3. Tax revenue effects of Pillar Two

3.4.1. Data source on profit location: The “profit matrix”

192. As in Chapter 2, the analysis in this chapter relies on the “profit matrix”, which combines data from different sources on the location of MNE profit (see stylised illustration in Figure 3.3 and detailed description in Chapter 5).

Figure 3.3. Profit matrix: Stylised overview and underlying data sources

Note: Aggregate CbCR data are used to fill columns of the profit matrix (e.g. profit of French MNEs across jurisdictions). ORBIS unconsolidated account data are used to fill rows of the profit matrix (i.e. MNE profit in France, split across ultimate parent jurisdictions). These two sources are used only where available, and in the case of ORBIS, where data coverage is sufficiently good. Other cells in the profit matrix are filled with extrapolations based on macroeconomic data, including FDI data.

Source: OECD Secretariat.

193. The profit matrix draws on three main sources of data, presented below in descending order of preference:

1. 1) Anonymised and aggregated data from Country-by-Country Reports (CbCRs), across 25 jurisdictions of ultimate parent (see list in Annex 5.A of Chapter 5);6

2. 2) ORBIS unconsolidated account data in 24 jurisdictions of affiliate with good ORBIS coverage (see list in Annex 5.A of Chapter 5);

3. 3) Extrapolations based on macroeconomic data (e.g. FDI data) in other cells.

194. Wherever possible, the data in the profit matrix (and in the tangible assets and payroll matrices described below) focus on MNE sub-groups with positive profits (i.e. entities belonging to an MNE group that is reporting an overall profit in the jurisdiction considered), rather than all MNE sub-groups (i.e. profit-making and loss-making sub-groups). Indeed, only MNE sub-groups with positive profits potentially face additional taxes under Pillar Two, while loss-making sub-groups generally do not pay corporate income tax. Subtracting the losses of loss-making MNE sub-groups when computing total profit in a jurisdiction could therefore lead to an underestimation of the amount of profit that could be subject to Pillar Two.

195. Still, a limitation of the approach is that the effect of a loss carry-forward mechanism under Pillar Two is not taken into account in the estimates. Such a mechanism could reduce revenue gains from Pillar Two in situations where an MNE sub-group in a low-tax jurisdiction makes a (low-taxed) profit and can use its past losses in that jurisdiction to offset it partly or fully. Estimating with precision the effect of the loss carry-forward mechanism would require firm-level information on the profitability of MNE sub-groups over several successive years. This information is not available to the OECD Secretariat in most jurisdictions.7 In practice, the effect of the loss-carry forward mechanism on Pillar Two revenue gains will depend on how frequently MNE sub-groups in low-tax jurisdictions switch from being loss-making to being profitable. This may depend on the economic situation – for example, the COVID-19 crisis is likely to make losses more frequent – but also on profit shifting and loss shifting behaviour by MNEs. For example, even taking into account a carry-forward mechanism under Pillar Two, it may still be more beneficial for MNEs to incur losses in higher-tax than in low-tax jurisdictions, as these losses can be used to offset future profits at a higher rate in higher-tax jurisdictions.

196. The profit matrix – at a relatively high level of aggregation – is displayed in Table 3.1. The full methodology underlying the construction of the profit matrix is presented in detail in Chapter 5. Chapter 5 also contains detailed information on the data sources used in the profit matrix, including a discussion of the caveats around their use. In addition, Chapter 5 displays a more disaggregated version of the profit matrix (i.e. with more detailed jurisdiction groups than in Table 3.1), as well as information on the relative importance of the different data sources underlying the matrix. Finally, Chapter 5 contains the results of the extensive benchmarking that was undertaken to assess the quality of the data and its consistency across sources.

3.4.2. Data sources on effective tax rates (ETRs)

197. The relevant ETR to assess if an MNE has low-taxed profit in a jurisdiction and would be subject to Pillar Two will be computed at the level of each MNE sub-group in each jurisdiction (assuming that Pillar Two would involve jurisdictional blending). Estimating with precision the effect of Pillar Two would therefore require data on ETRs at that level. However, the data are not available to the OECD Secretariat in most jurisdictions, and the approach therefore relies on average ETRs computed from aggregated data.

198. The average ETR on MNE profit in a jurisdiction is measured as the median estimate obtained across the three following data sources:

• Data from Tørsløv et al. (2018[5]). These data, themselves based on a combination of sources and assumptions, focus on the ETR on the profit of foreign-owned MNEs across a range of jurisdictions in 2015. The underlying data generally does not distinguish between profit-making and loss-making MNE sub-groups and can thus lead to higher measures of ETR than data that would focus only on profit-making sub-groups.8

• Data from the US Bureau of Economic Analysis (BEA) on US MNEs. In an annual report on the global activity of US MNEs, the BEA provides information on foreign taxes paid by affiliates of US MNEs across a set of jurisdictions. In each of these jurisdictions, the average ETR of US MNEs is computed by dividing foreign taxes paid by “profit-type return”, which is a measure of profit included in the BEA data that aims to approximate profit before tax and excludes various sources of financial income. To reduce the impact of potential volatility in the data, the ETR used in this chapter is computed as the average ETR over several years (2013-16), i.e. the sum of taxes paid in a jurisdiction over the period divided by the sum of profit-type returns in this jurisdiction over the same period.9 As with the data from Tørsløv et al. (2018[5]), the BEA data aggregates data relative to profit-making and loss-making MNE sub-groups, which leads to higher ETRs than when focusing only on profit-making sub-groups.

• Data from anonymised and aggregated CbCR reports. The anonymised and aggregated CbCRs from 25 jurisdictions of ultimate parent (see list in Annex 5.A of Chapter 5) have been used to compute average ETRs by jurisdiction of affiliate. The ETR in a given jurisdiction of affiliate is computed as the total taxes paid by foreign-owned MNEs (on an accrual basis) over total profit of foreign-owned MNEs, focusing only on profit-making sub-groups (contrary to the other two sources). Extreme outliers are eliminated (e.g. ETRs above 100%) and jurisdictions covered in the data of fewer than three ultimate parent jurisdictions are excluded. An important caveat is that due to potential inclusion of intracompany dividends in profit reported in CbCRs, ETRs may be underestimated, especially in jurisdictions with a large presence of parent companies (OECD, 2020[6]).

199. None of these three measures of ETRs is subject to the issue of profit double counting pointed out by Blouin and Robinson (2019[7]) (see also discussion in Clausing (2020[8])). In addition, using the median value across the three measures reduces the potential impact of limitations of individual data sources on the final revenue estimates (e.g. limitations related to the potential inclusion of dividends in CbCR data, which could lead to overestimating Pillar Two gains, or to the fact that data from Tørsløv et al. (2018[5]) and BEA data include loss-making sub-groups, which could lead to underestimating Pillar Two gains). A robustness analysis excluding CbCR data from the calculation and relying only on the average of the other two sources suggests that it could lead to lower estimates of Pillar Two gains, but without changing the broad order of magnitude of the results (see Annex 3.D).

200. All three data sources are available jointly for 42 jurisdictions representing 86% of world GDP (Table 3.2). At least one of the three data sources is available for another 99 jurisdictions, representing another 12% of world GDP. For the other 81 jurisdictions considered in the analysis, which are mainly lower income jurisdictions that together represent only 1% of world GDP, none of these three sources are available and the ETR is assumed to correspond to the statutory CIT rate, sourced from the OECD Corporate Tax Statistics (OECD, 2020[9]), complemented with other sources.10 This is a conservative assumption since ETRs are generally lower than statutory rates. However, the impact of this assumption on the global estimates is likely to be marginal as less than 1% of total MNE profit is found to be located in these jurisdictions.

3.4.3. Global low-taxed profit and global revenue gains (before carve-out)

201. Based on the profit matrix and the ETR data described above, the global amount of low-taxed profit (i.e. profit taxed below a potential minimum rate) is presented in Table 3.3 for an illustrative range of potential minimum rates. Estimated global gains from Pillar Two are then obtained by topping up the tax rate on the low-taxed profits up to the level of the minimum rate.11 A ±10% uncertainty range around the point estimates is applied to take into account data uncertainty. The results, which are presented in Table 3.3, correspond to the revenue gains in Scenario 1 (i.e. in a static scenario that does not take into account the interaction with Pillar One, nor behavioural reactions by MNEs and governments) with no formulaic substance-based carve-out. Consistent with the assumptions on GILTI discussed above, the table excludes the low-taxed profit of US MNEs and the corresponding revenue gains (revenue gains from GILTI are discussed in Section 3.8).

