Should a Global Minimum Tax Be Country-by-Country?

I. Introduction

Over the last several decades, large multinational enterprises have substantially reduced taxes on foreign income through complex parent-subsidiary structures that allow them to book that income in jurisdictions that have little or no connection to where value is produced or consumed and that impose little or no tax. One emerging response — proposed and partially adopted — is the multilateral implementation of a global minimum tax regime. Under that regime, an MNE whose effective tax rate would otherwise fall below a specified minimum would be required to pay whatever amount of additional tax would be necessary to raise its effective rate up to that minimum — that is, the MNE would incur top-up tax.

One of the key decision points for the design of a global minimum tax regime concerns how precisely to calculate the tax rate for the MNE that, when below the global minimum rate, triggers top-up tax. Is it enough that the taxpayer’s average tax rate across all countries exceeds the imposed global minimum? Or is it necessary that the rate imposed on it by each individual country exceed the minimum? That is, should the minimum tax have a global-average design (GAM) or a country-by-country design (CbCM)?

The United States has in place a kind of GAM for the U.S. shareholders of controlled foreign corporations: its tax on global intangible low-taxed income.¹ But the OECD, of which the United States is a key participant, has endorsed a CbC approach for its global anti-base-erosion (GLOBE) regime, also known as pillar 2.² The Biden administration along with congressional Democrats have proposed changing the U.S. system to conform in this respect to OECD proposals.³

Which is the best approach? Proponents of the CbCM approach put forward two arguments.

The first is that CbCM imposes a greater tax burden on MNEs than GAM. This first argument appears to rest on two premises. The first is the implicit supposition that the imposed global minimum tax rate will still, in some relevant sense, be too low — implying that collecting more revenue than would be collected under GAM at that imposed global minimum rate would be a good thing.

The second premise is, indeed, an arithmetic fact: All else the same, the amount of top-up tax that would be owed by a given taxpayer under GAM is never greater than, and in many cases strictly less than, the amount of top-up tax that would be owed under CbCM. This is because under CbCM, unlike under GAM, the fact that the taxpayer is subjected to a tax rate above the imposed minimum in some jurisdictions does not relieve it of the obligation to top up in other jurisdictions whose tax rates fall short of the imposed minimum.⁴

The second argument for CbCM concerns incentives — three sets of incentives to be precise: an MNE’s incentive to shift income to lower-tax jurisdictions, a small jurisdiction’s incentive to become a low-tax destination for income shifting, and a large jurisdiction’s incentive to prevent erosion of its tax base. The GAM is seen as giving MNEs with operations in high-tax jurisdictions an incentive “to invest in lower-tax jurisdictions, to take advantage of the automatic global averaging.”⁵ The result is that GAM “exacerbates the race to the bottom on corporate tax rates.”⁶ Importantly, this race appears to consist of two separate events: one in which low-tax jurisdictions race toward zero; and one in which high-tax jurisdictions race each other at least part of the way in the same direction. The CbCM regime is seen as dampening some or all of those harmful incentives.⁷

This report calls into question both arguments in favor of CbCM. It examines the incentives of taxpayers and taxing jurisdictions in a systematic light by grounding arguments in a rudimentary game-theoretic model. The model, the technical details of which are available online,⁸ is indeed elementary. It leaves out many important features of the problem. Nevertheless, it is at least an internally coherent and consistent account of several of the interconnected moving parts that appear to be central aspects of the issue.

The story that the model tells — albeit partial — is very different from the consensus view on GAM versus CbCM. The model indicates that, from the perspective of high-tax jurisdictions, GAM is superior to CbCM as a matter of both revenue raising and race-to-the-bottom incentives. Low-tax jurisdictions, on the other hand, ought to favor CbCM over GAM. MNEs should be indifferent.

