## 6. Regression Results – Tax Base Gainers, Breakdown of Income Balance

## Source details

**Canonical URL:** [6. Regression Results – Tax Base Gainers, Breakdown of Income Balance](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021041-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021041-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021041-print-pdf.pdf.json)

---

### Introduction — context and key literature findings
- Accurate balance of payments (BoP) data are essential for macroeconomic analysis; multinational profit-shifting strategies (transfer prices, corporate debt, location of intangible assets) directly affect international flows without necessarily changing production or employment.
- Example revision: Ireland revised up GDP growth from 7.8 to 26.3 percent as a result of some transactions involving multinational enterprises.
- Meta-study result: on average a country loses 1.5 percent of its corporate tax base for each percentage point increase in the corporate income tax rate.
- Reported aggregate global tax-base loss estimates in literature:
  - 2 percent of the global tax base (Beer et al., 2020)
  - over 8 percent (Tørsløv, Wier and Zucman, 2018)
  - 23 percent (Cobham and Janský, 2018)
- Crivelli, De Mooij, and Keen (2016): long-run revenue loss just below 1 percent of GDP for OECD members and 1.3 percent of GDP for other countries.

### Three main messages from the analysis
- Double-entry BoP accounting insulates the aggregate current account from distortion by profit shifting, but sub-accounts (trade, income, bilateral current account balances) are distorted.
  - Transfer-price example: underpricing related-party exports and overpricing imports understates the trade balance of the headquarters country; shifted profits recorded as dividends or reinvested earnings offset at the current-account aggregate level.
- Measurement errors can be sizable in practice, especially in the income balance due to challenges in recording retained earnings of affiliates abroad; export and import data (goods) tend to be higher quality.
  - If the offsetting income-account effect is harder to detect, there will be a resulting gap in the measured current account.
- Tax payments to the country gaining tax base create a real effect on the current account.
  - Quantifying this requires the (unobservable) net amount of shifted profits in each country.

### Empirical regression results and decomposition
- Regression summary results:
  - A 10-percentage point higher corporate income tax rate is, on average, associated with a reduction in the trade balance by 1 percent of GNI.
  - For the average country, this trade effect is roughly offset by an increase of the same magnitude in the net income balance.
  - For a subset of likely tax base gainers, the income balance offsets only about 4 percentage points of the improvement in their trade balances, implying nonlinearity.
  - In the group of tax base gainers, the trade account effect dominates and the current account strengthens by about 6 percent of GNI, on average.
- Decomposition for tax base gainers into real tax effect and measurement error:
  - Based on country-by-country estimates of profit shifting (Tørsløv, Wier, and Zucman, 2018), the genuine tax effect on the current account averages 2.1 percent of GNI in the group of tax base gainers.
  - The measurement error in the net income balance of tax base gainers is estimated at about 3.9 percent of GNI, on average.
- Related empirical magnitudes cited:
  - Carloni et al. (2019): the 2017 US tax reform improves the US trade balance by about 9 percent via tax-motivated related-party trade.
  - Guvenen et al. (2019): outbound profit shifting from US multinationals affects measured aggregate productivity in the United States by magnitudes ranging from 0.09 to 0.24 percent.

### Theory — channels and BoP accounting implications
- Transfer price manipulation
  - Arm’s length principle often hard to apply for unique or intangible intra-firm transactions, giving multinationals leeway to allocate profits to low-tax jurisdictions.
  - Theoretical neutrality: transfer price manipulation changes composition between trade and income balances but not aggregate current account, ignoring taxes.
  - When taxes are included, profit shifting to a foreign jurisdiction that taxes the profit causes a net weakening of the current account of the sending (high-tax) country by the tax paid to the foreign government.
  - Table summaries (conceptual): Examples show aggregate current account unchanged ignoring tax; with tax, current account changes by xtLT or -xtLT depending on country, and shifts via affiliates imply effects proportional to x(tHT-tLT).
- International debt shifting
  - Use of intercompany debt or third-party debt allocation can shift profits via deductible interest.
  - Related-party loans: interest payments affect income accounts bilaterally but offsetting retained earnings/dividend effects leave net income unchanged absent taxes.
  - Third-party borrowing bundled in high-tax jurisdictions can change bilateral current accounts while leaving aggregate current accounts unchanged, with real effects again arising via taxes paid on shifted profits.
- Location of intangible assets (IP)
  - Legal ownership of patents, trademarks in low-tax jurisdictions generates royalties recorded as trade in services; shifting the asset changes trade and income account composition.
  - Sales of patents/copyrights enter trade in R&D services (affect trade account and capital account offsets); non-produced non-financial assets enter the capital account without affecting balances.
  - Pricing such transfers is difficult due to private information about future returns.
- Treaty shopping
  - Use of conduit/intermediate entities can exploit bilateral withholding tax differences.
  - Example: if LT imposes no withholding tax, a $1 dividend outflow is fully distributed; if LT imposes withholding tax rate t, the dividend outflow reduces the income balance by $1 – t.
  - Treaty shopping that accesses zero-withholding treaty networks neutralizes withholding tax effects from perspective of sender and ultimate recipient; intermediate countries collect any nonzero withholding taxes, improving their income balance by the amount collected.

