![Kimberly A. Clausing](http://s3.amazonaws.com/binary.taxnotes.com/image/2019/Clausing_Kimberly_20190729_1_webs-2019-29083.png)
Kimberly A. Clausing (clausing@reed.edu, clausing@law.ucla.edu) is the Thormund A. Miller and Walter Mintz Professor of Economics at Reed College. In January 2021 she will be a professor of tax law and policy at the University of California — Los Angeles School of Law.
The author thanks Bijay Rai for research assistance and Thomas Horst and Tim Dowd for helpful suggestions.
In this article, the author analyzes 2017 country-by-country reporting data released by the United States and outlines five important lessons for scholars investigating international corporate tax avoidance.
Copyright 2020 Kimberly A. Clausing. All rights reserved.
One of the signature achievements of the base erosion and profit-shifting project is the collection of multinational enterprises’ country-by-country reporting data for government use in tax enforcement efforts. In late 2019 the United States became the first country to release a complete set of those data, in aggregate form, for 2017.
This article analyzes those data, demonstrating five important lessons for scholars investigating international corporate tax avoidance. First, profits and accumulated earnings are disproportionately reported in tax havens, and some previously unobservable havens play a sizable role. Tax havens have a dominant role in profit shifting; most of the misalignment between profits and economic activity occurs in haven countries. Second, unlike reported profits, economic activity is far less sensitive to tax differences across countries. Third, aggregate data sets that combine companies with profits and those with losses provide an incomplete and biased picture of international tax avoidance; CbC data are promising in their ability to distinguish profitable companies from those with losses. Fourth, analysis of the data indicates that profit shifting was a large problem in 2017, generating major revenue costs in many countries, including the United States. Fifth, relatively modest tax reform measures could substantially increase government revenue and stem tax competition pressure.
The CbC Data
Aggressive tax avoidance by multinational enterprises has generated years of press attention, advocacy, and on some occasions, public outrage. In response, the OECD launched the BEPS project in 2013, resulting in thousands of pages of guidelines by 2015 and several continuing multilateral initiatives.
One of the project’s signature achievements is the collection of MNE CbC data for government use in tax collection and enforcement. In late 2019 the United States became the first country to release a complete set of those data in aggregate form, covering the year 2017. In total, 90 countries have implemented CbC tax reporting rules, and 50 more have indicated that they plan to adopt.1
CbC data reporting was part of the action 13 BEPS recommendations, which developed a multifaceted approach for taxpayers to provide clearer information to tax authorities. Those data were intended to help tax authorities determine whether additional audit work would be appropriate.2 Companies were asked to provide a master file on the MNE’s structure, as well as consolidated financial statements, local files on each local entity, and a CbC report that indicated the quantities of key variables (such as profit, tax paid, employment, accumulated earnings, and tangible assets) in each location. According to the OECD, “The countries participating in the BEPS project agree that these new reporting provisions, and the transparency they will encourage, will contribute to the objective of understanding, controlling, and tackling BEPS behaviours.”3
These are new data, and some observers have cautioned that companies may be confused regarding how to comply with the requirements.4 Still, in the United States, companies had the option of voluntarily complying in 2016 (about 70 percent did so), and by 2017 they had some experience with the process.5 The IRS has provided detailed instructions for relevant Form 8975, including definitions of key terms and variables.
For researchers, the data have many strengths over prior data sets, but also a few weaknesses. In terms of strengths, it is immediately clear that there is far more country detail in the data. Prior IRS data (Form 5471, an informational return) was reported in aggregate form for only 44 economies. U.S. Bureau of Economic Analysis (BEA) surveys and balance of payments data include 57 economies. The U.S. CbC data show 140 economies in the full sample, and 94 economies in the positive profit sample.
As the following section shows, tax havens are particularly important in driving profit-shifting behavior, which makes intuitive sense. Why hire lawyers and accountants to move profits from 30 percent to 20 percent tax rate locations when you can move profits to jurisdictions with exceptionally low rates? One key advantage of the CbC data is that they reveal many important havens that were invisible in most prior data sets, including important havens such as the Isle of Man, Jersey, and Puerto Rico.
The data also allow a separate examination of the full sample, including both companies with profits and those with losses, as well as a sample that is confined to only companies with profits. The second sample is ideal for calculating effective tax rates, because taxes are presumably paid only by companies earning profits.
Which sample is best for considering the magnitude of profit shifting is less clear. If all companies experience periods with losses and periods with profits, over time, their profitability should reflect the sample as a whole. However, if the same companies consistently earn profits while another set consistently earns losses, then focusing on companies with profits would be better. Those two series may thus provide upper and lower bounds of the scale of profit shifting.
While the purpose of the CbC data, its broad country coverage, and the inclusion of the positive profit company series are all useful, there are downsides associated with the new data. First, there is limited release of the data. The United States has released full data for only 2017 and incomplete data for 2016. Many other countries have far less complete releases. Second, there is some evidence of confusion in terms of key variable definitions, as reported by the OECD.6 In some countries, companies reported profits in a way that may have led to double counting.
Whether double counting is present in the U.S. data is less clear, especially if stateless income is excluded. There is some income that is recorded as stateless on the form, and the full scope of stateless income is unclear. IRS instructions classify some passthrough income as stateless. Thus, the word “stateless” does not map neatly into international tax scholarship, where the term “stateless income” is used to describe income that is taxed nowhere, often because of exploitation of check-the-box rules that create disregarded entities for tax planning purposes.7 For the CbC data, however, IRS instructions state that a check-the-box election should not affect the foreign entity’s jurisdiction of tax residence.
There is ample evidence that untaxed foreign earnings are a large and increasing part of the profit-shifting problem. However, in the data, stateless income may instead be capturing types of passthrough income that should be characterized as U.S. or foreign income. This analysis completely omits that income to avoid possible double counting.
The IRS Form 8975 instructions clearly define revenue to exclude intracompany dividends, implying that profit should also exclude that source of income. Still, the definition of profit may be unclear, and companies are free to supply data as they see fit. Because the data are known to be used for transfer pricing risk assessment,8 it is unlikely that companies will have an incentive to overstate their income, especially in tax havens. It is also reassuring that in the U.S. data, foreign totals are similar to those reported from other sources that are known to exclude double counting. Appendix A provides comparisons with other series.