The structure of the model explicitly reflects a series of judgments about which of the myriad features of these complex real-world issues are irreducible, or at least deserve separate consideration.⁹ The model focuses on (pure) income shifting — that is, changes in where income is booked for tax purposes, as opposed to where economic activity actually takes place, the two being notoriously disconnected in the current international tax regime.¹⁰ It concentrates on the interaction between two kinds of country-actors: high-tax jurisdictions and low-tax jurisdictions. High-tax jurisdictions (think France or Germany) are the actual locations of a fixed amount of economic activity conducted by multinational companies that are not tax-resident in those jurisdictions (think Apple). “Economic activity” may refer simply to distribution and sales (think iPhone sales). The tax rate that high-tax jurisdictions impose on the unshifted income of those nonresident MNEs is tied, in some exogenous manner, to the general tax rate that they impose on domestic activity.

As a result of this constraint, were a high-tax jurisdiction to hypothetically ignore the incentives of nonresident MNEs to out-shift income, that jurisdiction would prefer to impose on these MNEs a tax rate that falls substantially above the imposed global minimum rate.¹¹ Low-tax jurisdictions (think Bermuda or Malta, possibly with Irish intermediation¹²) house little or no actual economic activity relative to the quantity of potentially shifted MNE income. Inducing income shifting from high-tax jurisdictions is the decisive factor in their decision-making regarding tax rate.¹³ Were it not for the possibility of paying less tax, MNEs would prefer not to shift income out of high-tax jurisdictions.¹⁴

All jurisdictions choose what tax rate to impose, and MNEs choose where to locate their income for tax purposes. High-tax jurisdictions move first. Low-tax jurisdictions respond to high-tax jurisdictions’ tax rates. And MNEs respond to the full array of tax rates. High-tax jurisdictions do not compete with each other in this model of pure income shifting — only with low-tax jurisdictions. Thus, for example, France and Germany do not compete with each other to increase the number of iPhones sold to their respective consumers. On the other hand, low-tax jurisdictions compete among themselves to become the destination for income shifted out of high-tax jurisdictions.

Under GAM, high-tax jurisdictions end up choosing the tax rate that is optimal for them domestically — that is, the rate they would choose if they ignored the prospect of out-shifting by foreign companies operating within their borders. Some out-shifting does occur, but high-tax jurisdictions still collect from those domestically operating foreign companies an amount of tax revenue equal to what they would collect were they to set their rates at the imposed global minimum and thereby prevent out-shifting altogether.

Under CbCM, to earn the same amount of tax revenue from foreign companies as under GAM, the high-tax jurisdiction would have to lower its tax rate so that it rested just above the imposed global minimum rate — likely a substantial deviation from the jurisdiction’s domestically optimal rate. If the high-tax jurisdiction resists this downward pressure, it collects no revenue from domestically operating foreign companies.

Therefore, what was dually achievable for high-tax jurisdictions under GAM — choosing a domestically optimal tax rate and effectively collecting the imposed global minimum rate on foreign company income earned within its borders — becomes a dilemma under CbCM: One of these, the revenue or the rate, must be forgone.

Low-tax jurisdictions, on the other hand, prefer CbCM to GAM. Under GAM, low-tax jurisdictions bid their tax rates down toward zero in an attempt to capture more of the out-shifting.¹⁵ Consequently, they end up with relatively little revenue. CbCM protects low-tax jurisdictions from this mutually harmful competition, and if high-tax jurisdictions resolve their aforementioned dilemma by forgoing revenue from foreign companies, low-tax jurisdictions are able to tax the companies’ income at the imposed global minimum rate rather than at near-zero rates.

Multinational companies are indifferent between the two regimes. Their global tax rate on foreign income that, but for shifting, would be sourced in high-tax jurisdictions equals the imposed global minimum rate under both regimes.

The following section provides an overview of the basic logic behind these predictions. A subsequent section considers shortcomings and extensions. A full account of the model can be found online.¹⁶

II. Basic Story

A. Two Key Differences

Decisive in the model are two of the several structural differences between the incentives created by a GAM regime and those created by a CbCM regime. One concerns the incentives of MNEs to shift income; the other concerns the incentives of low-tax jurisdictions to lower tax rates. Both have to do with what might be termed the “persistence” of incentives, as opposed to the magnitude of their force.

These two key differences between GAM and CbCM are explained in detail in subsequent parts of this section, especially subsection C. Nevertheless, it is worth highlighting them upfront.

