## Home or Away? Profit Shifting with Territorial Taxation ## Source details **Canonical URL:** [Home or Away? Profit Shifting with Territorial Taxation](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022177-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022177-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022177-print-pdf.pdf.json) --- ### Background and motivation - Tackling tax avoidance by multinational corporations has been on top of the international tax policy agenda since the global financial crisis of 2008. - The G20/OECD initiative on Base Erosion and Profit Shifting (BEPS) involves more than 130 countries under the Inclusive Framework. - Two broad systems for taxing foreign earnings: - Worldwide tax system: taxes resident corporations on worldwide earnings (with tax credits for foreign taxes). - Territorial tax system: exempts foreign earnings from domestic taxation. - Many countries have moved from worldwide to territorial systems over the last two decades; the US most recently with the 2017 Tax Cuts and Jobs Act (TCJA). - Compared to worldwide taxation, territorial taxation is associated with greater incentives for outward profit shifting. ### Research question and identification - Objective: provide causal evidence on the effect of territorial taxation on reported profits of multinationals. - Identification: exploit a 2009 UK policy reform that moved the UK from a worldwide to a territorial tax system by exempting foreign-source dividends received by UK multinationals from domestic taxable income. - Empirical approach: difference-in-differences using data on more than 50,000 multinational affiliates in Europe. - Treatment group: affiliates of UK multinationals. - Control group: affiliates of non-UK multinationals in the same country and year. - Mechanism expectation: affiliates active in countries with a lower tax rate than the UK should report higher pre-tax profit after 2009. ### Theoretical model — key expressions and predictions - Cost of misreporting: C_i = (휆/2) (휋_i)^2 p_i, where 휆>0 measures tax enforcement strength. - After-tax profits: - Π = (1−휏_h)[p_h − Σ_{i=1}^n 휋_i] + Σ_{i=1}^n (1−max(휏_h_repat, 휏_i))[p_i + 휋_i] − Σ_{i=1}^n (휆/2)(휋_i)^2 p_i. - Optimal shifted profit (first-order condition): - 휋_i / p_i = (휏_h − max(휏_h_repat, 휏_i)) / 휆. - Propositions: - Under a territorial system (휏_h_repat = 0), optimal reported profits increase in the tax differential 휏_h − 휏_i. - More profit shifting under a territorial system than under a worldwide system: - 휋_i / p_i (휏_h_repat > 0) < 휋_i / p_i (휏_h_repat = 0). ### Data and sample construction - Affiliate-level data: Bureau van Dijk's Amadeus database; unconsolidated balance sheets and ownership for European subsidiaries; zero-tax havens excluded. - Initial sample: 2006 to 2012; final unbalanced panel: 305,086 firm-year observations for 57,136 unique firms after exclusions (finance industry, public utilities, missing classifications, zero/negative assets or employees). - Definition relative to UK tax rate: - Low-tax countries (always lower than UK during observation): Austria, Bulgaria, Czech Republic, Estonia, Finland, Greece, Hungary, Ireland, Netherlands, Poland, Romania, Slovenia, Sweden, and some Portugal observations. - High-tax countries (always higher than UK during observation): Belgium, France, Germany, Italy, Luxembourg, Spain. - Main variables: EBT, operating profit, total assets, intangible assets, number of workers, turnover. - Winsorization: all firm-level ratio variables winsorized at 1 percent and 99 percent in full sample. ### Key affiliate-level summary statistics (exact figures) - Low-Tax Observations (Obs. = 155,076) - EBT (Euro): Mean 870,979; Median 31,907; Std. Dev. 4,802,172; Min -12,329,598; Max 46,736,000. - Operating Profit (Euro): Mean 889,759; Median 49,091; Std. Dev. 4,149,189; Min -9,870,000; Max 38,712,365. - EBT/Total assets: Mean 0.04; Median 0.04; Std. Dev. 2.56; Min -1.14; Max 0.73. - Total assets (Euro): Mean 16,075,129; Median 1,661,694; Std. Dev. 63,997,889; Min 13,594; Max 754,076,615. - Intangible assets (Euro) (Obs. = 152,169): Mean 264,648; Median 94; Std. Dev. 2,005,064; Min 0; Max 28,187,285. - Workers: Mean 97; Median 17; Std. Dev. 248; Min 1; Max 2,156. - Turnover (Euro): Mean 19,468,500; Median 2,032,027; Std. Dev. 67,916,729; Min 5,148; Max 675,585,141. - High-Tax Observations (Obs. = 150,010) - EBT (Euro): Mean 1,622,421; Median 144,026; Std. Dev. 7,044,888; Min -12,329,598; Max 46,736,000. - Operating Profit (Euro): Mean 1,514,149; Median 178,666; Std. Dev. 5,816,367; Min -9,870,000; Max 38,712,365. - EBT/Total assets: Mean 0.03; Median 0.04; Std. Dev. 0.20; Min -1.14; Max 0.73. - Total assets (Euro): Mean 39,501,772; Median 6,692,006; Std. Dev. 111,567,582; Min 13,594; Max 754,076,615. - Intangible assets (Euro) (Obs. = 148,473): Mean 1,039,982; Median 