Winners and Losers: The OECD’s Economic Impact Assessment of Pillar One

A LEAP OF FAITH
The economists who prepared the Economic Impact Assessment (EIA) of the Pillar One Blueprint and Pillar Two Blueprint were assigned a herculean task, made much more difficult by the lack of data to answer the research question they were posed and by the large number of unsettled policy decisions. In an earlier article, [Link to TPR] I explain the general difficulties faced by the EIA and assess its results.

In this follow-up, I explore the EIA’s estimates of Amount A, the amount of global pre-tax profit of MNEs that is to be reallocated to the so-called Market jurisdictions under the Pillar One Blueprint. I show how Amount A is calculated and reallocated among Market, Residence, and Source jurisdictions, and explore some of the complexity and data problems involved in the estimates. I highlight the key roles in the Amount A formula played by Global In-Scope Destination-based Sales (component C) and Global Residual In-scope Profits (component E), and then use estimates of the gap between components C and E to provide more fine-grained estimates of the jurisdictional impacts of Amount A than are available in the EIA. While the EIA predicts that Amount A will have only a small net global impact, which may suggest that Pillar One is of little importance, I find that at the jurisdictional level the impacts on tax revenues can be surprisingly large. My results show that middle-income jurisdictions, notably in East Asia and the Pacific, are the likely winners from Amount A; investment hubs, especially the small Caribbean island tax havens, suffer the largest losses.

THE CALCULATION OF PILLAR ONE AMOUNT A

The best way to understand the Amount A calculations is to start with its formula as visualized in Figure 1, which I have adapted from Figure 2.1 in the EIA (p. 29). Figure 1 explains how the net tax revenue each jurisdiction would receive from implementing the Amount A proposal is calculated.

Figure 1: Calculating Pillar One Amount A

The formula for Amount A is a (deceptively) simple calculation. For each jurisdiction J, the net tax revenue gain or loss is calculated as:

Jurisdiction J’s Net Revenue Tax Gain/Loss = A x B x (C x D – E x F) (1)

In equation (1), components A and B in the formula are global numbers that are identical for all tax jurisdictions. C, D, E, and F are jurisdiction-specific variables that vary for each jurisdiction depending on its roles as a Market jurisdiction (C x D) and as a Residence and/or Source jurisdiction (E x F). I explore the six components of Amount A below in more detail.

• A is Global Residual In-scope Profit (GRIP); that is, the amount of global residual profit before tax for in-scope businesses, i.e., for MNEs in particular sectors and above a specified size. MNEs with their primary activity in the consumer-facing businesses (CFB) and automated digital services (ADS) sectors and global revenues greater than $911 million (750 million euros) are defined as In-scope. The EIA ( pp. 61–62) assumes a residual profitability threshold (measured by profit before tax divided by turnover) of either 10% or 20%; above that threshold profits are deemed “residual” and available for reallocation. GRIP is therefore the potential amount of MNE global profit to be reallocated from Residence and Source jurisdictions to Market jurisdictions.
• B, the reallocation percentage, is a fixed percent to be set by the OECD/G20. The EIA (p. 37) assumes the percent could be 10%, 20%, or 30%.

Component A, when multiplied by component B, determines the amount of global residual in-scope profit reallocation; i.e., the actual amount of MNE global profit to be reallocated from Residence and Source jurisdictions to Market jurisdictions. All the OECD documents refer to the actual amount of reallocated GRIP as “Amount A,” the total profit pool for redistribution under Pillar One.

The last four components in equation (1) are specific to each jurisdiction, and determine that jurisdiction’s gross and net shares of Amount A. Components C and D are used to calculate the tax share of Amount A that is reallocated to a particular Market Jurisdiction (a “receiving” or “gaining” jurisdiction). Components E and F are used to calculate the tax share of Amount A that is reallocated away from (lost by) a particular Residence or Source Jurisdiction (a “relieving” or “losing” jurisdiction). The calculations can be summarized as follows:

• C is a “tax receiving” allocation key calculated by dividing the amount of In-Scope Destination-based Sales that take place in each Market jurisdiction by the amount of global in-scope destination-based Sales (GIDS) for all Market jurisdictions. Each allocation key is therefore specific to each Market jurisdiction and directly tied to its jurisdictional share of GIDS. As its allocation key (share of GIDS) rises, the jurisdiction gains more tax base, and vice versa. Component C therefore determines the jurisdiction’s gain in tax base under Pillar One.
• D is the corporate income tax (CIT) rate that each Market jurisdiction chooses to levy on its received share of Amount A, the global profit reallocation. The higher the tax rate, the more tax revenue generated on the Market jurisdiction’s received share of Amount A.
• E is a “tax relieving” allocation key calculated as each jurisdiction’s share of MNE GRIP currently booked in this jurisdiction under the existing international tax and transfer pricing rules. Allocation key E is specific to each jurisdiction and directly tied to its share of GRIP; as its share of GRIP rises, the jurisdiction must provide more tax relief to Market jurisdictions. Component D therefore determines the jurisdiction’s loss in tax base.
• F is the rate of tax relief that this jurisdiction, in its role as a Residence or Source country, is expected to provide to other jurisdictions on its share of Amount A that has been reallocated to other (Market) jurisdictions.

It is possible that components D and F (the two tax rates) may be the same (e.g., both set at the statutory CIT rate). In that situation, jurisdiction J, in its role as a Market jurisdiction, taxes its newly found tax base at the same rate as it provides tax relief, in its role as a Residence or Source jurisdiction, on the tax base it loses to other Market other jurisdictions. The EIA calculations assume that in the typical case components D and F are identical and set at the statutory CIT rate (p. 30). An alternative scenario assumes component F is five points lower than the statutory rate, except for investment hubs where the CIT is set at the minimum effective tax rate (p. 60).

If components D and F are set at the same tax rate, I can rewrite equation (1) in a simpler fashion as equation (2), where D* = D = F:

Jurisdiction J’s Net Tax Revenue Gain/Loss =

[ A x B ] x D* x [ C – E ] = Amount A x J’s tax rate x [J’s GIDS share – J’s GRIP share] (2)

Simplifying the Pillar One formula in this manner yields a straightforward result. Assume that jurisdiction J provides the same rate of tax relief on its “lost” tax base (component E, its share of GRIP) as the jurisdiction applies to its “found” tax base (component C, its share of GIDS) so that D = F. Then, whether jurisdiction J wins or loses net tax revenues under Amount A depends only on the gap between components C and E; the jurisdiction gains if C > E and loses if C < E. The amount of the net gain or loss in tax revenue depends on all the components in equation (2), some of which jurisdiction J may be able to influence, but the C-E gap determines who wins and who loses from Amount A.

This result also suggests a simple extension. If jurisdiction J can set its “tax relieving” rate below its “tax receiving” rate (D > F), the jurisdiction gains net tax revenue, with the maximum gain where F equals zero (no tax relief is provided). Equations (1) and (2) also point to the importance of the three roles that a tax jurisdiction can play in Amount A: Residence (home of the MNE ultimate investor), Source (host to foreign-owned MNE affiliates) and Market (the destination where MNE products are sold). The fourth role—origin (where the MNE products are made)—is ignored by Amount A. I explore these roles below.

THE ‘ELEPHANT IN THE ROOM’: HOW BIG IS THE PROBLEM?

The premise behind Pillar One is that Market jurisdictions are not receiving their “fair share” of MNE profits and are therefore being deprived of tax revenues on this base. What has never been clear to me, and is still not clear from the EIA, is how big is this problem and is it a big enough problem to replace the existing international tax rules with a new taxing right for Market jurisdictions? The various Pillar One documents all assert that there is a problem—but how big is the problem?

A structured way to answer this question in terms of the Pillar One Blueprint would have been to identify—for each jurisdiction where an MNE has GIDS—whether the MNE is already effectively connected to that jurisdiction by having an MNE parent or affiliate in that jurisdiction. If the answer is yes, then the Market jurisdiction already has taxing rights under existing Source and Residence tax principles and does not need a “New Taxing Right.”

If the answer is no, more follow-up information is needed to assess the size of the problem. For example, perhaps the answer is no because the MNE does not have a permanent establishment (PE) in that jurisdiction. That might be remedied by the jurisdiction adopting the broader PE definitions agreed to in first BEPS round. Alternatively, perhaps the answer is no because the MNE provides automated digital services (ADS) such as a Facebook-like platform where foreign consumers receive free access in exchange for personal data that the MNE sells to advertisers. An appropriate solution could also be a broader PE definition (see Pillar One Blueprint, pp. 68–69) such as a group PE test.