First, shifting incentives for MNEs are more persistent under CbCM than under GAM. Under GAM, companies have no incentive to shift income from high- to low-tax jurisdictions — more precisely, from jurisdictions whose tax rates are above the imposed global minimum to those whose tax rates are below it — beyond the point at which the MNE’s global average tax rate is reduced to the imposed global minimum. The result is that MNEs — which, all else the same, would prefer not to shift income — stop short of shifting all their income out of high-tax jurisdictions. Under CbCM, the incentive to shift does not stop when the MNE’s global average tax rate hits the imposed global minimum but rather continues until all income is shifted out of the high-tax jurisdiction.

To be sure, it is also true that — while the MNE’s global average tax rate is still above the imposed global minimum — the magnitude of the incentive to shift under GAM is larger than under CbCM. Shifting a dollar of income to a low-tax jurisdiction under CbCM can at best reduce the tax rate on that dollar to the imposed global minimum, whereas under GAM it can reduce the tax rate on that dollar to the tax rate imposed by the low-tax jurisdiction, which may be substantially lower than the imposed global minimum. However, at least in this model, the persistence of the shifting incentive is more important than its magnitude.

Second, a low-tax jurisdiction’s incentive to undercut competing low-tax jurisdictions is more persistent under GAM than under CbCM. A CbCM regime puts a brake on the incentives of low-tax jurisdictions to undercut each other’s tax rates. Under CbCM, a low-tax jurisdiction has no incentive to undercut another low-tax jurisdiction if that requires taxing below the imposed global minimum rate. Doing so would not lower the effective rate paid by MNEs in that jurisdiction — given CbC top-up tax — but would lower the revenue collected by the jurisdiction. Under GAM, however, a low-tax jurisdiction’s incentive to undercut persists until its tax rate is near zero. This is because MNEs that, all else the same, prefer to shift less rather than more, will find a way to lower their global average tax rate to the imposed global minimum with the least amount of shifting, and this will mean shifting to the low-tax jurisdiction with the lowest tax rate.

To see how these differences play out, consider first GAM and then CbCM.

B. A GAM Regime

It is worth saying more about the assumption that, all else the same, an MNE would prefer to shift less income rather than more out of the jurisdiction in which the income naturally arises. This preference, potentially slight, may arise from a variety of sources, including (1) the legal cost of shifting income — inclusive of the ongoing legal costs of adjusting tax structures in the face of ever-shifting laws, regulations, and treaties in the wide variety of jurisdictions that may be implicated by the intricate tax structures necessary for shifting;¹⁷ (2) the necessity of meeting extraneous brick-and-mortar or employment requirements imposed by low-tax jurisdictions or their intermediate enabling jurisdictions; (3) a loss of political goodwill with real-residence or real-source jurisdictions, perhaps manifest in public statements or high-profile legislative hearings, and perhaps rendering other interactions with those jurisdictions more difficult; and (4) damage to the company’s attempts to present itself to consumers as attractive from the perspective of social conscience and positive lifestyle.¹⁸

Assuming that those costs, though real, are relatively small, under GAM, MNEs shift income from high- to low-tax jurisdictions up to the point at which their average global rate has fallen to the imposed global minimum. The company has no incentive to shift beyond that point: Top-up tax under GAM would negate the effect of doing so and, all else the same, the company prefers not to shift.

The fact that the company prefers to shift as little income as possible is also important in precisely how the company chooses to reduce its global average tax rate to the imposed global minimum. It shifts to the low-tax jurisdiction with the very lowest tax rate. For instance, if one low-tax jurisdiction had a tax rate of 10 percent and the other 5 percent, and if the high-tax jurisdiction’s tax rate were 25 percent, while the imposed global minimum rate were 15 percent, the company could push its global average down to 15 percent by shifting only half its income to the 5 percent jurisdiction ((1/2) * 5 + (1/2) * 25 = 15). Reaching 15 percent with the 10 percent jurisdiction would require that the company shift two-thirds of its income ((2/3) * 10 + (1/3) * 25 = 15). The company prefers to shift as little as possible, so it shifts entirely to the 5 percent jurisdiction.