202. These estimates do not take into account potential gains related to pockets of low-taxed profit in higher-tax jurisdictions (i.e. jurisdictions with an average ETR above the minimum rate). These pockets, while difficult to assess with the available data, may be substantial, as discussed in Box 3.1. Not taking them into account could lead to significantly underestimating the revenue gains from Pillar Two. In light of this, the upper bound of the uncertainty ranges surrounding the estimates of the direct gains from Pillar Two in Scenario 1 is increased by 50% (last row in Table 3.3). Such an increase in revenue gains would correspond to a situation where pockets of low-taxed profit would represent close to 10% of total profit in higher-tax jurisdictions (Box 3.1). The lower bound of the uncertainty ranges is not changed and therefore assumes conservatively no revenue gains from these pockets. In scenarios with a formulaic substance-based carve-out (discussed below), the uncertainty around these pockets is reduced in proportion to the share of carved-out profit in higher-tax jurisdictions.12

203. The estimates in Table 3.3 are broadly consistent with recent estimates by the Oxford University Centre for Business Taxation (Devereux et al., 2020[10]). Assuming a 10% minimum tax rate, no formulaic substance-based carve-out, and including MNEs with an ultimate parent in the United States, they assess that global Pillar 2 revenue gains in a static scenario could reach USD 20 billion (1.1% of global CIT revenues) in an approach excluding pockets of low-taxed profit in higher-tax jurisdictions, or USD 32 billion (1.8% of global CIT revenues) in an approach including them.13 This is higher than the estimates with corresponding assumptions in Table 3.3 (i.e. estimated gains of 0.8%-1.0% of CIT revenues excluding pockets of low-taxed profit, and 0.8%-1.5% of CIT revenues including them) but the difference is likely explained in large part by the fact that the estimates in Table 3.3 exclude MNEs with an ultimate parent in the United States, while the estimates by Devereux et al. (2020[10]) include them.

3.4.4. Modelling the implications of a formulaic substance-based carve-out

General approach to model a formulaic substance-based carve-out

204. The Pillar Two Blueprint report envisages that the GloBE rules (IIR and UTPR) could include a formulaic substance-based carve-out based on a fixed percentage of payroll plus a fixed percentage of depreciation expenses on a broad range of tangible assets (OECD, 2020[1]). These two percentages could be identical or different. The illustrative results presented in this chapter assume identical percentages.

205. For example, if one assumes illustratively a 10% carve-out percentage (or "mark-up”) for both payroll and depreciation expenses, the GloBE rules would only apply to a given MNE sub-group in a given jurisdiction on the profits that exceed 10% of the sum of payroll plus depreciation expenses of that MNE sub-group in that jurisdiction. For example, if the profit of the sub-group is 100, its payroll 150 and its depreciation expenses 50, the profit on which Pillar Two applies would be 100-10%*(150+50)=80.

206. On this profit, Pillar Two would apply, as before, by topping-up the average ETR paid by the MNE sub-group in that jurisdiction up to the level of the agreed minimum tax rate. The Inclusive Framework will define at a later stage the exact rules to calculate this ETR in the presence of a carve-out, and in particular whether an MNE group that claims the benefit of the carve-out should be required to make a corresponding and proportional adjustment to the covered taxes. For example, if before the application of the carve-out a taxpayer has EUR 100 of profit and EUR 20 of covered taxes, the ETR in absence of carve-out is 20% (EUR 20 divided by EUR 100). If the carve-out reduces the taxpayer’s profit to EUR 80, then the ETR for the purposes of the GloBE rules could be either (i) EUR 20 of covered taxes divided by EUR 80 of profit (i.e. 25%) if the MNE is not required to make an adjustment to the covered taxes, or (ii) EUR 16 of covered taxes divided by EUR 80 of profit (i.e. 20%) if the MNE is required to make an adjustment to the covered taxes that would be corresponding and proportional to the effect of the carve-out on profit. The approach modelled in this chapter is the second, since the modelling relies on the same ETRs in the scenarios with and without the carve-out. This reflects that modelling the situation without adjustment would be difficult with the available data, and does not prejudge future decisions by the Inclusive Framework on this question. Since the first approach (i.e. without adjustment) would result in higher ETRs than the second approach, it would lead to lower revenue gains (for a given level of the minimum rate and the carve-out percentage).

207. Modelling with precision the impact of a carve-out would require firm-level information on tangible assets depreciation and payroll across all jurisdictions. Sufficiently detailed firm-level data at the unconsolidated level is available in the ORBIS database with good coverage for only 18 to 24 jurisdictions depending on the variable considered (see list in Annex 5.A of Chapter 5).14 In these jurisdictions, the share of carved-out profit is computed directly with ORBIS data, at the level of each MNE sub-group, as discussed below. In the other jurisdictions, the approach relies on more aggregate data, in combination with an analysis based on ORBIS data on the average relationship between firm-level and aggregate data, which is also described below.

208. The approach in this chapter is to estimate the share of carved-out profit in all jurisdictions, even those with average ETRs above the minimum rate. This offers the benefit of helping to gauge the effect of the carve-out on potential pockets of low-taxed profit in these jurisdictions. For practical reasons, the payroll carve-out and the depreciation carve-out are modelled separately, and their effects are summed to obtain the effect of a combined carve-out.15 This represents an approximation compared to the actual effect of a combined carve-out, because it could lead to some ‘double counting’ of carve-out effects in cases where the sum of the amounts carved out under each carve-out taken individually would exceed the total profit of the MNE sub-group considered.16 Computations based on ORBIS firm-level data suggest that this double counting is not quantitatively significant under the carve-out percentages considered in this chapter, as it leads to an overestimation of the effect of the carve-out by less than 4% on average across jurisdictions.

209. The coverage of depreciation expenses in ORBIS is less extensive than the coverage of tangible assets. In addition, depreciation is generally aggregated with amortisation of intangible assets. In aggregate data as well, the level of tangible assets is generally better covered than depreciation expenses. Against this background, the approach to approximate depreciation expenses is to use data on tangible assets combined with an assumption on the average depreciation rate. Evidence from the available data from the US BEA and the ORBIS database suggests that the average depreciation rate of tangible assets (i.e. property, plant and equipment) is about 5-10%,17 and this percentage is conservatively assumed to be 10% in the estimates of this chapter.

The aggregate data in the tangible assets and payroll matrices

210. Aggregate data on the location of MNE tangible assets and payroll across jurisdictions are based on a ‘tangible assets matrix’ and a ‘payroll matrix’ that combine various data sources and extrapolations, in the same spirit as the profit matrix. These data sources and methodology underlying these two matrices, as well as extensive benchmarking and checks to assess their quality are presented in Chapter 5. These matrices are presented at an aggregate level in Table 3.4.

Table 3.4. The tangible assets and payroll matrices: Results aggregated by broad jurisdiction groups
Panel A: The tangible assets matrix
		Jurisdiction of ultimate parent
	(USD billion of 2016)	High income	Middle income	Low income	Investment Hubs	Total
Jurisdiction of affiliate	High income (64 jurisd.)	11463.1	314.5	6.2	614.8	12398.7
	Middle income (105)	1320.4	4357.9	5.0	757.4	6440.7
	Low income (29)	20.5	11.2	17.1	4.2	53.1
	Investment Hubs (24)	437.8	69.5	0.9	422.1	930.3
	Total	13241.8	4753.2	29.2	1798.5	19822.8
Panel B: The payroll matrix
		Jurisdiction of ultimate parent
	(USD billion of 2016)	High income	Middle income	Low income	Investment Hubs	Total
Jurisdiction of affiliate	High income (64 jurisd.)	6967.3	153.6	3.0	472.2	7596.2
	Middle income (105)	497.8	1495.6	1.5	186.5	2181.4
	Low income (29)	7.0	3.1	6.8	1.8	18.7
	Investment Hubs (24)	225.3	18.2	0.4	170.3	414.2
	Total	7697.5	1670.5	11.8	830.7	10210.5
Note: Groups of jurisdictions (high, middle and low income) are based on the World Bank classification. The number of jurisdictions in each group is indicated in parentheses. Investment hubs are defined as jurisdictions with a total inward FDI position above 150% of GDP. MNEs with an ultimate parent in the United States are included in these tables.
Source: OECD Secretariat (see details in Chapter 5).