The EIA’s failure to at least estimate and publish the gaps—and the overlaps—among Residence, Source, and Market makes it impossible to determine the size of the problem Pillar One is supposed to address. I suspect that the “elephant in the room” is overstated for most jurisdictions for two reasons. First, the EIA estimates are likely overestimated because they are built on 2016 data when MNE tax structures were far more aggressive. The BEPS reforms launched in 2017 have significantly reduced the incentives to engage in tax-abusive structures, which the July 2020 OECD report on BEPS outcomes clearly demonstrates. Tax avoidance loopholes should continue to shrink as more jurisdictions sign onto the Multilateral Instrument and adopt the BEPS proposals. Elsewhere, I have argued that these early results on implementation of the first BEPS round changes are a “good news story” that reduces the rationale for Pillar One.

Second, economists are well aware that the overlap between Residence and Source jurisdictions is very high. OECD member countries are both home (Residence) and host (Source) to MNEs and many have roughly similar inward and outward FDI positions. A table that listed jurisdictions with their three-year average shares of GIDS (Market), GRIP (Residence) and GRIP (Source) would have been useful in assessing how big is the problem that Pillar One was designed to address.

HOW TRUSTWORTHY ARE THE EIA NUMBERS FOR AMOUNT A?

Chapter 5 of the EIA provides a detailed overview of the complex and painstaking process that the EIA economists used to estimate the global and jurisdictional impacts of Amount A. The thoroughness and ingenuity evidenced in the EIA must be highly commended as this was truly a herculean task. Their research was also innovative, as evidenced by the creation of a brand-new “full matrix of FDI by jurisdiction of ultimate investor” (which unfortunately was also not made available; see EIA, p. 224). The importance of country-by-country reporting (CbCR) as a new source of data on MNE profits in investment hubs is also noteworthy.

Whether the EIA results can be trusted probably depends on two core issues. I assume the probability of a third issue—that the empirical methods were flawed in some way—is zero and so ignore that possibility. The details provided in Chapter 5 about how data were carefully matched across the various datasets and the variety of robustness checks provide additional confidence in the quality and reliability of the empirical work. First, hiding behind equation (1) are dozens—maybe hundreds—of earlier choices that are not visible unless you “unpack” the calculations and work through them step by step. The process looks simple to the uninitiated but each step is showing only the proverbial “tip of the iceberg”; at least 90% of the decisions are unobservable, even with the authors providing a code book and examples in Annex A to the Pillar One Blueprint (pp. 212–230).

Second, the EIA is careful to point out the major data problems and their solutions. Perhaps the most important data problem is that a significant percent of the nearly 50,000 observations in the 222 x 222 data matrices—which underlie all the calculations for both the Pillar One and Pillar Two Blueprints—are based on extrapolations from a small number of observations, primarily for high-income jurisdictions. The extrapolations use gravity-type models with macroeconomic data such as FDI, GDP, GDP per capita (EIA , p. 224). The higher the share of extrapolations, the “softer” the data. The EIA notes, for example, on page 243 that the MNE turnover matrix “may be the most reliable of the four matrices” (i.e., profit, turnover, payroll, and tangible assets, see pp. 267–270) used to conduct its assessment of the two Blueprints. The reason is that the turnover matrix involves the least extrapolations to fill in missing values (i.e., only 4% of the turnover matrix cells compared to 27% for profit, 28% for tangible assets, and 36% for payroll; see Table 5.7 on p. 243).

In addition, drilling down to the jurisdiction groupings in Table 5.7, there are wide disparities in the degree of “soft” versus “hard” data across the four groups. Generally, high-income jurisdictions have the least extrapolations. Low-income jurisdictions, on the other hand, have the most: more than 90% of the data are extrapolated for turnover, tangible, assets, and payroll, and nearly 80% for profit. For investment hubs, only the turnover data has high reliability (4% extrapolation); for the other three matrices, more than one-third of the cells were extrapolated.

A related issue is that country-by-country reporting (CbCR) data were used (EIA, p. 243) as the primary data source for three of the matrices (profit, turnover, and tangible assets). While CbCR is clearly going to be an important source of data on MNEs activities going forward, the 2016 reports suffer from several problems:

• The CbCR data are for one year (2016), which was the first year that CbCR reports were collected. They were therefore more likely to have flaws and inconsistencies (p. 232). U.S. data are for 2017 since CbCR reporting was not mandatory in 2016 (p.232).
• CbCR reports were available only for 26 jurisdictions and only 25 were used. The CbCR report from China was omitted from the matrices (see n. 11, p. 278) because data were available for only 82 firms (presumably all state-owned Chinese MNEs).
• The EIA included only CbCR data for MNE sub-groups with positive profits were included; any sub-groups with losses were excluded from the matrices (p. 232).
• Because CbCR data are collected only for firms with global revenues above $911 million 750 million euros, data for small- and medium-sized MNEs are missing from the matrices (p. 233), which may have negatively affected the number of emerging market multinationals in the database.
• CbCR data use a 50% ownership minimum threshold for determining which MNEs belong to a corporate group, rather than the 10% standard used in foreign direct investment (FDI) calculations (p. 234). Using CbCR data can therefore also reduce the number of emerging market multinationals in the database. Korean and Latin American MNEs, for example, tend to use family or conglomerate (chaebol, grupa) structures with lower ownership rates than U.S. and Western European MNEs.