This dynamic sets up a competition among low-tax jurisdictions to be the lowest low-tax jurisdiction and thus grab more of the company’s out-shifting. In the example above, the 10 percent low-tax jurisdiction receives none of the company’s out-shifting, whereas if it slightly undercut the 5 percent jurisdiction by choosing a rate of, say, 4 percent, it would receive all of it and collect tax at that rate. A low percentage of something is better than any percentage of nothing. The 5 percent jurisdiction would then, for the same reason, have an incentive to undercut 4 percent with 3 percent. Thus, low-tax jurisdictions can be expected to bid themselves down to near-zero tax rates.¹⁹

Therefore, when the company out-shifts income to the point at which its global average rate falls to the imposed global minimum, the bulk of the revenue that is still collected from it goes to the high-tax jurisdiction. Moreover, the total amount of revenue collected from the company, and thus the total amount going to the high-tax jurisdiction, is not affected by the precise rate chosen by the high-tax jurisdiction itself. The following example will help clarify this last point, a central dynamic of a GAM system.

If the two low-tax jurisdictions have bid themselves down to 0 percent tax rates, and if the high-tax jurisdiction has a tax rate of 25 percent while the imposed global minimum tax rate is 15 percent, the company would leave 60 percent of its income in the high-tax jurisdiction. It would shift the rest into either or both of the low-tax jurisdictions, thus bringing its global average rate to 15 percent (60 percent of 25 percent being 15 percent). And it would pay tax to the high-tax jurisdiction at a rate of 25 percent on 60 percent of its income before shifting — which is the same as paying 15 percent on 100 percent of that income. If the high-tax jurisdiction’s rate were alternatively 30 percent, the company would leave 50 percent of its income there, again bringing its global average rate down to 15 percent (50 percent of 30 percent being 15 percent). It would then pay tax to the high-tax jurisdiction at a rate of 30 percent on 50 percent of its pre-shifting income — which is, again, the same as paying 15 percent on 100 percent of that income. Thus, a higher tax rate in the high-tax jurisdiction generates more out-shifting, but also collects more from still unshifted income. The two effects offset each other.²⁰

Because the high-tax jurisdiction receives the same amount of tax revenue from the foreign company regardless of the rate it sets, the shifting behavior of the foreign company generates no downward pressure on the tax rate of the high-tax jurisdiction. Thus, the high-tax jurisdiction might as well set its tax rate according to domestic policy considerations.

Therefore, GAM offers to the high-tax jurisdiction an amount of tax revenue from the foreign companies operating within its borders equal to the imposed global minimum rate times the pre-shifting income of the foreign companies. It obtains this amount no matter where it positions its own tax rate²¹ and thus without having to deviate from its domestically optimal rate. The low-tax jurisdictions compete themselves out of most of the revenue from these foreign companies. The foreign companies pay tax on their natively high-tax jurisdiction income at a rate equal to the imposed global minimum.

C. A CbCM Regime

In accord with the arithmetic point mentioned in the introduction, it is indeed true that switching from GAM to CbCM — holding fixed the behavior of all parties — would increase the total revenue extracted from multinational companies. The low-tax jurisdictions’ near-zero rates would now be topped up to the imposed global minimum (with the top-up revenue going perhaps to the company’s jurisdiction of residence).²² The fact that the company is paying at a rate above the imposed global minimum in the high-tax jurisdiction would no longer shelter it from top-up tax liability.

However, the effect of switching to CbCM should not be judged holding fixed the behavior of the parties at their choices under GAM. CbCM induces different behavior, and this should be taken into account.

