent category, if so what is it about, etc.
What makes it a gripe, on my part, is that I have actually solved the problem. Or more specifically, I have (ahem, if I do say so myself) developed the most coherent way to think about a concept that, in the end, is not entirely coherent. The underlying problem is that it says that certain "tax" rules are actually "spending" rules, whereas these two categories are not well-defined to begin with. I have explained, however, that in context what it often means is an "allocative" rule that has been smuggled into a mainly "distributional" instrument, such as the U.S. federal income tax, which discourages earning income but is not actually designed to do so (the aim, rather, is distributional, with income being used as an imperfect metric for "ability" or "ability to pay" or "low marginal utility of a dollar" or some such thing.)
Anyway, you can see a version of my paper on the subject, "Rethinking Tax Expenditures and Fiscal Language," here. The publication cite is 57 Tax Law Review 187 (2004). A shortened version appears as chapter 8 of my book on fiscal language, which is available here. But to quote the hilarious Will Ferrell character from Zoolander: "Doesn't anyone notice this? I feel like I'm taking crazy pills."
Does it always matter? No. Suppose we are considering which "tax expenditures" to repeal, and it simply means we have a list of items in front of us, such as those officially listed by the Joint Committee on Taxation as tax expenditures, such as in their 2013 estimates, available here. Then it's just a list of items. The rationale for assembling this particular list doesn't necessarily matter, so long as we look at each item on its merits and are willing to look at items that aren't on the list.
But there is a persistent delusion, which use of the lists in this way may unduly eccourage, that the tax expenditure concept as traditionally used has more of a coherent core than it actually does. And even very good economists may be subject to this problem.
Case in point, a very interesting (and otherwise meritorious) paper just posted by Raj Chetty, Nathaniel Hendren, Patrick Kline, and Emmanuel Saez, entitled "The Economic Impacts of Tax Expenditures: Evidence from Spatial Variation Across the U.S.," and available here.
As per its abstract, the paper "stud[ies] the effects of tax expenditures on intergenerational mobility, using spatial variation in tax expenditures across the United States."
Chetty et al take "tax expenditures" as given, such as from the JCT lists, but do appear to be convinced that it's a coherent category as to which one can reach general conclusions about tradeoffs, costs and benefits, etcetera. They study three in particular:: deductions for state and local income taxes, home mortgage interest deductions, and the earned income tax credit. For all three, they find "consistent and fairly robust relationships between higher local tax expenditures and lower intergenerational elasticity (IGE), i.e. higher economic mobility."
OK, that's an interesting result concerning these items that doesn't necessarily turn on broader generalizations about a "tax expenditure" concept. But what is the concept? All they say, in the first paragraph of the article, that tax expenditures are "spending" despite their placement in the tax code.
OK, fair enough with regard to home mortgage interest deductions, which are naturally thought of as allocative subsidies delivered through the tax code. (Although more precisely, the subsidy is excluding imputed rent from income, and home mortgage interest deductions merely preserve the net exclusion for people with mortgages.)
On deducting state and local income taxes, it's a little trickier. You could get exactly the same end result, without "spending," by adjusting state and local income tax rates to reflect directly the impact that we currently get from the federal deduction, accompanied by an intergovernmental transfer so that each government unit comes out the same. (But is that "spending?" What if we integrated the units of government but they were all doing exactly the same thing? By analogy, the Defense Department spends lots of money, but we don't have a second tier of "spending" when the Treasury sends over money that the Defense Department can use to pay its bills.)
Then we get to the earned income tax credit. Is this "spending"? All it actually does is adjust net income tax rates that apply to labor income in certain ranges for certain taxpayers. That isn't "spending" unless the rate structure is generally an aspect of "spending," which is not how traditional tax expenditure analysis is done.
These are certainly good enough economists that it would be nice if they were more aware of what the "tax expenditure" category can and cannot be coherently used to mean. Again, the results about these items that they reach are good contributions to our state of knowledge. But generalizing about "tax expenditures" should be done with a bit more care.