211. An important caveat to the analysis of carve-outs is that the definition of tangible assets and payroll in the data underlying the matrices may not necessarily match the exact definitions of the variables considered for a formulaic substance-based carve-out. The tangible assets matrix focuses on property, plant and equipment, net of accumulated depreciation, while the payroll matrix focuses on expenditures for salaries and wages, including bonuses, social contributions and other employee benefits (see Chapter 5). While this is broadly consistent with the variables considered for a carve-out, which are described in the Pillar Two Blueprint report, there may be differences related, for example, to the treatment of land or subcontracted labour expenses, which, depending on the exact definition of the carve-out, could affect the accuracy of the estimates.

Figure 3.4. Share of carved-out profit, as estimated with firm-level data, and relationship with aggregate profitability

Note: Each dot corresponds to one jurisdiction. Non-carved-out profit is computed for each MNE sub-group as profit before tax in excess of 1% of tangible assets (panel A) or 10% of payroll (Panel B). These figures focus on all profit-making MNE sub-groups in the jurisdictions considered, regardless of their ETR. Loss-making MNE sub-groups in the jurisdiction are not included. The sample consists of jurisdictions with relatively good coverage of unconsolidated accounts in ORBIS on both foreign and domestic MNE entities for the variables considered (tangible assets in Panel A, payroll in Panel B), see list in Annex 5.A of Chapter 5.

Source: OECD Secretariat calculations based on ORBIS data.

3.4.5. Methodology to estimate jurisdiction-level revenue gains

221. Revenue gains at the jurisdiction level and, in turn, for jurisdiction groups are derived using the following stylised modelling assumptions, also summarised in Figure 3.5, which do not pre-judge jurisdictions’ actual implementation decisions:

• Group 1: Jurisdictions with an average ETR above the minimum rate. These jurisdictions are assumed to implement an income inclusion rule (IIR), in the sense that they would apply a top-up tax to ensure that the profit of MNE entities (blended at the jurisdictional level) with an ultimate parent in their jurisdiction is taxed at least at the minimum rate.20 They are also assumed to implement an undertaxed payments rule (UTPR). Consistent with the Pillar Two Blueprint report, the IIR is assumed to apply in priority to the UTPR.21

• Group 2: Jurisdictions with a zero corporate tax rate. For the purposes of modelling, these jurisdictions are assumed not to introduce an IIR nor a UTPR, as it would also require introducing a corporate income tax (CIT) system, which many of these jurisdictions do not have. As a result, the low-taxed profit of MNE entities with an ultimate parent in these jurisdictions would not be taxed by these jurisdictions (since they would not introduce an IIR). If an intermediate-level parent (i.e. a parent entity that is not the ultimate parent) in the MNE group is located in a jurisdiction introducing an IIR, this low-taxed profit would be taxed by the jurisdiction of this intermediate parent (if there are several intermediate parents in this case, the highest one in the ownership chain would have priority according to the top-down principle described in the Pillar Two Blueprint report). If the low-taxed profit is not in scope of an applicable IIR (e.g. if there is no intermediate-level parent, or if intermediate parent jurisdictions are not introducing an IIR, or if the ultimate parent jurisdiction is low tax), the low-taxed profit could be taxed under the UTPR of a jurisdiction introducing a UTPR and from which intra-group payments originate.

  • In practice, it is difficult with the available data to model these rules with precision, as little information is available on the location of intermediate parents and transaction origins. Reflecting this, it is assumed for the purposes of modelling in this chapter that the profits of MNEs with an ultimate parent in Group 2 would be subject to a top-up tax imposed by other jurisdictions, in proportion to the amount of economic activity located in these jurisdictions (as a proxy for the location of intermediate parents in the case of IIR, and of the transactions-based nature of the UTPR).

  • Economic activity is proxied by MNE turnover, sourced from the “turnover matrix” described in Chapter 5.22^,23 Results are broadly robust to using tangible assets or payroll instead of turnover as a proxy (see Annex 3.C). When economic activity is located in a jurisdiction that does not implement an IIR nor a UTPR (e.g. a jurisdiction in Group 2), the corresponding low-taxed profit is assumed not to be subject to the top-up tax. However, this represents a relatively small fraction of total low-taxed profit under the assumptions considered in this chapter.

• Group 3: Jurisdictions with an average ETR below the minimum rate, but greater than zero. Among jurisdictions in Group 3, for the purposes of this modelling scenario, half are assumed to implement an IIR and a UTPR (as jurisdictions in Group 1), and the other half are assumed not to implement them (as jurisdictions in Group 2). One reason why it is likely that not all jurisdictions in this group would implement an IIR and a UTPR is that some of these jurisdictions may decide that imposing a minimum tax rate on foreign profits could seem inconsistent with maintaining an average ETR below this minimum rate on local profit. As a result, this choice may be linked to choices relative to other tax policy parameters, and more specifically whether the ETR on local profit is increased, which not all jurisdictions in this group may be willing to do (see assumptions underlying Scenario 4 in section 3.7.1 below, with which these assumptions aim to be consistent). In practice, identifying the jurisdictions in this group that would implement an IIR and a UTPR is not straightforward. Instead of arbitrarily selecting half of the jurisdictions in the group, a simplifying assumption is made that all jurisdictions in this group apply an IIR and a UTPR on half of the low-taxed profit on which they could apply them. This is by no means realistic in itself, but it aims to be a representative and neutral proxy for a situation where half of the jurisdictions in this group would implement and the other half would not.

3.5.1. Rationale for taking the interaction between both pillars into account

222. Scenario 1 considers Pillar Two in isolation, without taking into account its potential interaction with Pillar One. Scenario 2 aims to take this interaction into account, by assuming illustratively that both pillars would be introduced together and that Pillar Two would apply after the reallocation of profit induced by Pillar One.

223. For example, in an extreme case where an MNE group would have all its profit in a jurisdiction where it is taxed at a rate below the Pillar Two minimum rate, all its profit (after potential application of a formulaic substance-based carve-out) would be subject to Pillar Two if Pillar Two was introduced in isolation (Figure 3.6). However, if the reallocation of profit induced by Pillar One was applied first, a portion of these profits (or, more precisely, the taxing rights corresponding to these profits) would be reallocated to market jurisdictions (assuming that the MNE group is in scope of Pillar One and above the residual profit threshold). If the tax rate in these market jurisdictions is above the minimum rate under Pillar Two, these reallocated profits would not be subject to Pillar Two, reducing the overall revenue gains from Pillar Two.

Figure 3.5. Stylised modelling assumptions on Income Inclusion Rule (IIR) and Undertaxed Payments Rule (UTPR)

Note: These assumptions are stylised modelling assumptions and do not pre-judge actual implementation of IIR and UTPR. The United States (which is in group 1) is assumed to apply GILTI instead of an income inclusion rule.

Source: OECD Secretariat.

Figure 3.6. Stylised example on Pillar Two interaction with Pillar One

Source: OECD Secretariat.

224. In this example, computing the revenue gains from Pillar One and Pillar Two independently of each other would overstate the overall gains compared to computing the joint effect of both pillars in a way that takes into account the interaction between them. It is possible to consider an opposite example, where the reallocation taking place under Pillar One would increase the amount of low-taxed profits and therefore increase the revenue gains from Pillar Two. This would be the case for example if an MNE group had most of its profit located in higher-tax jurisdictions and most of its sales in low-tax jurisdictions. However, in practice, MNE profit tends to be more concentrated in low-tax jurisdictions than MNE final sales, which is why the reallocation taking place under Pillar One is expected to reduce the global amount of low-taxed profit. Therefore, taking into account the interaction with Pillar One is expected to reduce the estimated revenue gains from Pillar Two.