A 2020 paper further discusses the possibility of double counting in the data. 9 The analysis suggests a 14 percent discrepancy between CbC income totals (for both the United States and foreign countries) and totals in financial reports when stateless income is excluded. There are several possible reasons for discrepancies, including the larger company coverage of the CbC data, that reporting and definitional differences exist between the series, and the possibility that the CbC totals are overstated because of confusion about the Form 8975 directions. In the last event, recent analysis confirms that it is most likely that U.S. income is overstated, which would not substantially affect the current analysis, but it is also conceivable that some foreign lines may be mismeasured.
Lesson 1: Havens’ Role in Profit Shifting
An examination of the location of U.S. multinational company profits in recent years shows that a small number of tax havens with rock-bottom tax rates are disproportionately important.10
For example, in 2017 U.S. companies reported $4.2 trillion in offshore accumulated earnings, $3 trillion of which is in tax havens. But just nine havens account for $2.8 trillion of that $3 trillion.11 Of those nine, four are independent European jurisdictions (Ireland, Luxembourg, the Netherlands, and Switzerland), four are island jurisdictions with close affiliations to the United Kingdom (Bermuda, Jersey, and the Caymans) or the United States (Puerto Rico), and the other is Singapore. Those tax havens also appear to be rich economies, typically reporting well-above-average GDP per capita.12
Figure 1 shows some remarkable facts about where U.S. multinational companies reported profits in 2017. Of the 20 top foreign profit locations for U.S. MNEs, accounting for 84 percent of the total foreign profit ($513 billion of $638 billion), 11 are tax havens (those above in dark blue bars) and nine are not. Those 11 haven countries account for 56 percent of all U.S. multinational foreign profits, using the full sample of CbC data.13
Table 1 defines havens as economies with effective tax rates below 10 percent, using the positive profit sample to calculate those rates.14 Ireland represents a difficult case, because it has an important historical role as a haven, yet its effective tax rate in 2017 was a bit over the haven threshold of 10 percent. Table 1 includes summary statistics with Ireland listed as a haven.15
| Profit of U.S. MNCs, in Billions USD | Effective Tax Rate, Positive Profit Companies | Population, in Millions | GDP, in Billions USD |
---|---|---|---|---|
11 havens | 355 | 3.9% | 48 | 2,720 |
9 non-havens | 158 | 26.9% | 3,250 | 32,260 |
Ratio: havens/non-havens | 225% |
| 1.5% | 8.4% |
Note: In the rest of the world (excluding the United States), U.S. MNEs report $126 billion in profit and an average effective tax rate (in the positive profit sample) of 22.5 percent in countries with 3.9 billion people and $26.8 trillion in GDP. |
Table 1 makes clear that the tax havens shown in Figure 1 are different from the other countries. The non-havens are economic powerhouses with large economies (such as China, India, and Germany), close ties to the United States (Canada, Mexico), or both (the United Kingdom). Together, those economies have 3.25 billion people and $32 trillion in GDP. Their average tax rate is 27 percent.
Havens, on the other hand, have small populations and small economies. As a group, they are just 1.5 percent as populous as the other countries, and they have less than a tenth of the other countries’ GDP. Yet U.S. multinational companies report $355 billion in profits in havens, more than twice what they report in the big economies.
The Cayman Islands show $58.5 billion in reported U.S. MNE profit, more than 10 times its reported GDP. Profits in the Caymans, a country of 63,000 people, are higher than those in China, a country with 1.38 billion people. Profits in Bermuda (64,000 people) exceed those in Brazil (207 million people); profits in the Isle of Man (83,000 people) exceed those in Italy (60 million people). Profits in Jersey (105,000 people) exceed profits in all but 16 countries of the world.
And as concentrated as profits are in the lowest tax rate countries, accumulated earnings are even more concentrated. The top 11 havens in Figure 1 account for 56 percent of total profits, but 71 percent of total accumulated earnings. As noted above, stateless income is excluded from these analyses.
Lesson 2: Havens’ Role in Economic Activity
One could argue that perhaps tax havens are simply investment hubs, locations where a lot of economic activity occurs for multinational companies, despite their small size. Yet an examination of the real operations of U.S. affiliates in tax havens reveals that economic activities in those locations are much smaller than the underlying profit suggests. In 2017 U.S. multinational companies reported $355 billion in the 11 havens shown in Figure 1, 56 percent of all their foreign profit.16 In those same 11 economies, U.S. multinationals have only 5.6 percent of their foreign employees, one-tenth the faction of profit (56 percent) reported in those locations. Profit per worker averages $488,000 in the 11 havens (taking total profit of those jurisdictions relative to total employment), $49,000 worldwide, and $22,900 in non-haven countries. Of the 15 highest profit-per-employee economies (aside from Libya, mysteriously), every one is a tax haven with rock-bottom tax rates.
Havens host more assets than employees, with the 11 havens in Figure 1 accounting for 24 percent of foreign assets. Unsurprisingly, they have a larger role in related-party sales (51 percent of the foreign total) than in unrelated sales (24 percent of the total). Sales are measured based on where the sales originate, not on the customer’s final destination, so if they are made to high-tax countries from haven locations, they are recorded in the haven. Selling from lightly taxed affiliates is one way to increase their profitability relative to other affiliates, so it is unsurprising that sales, especially related-party sales, are higher in havens than their employment would suggest.
Table 2 shows profit per employee and profit per asset in the top 20 economies with large U.S. MNE profits and for the world as a whole (excluding the United States and stateless income). Jurisdictions referred to as “low-activity havens” have particularly high profits per employee. There are, in fact, other minor havens that surpass even them. Gibraltar has $79 million in profit per employee; the British Virgin Islands has $46 million in profit per employee.
