The above paper, which we discussed at our session yesterday, is empirical but raises normative issues. I'll start with the latter, as it gives context and relevance to the paper's empirical contribution.
The paper's core idea (as it notes at page 3) is to examine the apparently conflicting views recently stated by two experts. Both views relate to the question of how deferral for the foreign source income (FSI) of U.S. multinationals' ("US MNEs") foreign subsidiaries ("CFCs," for "controlled foreign corporations") affects the extent to which the companies engage in profit-shifting - that is, engaging in tax planning manipulations to create FSI of the CFCs in lieu of U.S. source income of the parent companies.
Suppose we are thinking of repealing deferral. Now, the standard view is that this could be done in either of two ways: either by making the CFCs' FSI immediately taxable in full (say, as a deemed dividend) to the U.S. parents, or by adopting a territorial system, under which the CFCs' FSI generally would not be taxed - albeit, in practice likely subject to "CFC rules," making some such income immediately taxable to the parent, in keeping with actual practice in all major countries that have adopted "territorial" systems.
A third option, which I favor and have discussed at length in various places, would be to repeal deferral, generally tax FSI at a lower rate than U.S. source income but not at 0% as under a pure territorial system, and repeal foreign tax credits. However, foreign taxes might still get effectively better-than-deductible treatment in some settings, and this might indeed be achieved at least partly through the use of CFC rules that, like those in many territorial countries, treated actually or presumptively low-tax FSI less favorably than other FSI. But one could write a whole book on this, as indeed I have, so let's leave that to one side for now.
Anyway, suppose that what we mean by "repeal deferral" is "adopt a territorial system," albeit possibly with CFC rules that actually make certain FSI domestically taxable. Then one question that would arise is whether, at least as a standalone matter, this would increase outbound profit-shifting by U.S. companies to the detriment of the U.S. tax base. If it would, then we might either have an argument against repealing deferral, or else a concern that we might need to address by other means as a part of the deferral repeal package.
Now to the two apparently conflicting claims by leading experts. John Samuels of GE asserted in 2010 (at a symposium discussion that was published in Tax Notes) that deferral does NOT reduce profit-shifting by U.S. companies. In effect, he said, profit-shifting to take advantage of deferral is a free option. Suppose you profit-shift in Year 1 and then have to repatriate the shifted earnings in Year 2. Unless the repatriation tax rate has increased in the interim, you haven't actually lost anything. So far as the deferred principal is concerned, you just pay the same tax in Year 2 that you would have otherwise paid in Year 1. (Also, while you got to defer it for a year, the amount that's out there also grows at your foreign rate of return - still leaving you, however, with a net time value benefit if the after-tax rate of return available to funds that are classified under tax rules as sitting "abroad" exceeds that which you could have garnered if they were classified as sitting "back home.")
Samuels has a nice point here, whether or not one entirely agrees with it. If profit-shifting imposes zero costs on the taxpayer, it's a free option and there's usually no reason NOT to do as much of it as possible. The worst-case scenario - so long as the repatriation tax rate doesn't increase, and the after-tax rate of return that you get "abroad" at least equals that which is available "domestically" - is that there's nothing lost, albeit not much gained if it was just for a year.
If this point is wholly true, how should it affect our thinking about the repeal of deferral in the context of shifting to a "territorial" system? I would say, possibly not so much (although it would be a relevant input to one's thinking).
Here's why. Suppose one were to say: Ah, Samuels has shown us that deferral doesn't actually reduce profit-shifting. So we don't need to think about the repeal of deferral, in the context of going more territorial (but possibly subject to revised source and CFC rules!), as a step that makes the profit-shifting problem worse.
But here's the thing. In this thought experiment, we haven't just repealed deferral - we have also repealed the tax on FSI that previously arose upon repatriation. In Samuels' scenario, companies are sometimes paying the repatriation tax. If we're eliminating that, it matters. Even just looking at the direct revenue effects, the U.S. loses tax revenue, if it either was collecting positive repatriation taxes, or there was a positive present value for future expected repatriation taxes. There may also be indirect positive effects on U.S. tax revenues.
Now, from a U.S. budgetary standpoint, after all, the primary problem isn't profit-shifting as such - it's the reduction in U.S. tax collections.
There also is the flip side to think about. Samuels has long argued that the U.S. should shift to a territorial system, by reason of the effects that he attributes to our imposing expected tax liabilities on U.S. MNEs with respect to their FSI. Obviously (and as I am sure he agrees), the problems that concern him are still there, even if deferral is a free option, if forced repatriation might (and sometimes does) occur at some point. So his main arguments for going territorial are still there - the free option to profit-shift for now doesn't make his arguments about disadvantaging U.S. companies, etc., go away.