3.5.2. Methodology on the interaction between both pillars

225. To take the interaction between both pillars into account, Scenario 2 applies Pillar Two in exactly the same way and with the same assumptions as in Scenario 1, but after adjusting the location of profit for the reallocation induced by Pillar One. In practice, this is done by computing an adjusted profit matrix post Pillar One reallocation. Each matrix cell ${Profit}_{ij}$ (corresponding to the profit in jurisdiction i of MNE groups with an ultimate parent in jurisdiction j) is adjusted in the following way:

${ProfitAdjustedForPillar1}_{ij}={Profit}_{ij}+ProfitReceive{d}_{ij}-ProfitRelie{ved}_{ij}$

226. The amount of profit received and relieved under Pillar One in each matrix cell is based on the Pillar One estimates described in Chapter 2.24 In theory, this adjustment can be done for any combination of Pillar One parameter and design options. For simplicity, only one illustrative set of Pillar One design and parameter assumptions among those explored in Chapter 2 is considered in this chapter (i.e. residual profit threshold percentage of 10%, reallocation percentage of 20% and global revenue threshold of EUR 750 million).

3.5.3. Results on Scenario 2

227. The resulting profit matrix, adjusted for Pillar One reallocation is presented in Table 3.7. Compared to the original profit matrix (Table 3.1), the total in each column is the same (by construction), but there has been some reallocation across rows, away from investment hubs and into other jurisdiction groups. The scale of this reallocation is limited, which implies that the global amount of low-taxed profit and the estimated revenue gains from Pillar Two are reduced only slightly by taking into account the interaction with Pillar One (Table 3.8). To the extent that the interaction with Pillar One has only a modest effect on Pillar Two results, considering different assumptions regarding Pillar One would likely affect the Pillar Two estimates only at the margin.

3.6.1. Assessing current profit shifting patterns

230. To evaluate the effect of Pillar Two introduction on MNE profit shifting patterns, the first step is to assess current profit shifting patterns. For consistency with the methodology applied to quantify the revenue effects of Pillar Two in this chapter, profit shifting needs to be assessed on a “trilateral” basis, i.e. for each combination of (i) jurisdiction where profit was located before shifting (“profit origin”), (ii) jurisdiction where profit has been shifted (“profit destination”), and (iii) jurisdiction of ultimate parent of the profit shifting entity (“ultimate parent”).25 This third jurisdiction may or may not be the same as the “profit origin” jurisdiction. This level of granularity is necessary to create an adjusted profit matrix taking into account the effect of Pillar Two on profit shifting intensity, so that Pillar Two can then be applied to this adjusted profit matrix (with assumptions consistent with those used in Scenarios 1 and 2). This requires going beyond most studies on profit shifting, which often focus on measuring an average profit shifting semi-elasticity across a range of destination jurisdictions – for recent reviews of these studies, see for example Bradbury et al. (2018[11]) and Beer et al. (2019[12]).

Profit shifting: Overview of the approach

231. To obtain this “trilateral” profit shifting intensity, the approach in this chapter is based on the following steps, which are further detailed in the following sections and presented in a stylised way in Figure 3.7:

1. (i) Identifying “profit destination” jurisdictions, i.e. jurisdictions where profit may have been shifted to, based on foreign direct investment (FDI) and ETR data;

2. (ii) Computing the amount of deemed shifted profit in these jurisdictions, assuming that profit up to a certain “normal” profitability rate may not have been shifted, but may instead reflect real economic activity in these jurisdictions. The share of shifted profit is measured on a “bilateral” basis, i.e. for each pair of “profit destination”-“ultimate parent” jurisdictions;

3. (iii) Identifying where the shifted profit originates from. For each “profit destination”-“ultimate parent” pair of jurisdictions, shifted profits need to be reattributed to “profit origin” jurisdictions. This is done based on tax rate differentials vis-à-vis the “profit destination” jurisdiction (assuming that a higher tax rate differential leads to more profit shifting, all else equal) and on the geographic distribution of economic activity of MNEs from the “ultimate parent” jurisdiction considered (with the idea that profit is more likely to originate from jurisdictions where these MNEs have more economic activity).

Figure 3.7. Stylised illustration on the methodology to assess profit shifting

Note: In this stylised example, only two “ultimate parent” jurisdictions, two “profit origin” jurisdictions and two “profit destination” jurisdictions are represented. In reality, all jurisdictions in the profit matrix (more than 200 jurisdictions) are considered as “ultimate parent” jurisdictions, and all jurisdictions in the profit matrix are either “profit origin” or “profit destination”, based on the criteria described in the next section (using FDI and ETR data). For each pair of “ultimate parent”-“profit destination” jurisdictions, profit up to a certain profitability rate is deemed ‘normal’ (orange bars), and the rest, if any, is deemed shifted (blue bars). In this example, some profit in the k₁-j_1, k₁-j₂ and k₂-j₂ pairs is deemed shifted, but not in the k₂-j₁ pair. Deemed shifted profit is assumed to originate from “profit origin” jurisdictions in a way that depends on the location of the MNEs’ economic activity and tax rate differentials, as further described below. This is materialised by the blue arrows in the figure – in this example, profit in the k₁-j₁ pair is found to come predominantly from jurisdiction i₁ (thick arrow) and to a lesser extent from jurisdiction i₂ (thin arrow). The corresponding arrows identifying the origin of profit deemed shifted in jurisdiction pairs k₁-j₂ and k₂-j₂ are not represented to avoid overburdening the figure.

Source: OECD Secretariat.

232. This approach to measuring profit shifting is new, although it shares some common features with Tørsløv et al. (2018[5]), Cobham et al. (2019[13]) and Clausing (2020[8]). The approach is enabled by the level of detail offered by the profit matrix and the underlying data sources, including anonymised and aggregated CbCR data, which give a detailed account of the amount of profit located across low-tax jurisdictions for each ultimate parent jurisdiction. To benchmark the results against the vast existing literature on profit shifting, an average aggregate profit shifting semi-elasticity is computed based on the results and compared with existing estimates of this semi-elasticity.

Profit shifting: Identifying “profit origin”

237. Once shifted profit has been identified, further assumptions are required to identify where profit originates from, i.e. the jurisdiction where it was generated before being shifted. This is done at the level of each “profit destination”-“ultimate parent” pair, using the following formula:

${ProfitShifted}_{i,j,k}={\lambda }_{j,k}.Y_{i,k}.f\left({\tau }_i-\mathrm{ }{\tau }_j\mathrm{ }\right)$

238. In this formula,$\mathrm{ }{ProfitShifted}_{i,j,k}$ is the amount of profit shifted from jurisdiction i to jurisdiction j by MNEs with an ultimate parent in jurisdiction k. The intuition is that this profit is proportional to the economic activity in i of MNEs with an ultimate parent in k. For example, an MNE with very little economic activity in a jurisdiction is unlikely to have profit shifted away from this jurisdiction, whatever the tax rate differential with this jurisdiction. This economic activity is proxied by the turnover in i of MNEs with an ultimate parent in jurisdiction k (denoted $Y_{i,k}$) sourced from the turnover matrix.31 Results are broadly robust to using tangible assets or payroll instead of turnover (Annex 3.C).

Table 3.9. Estimated global profit shifted to low-tax jurisdictions
		Baseline estimate (‘normal’ profitability: 7.9%)	Robustness check (‘normal’ profitability: 5%)	Robustness check (‘normal’ profitability: 10%)
Estimated amount of shifted MNE profit at the global level	In USD bn	727	837	662
	In % of global MNE profit	11.3%	13.5%	10.7%
Share of shifted profits in total observed profit	In zero-tax “profit destination” jurisdictions	90.8%	94.1%	88.5%
	In other “profit destination” jurisdictions	61.7%	73.7%	54.7%
Note: The amount of profit shifted is estimated based on a “profit matrix” combining a range of data sources to map the location of profit (see Chapter 5), and the median of three different data sources on ETRs on MNE profit across jurisdictions, following a methodology described in this chapter. “Profit destination” jurisdictions are identified based on FDI and ETR data. Only profit in excess of ‘normal’ profitability in “profit destination” jurisdiction is deemed shifted. For example, assuming that ‘normal’ profitability is 7.9%, i.e. the global average profitability in ORBIS data (first column), the estimates suggest that USD 727 billion of profit is shifted, which represents 11.3% of global MNE profit. In “profit destination jurisdictions”, the share of observed profit that is shifted is 90.8% on average across zero-tax “profit destination” jurisdictions and 61.7% on average across other “profit destination” jurisdictions. Zero-tax jurisdictions are those with no CIT system or a zero statutory CIT rate. MNEs with an ultimate parent in the United States are included in this table, for the purpose of comparability with the economic literature.
Source: OECD Secretariat.