A second group of havens shows more real economic activity but also exceptionally high profits per worker, many multiples of the world average. The final group of non-haven high-profit countries does not show high levels of profits per employee; those countries generate high profits because their markets are large, rich, or both.
| Profit Rank | Effective Tax Rate, Positive Profit Companies | Profit/Employee, in USD | Profit/Asset, in USD |
---|---|---|---|---|
Low Activity Havens |
|
|
| |
Cayman Islands | 1 | 0.1% | 37,358,162 | 2.6 |
Bermuda | 6 | 2.4% | 59,371,993 | 2.7 |
Luxembourg | 11 | 0.7% | 1,980,842 | 0.1 |
Jersey | 17 | 0.8% | 22,010,539 | 19.8 |
Isle of Man | 18 | 0% | 11,465,660 | 42.2 |
Higher Activity Havens |
|
|
| |
Singapore | 2 | 4.5% | 337,259 | 0.9 |
Switzerland | 3 | 5.7% | 617,761 | 0.9 |
Netherlands | 4 | 4.9% | 251,016 | 0.6 |
Puerto Rico | 5 | 1.4% | 479,416 | 2.1 |
Ireland | 8 | 12.8% | 199,366 | 0.3 |
Hong Kong | 15 | 9.4% | 134,266 | 0.6 |
Big Activity Countries |
|
|
| |
Canada | 7 | 15.8% | 32,314 | 0.2 |
China | 9 | 24.5% | 20,247 | 0.3 |
Japan | 10 | 20.1% | 63,844 | 0.6 |
U.K. | 12 | 10.5% | 14,919 | 0 |
Mexico | 13 | 35.8% | 12,721 | 0.2 |
Australia | 14 | 15.7% | 54,548 | 0.1 |
India | 16 | 33.9% | 8,328 | 0.4 |
Italy | 19 | 25.1% | 39,181 | 0.3 |
Germany | 20 | 21.9% | 12,105 | 0.1 |
World Average | 17% | 48,750 | 0.28 | |
Note: This table uses the full sample for all data except the effective tax rate, which is based on the positive-profit sample. |
The nonlinearity of the relationship between tax rates and profit per employee is remarkable. Figure 2 shows the relationship between profits per employee and effective tax rates for those countries where aggregate profits are reported as positive in the full sample. Some very low tax rate economies have tens of millions of dollars of profits per employee; most other economies have comparatively little profit per employee.
Zeroing in on those economies with profits per employee below $1 million, the nonlinearity persists. When reporting profits, companies are particularly responsive to the jurisdictions with the lowest rates.
Lesson 3: Effect of Companies With Losses
As discussed above, companies with profits are those that are paying profits taxes, so it makes the most sense to calculate effective tax rates based on the positive profit sample. However, when considering the overall magnitude of profits to be taxed, one can argue for using either sample. The positive profit sample will provide an upper bound, because some companies will be able to use losses to offset tax burdens. The full sample will provide a lower bound, because companies likely have some persistence in their tendency to earn profits or losses, and thus one would expect the true corporate tax base to lie somewhere between those samples.
| Full Sample | Positive Profit Sample | |||||
---|---|---|---|---|---|---|---|
| Profit Rank | Profit, in billions USD | Effective Tax Rate | Profit Rank | Profit, in billions USD | Effective Tax Rate | |
Cayman Islands | 1 | 58.5 | 0.1% | 3 | 62.4 | 0.1% | |
Singapore | 2 | 54.6 | 4.8% | 6 | 56.8 | 4.5% | |
Switzerland | 3 | 49.4 | 7.1% | 5 | 59.2 | 5.7% | |
Netherlands | 4 | 40 | 10.1% | 2 | 70 | 4.9% | |
Puerto Rico | 5 | 34.3 | 1.6% | 9 | 35.2 | 1.4% | |
Bermuda | 6 | 32.5 | 2.7% | 8 | 35.4 | 2.4% | |
Canada | 7 | 31.7 | 20.5% | 7 | 40.1 | 15.8% | |
Ireland | 8 | 29.5 | 17.6% | 10 | 34.2 | 12.8% | |
China | 9 | 26.8 | 27.4% | 11 | 28.5 | 24.5% | |
Japan | 10 | 24.9 | 20.8% | 12 | 25.5 | 20.1% | |
Luxembourg | 11 | 24.9 | 5.1% | 4 | 60.4 | 0.7% | |
U.K. | 12 | 18.1 | 51.6% | 1 | 81.7 | 10.5% | |
Mexico | 13 | 15.6 | 35.8% | Country data not available | |||
Australia | 14 | 14.8 | 21.2% | 14 | 18.1 | 15.7% | |
Hong Kong | 15 | 12.3 | 11% | 18 | 13.6 | 9.4% | |
India | 16 | 11.8 | 41.9% | 17 | 13.7 | 33.9% | |
Jersey | 17 | 11.7 | 1.1% | 16 | 14.2 | 0.8% | |
Isle of Man | 18 | 7.4 | 0% | Country data not available | |||
Italy | 19 | 7.1 | 27% | 21 | 7.5 | 25.1% | |
Germany | 20 | 6.8 | 69.5% | 13 | 19.8 | 21.9% | |
| Nicaragua (15), Brazil (19), and France (20) are not top 20 in full sample | ||||||
Top 20 Profit | 512.8 | 676.4 | |||||
11 Haven Profit | 355.1 | 441.5 | |||||
All Foreign Profit | 638.5 | 873.6 | |||||
Haven Share | 55.6% | 50.5% |
Table 3 compares the two series, showing the top 20 profit economies (using the full sample). It also reports totals for foreign profits, the top economies’ profits, and the tax haven subset of those economies. For each economy and sample, profits, calculated effective tax rates, and profit ranks are shown.
Comparing the series, several noteworthy patterns are clear. First, effective tax rates are systematically lower, and more correct, in the positive profit sample. That is only logical in aggregate data, because taxes are paid by companies with profits, and companies with losses will distort effective tax rates (tax paid relative to income) down for the full sample. Thus, effective tax rates are more accurate in the positive profit sample. For some haven countries, that makes little difference, because few companies with losses operate in them. For others, such as Ireland, the Netherlands, and Luxembourg, the difference is far more substantial. There are also enormous differences in effective tax rates for some non-haven countries, including the United Kingdom (11 percent for companies with positive profits; 56 percent in the full sample) and Germany (22 percent for companies with positive profits; 70 percent for the full sample). Indeed, effective tax rates in the full sample data are too high.