Under CbCM, a low-tax jurisdiction has no incentive to tax at a rate below the imposed global minimum. This is the second of the two key differences between CbCM and GAM mentioned in the introduction to this section. Undercutting the imposed global minimum in a CbCM regime does not increase the amount of income shifted to the low-tax jurisdiction — from the company’s perspective, the low-tax jurisdiction’s rate reduction changes nothing of relevance: It still pays the imposed global minimum rate all told, though now partly to the offstage jurisdiction(s) collecting top-up tax. On the other hand, the low-tax jurisdiction collects less of that fixed amount of revenue: the top-up going elsewhere.²³

Therefore, a CbCM does in a sense prevent harmful tax competition — it prevents competition that is harmful from the perspective of low-tax jurisdictions taken as a group. They do not bid their rates down to near-zero, as under GAM. Rather, they bid their rates down only as far as the imposed global minimum.

Yet removing that “harmful” tax competition among low-tax jurisdictions ends up harming high-tax jurisdictions. Viewed against the baseline of how they fare under GAM, high-tax jurisdictions now face an unpleasant choice.

If a high-tax jurisdiction chooses to leave its rate at its domestically optimal level — which is presumed to be non-negligibly larger than the imposed global minimum rate — it will receive no tax revenue from the foreign companies operating within its borders. Whereas, under GAM, the incentive to shift stops when the company’s global average tax rate hits the imposed global minimum, under CbCM the incentive to shift persists until no income is left in the high-tax jurisdiction. This is the first of the two fundamental differences between CbCM and GAM mentioned in the introduction to this section.²⁴

To understand this point in greater depth, return to the example in which both low-tax jurisdictions’ rates are 0 percent,²⁵ the high-tax jurisdiction’s rate is 25 percent, and the imposed global minimum is 15 percent. Recall that under GAM, the company stopped shifting income when its global average rate hit 15 percent, leaving 60 percent of its income in the high-tax jurisdiction. Under CbCM, the company continues to have an incentive to shift income out of the high-tax jurisdiction even after its global average rate has fallen 15 percent. Regardless of the company’s global average rate, each dollar of income shifted to a low-tax jurisdiction is taxed at 15 percent (counting top-up tax) rather than 25 percent. It is true that the dollar-by-dollar shifting incentive under CbCM (25 percent versus 15 percent) is not as stark as it is under GAM before the global average rate hits the imposed global minimum (25 percent versus 0 percent). But the incentive lasts longer under CbCM. Rather than stopping once there is enough income shifting to cause the global average to hit 15 percent, as under GAM, the incentive persists until every dollar is shifted.

Further, the high-tax jurisdiction cannot prevent this dynamic by lowering its rate part of the way toward the imposed global minimum. Any rate non-negligibly larger than the imposed global minimum will produce full out-shifting. If, in the example discussed in the preceding paragraph, the high-tax jurisdiction lowered its rate from 25 percent to 20 percent, the company would still have a persistent incentive to out-shift, now on the order of 20 percent versus 15 percent.

Therefore, unless the high-tax jurisdiction is willing to reduce its rate to such extent that it lies just above the imposed global minimum — near enough to the minimum so that the company’s slight aversion to shifting becomes decisive — it might as well keep its tax rate at its domestically optimal level. Assuming the high-tax jurisdiction does choose its domestic optimum, and recalling that the low-tax jurisdictions set their rates at the imposed global minimum, it follows that in a hypothetical change of regime from GAM to CbCM, the high-tax jurisdiction essentially hands over the tax revenue it was collecting under GAM to the low-tax jurisdictions.²⁶

The high-tax jurisdiction’s only other relevant choice is to take the income-shifting bait and reduce its tax rate to some level just above the imposed global minimum. By doing so, it does indeed recapture from the low-tax jurisdictions essentially the same amount of tax revenue as it had under GAM — the imposed global minimum times pre-shifting income. But given that the global minimum rate is substantially below its domestically preferred rate, it gains this additional revenue at the cost of deviating substantially from what would otherwise be its best policy. This is the price, under CbCM, of the tax revenue it was getting “for free” under GAM.

The low-tax jurisdictions are no worse off under CbCM and may be better off. If the high-tax jurisdiction reduces its tax rate to retain the foreign company’s income, the low-tax jurisdictions receive none of the tax revenue from that income — just like under GAM. If the high-tax jurisdiction demurs, the low-tax jurisdictions share an amount of revenue equal to the imposed global minimum rate times pre-shifting income.