12,589; Std. Dev. 4,074,373; Min 0; Max 28,187,285. - Workers: Mean 121; Median 30; Std. Dev. 287; Min 1; Max 2,156. - Turnover (Euro): Mean 41,859,669; Median 9,015,245; Std. Dev. 103,072,763; Min 5,148; Max 675,585,141. ### Parent-level consolidated data (Datastream) — matched sample - Matches obtained: 595 ultimate owners (out of 1,883 parent companies). - Matched dataset: unbalanced panel 2005 to 2014 with 3,587 firm-year observations (non-missing main regression variables). - Selected consolidated summary statistics (Obs. = 3,587 unless noted) - Total assets (thous. Euro): Mean 2,258,277; Median 92,301; Std. Dev. 13,972,717; Min 13; Max 285,302,595. - Intangible assets (thous. Euro) (Obs. = 3,585): Mean 442,404; Median 12,713; Std. Dev. 1,766,721; Min 0; Max 22,357,000. - Foreign assets (thous. Euro) (Obs. = 1,544): Mean 84,396; Median 3,014; Std. Dev. 1,062,445; Min 0; Max 40,876,316. - Pre-tax income (thous. Euro): Mean 227,639; Median 3,800; Std. Dev. 1,831,377; Min -2,261,516; Max 39,823,617. - Employees: Mean 10,599; Median 526; Std. Dev. 46,760; Min 1; Max 648,254. - Note: 43 percent of consolidated firms have foreign assets. ### Empirical strategy - Baseline DiD regression (equation (5)): - Dependent variable: y_ikt (measure of subsidiary profitability). - Key regressor: UKParent_i × Post_t, with Post_t = 1 for years after 2009. - Controls: firm fixed effects (α_i); ln(number of workers) and ln(fixed assets) when appropriate; year dummies or host country-year and industry-year fixed effects; home-country controls Z_kt (inflation rate, GDP per capita, long-term unemployment rate, GDP growth). - Errors clustered by parent. - Dynamic DiD (equation (6)) tests for pre-trends with leads and lags (omitted category θ = −1). - Consolidated test (equation (7)) checks whether affiliate-level changes reflect overall group profitability changes. ### Baseline DiD results — affiliate-level profitability (exact coefficients and standard errors) - Main DiD (selected specifications, Table 3): - Column (1): UK-Parent × Post = 0.019*** (0.007) controlling only for firm labor inputs. - Column (2): UK-Parent × Post = 0.017*** (0.006) including ln(Turnover). - Columns (3) and (4): adding host country-year and industry-year FE: DiD coefficients 0.021 and 0.017. - Column (6): Operating Profit/Assets dependent variable: UK-Parent × Post = 0.022*** (0.007). - Column (7): interaction with High Intangibles shows stronger responses for firms with above-median intangible assets. - Column (8) (positive pre-tax earnings sample, ln(EBT)): UK-Parent × Post = 0.111** (0.045) — interpreted as an 11 percent increase in EBT among profitable affiliates. - Interpretation: - Affiliate-level profitability (EBT/Assets) increased by 1.7-2.1 percentage points for UK multinationals' affiliates in lower-tax countries after 2009. - The 11 percent increase in pre-tax earnings among profitable affiliates corresponds to this effect. - The magnitude is equivalent to a seven-percentage-point increase in the statutory corporate income tax rate in the UK (authors' comparison). ### Dynamic and placebo checks - Dynamic DiD (Figure 3): pre-reform coefficients close to zero, supporting parallel trends; positive and significant coefficients from 2009 onward; gradual increase after reform consistent with adjustment costs; post-reform coefficients not statistically different from each other. - High-tax placebo (Table 3, Column (9)): affiliates in high-tax countries belonging to UK MNC groups with no low-tax affiliates: UK-Parent × Post = 0.002 (insignificant). - Robustness to firm entry/exit (Table A.4, selected estimates): - UK-Parent × Post coefficients: - (1) Dropping firms observed only before reform: 0.022*** (0.007) - (2) Only firms observed after reform: 0.020*** (0.007) - (3) ≥ 2y before/after: 0.022*** (0.007) - (4) all 6 years: 0.012 (0.008) - ln(Workers) coefficients: 0.016*** (0.002), 0.016*** (0.002), 0.015*** (0.002), 0.010*** (0.002) - N: (1) 146,470; (2) 148,779; (3) 123,965; (4) 89,461. - R2: (1) 0.482; (2) 0.488; (3) 0.467; (4) 0.459. ### Where shifted profits originate (consolidated vs. affiliate evidence) - Consolidated regression (Table 5, Column (1)): DiD coefficient = 0.003 (insignificant); Sample: Parent Level, N = 3,587, R2 = 0.01. Interpretation: no change in consolidated profitability, consistent with profit shifting rather than higher overall multinational profitability. - High-tax affiliates within groups that also have low-tax affiliates (Table 5, Column (2)): DiD coefficient = -0.003 (insignificant); Sample: Affiliate Level, High tax, N = 142,319, R2 = 0.592. Interpretation: small negative effect in high-tax affiliates insufficient to explain most additional profits in low-tax affiliates. - Domestic (UK) affiliates of UK multinational groups (Table 5, Column (3)): DiD coefficient = -0.01 (significant at 5 percent); Sample: Affiliate Level, UK, N = 68,548, R2 = 0.578. Interpretation: consistent with profits shifted out of the UK to low-tax countries after the reform. ### Channels and heterogeneity - Stronger responses for firms with high intangible assets, consistent with profit shifting via royalty or intellectual property channels. - Transfer mispricing, relocation of intellectual property, and debt shifting are emphasized channels. - Operating Profit/Assets results (Column (6)) suggest internal debt is of minor importance relative to non-financial channels. ### Effects on measured productivity (TFP) - TFP measurement: - Primary: Levinsohn and Petrin (2003) approach (LP) using costs of goods sold as intermediate input proxy. - Robustness: factor share approach imposing constant returns to scale: ln TFP_it = ln(Y_it/L_it^k) − β^k ln(K_it/L_it^k). - Table 4 results (difference-in-differences on ln(TFP)): - Low-tax sample: - Column (1): UK-Parent × Post = 0.093** (0.046) for ln(TFP_FS) — implies ~9 percent increase. - Column (2): UK-Parent × Post = 0.046* (0.027) for ln(TFP_LP) — implies ~5 percent increase. - Columns (3)–(4): similar with host country-year and industry-year FE. - Columns (5)–(6): effect driven by firms with high intangible assets. - High-tax affiliates (Columns (7)–(8)): coefficients 0.003 and 0.005 (insignificant). - Sample sizes: examples include N = 65,698; 66,074; 65,696; 66,072; 90,681; 92,431. - R^2 examples: 0.989, 0.731, 0.989, 0.735, 0.989, 0.735, 0.989, 0.773. - Interpretation: measured TFP of low-tax UK affiliates increases by between 5 percent and 9 percent after the reform, consistent with measurement distortions due to profit shifting rather than real productivity gains in the short run. ### Policy implications - For reforming countries: countermeasures against profit shifting should accompany a switch to territorial taxation. - US example: concerns about outward profit shifting under territorial taxation motivated anti-avoidance measures accompanying the TCJA; the TCJA also reduced the federal corporate tax rate from 35 percent to 21 percent. - For host countries seeking to attract investment or book profits: territorial systems may intensify pressure to lower statutory tax rates. - Global scale: the increasing adoption of territorial taxation supports ideas to introduce a minimum tax on outbound investments (for example, similar in spirit to the Global Anti-Base Erosion (GloBE) tax proposed by the OECD) to alleviate profit shifting and tax competition. _italic: IMF Working Paper — Home or Away? Profit Shifting with Territorial Taxation_ ### 1. Introduction ........................................................................................................ ### 1. Introduction ### Background and motivation - Tackling tax avoidance by multinational corporations has been on top of the international tax policy agenda since the global financial crisis of 2008. - The G20/OECD initiative on Base Erosion and Profit Shifting (BEPS) involves more than 130 countries under the Inclusive Framework. - Two broad systems for taxing foreign earnings exist: - Worldwide tax system: taxes resident corporations on worldwide earnings (with tax credits for foreign taxes). - Territorial tax system: exempts foreign earnings from domestic taxation. - Many countries have moved from worldwide to territorial systems over the last two decades; the US most recently with the 2017 Tax Cuts and Jobs Act (TCJA). - Compared to worldwide taxation, territorial taxation is associated with greater incentives for outward profit shifting. ### Research question and identification - Research objective: provide causal evidence on the effect of territorial taxation on reported profits of multinationals. - Identification strategy: exploit a 2009 UK policy reform that moved the UK from a worldwide to a territorial tax system by exempting foreign-source dividends received by UK multinationals from domestic taxable income. - Empirical approach: difference-in-differences using data on more than 50,000 multinational affiliates in Europe. - Treatment group: affiliates of UK multinationals. - Control group: affiliates of non-UK multinationals in the same country and year. - Mechanism expectation: if an affiliate is active in a country with a lower tax rate than the UK, it should report higher pre-tax profit after 2009. ### Key empirical findings (affiliate-level) - Profitability (earnings before tax relative to total assets) increased by 1.7-2.1 percentage points for UK multinationals' affiliates in lower-tax countries after 2009. - This effect corresponds to an eleven percent increase in pre-tax earnings. - The magnitude is equivalent to a seven-percentage-point increase in the statutory corporate income tax rate in the UK. - Firms with high levels of intangible assets show especially strong responses, consistent with profit shifting