Second, profits are systematically higher in the positive profit sample, for both important high-tax countries and some low-tax countries. For many haven countries, profits are similar in the two samples, although there are some exceptions, such as the Netherlands and Luxembourg. Non-haven countries with large differences include the United Kingdom and Germany.
Third, in both samples, tax havens are important, with the most consequential 11 havens accounting for 51 percent of all foreign profits in the positive profit sample, and 56 percent of all foreign profits in the full sample. While havens have a higher share of all profits in the full sample (because of the greater importance of companies with losses in non-haven countries), they have higher total profits in the full sample, with about $85 billion more reported profits in the most consequential havens in 2017.
Finally, overall, the total amount of foreign profits is about $235 billion higher in the positive profit sample.
As shown in the following sections, those differences have implications for estimates of the magnitude of profit shifting, as well as the likely revenue benefits from measures that reduce profit shifting.
Lesson 4: U.S. MNEs’ Profit Shifting Large in 2017
The data reviewed in the prior sections indicate that profit shifting was undoubtedly a large problem in 2017. U.S. multinationals booked $355 billion in just 11 havens — $441 billion if using the positive profit sample. Even ignoring less important havens and all profit shifting among non-havens, the magnitudes are large. If 90 percent of those haven profits truly belong in some other high-tax country, that alone would imply that between $320 billion and $397 billion is misallocated. For U.S. MNEs, about two-thirds of their real activity (sales, employment, payroll, and assets) is in the United States, with the other one-third located abroad.17 Therefore, if two-thirds of the haven income should have been taxable in the United States, that would imply a U.S. corporate tax base between $213 billion and $265 billion higher. At the 35 percent statutory rate in effect in 2017, that would imply a loss of revenue of between $75 billion and $93 billion.
Of course, there are more sophisticated ways of assessing the revenue loss from profit shifting, but even those back-of-the-envelope measures indicate large quantities. A prior study undertakes three more sophisticated measures of profit shifting.18 In the first, all haven affiliates are assigned the worldwide average profit/employee ratio, with havens defined as those economies with tax rates under 10 percent. (Ireland is excluded.) Excess profits in havens are then reassigned to the United States and other foreign countries at the two-thirds to one-third ratio.
For example, using the CbC full sample series, profit per worker in Singapore is $337,000. If there were instead $50,500 of profit per worker (the average across all foreign countries), that would imply about $46 billion less earned in Singapore.19 Two-thirds of the excess Singapore income is then assigned to the U.S. tax base. That method generates a U.S. revenue loss of $64 billion using the full sample and $93 billion using the positive profit sample.
The second and third techniques use regression analysis to estimate the tax sensitivity of foreign profits, controlling for the scale of foreign operations (measured by assets, employment, and employee compensation) and country-specific factors (captured by country-specific fixed effects).20 The tax sensitivity is then removed, resulting in fewer profits in low-tax countries. Any reduction in profits in low-tax countries is capped at existing profit, so overall profits are unchanged and then reallocated to higher-tax countries by formula. That technique generates U.S. revenue losses of between $64 billion (using the full sample) and $99 billion (using the positive profit sample).
Using the same method, I also allow for nonlinear elasticities, which are important to include, as persuasively argued in one 2017 study21 and as abundantly clear from Figure 2. Nonlinear elasticities fit the data better than linear elasticities because of the clearly disproportionate clustering of profit in countries with the lowest tax rates. Accounting for nonlinearity raises the revenue loss to $73 billion (for the full sample) or $115 billion (for the positive profit sample).
A forthcoming article discusses those estimates in greater detail and provides companion estimates using other data series for comparison.22 What is clear, with any data series, is the large U.S. revenue losses caused by the profit shifting of U.S. multinationals. Even the lowest possible revenue losses exceed 20 percent of U.S. corporate tax revenues, and the highest numbers are more than twice that size.
The assumption that two-thirds of the excess profits in haven countries belong to the United States (absent profit shifting) naturally implies that other non-haven countries are losing about half the U.S. revenue loss because of profit shifting of U.S. MNEs out of their tax bases, although foreign tax rates differ. Thus, efforts to stem U.S. MNE profit shifting will partially benefit other foreign countries. Likewise, foreign countries’ efforts to stem profit shifting will also benefit the United States, because U.S. multinationals are not the only companies that shift profits.
Indeed, foreign-headquartered multinational companies surely shift profit out of the U.S. tax base, and foreign MNE operations in the United States are sizable. In 2019 the most recent year with available data, the profits of foreign multinational companies operating in the United States were about 42 percent the size of U.S. MNE profits abroad. (A similar fraction is found in 2017 and 2018.23) Therefore, if we assume that foreign multinational companies are equally aggressive in profit shifting, that implies that the total U.S. revenue loss resulting from profit shifting by both U.S. and foreign MNEs would be 1.42 times the above estimates, ranging from $91 billion to $163 billion. Opinions differ regarding the relative tax aggressiveness of foreign multinationals, but it is certainly not the case that U.S. multinational companies are the only one shifting profits.