A CLOSER LOOK AT COMPONENTS C AND E

Some examples of the problems faced by the EIA and the solutions that were used are briefly outlined below. I focus on problems with components C and E since they drive the jurisdictional net revenue effects of Amount A (see equation (1)).

Measuring Component C (GIDS): Problems and Solutions

• For consumer-facing businesses (CFB), remote sales (both digital and physical) are not included in the proxy measure of CFB destination-based sales, neither at their point of origin nor destination. The EIA solution was to omit remote sales from the CFB calculations (EIA , p. 40). As a result, CFB destination-based sales are likely underestimated in the calculation of component C; by how much, we do not know.
• Destination-based sales by MNEs at the jurisdiction level are typically measured as MNE turnover minus MNE exports. The calculations should include locally owned MNEs (i.e., ultimate parents and their domestic affiliates) and domestic affiliates (branches and subsidiaries) of foreign-owned MNEs; non-MNEs should be excluded. However, sufficiently fine-grained data for this calculation in the EIA were available only for 16 jurisdictions (presumably all high-income OECD countries). More data, but only for 59 jurisdictions, were available but did not separate MNE from non-MNE entities. The EIA solution was to, first, set the MNE/non-MNE percentage for the other 206 jurisdictions at the average for the 16 jurisdictions (EIA, p. 41). This is problematic because other tests by the EIA suggest that smaller markets have fewer MNEs selling in their jurisdictions and many very small jurisdictions have no MNEs at all (p. 47). Thus, the MNE/non-MNE percentage may have been set too high for low-income jurisdictions; on the other hand, the MNE/non-MNE percentage is likely too low for investment hubs. Second, the economists then regressed logged Sales/GDP against logged values of GDP, GDP per capita and trade openness for the 59 available jurisdictions and used the results of this regression to create proxies for the other 163 jurisdictions. Without more information on the 59 jurisdictions for which data were available, the bias from this regression is not clear. In sum, for 206 of the 222 jurisdictions, the calculations for CFB destination-based sales are indirect proxies at best.
• For automated digital services (ADS), the data are even more problematic. Sixteen pages in the Pillar One Blueprint are devoted to defining what is in (and out) of ADS (pp. 22–38). Another 19 pages (pp. 71–81, 83–93) lay out a hierarchy of ADS resource sourcing rules with the first best options being the real-time location (GPS data or IP address) or ordinary residence (billing address) of the viewer or consumer. The problems with these sourcing rules are obvious and discussed to some extent in the report. For example, internet users use VPN to secure (and therefore hide) their IP addresses; billing information is unavailable because consumers use third-parties such as PayPal to pay for purchases; adult children (like me) in Texas buy gifts on Amazon.CA for delivery to their mothers in Canada). The rules become even more clearly wishful thinking once they are compared with the empirical technique used in EIA to proxy for ADS destination-based sales (pp. 43–46): the number of “regular internet users” multiplied by the per-capita consumption of all goods and services in that jurisdiction multiplied by a constant to proxy for the share of ADS consumption in total consumption. In other words, given that there are no data available on ADS sales, the EIA assumes they can be proxied by multiplying national levels of internet usage by per-capita consumption times a constant. A leap of faith, indeed! (This is the C from the definition of GDP as C + I + G + X – M, which you may remember from your Economics 100 course, divided by the country’s population.)