239. The amount of shifted profit is also assumed to be a function of the tax rate differential between jurisdictions i and j: $f\left({\tau }_i-\mathrm{ }{\tau }_j\mathbf{ }\right)$. The tax rates considered are the statutory CIT rates in “profit origin” jurisdictions, in line with most of the profit shifting literature that focuses on statutory rates, and the ETRs in “profit destination” jurisdictions (with the same data sources as in the rest of this chapter), as ETRs sometimes differ considerably from statutory rates in these jurisdictions. Several shapes of the relationship between profit shifting intensity and tax rate differentials (i.e. of the function $f$) are explored in this chapter, as further discussed below.

240. Finally, the amount of profit shifted depends on an array of ${\lambda }_{j,k}$ scaling factors, which are specific to each “profit destination”-“ultimate parent” pair of jurisdictions. These scaling factors capture that certain “profit destination” jurisdictions are more attractive to MNEs of a given “ultimate parent” jurisdiction – owing to, for example, geographic proximity or the legal environment – and that tax rate differentials alone do not predict where shifted profits are located. Instead of being taken from the literature, or set at arbitrary levels, these factors are set at the unique level that is consistent with the amount of profit that is deemed shifted in each “profit destination”-“ultimate parent” pair of jurisdictions, as computed in the previous section. Formally, these ${\lambda }_{j,k}$ factors are computed based on the following formula:

${\lambda }_{j,k}=\frac{{\sum }_i{ProfitShifted}_{i,j,k}}{{\sum }_i{Y}_{i,k}.f\left({\tau }_i-\mathrm{ }{\tau }_j\mathrm{ }\right)}$

241. This way of defining the ${\lambda }_{j,k}$ factors ensures the consistency of the approach, in the sense that the total profit attributed across “profit origin” jurisdictions – for a given “profit destination”-“ultimate parent” pair of jurisdictions – corresponds exactly to the total profit deemed shifted in that pair of jurisdictions. Ultimately, the average of these ${\lambda }_{j,k}$ factors can be compared to estimates of the profit sensitivity to tax rate differentials from the literature, as further discussed below.

242. A central question is the shape of the $f$ function, i.e. the relationship between profit shifting and tax rate differentials. This is important for the modelling of profit shifting in this section, but also because this assumption defines the way in which reduced tax rate differentials under Pillar Two will affect profit shifting intensity. The academic literature offers limited insights in this area. Most studies assume a linear relationship between profit shifting and statutory tax rate differentials and find evidence that this relationship is significant, but do not test empirically for other potential shapes (see for example Bradbury et al. (2018[11]) and Beer et al. (2019[12]) for recent reviews). Such a linear relationship is consistent with the theoretical framework of Huizinga and Laeven (2008[17]), which is based on the underlying assumption that the cost of profit shifting is quadratic. With this assumption, an “interior” solution to the MNE’s profit maximisation problem implies that profits are shifted in proportion to tax rate differentials.

243. However, some recent studies suggest that the position is more complex in reality. In particular, Bilicka (2019[18]) and Johannesen et al. (2019[19]) show that many MNE entities report zero profit in higher-tax jurisdictions. This would suggest that these MNEs are able to shift all their profit from these jurisdictions, in which case the solution to the MNE’s profit maximisation problem is not always “interior” and the relationship between profit shifting and tax rate differentials is no longer linear. Also, Dowd et al. (2017[20]) find a non-linear relationship according to which US MNEs shift more profits to jurisdictions with very low tax rates than a linear elasticity would imply.32 However, the question considered in that paper is somewhat different from the one considered in this chapter. Indeed, Dowd et al. (2017[20]) focus on the choice of potential “profit destination” jurisdictions for a given “ultimate parent” jurisdiction (in which case, the fact that MNEs shift as much profit as possible to jurisdictions with the lowest ETRs seems intuitive), while this chapter aims to identify “profit origin” jurisdictions for given “ultimate parent” and “profit destination” jurisdictions.

244. Another insight from the literature is that, all else equal, profit shifting tends to be more intense in lower income jurisdictions than in higher income ones. For example, Fuest et al. (2011[21]), based on German micro-data, find that profit shifting via intra-company loans is approximately twice as intense among the developing countries in their sample compared to other jurisdictions, which they suggest may be due to the limited capacity of these jurisdictions to enforce anti-tax avoidance policies. Johannesen et al. (2019[19]) find that a 10 percentage point decrease in foreign affiliates’ tax rates increases the likelihood that an MNE reports zero profits by 3 percentage points in low/middle income countries, but only by 1.7 percentage points in high income countries, based on firm-level data from ORBIS. Finally, Cobham and Janský (2018[22]) relying on macro data, find that “the intensity of losses is substantially greater in low‐income and lower middle‐income jurisdictions; and in sub‐Saharan Africa, Latin America and the Caribbean and in South Asia compared with other regions.”

245. Against this background, the baseline assumption in this chapter is that profit shifting is generally proportional to tax rate differentials, but more intense in lower income jurisdictions than in higher income ones. Formally, the function $f$is assumed to be defined as follows: $f\left({\tau }_i-\mathrm{ }{\tau }_j\mathrm{ }\right)={\alpha }_i.\left({\tau }_i-\mathrm{ }{\tau }_j\right)$ if $\left({\tau }_i-\mathrm{ }{\tau }_j\right)$ is positive, and zero otherwise. The coefficient ${\alpha }_i$ is equal to 1 in high income jurisdictions, 1.5 in middle income ones and 2 in low income jurisdictions (based on the World Bank classification of jurisdictions by income groups). This shape of the relationship between profit shifting and tax rate differentials is presented in Figure 3.8. As discussed above, the absolute amounts of shifted profit attributed to each “profit origin” jurisdiction will depend on the assumptions presented in this figure, but also on the ${\lambda }_{j,k}$ factors that capture the amount of deemed shifted profit in each “profit destination” jurisdiction (and for each “ultimate parent” jurisdiction).

Figure 3.8. Stylised shape of relationship between profit shifting intensity and tax rate differentials: Baseline shape

Note: Under the assumptions presented in this figure, the intensity of profit shifting to a jurisdiction is assumed to be proportional to the tax rate differential vis-à-vis this jurisdiction. Profit shifting intensity is assumed to be 1.5 times (resp. 2 times) higher for profit shifted from middle income (resp. low income) jurisdictions compared to high income jurisdictions. The amount of shifted profit (i.e. scale of the Y-axis) will depend on the ${\lambda }_{j,k}$ factors defined above, which capture the amount of deemed shifted profit in each “profit destination” jurisdiction $j$ (and for each “ultimate parent” jurisdiction $k$).

Source: OECD Secretariat.

Profit shifting: Alternative shapes of the relationship between profit shifting and tax rate differentials

246. To account for potential non-linearities in the shape of the relationship between profit shifting and tax rate differentials, alternative shapes are considered as robustness checks. These shapes are used for the purpose of creating uncertainty ranges around the estimates.

247. Two alternative shapes are considered. The first one is based on the idea that lower income jurisdictions may suffer from profit shifting for tax but also non-tax-related reasons. For example, investors may shift profits away from these jurisdictions due to fears of political instability and/or to circumvent capital controls (this may also be the case in high income jurisdictions, but it seems likely to be less frequent than in lower income jurisdictions). As a result, profit shifting may exist in these jurisdictions even for relatively low tax rate differentials and the subsequent slope of the relationship between profit shifting and tax rate differentials may be less steep than envisaged in the baseline (Figure 3.9, Panel A). This shape also takes into account the assumption that under relatively small tax rate differentials (i.e. less than 5 percentage points), MNEs may generally consider that the costs of profit shifting would exceed the gains (in terms of tax savings). This is why in Figure 3.9 (Panel A) profit shifting intensity does not increase with the tax rate differential as long as this differential remains below 5 percentage points.