Table 4 summarizes those estimates. There is a wide range of possibilities, depending on the data set and method used.24
| Full Country-by-Country Sample | Positive Profit Country-by-Country Sample |
---|---|---|
1. Back of Envelope: Remove 90% of haven income from 11 prominent havens | Excess Haven Tax Base: $320 billion Implied U.S. Revenue Loss: $75 billion | Excess Haven Tax Base: $393 billion Implied U.S. Revenue Loss: $93 billion |
2. Assign all havens the world average profit/employee ratio | Excess Haven Tax Base: $274 billion Implied U.S. Revenue Loss: $64 billion | Excess Haven Tax Base: $399 billion Implied U.S. Revenue Loss: $93 billion |
3. Remove tax elasticity; reallocate profits, linear elasticity | Implied U.S. Revenue Loss: $64 billion | Implied U.S. Revenue Loss: $99 billion |
4. Remove tax elasticity; reallocate profits, nonlinear elasticity | Implied U.S. Revenue Loss: $73 billion | Implied U.S. Revenue Loss: $115 billion |
(all above) Add Foreign MNC Profit Shifting | Scale up by 42 percent (less if foreign MNCs are less tax-sensitive) | Scale up by 42 percent (less if foreign MNCs are less tax-sensitive) |
Note: This table shows the U.S. revenue costs due to the profit shifting of U.S. MNCs, assuming that two-thirds of the excess haven income belongs in the United States tax base (absent profit shifting), and the remainder belongs in other non-haven countries. I assume shifted profit would have been taxed at pre-TCJA tax rates of either the U.S. statutory rate (rows 1 and 2) or 30 percent (rows 3 and 4). |
What is the best estimate of U.S. revenue lost to U.S. multinational company profit shifting? It depends on one’s judgements about the persistence of company losses, as well as the techniques above. Still, it likely exceeds $64 billion. Worldwide, adding in revenue losses for foreign non-haven economies, the total revenue loss resulting from U.S. MNE profit shifting is about 50 percent larger than that (although foreign tax rates may differ). For the United States, the total cost of profit shifting also includes that of foreign-headquartered multinationals, and thus likely exceeded $100 billion per year by 2017. In comparison, U.S. corporate tax revenues for fiscal 2017 were about $297 billion, using Congressional Budget Office data.
Lesson 5: Policy Interventions to Raise Revenue
The large magnitude of revenue lost from profit shifting implies large amounts of revenue to be gained from measures that weaken profit-shifting incentives. One such option is a minimum tax. Since passage of the Tax Cuts and Jobs Act, the United States has implemented a weak minimum tax on U.S. resident multinational companies, known as GILTI (for global intangible low-taxed income).
Several features make GILTI weak, and the Joint Committee on Taxation estimated that it would raise only about $9 billion per year from 2021 to 2025. First, the minimum tax applies only to some income. A 10 percent return on assets is excluded. Second, a 50 percent deduction is allowed, reducing the headline rate on GILTI to half the normal U.S. corporate tax rate.25
Third, and importantly, the minimum tax is based on average global tax burdens, so payments to higher-tax countries can offset minimum tax due on haven income. That feature perversely makes the United States the least desirable place to book income for companies paying GILTI tax. If income is earned in a zero-tax rate haven, it is taxed at half the U.S. rate. If income is earned in a higher-tax location, it generates tax credits that can offset the GILTI tax due on haven income. But if the income is earned in the United States, it is taxed at the full U.S. rate without the offsetting benefit of lighter minimum tax burdens.26
However, it is straightforward to design a less feeble minimum tax. All foreign income can be included, it can apply on a country-by-country basis, and it can be levied at a higher rate. Indeed, all three changes would make for a far more effective response to the problem of corporate tax base erosion from profit shifting.27
While companies will argue that competitiveness concerns reduce the merits of strong minimum taxes, those concerns can be countered through coordination with other countries and the inclusion of tough anti-inversion measures.28
Table 5 shows that a country-by-country minimum tax at 21 percent can raise substantial revenue — between $570 billion and $910 billion over the next 10 years, if foreign profits grow at a rate of 4 percent. Those revenues are far in excess of JCT projections for the current minimum tax.
| Full Sample | Positive Profit Sample | ||||
---|---|---|---|---|---|---|
Economy | Profit in 2017 | Effective Tax Rate | Implied Minimum Tax Revenue | Profit in 2017 | Effective Tax Rate | Implied Minimum Tax Revenue |
Barbados | 6.2 | 0.1% | 1.3 | Country data not available | ||
Bermuda | 32.5 | 2.7% | 5.9 | 35.4 | 2.4% | 6.6 |
Cayman Islands | 58.5 | 0.1% | 12.2 | 62.4 | 0.1% | 13 |
Puerto Rico | 34.3 | 1.6% | 6.7 | 35.2 | 1.4% | 6.9 |
Hong Kong | 12.3 | 11% | 1.2 | 13.6 | 9.4% | 1.6 |
Singapore | 54.6 | 4.8% | 8.8 | 56.8 | 4.5% | 9.4 |
Gibraltar | 5.1 | 0.3% | 1 | Country data not available | ||
Ireland | 29.5 | 17.6% | 1 | 34.2 | 12.8% | 2.8 |
Isle of Man | 7.4 | 0% | 1.6 | Country data not available | ||
Jersey | 11.7 | 1.1% | 2.3 | 14.2 | 0.8% | 2.9 |
Luxembourg | 24.9 | 5.1% | 4 | 60.4 | 0.7% | 12.2 |
Netherlands | 40 | 10.1% | 4.3 | 70 | 4.9% | 11.2 |
Switzerland | 49.4 | 7.1% | 6.9 | 59.2 | 5.7% | 9.7 |
All Economies | 638.5 |
| 60.8 | 873.6 |
| 97 |
U.S. Revenue (2/3 of total) | 40.5 |
|
| 64.6 | ||
Implied U.S. Revenue, 2021-2030 | 569.4 |
|
| 907.5 | ||
Less Expected GILTI revenue | -137 |
|
| -137 | ||
Net Additional U.S. Revenue, 2021-2030 | 432.4 |
|
| 770.5 |
Unsurprisingly, given the above analysis, estimates of the amount of revenue raised depend on the sample used. Using the full sample, revenues are smaller, in part because of the overestimation of the effective tax rate when companies with losses are included in the totals, and in part because of the lower total tax base.
In the full sample, there are 13 jurisdictions that raise more than $1 billion in minimum tax revenue. In the positive profit sample, many of the smaller havens show similar amounts of minimum tax revenue as in the full sample analysis, but in some cases, such as the Netherlands and Luxembourg, the estimates are different.
There are also some countries, such as the United Kingdom, that appear to have tax rates well above the minimum in the full sample (51.6 percent) because of the presence of companies with losses. But once those companies are excluded, the positive profit U.K. affiliates pay only 10.5 percent in tax, and therefore would generate $8.5 billion in minimum tax revenue.