### Data collection and measurement issues
- Trade data (goods) are generally based on administrative records (IMTS) and typically more reliable; trade in services often relies on ITRS or surveys with variable coverage.
- Income account data are more likely survey-based; ITRS may not capture ultimate counterpart or distinguish portfolio vs direct income; reinvested earnings often collected via surveys with variable quality.
- Given data differences, profit shifting may affect measured but not true current accounts if income-account offsets (especially reinvested earnings) are missed.
- GDP ratio sensitivity: presentation of current account as share of GDP amplifies sensitivity to profit shifting because GDP includes net exports but not foreign income; hence transfer-price-induced trade changes affect GDP while offsetting income-account entries do not.
- Merchanting: merchanting is service trade (buying and selling without processing). If undertaken by residents abroad, it is an export service. Merchanting among related parties can enable profit shifting similar to transfer-price manipulation; merchanting among third parties generates genuine profits without profit-shifting scope.

### Empirical analysis — methods and key findings
- Graphical evidence:
  - Testable hypotheses: transfer price manipulation → stronger trade balances and weaker income balances; profit shifting through debt → no trade or net income balance effect; current accounts should not differ by attractiveness for profit shifting except for real tax effect and measurement errors.
  - Empirical pattern: countries likely to benefit from profit shifting (tax base gainers) tend to have stronger trade balances and weaker income balances across multiple definitions and time averages (2010-2018, last year, last five years, since 1990).
- Regression specification:
  - Benchmark panel: B/GNI_it = β τ_it + γ′ x_it + c_i + λ_t + ε_it (τ = statutory corporate income tax rate).
  - GFE extension: B/GNI_it = β τ_it + γ′ x_it + c_i + η_gi + ε_it (η_gi = group-period fixed effects; groups set to 2).
  - Data: panel of 81 countries from 1990-2018 based on IMF BoP, WEO, and updated Chinn and Ito (2006) data.
- Key regression results:
  - Column (2) of Table 4: Corporate tax rate coefficient on Trade Balance = -0.103***.
  - Column (3) of Table 4: Corporate tax rate coefficient on Income Balance = 0.109***.
  - Interpretation: a 10-percentage point higher tax rate weakens the trade balance to GNI-ratio by about 1 percentage point and improves the income balance by about 1 percentage point, on average.
  - GFE specification shows the income-balance offset does not fully offset the trade effect, implying a positive effect on the current account for some groups.
- Tax-base-gainer group analysis (Table 5):
  - Net positive effect on the current account for base gainers ranging from 4.8 to 7.2 percent of GNI across different groupings (Tørsløv, Clausing, Cobham groupings), indicating non-linearity concentrated in major hubs.
- Income balance breakdown (Table 6):
  - Tax-base-gainer groups show significantly higher net interest income; net equity income is significantly weaker for one grouping and insignificant for others.
- Model fit and sample sizes:
  - Observations in Table 4: 1,535 (columns 1–3) and 911 (columns 4–6).
  - R^2 range in Table 4: 0.668 to 0.845 (first three columns) and 0.735 to 0.836 (last three columns).
  - Observations in Table 5: 1,588; R^2 range: 0.285 to 0.406.
  - Observations in Table 6: reported per subtable (1,477 and 1,232) with R^2 values reported per regression.

### Quantification of current account boost for tax base gainers
- Average estimated boost of current account for tax base gainers ≈ 6 percentage points of GNI (average of coefficients in specified columns).
- Decomposition (equation (3) in source):
  - ΔCA ≈ real tax effect + measurement error
  - Empirical decomposition:
    - Real tax effect ≈ 2.1% of GNI (computed from Tørsløv et al. (2018) revenue estimates divided by GNI).
    - Measurement error ≈ 3.9% of GNI.
  - Aggregate statement in source: ΔCA ≈ 6% of GNI = 2.1% of GNI (real tax effect) + 3.9% of GNI (measurement error).