via royalty or intellectual property channels. - Placebo test: UK multinationals with only affiliates in countries with a higher tax rate than the UK show no effect. ### Where shifted profits originate - Consolidated data on global profitability of UK multinationals (with domestic UK companies as control) show no effect of the reform, implying the affiliate-level effect is not due to higher overall profitability of UK multinationals. - Affiliate-level evidence indicates most additional profits shifted to low-tax countries originate in the UK headquarter and its domestic affiliates, rather than being shifted away from other high-tax foreign affiliates. ### Measurement distortions and productivity - Profit shifting distorts firm-level productivity measures. - Measured total factor productivity (TFP) of affiliates of UK multinationals in low-tax countries increases by 5-9 percent after the reform. - No significant effects on TFP are found in high-tax countries, suggesting the TFP increases in low-tax countries reflect profit shifting rather than real productivity gains. ### Institutional detail on the UK reform (2009) and related features - Pre-2009: UK taxed worldwide profits with deferral; foreign profits taxed upon repatriation with a foreign tax credit capped at the UK tax liability. - Post-2009: UK excluded profits that a British firm receives from foreign subsidiaries from domestic taxable income (territorial system); the exemption is 100 percent for a wide range of foreign-source dividends, including profits accumulated before the legislation. - Unlike the TCJA, the UK reform did not impose any deemed tax on unrepatriated profits or any minimum tax without deferral. - The UK tightened its controlled foreign company (CFC) regime after the reform, but the new CFC regime took effect in January 2013 (after the study period). Under the pre-2013 CFC regime, exemptions (including for actively trading subsidiaries) left scope for avoiding UK taxation by mixing passive and active income. ### Channels of profit shifting emphasized in the literature and relevant here - Transfer mispricing: manipulating internal transaction prices so low-tax affiliates report higher profits. - Relocation of intellectual property: moving IP to low-tax jurisdictions and charging royalties. - Debt shifting: borrowing from low-tax affiliates so interest payments relocate profit. ### Policy implications - For reforming countries: countermeasures against profit shifting should accompany a switch to territorial taxation. - US example: concerns about outward profit shifting under territorial taxation motivated anti-avoidance measures accompanying the TCJA; the TCJA also reduced the federal corporate tax rate from 35 percent to 21 percent. - For host countries seeking to attract investment or book profits: territorial systems may intensify pressure to lower statutory tax rates. - On a global scale: the increasing adoption of territorial taxation supports ideas to introduce a minimum tax on outbound investments (for example, similar in spirit to the Global Anti-Base Erosion (GloBE) tax proposed by the OECD) to alleviate profit shifting and tax competition. ### Contribution to the literature - First causal evidence on how adopting a territorial tax system affects the level of reported profits by multinationals. - Adds to theoretical and empirical debates on worldwide vs territorial systems, tax-induced distortions to investment and allocation, and the measurement of multinational firms' real activity. _italic: IMF Working Paper — Home or Away? Profit Shifting with Territorial Taxation_ ### 2.2 A Simple Model of Profit Shifting with Tax Regime Change ### 2.2 A Simple Model of Profit Shifting with Tax Regime Change ### Model setup and assumptions - MNC with headquarter in country h and affiliates in n countries. Economic profits in country i denoted p_i. - Profit shifting amount from headquarter to affiliate i (or vice versa) denoted 휋_i. - Cost of misreporting (quadratic in shifting degree): - C_i = (휆/2) (휋_i)^2 p_i, where 휆>0 measures tax enforcement strength. - Under worldwide system, dividend repatriation tax is 휏_h_repat; foreign taxes are credited so total tax on foreign profits is max(휏_h_repat, 휏_i). - Under territorial system, 휏_h_repat = 0. - Profit shifting costs are not tax deductible (results qualitatively same if deductible — see Appendix A.1). ### After-tax profits expression - After-tax profits (Π): - Π = (1−휏_h)[p_h − Σ_{i=1}^n 휋_i] + Σ_{i=1}^n (1−max(휏_h_repat, 휏_i))[p_i + 휋_i] − Σ_{i=1}^n (휆/2)(휋_i)^2 p_i. ### Optimal profit shifting (first-order condition) - Optimal shifted profit relative to economic profit: - 휋_i / p_i = (휏_h − max(휏_h_repat, 휏_i)) / 휆. ### Propositions and comparative statics - Proposition 1 (Optimal Reported Profitability): - (i) Under a territorial tax system, optimal reported profits increase