One additional estimation option is to use the positive profit sample to calculate the effective tax rates, as seems only sensible, but use the full sample to calculate the tax base (under the presumption that companies with losses will reduce the size of the tax base). That hybrid technique implies $48 billion in U.S. revenue using 2017 data, or $674 billion from 2021 to 2030. That would be my preferred revenue estimate. It errs on the side of caution by assuming a lower tax base but avoids the biased effective tax rates from the full sample.
These estimates are meant to capture both revenue actually collected under a minimum tax and reduced profit shifting resulting from the tax. For example, if a company decides that the minimum tax makes profit shifting less lucrative and stops shifting profit to havens, that will lead to more corporate tax base revenue in non-haven countries, but the revenue will not show up as minimum tax revenue.
Similarly, if Bermuda, for example, raises its tax rate to 15 percent, that will reduce minimum tax revenues at the same time that it reduces profit shifting to Bermuda, thus raising the corporate tax base in other countries accordingly. That tendency of minimum taxes to encourage higher tax rates abroad is a feature (not a bug) of the policy because it should work to counter the race to the bottom in corporate rates. Indeed, the pressures of tax competition and corporate base erosion have led to steadily decreasing corporate rates in recent decades. In 1985 the average statutory tax rate among OECD countries was 43 percent; by 2000 it was 30 percent; and in 2019 it was 21.7 percent.29
Should Public CbC Disclosure Be Mandatory?
Very few companies release public CbC reporting information. When a research assistant combed through the first 900 entries in the Forbes Global 2000 list of largest companies in the summer of 2020, only 20 companies had released what looked like CbC reports. Most were finance companies based in the United Kingdom (with a few elsewhere in Europe) that have similar reporting requirements within the scope of the capital requirements directive (2013/36/EU).
Beyond European finance company disclosures, only seven companies had released detailed CbC data. Six are in extractive industries and are known for having higher worldwide tax burdens because of the importance of local operations in the countries of extraction. The remaining company is a U.K. telecommunications company. Their statistics follow:
Anglo-American shows a worldwide effective tax rate of 22.5 percent, with profit per employee of $103,000. It reports activities in seven of the Figure 1 havens, with an average effective tax rate in those havens of 4.1 percent and average profit per employee of $2.3 million.
Shell shows a worldwide effective rate of 29.7 percent and profit per employee of $408,000. It operates in 10 of the Figure 1 havens, with an average effective tax rate in those havens of 5.6 percent and average profit per employee of $610,000.
Rio Tinto shows a global effective rate of 19.5 percent, with profit per employee of $401,000. It operates in seven of the Figure 1 havens, with an average effective rate in those havens of 7 percent, and average profit per employee of more than $1 million.
Eni shows a worldwide effective tax rate of 106 percent and an average profit per employee of only €120. Although it operates in six of the Figure 1 havens, it earns a loss on its haven income.
Repsol shows a global effective rate of 54.2 percent, with profit per employee of €93,000. It operates in six of the Figure 1 havens, with an average effective tax rate in those havens of 19.6 percent, and average profit per employee of more than €1.5 million.
Vodafone made losses both worldwide and in the havens in which it operates.
Clearly, those companies are not representative of typical multinationals. Extractive industries are more heavily taxed, and Vodafone did not make a profit.
Even for the companies that have voluntarily released CbC data, they often operate in prominent tax havens, and there is evidence indicative of profit shifting. While some haven affiliates show losses or below-average profits, overall, profit per employee tends to be substantially higher than the worldwide average, while effective tax rates are substantially lower.
However, in general, CbC data have the potential to provide useful information to stakeholders on the distribution of a company’s basic economic indicators. Indeed, as I have argued elsewhere, there is a strong argument for public disclosure of CbC reporting data, similar to the tables that are reported in aggregate form by the IRS Statistics on Income division.30
Public disclosure can provide useful information to potential customers, investors, and employees, as well as community members. By requiring publicly available information in a standardized format, governments can ensure that stakeholders have a complete and representative picture. We rely on investigative journalism to uncover particularly glaring examples of tax avoidance, and the companies in the spotlight are more likely to be household names. With public disclosure, all companies will be treated similarly. If the public, including a company’s potential customers, investors, and employees, value socially responsible corporate behavior, companies may pay a reputational penalty if they are too aggressive in their international tax avoidance. That will act as an incentive for managers to consider the stakeholders in their communities, instead of focusing solely on after-tax profits.
Reputational motives may cause some companies to be less aggressive in their tax avoidance efforts, or even less interested in advocating for loopholes that allow effective tax rates to drop far below the headline corporate rates that are more familiar to the public. Indeed, improved transparency can better align the reality of our tax system with how it is advertised. For example, pre-TCJA, the U.S. government supposedly had a worldwide tax system that taxed both foreign and domestic income at 35 percent. However, as international tax experts know well, the reality was quite different. Little foreign income was taxed by the U.S. government, and companies paid effective tax rates far below the statutory rate.
Companies may argue against public disclosure by raising fears that giving away their CbC information is akin to giving away secret intellectual property, because competitor companies will learn valuable operational information. Those claims are both self-serving and exaggerated. However, if a company’s nontransparent tax-minimization strategies really do give it a competitive advantage in the marketplace, that itself is a problem that the public and policymakers should fully understand.31
If international corporate taxation were substantially reformed in a manner that weakened the incentives driving international profit shifting, disclosure would become less necessary. However, in the meantime, shedding sunlight on corporate tax avoidance will help build the political will to address the problem. Good public information may stiffen policymakers’ resolve in the face of powerful interests that advocate for light tax treatment of mobile multinational income.
Conclusion
The new CbC reporting data provide useful information that other sources lack, including a large expansion of country coverage and important data on the subset of companies earning positive profits. While the data are not perfect, and there have been some difficulties implementing reporting requirements, in the coming years, the data should provide comparable, consistent information on key variables for numerous countries hosting multinational companies.
This article takes a careful look at the release of the U.S. CbC data for 2017, which is more complete than the 2016 release. In my analysis of those data, I exclude income reported as stateless. The remaining aggregate total is comparable to totals reported by other reliable data sources.