248. The second alternative shape takes the same starting point, but also assumes that the slope of the relationship becomes less steep – across all income groups – above a certain level. The idea is that above a certain tax rate differential, MNEs’ profit shifting incentives are high anyway and, as a result, these incentives would not be greatly affected by marginal changes in tax rate differentials (Figure 3.9, Panel B).

Figure 3.9. Alternative shapes of the relationship between profit shifting intensity and tax rate differentials

Note: In Panel A, profit shifting intensity is assumed to be higher among lower income jurisdictions than higher income jurisdictions, as in the baseline (Figure 3.8). However, profit shifting in lower income jurisdictions is assumed to be partly driven by non-tax factors, which implies that lower income jurisdictions face profit shifting even under low (or even zero) tax rate differentials. In turn, reducing tax rate differentials in lower income jurisdictions tends to reduce profit shifting less than in the baseline scenario. In Panel B, this assumption is kept, and, in addition, profit shifting intensity is assumed to become less sensitive to tax rate differentials above a certain tax rate differential. This could reflect that for relatively high tax rate differentials, the incentive to shift profit is high anyway, and therefore less influenced by the exact level of the rate differential. In both panels, the amount of shifted profit (i.e. scale of the Y-axis) will depend on the ${\lambda }_{j,k}$ factors defined above, which capture the amount of deemed shifted profit in each “profit destination” jurisdiction $j$ (and for each “ultimate parent” jurisdiction $k$).

Source: OECD Secretariat.

3.6.2. Impact of Pillar Two on tax rate differentials and profit shifting

250. Pillar Two is expected to reduce tax rate differentials between jurisdictions. More precisely, it is expected to reduce the tax rate differential vis-à-vis jurisdictions where the ETR is currently below the minimum rate, by increasing this ETR up to the level of the minimum rate. In a scenario where governments in these jurisdictions do not react, as in Scenario 3, this minimum rate would be paid in another jurisdiction (e.g. the jurisdiction of the ultimate parent in the case of the IIR). If some of the jurisdictions with an ETR below the minimum rate increase their ETR up to the minimum tax rate, as envisaged in Scenario 4, then the tax would be paid in these jurisdictions.

251. In both of these cases, the incentives of MNEs to shift profit to low-tax jurisdictions would be reduced compared to a scenario without Pillar Two. It is also useful to note that these incentives would be the same regardless of whether governments in low-tax jurisdictions increase their ETR or not, since the MNE would face the same amount of tax on its shifted profit (the only difference being where this tax is paid). This is why the order in which the MNE and government reactions are considered in this chapter has relatively little impact on the final outcome, as mentioned earlier.34

252. Profit shifting intensity is assumed to be substantially reduced by the decline in tax rate differentials induced by Pillar Two. However, profit shifting is not expected to be completely eliminated by Pillar Two, since some tax rate differentials would persist between the minimum rate and the (higher) rate that applies in the jurisdictions with an ETR above the minimum rate (see stylised example in Figure 3.10). In practice, the decision of an MNE group to continue or not with a certain profit shifting scheme will depend on the tax rate differential after the implementation of Pillar Two and the costs of the profit shifting scheme (e.g. financial and advisory costs associated with setting up the scheme, reputational costs). In some cases, the MNE group may continue with the scheme, while in others it may find that the costs will outweigh the benefits after Pillar Two is introduced. It is difficult to assess how frequent the decision to reduce or even stop profit shifting will be, since the costs of profit shifting are not known with precision and will vary across MNE groups and jurisdictions.

253. In addition, MNE decisions may be more complex than a simple choice between continuing or stopping a profit shifting scheme, as a possible alternative may be to create a different scheme to shift profit to another jurisdiction (e.g. a jurisdiction with a higher ETR but to which the costs of shifting profit are lower) including potentially through changes in corporate structure.35 These complex reactions will depend on the particular circumstances and choices of each MNE group, and modelling them goes beyond the ambition of this chapter. If MNE reactions depart significantly from the stylised modelling assumptions used in this chapter, which are described below, this could significantly affect the estimated Pillar Two revenue effects, especially at the level of individual jurisdictions. At the global level, or for broad jurisdiction groups, the effect of different profit shifting reactions could partly cancel each other (e.g. if one jurisdiction in the group receives more profit, another jurisdiction in the group may receive less).

Figure 3.10. Stylised example of the impact of Pillar Two on profit shifting intensity

Note: This stylised example, based on fictitious numbers, illustrates how Pillar Two may reduce profit shifting but not eliminate it. In the current situation, most of the profit of the MNE group considered is located in Jurisdiction 1, which has a 0% tax rate. Most of this profit is deemed shifted, which is why little profit remains in Jurisdiction 1 in the counterfactual scenario without profit shifting. This profit is reallocated to Jurisdictions 2 and 3 based on the modelling on profit shifting described above. In the scenario where profit shifting is reduced by Pillar Two (assuming illustratively a 12.5% minimum tax rate), a greater share of profit remains in Jurisdiction 1 as the tax rate differential vis-à-vis Jurisdictions 2 and 3 is reduced but not eliminated.

Source: OECD Secretariat

254. Against this background, the stylised baseline assumption in this chapter is that profit shifting would be reduced proportionally to the reduction in tax rate differential induced by Pillar Two. For example, if Pillar Two reduces the tax rate differential between two jurisdictions by a third, profit shifting between these two jurisdictions is assumed to be reduced by a third. This assumption is consistent with the baseline modelling of profit shifting above.

255. However, this assumption overlooks the fact that (i) some profit shifting in lower income jurisdictions is related to non-tax factors and may persist after the introduction of Pillar Two, and (ii) that for high tax rate differentials, a marginal change in the tax rate differential may not lead to much of a change in the profit shifting intensity, since the profit shifting incentive is high anyway (put differently, the gain from profit shifting outweighs the costs in this case by such a large amount, that a marginal reduction in this gain would not change the choice of the MNE group to engage in profit shifting). These considerations are taken into account in the robustness check scenarios based on the two alternative shapes of the relationship between profit shifting intensity and tax rate differentials (Figure 3.9). In particular, with the second shape, profit shifting incentives become less sensitive to tax rate differentials when these differentials are relatively high. These two robustness check scenarios are used to build the uncertainty ranges around the estimates.

256. Based on the assumptions described above, an adjusted profit matrix taking into account the reduced profit shifting intensity is presented in Table 3.10 for an illustrative minimum tax rate of 12.5%. This is done for the baseline shape of the relationship between profit shifting intensity and tax rate differentials (Panel A) and for the two alternative shapes from Figure 3.9 (Panels B and C). Compared to Table 3.7, the total amount of profit in each column is the same (by construction), but the location of profit has shifted across matrix rows. Broadly speaking, profit has been reallocated away from investment hubs and into the other jurisdiction groups. The amount of profit in investment hubs is reduced by about 9-10% compared to Table 3.7 (depending on the panel considered) and the amount of profit in other jurisdiction groups is increased by 1-8% depending on the group and the panel considered. In particular, middle and low income jurisdictions tend to regain less profit under the alternative shapes of the relationship between profit shifting intensity and tax rate differentials (Panels B and C) than under the baseline (Panel A), which is consistent with the assumptions underlying these alternative shapes and discussed in the previous section.

3.6.3. Implications of a formulaic substance-based carve-out on profit shifting behaviour

257. The estimates in the previous section are based on a scenario where Pillar Two would not include a formulaic substance-based carve-out. This section discusses how the inclusion of a potential formulaic substance-based carve-out could modify the effect of Pillar Two on MNE profit shifting incentives.