Whichever choice the high-tax jurisdiction makes, the multinational company pays the same amount of tax under CbCM as under GAM: the imposed global minimum rate times pre-shifting income. The high-tax jurisdiction’s decision under CbCM determines only whether this goes into the high-tax jurisdiction’s coffers or those of the low-tax jurisdictions.

III. Discussion

The foregoing analysis rests on a series of assumptions. Some of these were mentioned in the introduction but deserve additional consideration.

The first set of assumptions concerns the differing underlying characteristics of high- and low-tax jurisdictions — parametric characteristics that cause them to choose, respectively, high and low tax rates. High-tax jurisdictions have three main characteristics. First, their economies are large enough to contain the actual economic activity of the multinational company. For example, they present to the multinational a substantial demand curve for the company’s products. Second, in parallel with their sizable internal economies, their governments are large and active, with formidable revenue needs. Third, the tax rate that they impose on foreign multinationals operating within their borders is linked to the tax rate that they impose on purely domestic activity. They are perhaps constrained by treaty from taxing foreign multinationals less favorably than domestic companies, and by domestic politics from taxing them substantially more favorably. Fourth, the tax rate that a high-tax jurisdiction would prefer to set, taking into account only domestic linkages and not the revenue collected from the multinational, is substantially greater than the imposed global minimum tax rate.²⁷

A low-tax jurisdiction, on the other hand, is motivated solely (or at least predominantly) by capturing tax revenue from the multinational company, which it can accomplish only (or at least mainly) by inducing the purely legal-formalistic relocation of profits into its territory. A substantial relocation of actual economic activity is not an option because of the small size of the low-tax jurisdiction’s economy.

Continuing in this vein, the model concerns solely multinationals’ incentives to book income outside the high-tax jurisdiction. The real economic activity of the companies that produces this income remains always inside the high-tax jurisdiction. Indeed, it remains fixed in magnitude no matter what tax rates are set. Thus, the model does not capture competition among high-tax jurisdictions for real economic activity.

Alternatively, one can broaden the interpretation of the high-tax jurisdiction’s “domestically optimal tax rate” so that it is the rate that it would choose if the kind of purely formalistic income shifting studied here were taken out of the picture. The rate would then take into account not only domestic considerations but also the competition among high-tax jurisdictions for real economic activity. One would then need to broaden the aforementioned assumption regarding the relative smallness of the imposed global minimum rate: It would have to be substantially smaller than this potentially lower rate.

It is, of course, possible that those who emphasize the “race to the bottom” in corporate tax rates have in mind some combination of real and nominal relocation. However, it should be kept in mind that the global minimum tax regime, as it is emerging, appears to be directed mostly, if not solely, at purely nominal profit shifting. Carveouts for real relocation have been proposed, at least in the context of emerging (though already substantial) economies.²⁸

The model assumes that any top-up tax that may be imposed — in fact, in the end none is imposed — is diverted away from the high- and low-tax jurisdictions. One may imagine that such revenue flows to the residence jurisdiction of the multinational, just as revenue from GILTI flows to Treasury. However, it is not inconsistent with the model to imagine that top-up tax revenue is shared among a group that may include one or more of the jurisdictions in the model, as long as their fractional share is relatively small.

It is also worth noting in this regard that proposals to allow low-tax jurisdictions to collect top-up tax — so-called qualified domestic minimum top-up taxes (QDMTTs) — are irrelevant in this model.²⁹ It makes no difference whether a low-tax jurisdiction — which here houses little or no real economic activity — adds a prioritized top-up tax or simply raises its rate directly.³⁰

The model assumes that low-tax jurisdictions compete among themselves for profit shifting out of the high-tax jurisdiction. The model generates this dynamic by this supposing that, but for the tax benefits, the company would prefer to out-shift as little as possible. In the formal model, it is assumed that the company incurs a small unit cost for each dollar of income that is out-shifted. Some arguments supporting this kind of assumption have already been listed. It is worth mentioning several other points in defense of this modeling choice.

First, the very low rates imposed by so-called tax haven jurisdictions over the last several decades seem consistent with competitive forces: Rates have indeed been near zero rather than merely substantially below those of high-tax jurisdictions.