Early analysis of the data suggests five important lessons for scholars of international taxation. First, tax havens play a dominant role in the profit shifting of U.S. multinationals. Of $4.2 trillion in accumulated foreign earnings, $3 trillion is in tax havens. In 2017 just 11 tax havens accounted for 56 percent of U.S. MNE foreign profits.
Second, while havens play an important role in reported profits, they are far less important in terms of real economic activity. As a consequence, profit ratios (profits per employee) are far higher in haven countries. Still, it is useful to distinguish two types of tax havens: those that have little economic activity and those that host substantial employment. For low-activity havens, profits per employee are typically more than two orders of magnitude higher than those in typical foreign countries. For higher-activity havens, profits per employee are between 2.8 and 12 times higher than in typical foreign countries. In the data, there is a highly nonlinear (negative) relationship between reported profits and tax rates.
Third, companies with losses distort data on country effective tax rates. Many commonly used data series aggregate companies with losses and those with profits, increasing measures of foreign effective rates because taxes are typically paid solely by companies with profits, whereas companies with losses reduce the effective rate denominator. Likewise, the total amount of foreign taxable income will be understated in those series. Those considerations are important for estimates of the magnitude of profit shifting and the possible revenue gains from minimum taxation.
Fourth, the 2017 CbC data indicate that profit shifting is a large problem, likely costing the U.S. government more than $100 billion in forgone revenue by 2017 (at 2017 tax rates). The precise estimate of revenue loss is ambiguous because of uncertainties about method and the ideal data series, but all estimates are quite large. Fifth, stronger minimum taxes therefore have large revenue potential, raising more than $530 billion from 2021 to 2030, beyond the revenue from the current minimum tax regime.32 Stronger minimum taxation can also help stem tax competition pressures.
A final and more minor lesson is that there is little to be learned from companies’ voluntary disclosures of their CbC data. Few companies have disclosed, and those that have are not representative. However, there is a strong case for requiring public disclosure.
As I have argued,33 global flows of trade and investment have the potential to raise living standards worldwide, but it is also essential to build tax systems that can work despite the mobility of capital. For our modern global economy to benefit everyone, we need a tax system that can ask winners to pay reasonable taxes, while lightening burdens on those facing disruption. Improving the taxation of cross-border multinational income is an essential step in that direction, and toward that end, the CbC data provide important lessons about possible policy responses.
Appendix
Table 6 compares the CbC data series with two series produced by the BEA that are known to exclude double counting. The BEA adjusted series begins with net income, adds back foreign taxes, and subtracts equity income. The direct investment income series is provided posttax but is also calculated pretax. While those two series match well for 2017, they are further apart in other recent years. Both BEA series include companies with losses in the aggregate totals.
| BEA Adjusted Method | BEA Balance of Payments Direct Investment Income | IRS Country-by-Country Data | |||
---|---|---|---|---|---|---|
| Net Income + Foreign Tax-Equity Income | After Tax | Before Tax | Full Sample | Positive Profit | Accumulated Earnings |
All countriesb | 571,007 | 470,933 | 574,958 | 638,467 | 873,621 | 4,240,635 |
Stateless (omitted from totals and subtotals) | 203,571 | 215,170 | 690,583 | |||
Puerto Rico |
|
|
| 34,335 | 35,236 | 114,439 |
Ireland | 82,519 | 51,804 | 55,930 | 29,478 | 34,221 | 103,961 |
Luxembourg | 6,484 | 36,825 | 38,734 | 24,866 | 60,438 | 357,328 |
Netherlands | 58,676 | 76,083 | 81,120 | 40,010 | 69,964 | 461,814 |
Switzerland | 37,696 | 30,474 | 34,332 | 49,376 | 59,204 | 374,797 |
Bermuda | -10,431 | 32,341 | 33,215 | 32,476 | 35,433 | 634,413 |
U.K. Caymansc | 20,675 | 33,235 | 33,888 | 58,540 | 62,369 | 142,467 |
Singapore | 35,270 | 24,496 | 27,529 | 54,642 | 56,788 | 174,888 |
Big Haven Total | 230,889 | 285,258 | 304,748 | 323,723 | 413,653 | 2,364,107 |
Big Haven Share | 40% |
| 53% | 51% | 47% | 56% |
Note: Big havens include only the specific havens listed above, although the CbC data reveal many other small havens. For example, in the 2017 data, Jersey emerges as a big haven with $461 billion in accumulated profits; however, Jersey is not included as a big haven in this table. aThis calculation adds back foreign taxes paid from the income statement to the direct investment earnings series. There may be imperfect country matching if direct investment income is distributed across countries differently from net income, but it gives plausible relative magnitudes, especially for the totals. bThis total excludes stateless income. cBEA data lists as “U.K. Caribbean Islands,” but other sources list as “Caymans.” |
FOOTNOTES
1 Sean Foley et al., “The State of Country-by-Country Reporting,” Tax Notes Int’l, Aug. 31, 2020, p. 1163.
2 The final regulations implementing CbC reporting (T.D. 9773) indicate that “the information reported on the [CbC report] will be used for high-level transfer pricing risk identification and assessment, and that transfer pricing adjustments will not be made solely on the basis of a [CbC report], but that the report may be the basis for further inquiries into transfer pricing practices or other tax matters which may lead to adjustments.” See also section 5 of Draft U.S. Model CbC CAA (on the Basis of a Double Tax Convention) (June 30, 2017).
3 OECD, “Transfer Pricing Documentation and Country-by-Country Reporting, Action 13 — 2015 Final Report” (2015).
4 See, e.g., Michelle Hanlon, “Country-by-Country Reporting and the International Allocation of Taxing Rights,” 72(4/5) Bull. Int’l Tax’n (Mar. 22, 2018).
5 In the 2016 data, 1,101 MNEs reported $1.4 trillion in worldwide profit. In the 2017 data, 1,575 MNEs report $2 trillion in profit.
6 OECD, “Guidance on the Implementation of Country-by-Country Reporting — BEPS Action 13” (2019).
7 For more on the stateless income problem, see Edward D. Kleinbard, “The Lessons of Stateless Income,” 65(1) Tax L. Rev. 99 (2011).