Measuring Component D (GRIP): Problems and Solutions

• The EIA (p. 224) notes that, of the four matrices, the profit matrix was “arguably the most difficult to extrapolate” because the location of MNE profit may bear little relation to where the MNE’s economic activity takes place. There are at least two reasons for this. First, intangible assets typically provide the greatest contribution to MNE profit; the cost of developing and exploiting intangibles is not necessarily tied to their value; and the easy costs to measure—labor and physical capital costs in CbCR reports—may not be highly correlated with intangible value. Second, incentives (e.g., CIT differentials, foreign exchange controls, political risk) to engage in base erosion and profit shifting behaviors encourage parking MNE profits in tax havens and/or routing profits through investment hubs and special-purpose entities.
• Given the absence of micro-level, location-specific data on MNE profit, the EIA was forced to engage in a highly complex set of calculations based on macroeconomic data on foreign direct investment (FDI). As EIA Annex 5.C. explains (pp. 257–266), the economists started with bilateral FDI statistics from the OECD and International Monetary Fund covering 98% of bilateral FDI. Extrapolation using gravity models based on GDP and per capita GDP was used to fill in the missing 2% of observations. This created a matrix of 222 x 222 countries for 2016, in four country groups, as shown in Annex Table 5.C.3 on page 259.
• FDI data are reported for direct investors; however, the Amount A calculations need profit data organized by ultimate investors. Direct and ultimate investors are unlikely to be the same because profit shifting incentives encourage pass-through (conduit) FDI using multiple jurisdictions (especially investment hubs). Double counting is therefore likely to occur when FDI is measured by direct investors given the number of intermediate jurisdictions.
• Missing data are also a problem. Data on FDI by ultimate investor are available only for 15 receiving (Source) jurisdictions that represent 23% of global FDI (EIA, p. 260). The EIA outlines in some detail how existing FDI data were cleaned to eliminate pass-through jurisdictions, identify the ultimate owner (Residence) jurisdiction, and extrapolate for missing observations (see pp. 235–237 and 259–265). The source jurisdictions are Austria, Canada, Switzerland, Czech Republic, Germany, Estonia, Finland, France, Hungary, Iceland, Italy, Lithuania, Poland, Turkey, and the United States (EIA, n. 31, p. 280).
• To get from FDI to MNE profit also requires estimating a rate of return on FDI and then multiplying that rate by the amount of FDI (EIA, pp. 265–266). The economists computed an average rate of return of 7.8%, based on comparisons of FDI position and FDI income for 2016, and then used the median differential for 2013 to 2016 to compute jurisdiction specific rates of return (p. 266).

My conclusion is that the Amount A formula, illustrated in Figure 1 and equation (1), is simple—and deceptive because the calculations involved in estimating Amount A are so complex.

IDENTIFYING THE LIKELY WINNERS AND LOSERS FROM AMOUNT A

I turn to analyzing the likely winners and losers in terms of net tax revenue from Pillar One, building the model in equation (1). In an earlier article, I discuss the OECD Secretariat’s decision to not publish the Amount A results for individual—or even less finely grouped—jurisdictions because of the political difficulties. The Pillar One Blueprint and the EIA, by my count, use the word political (as in political agreement, pressures or decisions) 25 times, suggesting the level of political sensitivity of the EIA results. The missing data in the EIA make it difficult, but not impossible, to perform a more fine-grained analysis. I start first with the EIA estimate and then show how it is possible to use the EIA data tables to craft more fine-grained estimates of the impacts, using three examples: (1) all jurisdictions, (2) investment hubs, and (3) low- and middle-income countries. Many developing countries such as Nigeria are neither tax havens nor investment hubs but may have a large deficit with respect to in-scope entities (ADS and CFB), so I consider low- and middle-income jurisdictions separately. I have omitted the high-income jurisdictions from these calculations partly because the list, by geographic group (see EIA, p. 271), is a mixture of OECD countries, developing countries, and small islands.

The EIA Estimates of Winners and Losers From Amount A

I start with the EIA’s estimates of Amount A, which are presented in ranges based on an assumed set of possible policy choices, for four jurisdiction groupings (high-, medium-, and low-income, and investment hubs). Figure 2.14 (EIA, pp. 61–62) shows the estimated impacts on corporate income tax (CIT) revenues by jurisdiction groups, based on the following policy choices:

• Component A (GRIP)
  • Global revenue threshold of $911 million 750 million euros
  • Residual profitability threshold of 10% or 20% (shown as black or blue boxes)
• Component B (Reallocation Percentage) set at 10%, 20% or 30% (shown in panels, A, B, and C)
• Component C (GIDS)
  • ADS: revenue threshold of $1.21 million (1 million euros) for nexus
  • CFB: revenue threshold of $3.64 million (3 million euros) for nexus

(The EIA assumes Market jurisdictions have nexus simply based on a minimum threshold of foreign sales. My view is that nexus for CIT purposes should not be based on imported goods and services, given that they are taxable under the GATT and GATS agreements. Nexus should be based on a sustained and effective connection that involves control over the investment in the Market jurisdiction.)