in the tax differential 휏_h − 휏_i. - (ii) Under a worldwide tax system, optimal reported profits increase in the adjusted differential 휏_h − max(휏_h_repat, 휏_i). If 휏_h_repat = 휏_h > 휏_i, no profit is shifted abroad. - Intuition: Repatriation tax under worldwide taxation reduces the tax-efficiency of shifting abroad (permits deferral only), thus lowering incentives to distort reported profits compared with a territorial system. - Proposition 2 (Tax Reform and Reported Profitability, Main Prediction): - More profit shifting under a territorial system than under a worldwide system: - 휋_i / p_i (휏_h_repat > 0) < 휋_i / p_i (휏_h_repat = 0). --- ### 3. Data and Descriptive Analysis ### 3.1 Affiliate-level data (Amadeus) - Data source: Bureau van Dijk's Amadeus database; unconsolidated balance sheets and ownership for European subsidiaries. Zero-tax havens are not included. - Sample construction and exclusions: - Initial sample: 2006 to 2012. - Exclude finance industry and public utilities; drop observations with missing industry classification or zero/negative assets or employees. - Final sample: unbalanced panel with 305,086 firm-year observations for 57,136 unique firms. - Definition of “low-tax" and “high-tax" observations (relative to UK in years observed): - Low-tax countries (always lower than UK during observation): Austria, Bulgaria, Czech Republic, Estonia, Finland, Greece, Hungary, Ireland, Netherlands, Poland, Romania, Slovenia, Sweden, and some Portugal observations. - High-tax countries (always higher than UK during observation): Belgium, France, Germany, Italy, Luxembourg, Spain. - Main variables: EBT, operating profit, total assets, intangible assets, number of workers, turnover. - Winsorization: all firm-level ratio variables winsorized at 1 percent and 99 percent in full sample (low-tax and high-tax combined). ### Table 1 — Summary statistics: Affiliate-level (exact figures) - Low-Tax Observations (Obs. = 155,076) - EBT (Euro): Mean 870,979; Median 31,907; Std. Dev. 4,802,172; Min -12,329,598; Max 46,736,000. - Operating Profit (Euro): Mean 889,759; Median 49,091; Std. Dev. 4,149,189; Min -9,870,000; Max 38,712,365. - EBT/Total assets: Mean 0.04; Median 0.04; Std. Dev. 2.56; Min -1.14; Max 0.73. - Total assets (Euro): Mean 16,075,129; Median 1,661,694; Std. Dev. 63,997,889; Min 13,594; Max 754,076,615. - Intangible assets (Euro) (Obs. = 152,169): Mean 264,648; Median 94; Std. Dev. 2,005,064; Min 0; Max 28,187,285. - Workers: Mean 97; Median 17; Std. Dev. 248; Min 1; Max 2,156. - Turnover (Euro): Mean 19,468,500; Median 2,032,027; Std. Dev. 67,916,729; Min 5,148; Max 675,585,141. - High-Tax Observations (Obs. = 150,010) - EBT (Euro): Mean 1,622,421; Median 144,026; Std. Dev. 7,044,888; Min -12,329,598; Max 46,736,000. - Operating Profit (Euro): Mean 1,514,149; Median 178,666; Std. Dev. 5,816,367; Min -9,870,000; Max 38,712,365. - EBT/Total assets: Mean 0.03; Median 0.04; Std. Dev. 0.20; Min -1.14; Max 0.73. - Total assets (Euro): Mean 39,501,772; Median 6,692,006; Std. Dev. 111,567,582; Min 13,594; Max 754,076,615. - Intangible assets (Euro) (Obs. = 148,473): Mean 1,039,982; Median 12,589; Std. Dev. 4,074,373; Min 0; Max 28,187,285. - Workers: Mean 121; Median 30; Std. Dev. 287; Min 1; Max 2,156. - Turnover (Euro): Mean 41,859,669; Median 9,015,245; Std. Dev. 103,072,763; Min 5,148; Max 675,585,141. ### Pre-tax profitability measures and distributional features - Preferred profitability measure: EBT/Total assets (to retain zero and negative EBT observations). - Alternative measure used in robustness: ln(EBT) (but excludes zero/negative EBT). - Data features: - Around 31 percent of observations report negative or zero EBT. - Bunching at zero is evident: in both low-tax and high-tax groups, more than 5 percent of companies report return to assets between 0 and 0.01 percent. - Implication: excluding loss-making affiliates biases down estimates of profit shifting. ### Total Factor Productivity (TFP) measurement - Production function specification (log Cobb-Douglas): - ln Y_it^k = β_0^k + β_K ln K_it^k + β_L ln L_it^k + β_M ln M_it^k + ω_it + ε_it. - Primary TFP measure: Levinsohn and Petrin (2003) approach (LP), using costs of goods sold as intermediate input proxy. - Robustness TFP measure: factor share approach imposing constant returns to scale: - ln TFP_it = ln(Y_it/L_it^k) − β^k ln(K_it/L_it^k), where β^k is cost-share of capital computed as mean of one minus labor share in value added for each country-industry pair. ### 3.2 Data on UK parents (Datastream consolidated data) - Merge: unconsolidated Amadeus data merged to consolidated UK parent data by ultimate owner's ISIN. - Matches obtained: 595 ultimate owners (out of 1,883 parent companies). - Matched dataset: unbalanced panel 2005 to 2014 with 3,587 firm-year observations (non-missing main regression variables). - Table 2 — Summary