8 See supra note 2.
9 Thomas Horst and Alex Curatolo, “Assessing the Double Count of Pretax Profit in the IRS Summary of CbC Data for Fiscal 2017,” Tax Notes Int’l, Apr. 27, 2020, p. 427. Horst has done recent calculations that show the U.S. share of profit in the CbC data is higher than the U.S. share of profit in financial reporting data, a pattern that suggests the U.S. income is more likely to be overstated in the CbC data.
10 This analysis uses a data-driven definition of tax havens, focusing on those jurisdictions with low effective tax rates, typically using a cutoff of 10 percent. Other organizations and scholars have used other definitions, including analyses of countries’ secrecy laws.
11 The $3 trillion figure includes a broader group of smaller havens beyond those including Isle of Man, Gibraltar, Macau, St. Lucia, St. Kitts and Nevis, Barbados, and Mauritius.
12 Puerto Rico had a lower GDP per capita of $31,000 in 2017. All the other listed economies had per capita GDP between $49,000 and $108,000 in 2017. Of course, those statistics are distorted by international tax avoidance.
13 Again, those data are not intended to include lower-tier income either through ownership or related-party dividend payments.
14 The Isle of Man and Mexico are not included in that sample, so the full sample is used for those economies.
15 Without Ireland as a haven, 10 havens have 174 percent the profit of the 10 non-havens. Haven tax rates average 3 percent, and non-havens average 24.3 percent. The population of havens is 1.3 percent that of non-havens, and the GDP of havens is 7.3 percent that of non-havens. The 10 havens account for 53 percent of U.S. multinational foreign profits.
16 Including the other tiny jurisdictions with an effective tax rate under 10 percent would add $22 billion to that total, bringing the haven share to 59 percent.
17 I use several different data series to assess the U.S. share of real activity, but they all generate a roughly two-thirds share. For the 2017 U.S. CbC data, the average of the U.S. share of employees, assets, and sales is 68.8 percent, excluding stateless income from the denominator. In the U.S. BEA data for 2017, the average of employee and sales shares is 67.1 percent, and the average including employment compensation is 70.9 percent.
18 Kimberly A. Clausing, “Profit Shifting Before and After the Tax Cuts and Jobs Act,” Nat’l Tax J. (forthcoming Dec. 2020).
19 Careful readers will note that $50,500 is higher than the $48,750 reported for the world in Table 2. The first number includes the United States, while the second considers just the average among foreign jurisdictions. Both calculations exclude stateless income.
20 I use a different data series to estimate elasticity because the CbC data set does not allow time variation. I chose a benchmark elasticity of 3, slightly below that of the preferred specification (3.2), which regresses the natural log of direct investment income on the effective tax rate, log employment, log employee compensation, and log assets, controlling for country-specific fixed effects. That elasticity is a simple benchmark, and readers can scale the estimate up and down accordingly with simple multiplication. Any elasticity from a pooled or cross-section analysis of any one of four data sets discussed in Clausing, supra note 18, would be higher. For the elasticity methods, in line with prior work, I assume a 30 percent effective U.S. tax rate, allowing some base narrowing relative to the statutory rate. Using the statutory rate instead would increase the estimates for two reasons: There would be a higher rate on the reassigned income, and the calculated amount of excess profits abroad changes because of the larger discrepancy between the U.S. and foreign rate.
21 In “Profit Shifting of U.S. Multinationals,” 148 J. Pub. Econ. 1 (2017), Tim Dowd, Paul Landefeld, and Anne Moore argue that tax responsiveness is nonlinear, so that elasticities are highest for haven countries. They use U.S. tax data, which is excellent for studying that question, finding large elasticities for haven data.
Indeed, in my estimations, nonlinear elasticities typically fit the data better than linear ones. That makes intuitive sense. When shifting profits, it is most advantageous to achieve the lowest tax rate possible. For the estimates of Table 4, I use synthetic nonlinear elasticities that are similar to those reported by Dowd, Landefeld, and Moore, but substantially smaller than those estimated by those data series.
22 Clausing, supra note 18.
23 The after-tax ratio is 39 percent, but because those numbers are after tax, adjusting for tax rate differences suggests a pretax ratio of about 42 percent. That uses the income without current cost adjustment series from the BEA, “U.S. Direct Investment Abroad: Balance of Payments and Direct Investment Position Data,” for U.S. MNE income abroad, and a parallel series, BEA, “Foreign Direct Investment in the U.S.: Balance of Payments and Direct Investment Position Data,” for foreign MNE income.
24 Further, if the CbC data overstate foreign income totals beyond the stateless income line (which is excluded), that may mean that those numbers could be as much as 14 percent overstated. However, there are other possible reasons for the discrepancies analyzed by Horst and Curatolo, supra note 9. Appendix A indicates that other sources of data (without double counting) show smaller differences with the CbC totals once stateless income is excluded. Finally, recent analysis by Horst indicates that the U.S. income line is more likely to be overstated than the foreign income lines.
25 Many subtleties affect the true tax rate. For example, because foreign tax rates are only 80 percent creditable, that can raise the GILTI rate from 10.5 percent to 13.125 percent. Also, interactions with other foreign tax provisions can affect the tax’s true burden.
26 For more on the incentives created by GILTI and its effects on profit shifting, see Clausing, supra note 18.
27 Another option that might be administratively simple is to implement a mandatory high-tax kickout, excluding averaging with higher-tax countries.
28 For further discussion, see Clausing, Emmanuel Saez, and Gabriel Zucman, “Ending Corporate Tax Avoidance and Tax Competition: A Plan to Collect the Tax Deficit of Multinationals” (July 19, 2020).
29 For additional detail on the arguments for minimum taxation, see id.
30 For a more detailed argument about the importance of public disclosure, see Clausing, Open: The Progressive Case for Free Trade, Immigration, and Global Capital, ch. 11 (2019).
31 It would also be useful for financial reporting to include more specific information about the tax payments associated with particular international tax provisions, including GILTI, the base erosion antiabuse tax, and any future minimum taxes.
32 My preferred estimate of $674 billion is reported under Lesson 5, supra; $137 billion is my forecast of GILTI revenue for the same period.
33 Clausing, supra note 30.
END FOOTNOTES