The EIA figure has three panels—A, B, and C—which nicely show the range of impacts expected on tax revenues of these choices. The left panels show country groupings by income levels; the right panels show groupings by statutory CIT rates. From the left, reading down the panels: High-, middle-, and low-income countries gain more CIT revenues with the lower (10%) residual profitability threshold (because component A rises); low-income countries gain the most but the gains are small (1% to 2%). Investment hubs, on the other hand, lose CIT revenue (–2% to –6%). As component B, the reallocation percentage, increases, the revenue gains to high-, middle- and low-income countries also rise, with the largest gains to low-income countries. Investment hubs, on the other hand, face much larger losses as component B rises. From the right, reading down the panels: the range of potential impacts by tax grouping, is small, typically less than +/–1%. The only clear winners appear to be jurisdictions with CIT rates above 30%, which gain roughly .5% in the first panel rising to 1.5% in the third panel.

A More Fine-Grained Analysis of Winners and Losers From Amount A

Table 1 provides my “guesstimate” of winners and losers, in terms of equation (1), based on data extracted from various EIA tables as proxies for the gap between components C and E. The EIA unfortunately does not include a matrix for GIDS (component C). Data on MNE turnover are available (EIA, p. 268) but, as my earlier discussion makes clear, these results should be taken with caution.

Table 1: Guesstimate of the Amount A Impacts (US$ Billion)

Note: MNE Turnover used as proxy for Component C, Global In-scope Destination-based Sales (GIDS). Source: Author’s calculations using data from EIA tables, pp. 58–59 and 267–270.

In terms of potential losers from Pillar One, Table 1 suggests that the gap between components C and E is large and negative (C < E) for two groups: investment hubs that are Source jurisdictions (–22% to –32%) and high-income Residence jurisdictions (–13% to –15%). These two groups are potentially the largest losers from Amount A. High-income jurisdictions, however, are both home and host countries to FDI. As host (Source) countries, high-Income jurisdictions show gains (6% to 13%). Thus, at least for high-income countries, their loss as home (Residence) countries is partly offset by their gain as host (Source) countries. This is not the case for investment hubs where, as Residence countries, their gains are miniscule (less than 1%). Investment hubs are clearly losers from Amount A.

In terms of potential winners, the gap is large and positive (C > E) for middle-income jurisdictions, and this is in both their roles as Residence (13% to 14%) and Source (16% to 19%) jurisdictions. Thus middle income countries are potentially the largest gainers from Amount A.

Still, these gains and losses in net tax revenue at the jurisdiction level are mostly in the 10% to 20% range, which are not that large. Is it worth upending the international tax rules for these results? The only group where the effects of Pillar One are larger are investment hubs, especially those that are primarily Source not Residence jurisdictions. To address these issues, I explore the impacts of Amount A on two jurisdictional groups—investment hubs and low-/middle-income jurisdictions—disaggregated into geographic regions, using data from the EIA matrices (pp. 267–270).

A Closer Look at Investment Hubs

The 24 investment hubs (Hubs, for short) in the EIA are defined as jurisdictions with an inward FDI position (stock) greater than 150% of GDP in 2016. Perusing the list of Hubs (see Panel B of Annex Table 5.D.2, p. 272) and fact checking reveal some interesting statistics, and point to some reasons why the Amount A results are so politically sensitive.

• All but two of the 24 Hubs (Hungary and Mozambique) appear on existing lists of tax havens. (See Table 2 in Lorraine Eden and Robert T. Kudrle, Tax Havens: Renegade States in the International Tax Regime? Law and Policy, 27.1: 100–127 (2005). A more recent list is provided by the Tax Justice Network at https://www.taxjustice.net/press/new-ranking-reveals-corporate-tax-havens-behind-breakdown-of-global-corporate-tax-system-toll-of-uks-tax-war-exposed/.)
• Two-thirds (16) of the Hubs are British territories, Commonwealth members or former colonies; all but two (Singapore and Hong Kong) are tiny tax havens.
• Five Hubs are OECD members: Hungary, Ireland, Luxembourg, the Netherlands, and Switzerland.
• Two Hubs (the Netherlands and Switzerland) are in the top quintile by GDP size (i.e., GDP > $400 billion); another four Hubs (Hungary, Ireland, Hong Kong, and Singapore) are in the second highest quintile (see EIA, p. 49).
• My fact-checking, using UNCTAD data on inward FDI position as a % of GDP for 2014-2018, generated a list of investment hubs different from the EIA list, both adding and removing jurisdictions. (My calculations would add four more countries to the investment hub list: Aruba, Saint Kitts and Nevis, Saint Vincent and the Grenadines, and the Seychelles. My estimates would exclude five jurisdictions: Bermuda, Hungary, Marshall Islands, Mauritius, and (possibly) Barbados.)