statistics: Consolidated firms (exact figures; Obs. = 3,587 unless noted) - Total assets (thous. Euro): Mean 2,258,277; Median 92,301; Std. Dev. 13,972,717; Min 13; Max 285,302,595. - Intangible assets (thous. Euro) (Obs. = 3,585): Mean 442,404; Median 12,713; Std. Dev. 1,766,721; Min 0; Max 22,357,000. - Foreign assets (thous. Euro) (Obs. = 1,544): Mean 84,396; Median 3,014; Std. Dev. 1,062,445; Min 0; Max 40,876,316. - Pre-tax income (thous. Euro): Mean 227,639; Median 3,800; Std. Dev. 1,831,377; Min -2,261,516; Max 39,823,617. - EBIT (thous. Euro) (Obs. = 3,537): Mean 252,244; Median 5,108; Std. Dev. 1,885,164; Min -2,112,087; Max 40,265,783. - Total Debt (thous. Euro): Mean 487,068; Median 6,729; Std. Dev. 2,235,239; Min 0; Max 37,661,580. - Employees: Mean 10,599; Median 526; Std. Dev. 46,760; Min 1; Max 648,254. - Additional note: 43 percent of consolidated firms have foreign assets. ### 3.3 Descriptive analysis of UK parent firms around 2009 reform - Empirical question: Did UK firms shift more profits abroad after UK's 2009 change to a territorial tax system? - Figure 2 findings (parent-firm-level, Datastream 2005-2014): - Panel A (Effective foreign tax rates): Before 2009, many UK multinationals had effective foreign tax rates clustered near the UK tax rate of 29 percent. After 2009, this bunching partially diffuses, with more firms having lower effective foreign tax rates (around 20 percent). - Panel B (Statutory foreign CIT rates): A gradual decline in statutory CIT rates could shift distributions left, but the diffusion of bunching around the pre-reform UK tax rate is more clearly associated with the tax regime change. - Interpretation: The repatriation-tax removal under the territorial regime is associated with increased incentives and observed patterns consistent with profit shifting to lower-tax jurisdictions; not all profits shifted to tax havens due to requirements for substantive activities and uses of retained earnings for subsidiary investment. *Source: IMF Working Paper — section "2.2 A Simple Model of Profit Shifting with Tax Regime Change" and related data description in Sections 3.1–3.3.* ### 4. Empirical Strategy ### 4. Empirical Strategy ### Identification and baseline specification - Identification relies on variation in the incentive for outward profit shifting arising from the UK's switch to a territorial tax system in 2009. - Main empirical approach: difference-in-differences (DiD) comparing affiliates of UK-based multinationals (treated) with affiliates of non-UK-owned multinationals (control) in the same host country. - Expected sign: positive DiD coefficient on reported profitability in low-tax countries. - Baseline regression (equation (5)): - Dependent variable: y_ikt = measure of profitability of subsidiary i in country k in year t. - Key regressor: UKParent_i × Post_t, where UKParent_i = 1 for UK-owned affiliate, Post_t = 1 for years after 2009. - Controls: - Firm fixed effects (α_i) to control time-invariant subsidiary and parent characteristics. - Time-varying input factor proxies X_ikt: typically ln(number of workers) and ln(fixed assets) when not part of dependent variable. - Year dummies (d_t) or time-varying host country and industry fixed effects in alternative specifications. - Home-country controls Z_kt: inflation rate, GDP per capita, long-term unemployment rate, GDP growth. - Error term: ε_ikt. ### Common trends and dynamic specification - Identifying assumption: absent the reform, treated and control firms would have similar trends in reported profitability. - Dynamic DiD (equation (6)) estimated to test pre-trends: - y_ikt = Σ_{θ=-3}^{3} δ_θ 1[t=θ] × UKParent_i + β_x X_ikt + β_z Z_kt + α_i + d_t + ε_ikt. - Omitted category: θ = −1; leads and lags allow testing for pre-existing trends and dynamic responses. ### Potential group-level (consolidated) effects - Concern: affiliate-level profitability changes could reflect overall group profitability changes post-reform (e.g., lower cost of capital). - Consolidated test (equation (7)): - y_jt = γ_1 MNC_j × Post_t + γ_x X_jt + η_j + d_t + ε_jt. - MNC_j = 1 if the UK company group has foreign affiliates. - Expectation: γ_1 should be insignificant if affiliate-level changes reflect profit shifting (which cancels out at consolidated level). --- ### 5. Results ### Baseline results (main DiD; specification (5)) - Data: Amadeus for 2006-2012. - Dependent variables used across specifications include EBT/Assets, Operating Profit/Assets, ln(EBT), and EBT/Assets for various subsamples. - Key findings from Table 3: - Column (1): UK-Parent × Post = 0.019*** (standard error 0.007) when controlling only for firm labor inputs. - Column (2): including ln(Turnover) yields a slightly smaller coefficient on UK-Parent × Post (0.017***; se 0.006). - Columns (3) and (4): adding host country-year and industry-year fixed effects: - DiD coefficients: 0.021 (without turnover) and 0.017 (with turnover). - Column (5): omitting employment does not affect the estimated coefficient relative to Column (3). - Column (6): using Operating Profit/Assets as dependent variable: UK-Parent × Post = 0.022*** (se 0.007). - Interpretation: robustness to potential double counting of profits and indication that profit shifting via internal debt is of minor importance. - Column (7): interaction with High Intangibles shows firms with above-median intangible assets reacted more strongly to the reform. - Column (8): among affiliates with positive pre-tax earnings, ln(EBT) dependent variable: - UK-Parent × Post = 0.111** (se 0.045). - Interpretation: among profitable affiliates, EBT increases by 11 percent following the territorial tax reform. - Column (3) implies for the full sample: pre-tax EBT/Assets (pre-reform mean 0.06 for UK-owned firms) increases by 35 percent for the full sample (effect three times larger than when focusing on profitable affiliates). - Additional table statistics and sample sizes (selected): - N in many specifications: 155,076 (Columns (1)–(7)); Column (8) N = 97,357; Column (9) N = 150,010. - R^2 ranges reported (examples): 0.491, 0.526, 0.495, 0.529, 0.494, 0.485, 0.495, 0.851, 0.591. - Notes on estimation: - Specifications (3)–(9) include home-country characteristics: inflation, GDP per capita, long-run unemployment rate, GDP growth. - Standard errors clustered by parent. Significance: ***, **, * at 1 percent, 5 percent, 10 percent. ### Placebo and robustness checks - Dynamic DiD (equation (6)) results (Figure 3): - Pre-reform coefficients are close to zero, supporting the parallel trends assumption. - Positive and statistically significant coefficients from 2009 onward indicate the reform induced low-tax affiliates of UK multinationals to report more profits. - Gradual increase in profitability after reform consistent with delayed responses due to adjustment costs. - Caution: post-reform coefficient estimates are not statistically different from each other. - High-tax placebo (Table 3, Column (9)): - For affiliates in high-tax countries that belong to UK MNC groups with no low-tax affiliates: - UK-Parent × Post = 0.002 (statistically insignificant). - Interpretation: no incentive and no evidence of shifting profits to high-tax countries. - Consolidated (group-level) test expectation: - With consolidated data capturing overall group profitability, which is invariant to inter-company profit shifting, γ_1 should be insignificant if affiliate-level effects reflect profit shifting. ### Effects on measured TFP - Motivation: profit shifting through transfer pricing and licensing can inflate measured productivity (sales and/or capital), while internal financing should not affect TFP measures absent changes in operating profit or turnover. - Expectation based on literature: non-financial inter-company transactions account for about 70-80 percent of the response to taxation; thus measured TFP could increase by about 8-9 percent given an overall pre-tax profit increase of 11 percent. - Results from Table 4 (difference-in-differences on ln(TFP)): - Dependent variables: ln(TFP_FS) (factor-share measure) and ln(TFP_LP) (Levinsohn-Petrin measure). - Low-tax sample: - Column (1): UK-Parent × Post = 0.093** (se 0.046) for ln(TFP_FS). - Column (2): UK-Parent × Post = 0.046* (se 0.027) for ln(TFP_LP). - Interpretation: increase in measured TFP between 5 percent and 9 percent for low-tax UK affiliates after the territorial reform. - Columns (3) and (4): replacing year FE with host country-year and industry-year FE yields very similar results. - Columns (5) and (6): effect driven by firms with high intangible assets. - High-tax affiliates (Columns (7) and (8)): no similar increase in observed TFP for UK affiliates (coefficients 0.003 and 0.005, insignificant). - Sample sizes: examples include N = 65,698; 66,074; 65,696; 66,072; 90,681; 92,431. - R^2 examples: 0.989, 0.731, 0.989, 0.735, 0.989, 0.735, 0.989, 0.773. - Specifications (3)–(8) include home-country characteristics: inflation, GDP per capita, long-run unemployment rate, GDP growth. - Standard errors clustered by firm. - Interpretation and caveats: - Results consistent with profit shifting being concentrated in firms with high intangible assets and with non-financial channels (transfer pricing, licensing) being dominant. - Long-term real TFP effects (e.g., via R&D and innovation) may appear with significant delay; immediate measured TFP increases likely reflect mismeasurement from profit shifting. *Source: IMF Working Paper — “Home or Away? Profit Shifting with Territorial Taxation” (Sections 4–5, empirical strategy and results; data: Amadeus for 2006-2012).