Now suppose instead that profit-shifting, within a deferral regime, is costly rather than a free option. There might be fixed up-front costs of setting it up - for example, putting enough real stuff abroad to be able to profit-shift, paying lawyers and accountants to set up all the shell entity mechanisms and such, etc. Then companies would have to ask themselves whether profit-shifting's benefits would be worth these costs. All else equal, they'd be more likely to say no if they were anticipating only very short deferral periods.
Suppose also that there are ongoing marginal costs of profit-shifting. An example would be having to make sub-optimal interim investments (defined in terms of risk-adjusted after-tax returns) with the "trapped earnings." This would likewise undermine the "free option" argument, although it's not necessarily related in the same way to short-term versus long-term deferral periods.
Based mainly on the fixed costs, Patrick Driessen has argued (as cited at page 3 of the Dyreng-Markle paper, right after the Samuels quote) that some U.S. companies might have profit-shifted a lot more, but for the fixed costs that made them think it wouldn't be worth if taxable repatriation might be likely fairly soon. For such companies, adopting a territorial system under which repatriating the supposedly foreign-source profits had no adverse U.S. tax consequences would indeed, Driessen argues, induce additional profit-shifting out of the U.S.
Now, once again, what matters the most here is the prospect of paying a U.S. tax at some point - not the deferral period as such. Again, let them profit-shift all they like, but if we always got to tax the full profits in the very next year (at the same rate) it wouldn't matter so much.
Despite all this, one's choice between the Samuels and Driessen views, as descriptions of some underlying reality, does, to a degree, matter substantively all on its own. For example, the incentive to defer both has greater efficiency costs, and greater efficacy in discouraging profit-shifting - so it is both bad and good - under the Driessen view. Also, the amounts and periods at stake might be large enough to matter on their own. And, if US MNEs differ in their capacity to benefit from incurring the fixed costs of setting up profit-shifting mechanisms, we might be taxing them unequally - to the relative benefit of those that CAN benefit from incurring the fixed costs - in a way that is independently undesirable.
So the question is certainly interesting and important - one just wants to avoid misstating (and possibly overstating) why, how, and how much it matters.
Anyway, Dyreng and Markle explore a particular hypothesis that is pursuant to the Driessen view, and in tension with the Samuels view. Suppose, they say, that financially constrained firms expect shorter deferral periods than financially unconstrained firms. E.g., they anticipate bringing the funds home sooner rather than later, because they are less well-positioned to meet anticipated short-term cash needs, such as by borrowing. Then, by reason of the fixed cost problem, one might expect less profit shifting from the constrained firms than from the unconstrained ones.
Dyreng and Markle therefore attempt the following. First, devise (a) a measure of financially constrained firms. Second, devise (b) a measure of profit-shifting. Third, determine (a) is negatively correlated with (b), and in such a way that we can reasonably conclude that (a) indeed caused (b).
The problem here is that both (a) and (b) are hard measures to devise. As to (a), note that "financially constrained" need not mean distressed, losing money, etc. It just refers to difficulty in cheaply raising funds commensurate with one's appetite for same. They look at three preexisting measures of financially constrained firms in the literature - all of which have serious problems - but figure that, if they get a positive finding and alternative explanations aren't good enough, then they must be onto something.
As to (b), their task is made easier by the fact that they don't need an underlying measure of "correct" linkages and particular countries - just differences between the two types of firms identified by (a). Thus, suppose a U.S. company develops valuable IP in California, leading to $2X of worldwide sales, exactly half of which ($X) are in the U.S. Under an origin-basis approach to defining the source of income, arguably all of the income ought to be classified as U.S. source. Under a destination-basis approach, such as that which would arise under sales-based formulary apportionment, arguably half of the income ought to be classified as FSI.
Dyreng and Markle need not resolve this conundrum, in order to apply their test. Suppose for convenience that the company had basis and expenses of zero, such that its worldwide income was $2X. And suppose this meant that a financially constrained firm that couldn't profit-shift had a U.S. to foreign income breakdown of $2X / 0 under the origin method, or $X / $X under the destination method. Either way, an identical, except financially unconstrained, firm that could profit-shift in the amount of $Y would be visibly different. Its breakdown would ($2X - $Y) / $Y if everyone used the origin basis, and ($X - $Y) / ($X + $Y) under the destination basis, So the difference - not the "correct" answer absent profit-shifting - is the point of interest here.
The paper finds that being financially constrained does appear to reduce profit-shifting, in tension with the "free option" argument. This suggests that deferral does tend to reduce profit-shifting, albeit just for a subset of firms, although it also might support inferring that deferral has greater year-by-year efficiency costs than one might have inferred under the "free option" view.