258. A formulaic substance-based carve-out could have two main effects regarding MNE profit shifting behaviour. First, a carve-out might in theory reduce the effect of Pillar Two on profit shifting in specific situations where an MNE group would shift profit to a jurisdiction where it already has a substantial amount of tangible assets and/or payroll. In practice, this is not likely to be the case very often, as profit is frequently (and more likely) shifted to jurisdictions where MNE groups have minimal levels of substantive activity (see Table 3.5 for example). For example, assuming that, after the introduction of Pillar Two, (i) incentives to shift profit are not modified by Pillar Two as long as the shifted profit is ultimately carved out under Pillar Two, and that (ii) Pillar Two has the same effect on the shifting of ‘non-carved-out profit’ as in the no-carve-out case,36 then the effect of Pillar Two on the global amount of shifted profit would be reduced by less than 1% compared to a situation where Pillar Two is applied without carve-out.37

259. Second, the existence of a formulaic substance-based carve-out could encourage MNE groups to relocate tangible assets or employees to low-tax jurisdictions where they already shift profit (or could shift profit in the future) in order to benefit from the carve-out. This effect is difficult to model as it depends on the cost of relocating tangible assets or employees across jurisdictions, which is specific to each MNE group and each jurisdiction. In practice, tangible assets and employees are generally less mobile than intangible assets. Under the carve-out percentages considered in this chapter, it seems unlikely that a substantial amount of relocation would take place.38 Still, a formulaic substance-based carve-out may trigger attempts by tax planning MNEs to reclassify the location of tangible assets or employees for tax purposes without changing significantly their actual physical location, or to resort to other schemes to try to benefit from the carve-out without actually relocating tangible assets or employees. Reflecting this, it is important that the design of the carve-out is ‘abuse-proof’, in the sense that it opens carve-out rights only when justified by the actual location of tangible assets and/or employees.

260. Against this background, the estimates in this chapter assume that the effect of Pillar Two on profit shifting intensity are the same in the scenarios with a formulaic substance-based carve-out as in the no-carve-out scenario.

3.6.4. Global Pillar Two revenue gains under Scenario 3

261. Revenue gains from Pillar Two under Scenario 3 are computed based on the assumptions described in the previous sections and presented in Table 3.11. For simplicity, only point estimates are presented, and ranges reflecting uncertainty relative to pockets of low-taxed profits are not included in this table (these ranges are included in the summary tables presented in Section 3.9). The revenue gains consist of two components:

• Effect of the reduced profit shifting intensity: Profit that is no longer shifted is assumed to be taxed at the statutory CIT rate in the “profit origin” jurisdictions where it was generated. This assumption is consistent with the literature on profit shifting, which generally assumes that the marginal tax rate that would otherwise be applied to shifted profit is the statutory rate.39 In contrast, “profit destination” jurisdictions would receive less shifted profit, leading them to lose tax revenues. For these jurisdictions, it is assumed that the shifted profit was taxed at the average ETR on MNE income, reflecting that some of these jurisdictions tax MNE profit at a rate that is well below the statutory rate.40 Overall, at the global level, the net result is a tax revenue gain, since the tax rate applied in “profit origin” jurisdictions tends is higher than in “profit destination” ones.

• Revenues collected via the minimum tax: Even if MNE profit shifting intensity is reduced, some profit would remain located in jurisdictions where the ETR is below the minimum rate. Profit in these jurisdictions would be subject to the IIR and UTPR (except regarding US MNEs, where they would be subject to GILTI), which are modelled with the same assumptions as in Scenarios 1 and 2. The interaction with Pillar One is taken into account in the same way as in Scenario 2, based on the profit matrix adjusted for the new profit shifting patterns.

262. To reflect the uncertainties around the estimates, results in the final section of this chapter are presented as ranges. In the case of the effect of reduced profit shifting, the range is defined in the following way: for each jurisdiction, the bottom (top) point of the range is the minimum (maximum) value across the results obtained with the three shapes of the relationship between profit shifting intensity and tax rate differentials considered in this chapter (baseline and alternative shapes No. 1 and No. 2, as described in Figure 3.8 and Figure 3.9).

3.7.1. Assumptions on government reactions

266. The modelling of Scenario 4 in this chapter is based on the following assumptions on government reactions. These assumptions, which are summarised in Table 3.12, are consistent with those made on Pillar Two implementation in Section 3.4.5 above. They do not depend on the assumption made on a potential formulaic substance-based carve-out, even though, as discussed above, such a carve-out would tend to reduce the incentives for low-tax jurisdictions to increase their ETR.

• Group 1: Jurisdictions with an average ETR above the minimum rate. No reaction is modelled. This is because, as discussed above, the potential policy reaction of jurisdictions with an ETR above the minimum rate is a priori ambiguous.44

• Group 2: Jurisdictions with a zero corporate tax rate. For the purposes of the modelling in this chapter, jurisdictions in this group are not assumed to increase their ETR. The reason is that these jurisdictions generally do not have a CIT system. Introducing a corporate tax system from scratch would generate significant administrative costs, could lead these jurisdictions to be seen as less attractive investment destinations, and could also have spillovers on other sectors of their economies since non-MNE firms would possibly have to be subject to the newly-introduced CIT as well. Assuming that Pillar Two includes a formulaic substance-based carve-out, an additional argument is that an ETR increase would no longer be neutral for the overall amount of tax paid by MNEs, as discussed above. Under the assumption that GILTI would coexist with Pillar Two, the ETR increase would not be neutral for US MNEs either.

• Group 3: Jurisdictions with an average ETR below the minimum rate, but greater than zero. Jurisdictions in this group have a CIT system, so it is easier for them to increase their ETR than for jurisdictions in Group 2. Still, it seems plausible that not all of them will increase their ETR, including for the reasons discussed above. In addition, the ETRs faced by MNEs may not be homogeneous, as they reflect the application of different tax provisions (e.g. different kinds of tax deductions and preferential regimes), in which case it may not be straightforward to bring the ETR of all MNE entities up to the level of the minimum rate. Against this background:

  • The assumption in this chapter is that half of jurisdictions in Group 3 would increase their ETR on the profit of MNEs up to the level of the minimum rate under Pillar Two, while the other half would not. This is consistent with the assumptions on IIR and UTPR implementation, described in Section 3.4.5, where half of jurisdictions in this group are assumed to implement an IIR and a UTPR. Potential revenue gains related to an increase in the ETR of non-MNE entities are not taken into account in this chapter.

  • In practice, and as above, identifying the jurisdictions in Group 3 that would increase their ETR is not straightforward and a similar methodology as used for IIR and UTPR implementation in Scenario 1 is applied. Instead of arbitrarily selecting half of the jurisdictions in the group, a simplifying assumption is made that all jurisdictions in this group increase their ETR to close half of the gap vis-à-vis the minimum rate. Again, this does not aim to be realistic in itself, but rather to be a reasonable proxy for a situation where half of the jurisdictions in this group would increase their ETR to the level of the minimum rate and the other half would not change their ETR. For the purpose of creating uncertainty ranges around the estimates, it is assumed that between one-third and two-thirds of jurisdictions in this group would increase their ETR to the level of the minimum rate (instead of one-half in the baseline).

3.7.2. Scenario 4 methodology and results

267. The methodology to compute revenue gains from Scenario 4 is relatively straightforward, since the location of profit is the same as in Scenario 3 (Table 3.10). The main difference with Scenario 3 is the distribution of revenue gains across jurisdictions, as jurisdictions increasing their ETR capture a greater share of global revenue gains. In practice, jurisdiction-specific gains are computed consistently with Scenarios 1-3 above, with the additional assumption that profit in a jurisdiction increasing its ETR is taxed by this jurisdiction, instead of being subject to the IIRs and UTPRs of other jurisdictions (Figure 3.11).

Figure 3.11. Stylised illustration on Scenario 4 assumptions

Note: This stylised figure illustrates how the reaction of governments is taken into account in Scenario 4. For the sake of simplicity, this figure focuses on a scenario without formulaic substance-based carve-out. Compared to the similar illustration for Scenarios 1-3 (see Figure 3.5), half of the profit located in countries from group 3 is taxed by these countries as they increase their ETR. The other half of this profit remains subject to IIR and UTPR as in Scenarios 1-3.

Source: OECD Secretariat.

268. Global revenue gains in Scenario 4 are presented in Table 3.13. For simplicity, only point estimates are presented, and ranges reflecting uncertainty relative to pockets of low-taxed profits are not included in this table (these ranges are included in the summary tables presented in Section 3.9). Also for simplicity, results in Table 3.13 only focus on the baseline shape of the relationship between profit shifting intensity and tax rate differentials. Results for the two alternative shapes discussed above are presented in Annex 3.A.

3.9.1. Global revenue gains, excluding US MNEs

274. The estimates of net global revenue effects of Pillar Two across the four scenarios considered – excluding gains related to US MNEs – are summarised in Table 3.15. A range of illustrative minimum rates are included. Results are presented in absence of a formulaic substance-based carve-out (Panel A) or with a 10% carve-out on payroll and tangible asset depreciation (Panel B).