* ### 5.2 Where Is the Shifted Profit Coming From? ### 5.2 Where Is the Shifted Profit Coming From? ### Evidence from consolidated vs. affiliate-level data - Regression of eq. (7) on consolidated data (Column (1), Table 5) yields a DiD coefficient of 0.003 (insignificant). - Interpretation: profitability of consolidated accounts of UK MNCs remains approximately constant after the reform, indicating that the effects found on EBT at the affiliate level arise because of profit shifting rather than higher overall multinational profitability. - Table 5 sample and fit for Column (1): - Sample: Parent Level - N = 3,587 - R2 = 0.01 ### Evidence from UK-owned affiliates in high-tax countries (within groups that also have low-tax affiliates) - Column (2), Table 5 reports a DiD coefficient of -0.003 (insignificant at conventional levels). - Interpretation: at the margin, some profits may be relocated from high- to low-tax countries by UK multinationals after the reform, but the effect is too small to explain most of the additional profits in low-tax affiliates. - Table 5 sample and fit for Column (2): - Sample: Affiliate Level, High tax - N = 142,319 - R2 = 0.592 ### Evidence from domestic (UK) affiliates of UK multinational groups - Column (3), Table 5 compares UK affiliates of UK MNCs to UK affiliates of foreign MNCs and reports a DiD coefficient of -0.01, significant at the 5 percent level. - Interpretation: consistent with the notion that profits are shifted out of the UK to low-tax countries after the territorial reform. - Table 5 sample and fit for Column (3): - Sample: Affiliate Level, UK - N = 68,548 - R2 = 0.578 ### Summary of Table 5 estimation controls and covariates - Columns use combinations of controls including ln(Workers), additional controls, affiliate fixed effects, parent fixed effects, year fixed effects, host country-year fixed effects, and industry-year fixed effects as indicated in Table 5. - Coefficients on ln(Workers) in Table 5: - Column (1): -0.011 - Column (2): 0.009*** (standard error 0.002) - Column (3): -0.007* (standard error 0.004) *IMF Working Paper — Home or Away? Profit Shifting with Territorial Taxation* ### 2012. Specification (3) to (8) include the following characteristics of the multinational firm's home country: inflation ### wpiea2022177-print-pdf - 2012. Specification (3) to (8) include the following characteristics of the multinational firm's home country: inflation ### Key empirical specification details - Specification (3) to (8) include the following characteristics of the multinational firm's home country: inflation, GDP per capita, long-run unemployment rate, and GDP growth. - Standard errors in parentheses are clustered by parent. - ***, ** and * indicate significance at the 1 percent, 5 percent, and 10 percent levels. - Dependent variable: earnings before taxes divided by total assets. - Data on low-tax countries from Amadeus for 2006-2012. - Controls and fixed effects applied: - Add. Controls: ✓ in all reported specifications. - Affiliate FE: ✓ in all reported specifications. - Host Country-Year FE: ✓ in all reported specifications. - Industry-Year FE: ✓ in all reported specifications. ### Table A.4 — Profit Shifting after Territorial: Robustness Regarding Firm Entry & Exit (selected estimates) - Columns correspond to sample variations: - (1) Dropping firms observed only before reform - (2) Only firms observed after reform - (3) ≥ 2y before/after - (4) all 6 years - UK-Parent × Post coefficients: - (1): 0.022*** (0.007) - (2): 0.020*** (0.007) - (3): 0.022*** (0.007) - (4): 0.012 (0.008) - ln (Workers) coefficients: - (1): 0.016*** (0.002) - (2): 0.016*** (0.002) - (3): 0.015*** (0.002) - (4): 0.010*** (0.002) - Sample sizes (N): - (1): 146,470 - (2): 148,779 - (3): 123,965 - (4): 89,461 - R-squared (R2): - (1): 0.482 - (2): 0.488 - (3): 0.467 - (4): 0.459 ### Substantive findings and robustness implications - The positive and statistically significant UK-Parent × Post coefficients in columns (1)–(3) (0.022***, 0.020***, 0.022***) indicate an increase in earnings before taxes divided by total assets for UK-parented affiliates after the reform in those samples. - The UK-Parent × Post coefficient in the full 6-year panel (column (4)) is 0.012 and not marked as significant given the reported standard error (0.008). - The ln (Workers) coefficient is positive and statistically significant across all columns (0.016***, 0.016***, 0.015***, 0.010***), indicating a robust positive relationship between firm size (workers) and earnings before taxes divided by total assets in these samples. - Results are robust to dropping firms observed only before reform, keeping only firms observed after reform, and restricting to firms observed ≥ 2 years before and after reform, though significance in the full 6-year sample is attenuated. *Home or Away? Profit Shifting with Territorial Taxation — Working Paper No. WP/2022/177* --- _Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022177-print-pdf.pdf_