It is unfortunate that the EIA chose to report disaggregated data on investment hubs (see pp. 267–270) grouped into three categories based on geography (Americas, Europe, and Other). The bullet points above suggest other Hub groupings that might have been more illuminating, such as the biggest Hubs, the OECD Hubs or the U.K.-affiliated Hubs. For example, given their size and significant presence as Source and Residence jurisdictions, the largest Hubs are likely to be the most financially able to provide tax relief to Market jurisdictions—and the ones most targeted by Market jurisdictions to provide that relief given the Pillar One sourcing rules for tax relief. Politics may also matter. For example, two-thirds of the Hubs on the EIA list are U.K. affiliates; BREXIT is taking the U.K. out of the EU, reducing the U.K.’s political leverage. Also, five of the Hubs are OECD members and thus have a privileged “seat at the OECD table” in the negotiations over both Pillars.

Still, even with a geographic grouping, some comparisons are possible by reorganizing the EIA data for Hubs along the lines of Table 1, with the caveat discussed above re the “soft” data quality. In addition, my calculations for component C use the MNE turnover matrix and for component E use the MNE profit matrix because data for components C and E are not available broken out by Hub category in the EIA. For comparison with my estimates of Component E, which use pre-tax profit, I also report the Hubs share of Component E for all Jurisdictions, using data from Table 1 and 10% or 20% thresholds. Lastly, I report payroll and tangible asset data for Hubs, from the EIA matrices, as an indicator of the level of “hard” economic activity (i.e., labor and property, plant and equipment expenditure) in the Hubs. The results are reported in Table 2.

Table 2: Guesstimate of the Amount A Impacts on Investment Hubs, by Region (US$ Billion)

Notes: (1) Turnover Matrix data used as proxy for Component C (EIA, p. 268) (2) Profit Matrix data used as proxy for Component E (EIA, p. 267) (3) Americas = Anguilla, Bahamas, Barbados, Bermuda, British Virgin Islands, Cayman Islands, Turks and Caicos Islands (4) Europe = Bailiwick of Guernsey, Cyprus, Gibraltar, Hungary, Ireland, Isle of Man, Jersey, Luxembourg, Malta, Netherlands, and Switzerland (5) Other = Hong Kong (China), Liberia, Marshall Islands, Mauritius, Mozambique and Singapore Source: author’s calculations using data from EIA Tables, pp. 58–59 and 267–270.

For Residence jurisdictions, I find a small negative C-E gap of –1.1%, which is close to EIA estimates of 0.3% (10% threshold) and 0.6% (20% threshold). For Source jurisdictions, I find a large negative gap of –8.8%; compared to the much larger and more negative EIA estimates of –21.7% (10% threshold) and –32.1% (20% threshold). This suggests my Source estimates of the C-E gap, by region, are too low.

Three of the key results of Table 2 are the following. First, the European Hubs gain net tax revenue from Amount A. Their gains are less than 10% but are positive both as Residence and as Source jurisdictions, and they perform significantly better than the other Hubs. This is likely due to the dominance of four countries (Ireland, Luxembourg, the Netherlands, and Switzerland) in the European group. Their very high shares of labor and tangible assets for all Hubs, both as Residence and Source jurisdictions, is also an explanation.

Second, the Americas Hubs (all of which are all tiny island havens that are U.K.-affiliated jurisdictions) suffer especially large losses as Source jurisdictions, which is probably not surprising given their very small shares of total Hub payroll (1.9%) and tangible assets (6.1%). The C-E gap is much less negative for Americas Hubs, as Residence jurisdictions, reflecting their larger shares of payroll and tangible assets. Third, the Others grouping gains as Residence but loses as Source jurisdictions; I suspect this is because Hong Kong and Singapore (two of the largest Hubs) are included in the group.

Had the groupings been not by geography but rather by size, these results suggest that (1) the largest Hubs likely gain from Amount A whereas (2) the small island Hubs unambiguously lose from Amount A. My reasoning is as follows. While higher tax rates normally discourage inward FDI, the effect is only large when taxes are the main driver of FDI location (as is the case with small island tax havens). On the other hand, where the main purpose of FDI is market seeking and the Hub is located in a large and/or high-income market or has privileged access to a regional market (e.g., USMCA or the EU), tax rates have less effect on FDI location decisions.