275. Results are presented as ranges to reflect the uncertainty. These ranges take into account the uncertainty around three factors:

• In all four scenarios, the amount of low-taxed profit subject to the IIR or UTPR is assumed to be subject to an uncertainty factor of ±10% around the point estimate. In addition, the upper bound of the range in Scenario 1 is increased by 50% in the case without carve-out, or 40% in the case with a 10% carve-out, to account for the uncertainty around pockets of low-taxed profit in high-tax jurisdictions, as discussed in Section 3.4.3. The lower bound of the range is unchanged. In Scenarios 2-4, the upper bound of the range is increased by the same absolute amount as in Scenario 1, reflecting that the absolute level of uncertainty around these pockets is the same across the four scenarios.

• In Scenarios 3 and 4, the effect of reduced profit shifting is assumed to be in a range between the minimum and the maximum estimates obtained across the three shapes of the relationship between profit shifting intensity and tax rate differentials considered in the analysis (i.e. the baseline shape of Figure 3.8 and the two alternative shapes presented in Figure 3.9). When considering global estimates (as opposed to jurisdiction-group-level estimates), the effects from using these different shapes partly offset each other. For example, when a group of jurisdictions tends to gain more revenues with a shape, the others groups tend to gain less, and the global estimate does not vary strongly (as can be seen in Table 3.11). To avoid that global ranges on the effect of reduced profit shifting would be too narrow because of this effect, an additional ±10% is added on account of the uncertainty (i.e. when building global ranges, -10% is applied to the minimum result across the three profit shifting shapes and +10% to the maximum result).

• In Scenario 4, the share of low-tax jurisdictions (from Group 3) increasing their ETR is assumed to be in a range between one-third and two-thirds. As discussed in Section 3.7.1, the baseline estimate is that half of jurisdictions in this group increase their ETR.

276. A more detailed breakdown of the contribution of the different components to this outcome is presented illustratively in Figure 3.12, using the case with a 12.5% minimum rate as an example, either with no carve-out (Panel A) or with a 10% carve-out on payroll and tangible asset depreciation (Panel B). Overall, these results suggest that the interaction with Pillar One has only a slight impact on the Pillar Two revenue gains (Scenario 2). The reduction in MNE profit shifting intensity leads to a substantial increase in the global revenue gains (Scenario 3). Finally, ETR increases in certain jurisdictions modify global revenue gains only slightly,47 but change significantly the distribution of these gains across jurisdictions (Scenario 4). Across all four scenarios, the effect of a formulaic substance-based carve-out on the estimated gains is relatively small.

3.9.2. Global revenue gains, including US MNEs

277. For the sake of completeness, it is interesting to consider global gains including revenue gains related to US MNEs. The gains related to US MNEs correspond to the effect of GILTI, under the illustrative assumption that GILTI would coexist with Pillar Two, using the estimates from the US JCT described in Section 3.8. An overview of the global results including US MNEs is presented in Table 3.16 for several illustrative minimum tax rates (10%, 12.5% and 15%) in two illustrative scenarios assuming either no carve-out or a 10% carve-out on payroll and tangible asset depreciation. Results focus on Scenario 3 (i.e. including interaction with Pillar One and MNE reaction). For example, assuming a 12.5% minimum tax rate, total gains combining direct and indirect effects of Pillar Two and revenue gains from GILTI could reach about 2.0-3.8% of CIT revenues.

Figure 3.12. Global revenue gains from Pillar Two, in % of global CIT revenues (excluding US MNEs)

Estimates in these figures exclude gains related to MNEs with an ultimate parent in the United States

Note: Scenario 1 is a static scenario without behavioural reactions. Scenario 2 takes into account the interaction of Pillar Two with Pillar One. Scenario 3 adds the reaction of MNEs in the form of lower profit shifting intensity. Scenario 4 adds the reaction of governments, in the form of some jurisdictions with an ETR below the minimum rate increasing their ETR to the minimum rate. The results focus illustratively on a scenario assuming a 12.5% minimum rate, with no formulaic substance-based carve-out (Panel A) or a 10% carve-out on payroll and tangible asset depreciation (Panel B). Consistent with the assumption that GILTI would coexist with Pillar Two, the estimates in these figures exclude revenues gains related to MNEs with an ultimate parent in the United States. The ranges between low and high estimates reflect data uncertainty, and the upper bound of the ranges also takes into account uncertainty related to pockets of low-taxed profit in higher-tax jurisdictions.

Source: OECD Secretariat.

3.9.3. Revenue gains for broad jurisdiction groups

278. Results by broad jurisdiction groups across the four scenarios and for several illustrative minimum tax rates (10%, 12.5% and 15%) are presented in Figure 3.13. The results exclude revenue gains related to US MNEs, and the group of high income jurisdictions excludes the United States. Panel A includes all four groups considered in the analysis (high, middle and low income, and investment hubs), while Panel B presents the same results excluding investment hubs (Panel A). Excluding investment hubs makes it possible to use a different scale and increases the readability of the results for the other groups.

279. High, middle and low income jurisdictions would gain revenues from Pillar Two under all four scenarios considered. Revenue gains tend to be larger among high income jurisdictions than lower income ones, reflecting that most MNE group ultimate parents are located in high income jurisdictions, implying that gains from the IIR primarily accrue to these jurisdictions. Still, gains in middle and low income jurisdictions are also significant, especially once the reduced profit shifting intensity of MNEs is taken into account (Scenarios 3 and 4).

280. Investment hubs would gain revenues in Scenarios 1 and 2, while they would gain less revenues on average in Scenario 3. They could even lose revenues in this scenario, depending on the level of the minimum tax rate (Figure 3.13, Panel A). This is because reduced MNE profit shifting would reduce the size of the tax base in a number of investment hubs where profit is currently shifted. However, the magnitude of this average revenue loss is difficult to measure with precision, as it depends on the effective tax rate on this shifted profit. The loss may be compounded by potential knock-on effects of reduced investment in investment hubs on other tax bases, which are not taken into account in the estimates. Also, it is important to note that results vary across investment hubs, which are a relatively heterogeneous group. Finally, in Scenario 4, investment hubs would on average benefit from relatively large revenue gains, reflecting the assumption that a number of them would increase their ETR on low-taxed profit. As investment hubs capture a higher share of global revenue gains, gains in other jurisdiction groups tend to be lower in Scenario 4 than in Scenario 3.

281. Results for hybrid scenarios taking into account MNE and government reactions (as in Scenarios 3 and 4) but not the interaction with Pillar One are presented in Annex 3.B and are qualitatively similar to the results in Figure 3.13. The results in Figure 3.13 assume illustratively a 10% carve-out on payroll and tangible asset depreciation. Results in a scenario without a formulaic substance-based carve-out are presented in Annex 3.E. They are also qualitatively similar to Figure 3.13, but revenue gains across all groups are slightly higher than in the scenario with a carve-out.

Figure 3.13. Pillar Two revenue gains by broad jurisdiction groups

Note: Each facet in the figure correspond to one of four scenarios (in columns) and one of three illustrative minimum tax rates (in rows). Results in Panel B are identical to results in Panel A, except that the investment hubs group is excluded, which makes it possible to use a different scale and increases the readability of results for the other groups. Scenario 1 is a static scenario without behavioural reactions. Scenario 2 takes into account the interaction of Pillar Two with Pillar One. Scenario 3 adds the reaction of MNEs in the form of lower profit shifting intensity. Scenario 4 adds the reaction of governments, in the form of some jurisdictions with an ETR below the minimum rate increasing their ETR to the minimum rate. The results focus illustratively on a scenario with a 10% carve-out on payroll and tangible asset depreciation (results without carve-out are presented in Annex 3.E). Groups of jurisdictions (high, middle and low income) are based on the World Bank classification. Investment hubs are defined as jurisdictions with a total inward FDI position above 150% of GDP. The high income group excludes the United States, in light of the illustrative assumption that GILTI would coexist with Pillar Two. The ranges reflect data uncertainty, and the upper bound of the ranges also takes into account uncertainty related to pockets of low-taxed profit in higher-tax jurisdictions.

Source: OECD Secretariat.