Second, if companies did not have such an aversion to shifting, it is unclear why under GAM there would be a race to the bottom among low-tax jurisdictions, as CbCM proponents seem to assert. If a company does not care how much it shifts apart from the tax consequences of doing so, then it does not care whether it is shifting its average rate down to the imposed global minimum through shifting to a moderately low-tax jurisdiction or to a very low-tax jurisdiction. That is, if, in accord with an example from earlier, the high-tax jurisdiction’s tax rate is 25 percent, the imposed global minimum is 15 percent, one low-tax jurisdiction’s tax rate is 10 percent, and the other’s is 5 percent, the company does not care whether it drives its average rate to 15 percent by shifting half its income to the 5 percent jurisdiction, two-thirds of its income to the 10 percent jurisdiction, or some middling fraction of its income to some corresponding combination of the two low-tax jurisdictions. Thus, without at least some lexicographic desire on the part of the company to minimize the amount shifting, there seems to be no reason why, under GAM, all low-tax jurisdictions would not set their tax rates just below the imposed global minimum and split among themselves the shifting (of almost all) of the company’s income out of the high-tax jurisdiction.

Nevertheless, it is worth noting that as a mathematical matter, were the low-tax jurisdictions under GAM to choose tax rates that are not very small, there would indeed be a downward pull on the tax rate of the high-tax jurisdiction. Using the above example, if there is a single low-tax jurisdiction with a tax rate of 10 percent rather than 0 percent, then the company leaves one-third of its income in the 25 percent high-tax jurisdiction. The high-tax jurisdiction thus collects one-third of 25 percent, or 8.33 percent, on each dollar of pre-shifting income. If, under these conditions, the high-tax jurisdiction lowers its rate to 20 percent, then the company leaves half its income in the high-tax jurisdiction ((1/2) * 20 + (1/2) * 10 = 15), and its tax revenue is half of 20 percent, or 10 percent, which is greater than 8.33 percent. The force of this downward pull changes in complicated ways as high- and low-tax jurisdictions’ rates are varied.³¹

At the very least, the model’s assumption about competition among low-tax jurisdictions under GAM can be viewed as having been adopted for purposes of argument, so as to conform to the apparent premises of pro-CbCM arguments. In this context, a larger point emerges: The race to the bottom among low-tax jurisdictions must be separately considered from a race to the bottom among high-tax jurisdictions. As just noted, a race to the bottom among low-tax jurisdictions under GAM actually serves to prevent a race down to the imposed global minimum among high-tax jurisdictions. When low-tax jurisdiction tax rates are very low, the high-tax jurisdiction gains very little revenue from the multinational company by lowering its tax rate: Base and rate effects offset each other nearly perfectly.

It should always be kept in mind that there is an easy way to shield high-tax jurisdictions from the dilemma they face under CbCM, wherein those jurisdictions must choose between substantially reducing their rate to near the imposed global minimum or forgoing revenue from foreign companies operating within their borders. This is to raise the imposed global minimum so that it is closer to the rate that high-tax jurisdictions would choose were out-shifting by foreign companies not an issue.

Finally, the analysis in this report does not consider the impact of the OECD’s other major reform proposal, known as pillar 1 (the global minimum tax regime being pillar 2). This proposal would essentially allow market jurisdictions — here, high-tax jurisdictions — to prevent some out-shifting by fiat rule rather than rate reduction. Were pillar 1 widely adopted, it would change the operation of both GAM and CbCM. However, to the extent that pillar 1 does not render a global minimum tax superfluous, it is likely that at least some of the considerations highlighted in this report will continue to be relevant to the choice between GAM and CbCM.

IV. Conclusion

This report shows in the context of a simple game theoretic model that a global minimum tax regime that operates on a country-by-country basis is not necessarily superior to one that is based on global averaging — at least not from the perspective of the high-tax jurisdictions spearheading reform. The model suggests that, relative to a GAM, a CbCM forces high-tax jurisdictions to choose between less revenue or deleterious rate reductions.