Low- and Middle-Income Jurisdictions

Following the same approach, I also estimate the impacts of Amount A on low- and medium-income jurisdictions, grouped by geography. My results are reported in Table 3. The EIA matrices list 135 low- and middle-income jurisdictions, grouped by region; the total number in each group are reported in brackets in column 2. (The set includes one high-income country (South Africa) due to the way the EIA matrices were constructed. It is useful to peruse the list, noting which groups include big emerging economies (Brazil, Mexico, China, India, Russia).)

My estimates of the C-E gap for the 135 jurisdictions are net positive (4.5% Residence, 5.9% Source) but smaller than the EIA estimates, again suggesting that my C-E gap estimates, for each of the regions, may also be under-estimated. The only jurisdiction group with positive gains for both Residence (3.9%) and Source (5.7%) is the East Asia and Pacific group, which includes China. Note also its large share of payroll and tangible assets, both for its own MNEs (Residence) and foreign affiliates (Source). The South Asia group (which includes India) loses net tax base (–3.1% Residence, –2.9% Source). The net effects on the other regional groups, including Latin America and the Caribbean (which includes Brazil and Mexico), are small.

Table 3: Guesstimate of the Amount A Impacts on Low- and Middle-Income Jurisdictions, by Region (US$ Billion)

Notes: (1) Turnover Matrix data used for Component C (EIA, p. 268); Profit Matrix data used for Component E (EIA, p. 267). (2) For the complete list of countries in each jurisdiction group see EIA, pp. 271–272. Source: author’s calculations using data from EIA Tables, pp. 58–59 and 267–270.

Summary: Winners and Losers From Amount A

Table 1 suggests that high-income jurisdictions would lose tax base as Residence jurisdictions but gain tax base as Source jurisdictions should Amount A be adopted. Tables 1 and 3 suggest the apparent winners from Amount A are middle-income countries, both as Residence and Source jurisdictions. The largest net gains go to East Asia and the Pacific, in the 4% to 6% range. For other low- and middle-income jurisdictions, the effects are smaller and more mixed. The South Asia group suffers losses of –3% in both Residence and Source roles.

Tables 1 and 2 show that investment hubs are likely to suffer the largest losses should the Pillar One proposals be adopted. However, not all Hubs. The jurisdictions most likely to suffer are the Americas investment hubs, i.e., especially the Caribbean islands that have built their economic development strategies on being tax havens and financial conduits. Given their small size and lack of resources, their future prospects in a Pillar One world look very difficult.

Amount A has offsetting impacts on the Asian investment hubs; they lose as Residence but gain as Source jurisdictions. The 10 European investment hubs, on the other hand, do much better. They gain tax base, both as Residence and as Source jurisdictions. I attribute their relative success to their developed-country economic status and to their additional roles as gateways for market-seeking FDI, regional headquarters, and R&D and marketing centers. If the tax loopholes are closed, the European investment hubs have other ways to create employment, income, and economic growth.

CONCLUSION

While the total amount of net tax revenues expected from Amount A under the Pillar One Blueprint are small; the impacts can vary significantly across jurisdictions. A key factor affecting who wins and who loses from Amount A is the gap between components C and E in the formula. The Economic Impact Assessment of Pillar One groups jurisdictions into four categories, obscuring results at the jurisdiction level. My goal in this article has been to examine the EIA’s calculations of Amount A and provide more fine-grained estimates of its jurisdictional impacts. My results suggest that the Caribbean islands are likely to suffer the largest relative losses, and middle-income jurisdictions in East Asia and the Pacific the largest relative gains, from Amount A.

I hope my work will encourage other transfer pricing professionals to read the EIA and build their own economic models of Amount A. We need better estimates—and a clearer understanding—of its potential impacts, not only on tax revenues but on economic cooperation and development.

This column does not necessarily reflect the opinion of The Bureau of National Affairs, Inc. or its owners.

Author Information

Lorraine Eden is Professor Emerita of Management and Research Professor of Law at Texas A&M University, leden@tamu.edu. Helpful comments on earlier drafts were provided by Joan Hortalà Vallvé, Jeffrey Owens, Philippe Paumier, Niraja Srinivasan, Ognian Stoichkov, Oliver Treidler, and Charles Hermann. The opinions expressed herein are of the author, who accepts responsibility for any errors or omissions.