### Conclusion and policy implications
- Theoretical conclusion:
  - Profit shifting does not distort aggregate current account balances in theory; it affects components (trade and income balances) with offsetting accounting entries when shifting occurs via transfer price manipulation, and affects debit/credit sides of income balance for intragroup lending.
  - Bilateral balances between particular country pairs can be significantly affected.
  - Measured current account ratio to GDP can be affected by tax-motivated arrangements because net exports enter GDP while payments to nonresidents are not deducted.
  - Real tax effect: profits taxed and retained locally do not repatriate through the income account.
  - Measurement errors are more likely in the income balance than trade balance, so theoretical offsets may not appear in data.
- Empirical summary:
  - On average, profit shifting affects trade and income balances symmetrically, with no significant impact on current account balances.
  - Exception: countries that attract significant mobile profits (tax base gainers) show current account to GNI ratio raised by about 6 percentage points.
  - Of this 6 percentage points: approximately one third (2.1% of GNI) due to the real tax effect; approximately two thirds (3.9% of GNI) due to measurement errors.
- Policy implications:
  - In countries with significant tax-related external flows, macroeconomists should focus on aggregate current account balance and exercise caution interpreting its components and bilateral current accounts.
  - The aggregate current account should be treated cautiously when there are doubts about accurately monitoring items on the income account, notably reinvested foreign earnings.

*Source: wpiea2021041-print-pdf - 6. Regression Results – Tax Base Gainers, Breakdown of Income Balance*

### References________________________________________________________________________________________________26

### References

### Figures

- 1. Ireland: Trade and Income Balances (USD, billions)__________________________________________________ 5
- 2. Graphical Representation of the BoP Impact of Transfer Price Manipulation _____________________ 9
- 3. Graphical Representation of the BoP Impact of Profit Shifting via Intragroup Lending _________ 12
- 4. Profit Shifting through Merchanti ng _________________________________________________________________ 16
- 5. Trade and Income Account  Balances in  Countries Likely to Gain Tax Base, Average  
2010-2018 ______________________________________________________________________________________________ 18

### Tables

- 1. The Balance of Payments Impact of Profit Shifting, Ignoring Tax Payments ______________________ 9
- 2. The Full Balance of Payments Impact of Profit Shifting ____________________________________________ 10
- 3. Expected Signs of the Profit Shifting Effects_________________________________________________________ 20
- 4. Regression Results – Statutory Corporate Income Tax Rate and Current Account ______________ 21
- 5. Regression Results – Tax Base Gainers_______________________________________________________________ 22

*wpiea2021041-print-pdf - References________________________________________________________________________________________________26*

### 6. Regression Results – Tax Base Gainers, Breakdown of Income Balance __________________________ 23

### 6. Regression Results – Tax Base Gainers, Breakdown of Income Balance

### Introduction — context and key literature findings
- Accurate balance of payments (BoP) data are essential for macroeconomic analysis; multinational profit-shifting strategies (transfer prices, corporate debt, location of intangible assets) directly affect international flows without necessarily changing production or employment.
- Example revision: Ireland revised up GDP growth from 7.8 to 26.3 percent as a result of some transactions involving multinational enterprises.
- Meta-study result: on average a country loses 1.5 percent of its corporate tax base for each percentage point increase in the corporate income tax rate.
- Reported aggregate global tax-base loss estimates in literature:
  - 2 percent of the global tax base (Beer et al., 2020)
  - over 8 percent (Tørsløv, Wier and Zucman, 2018)
  - 23 percent (Cobham and Janský, 2018)
- Crivelli, De Mooij, and Keen (2016): long-run revenue loss just below 1 percent of GDP for OECD members and 1.3 percent of GDP for other countries.

### Three main messages from the analysis
- Double-entry BoP accounting insulates the aggregate current account from distortion by profit shifting, but sub-accounts (trade, income, bilateral current account balances) are distorted.
  - Transfer-price example: underpricing related-party exports and overpricing imports understates the trade balance of the headquarters country; shifted profits recorded as dividends or reinvested earnings offset at the current-account aggregate level.
- Measurement errors can be sizable in practice, especially in the income balance due to challenges in recording retained earnings of affiliates abroad; export and import data (goods) tend to be higher quality.
  - If the offsetting income-account effect is harder to detect, there will be a resulting gap in the measured current account.
- Tax payments to the country gaining tax base create a real effect on the current account.
  - Quantifying this requires the (unobservable) net amount of shifted profits in each country.

### Empirical regression results and decomposition
- Regression summary results:
  - A 10-percentage point higher corporate income tax rate is, on average, associated with a reduction in the trade balance by 1 percent of GNI.
  - For the average country, this trade effect is roughly offset by an increase of the same magnitude in the net income balance.
  - For a subset of likely tax base gainers, the income balance offsets only about 4 percentage points of the improvement in their trade balances, implying nonlinearity.
  - In the group of tax base gainers, the trade account effect dominates and the current account strengthens by about 6 percent of GNI, on average.
- Decomposition for tax base gainers into real tax effect and measurement error:
  - Based on country-by-country estimates of profit shifting (Tørsløv, Wier, and Zucman, 2018), the genuine tax effect on the current account averages 2.1 percent of GNI in the group of tax base gainers.
  - The measurement error in the net income balance of tax base gainers is estimated at about 3.9 percent of GNI, on average.
- Related empirical magnitudes cited:
  - Carloni et al. (2019): the 2017 US tax reform improves the US trade balance by about 9 percent via tax-motivated related-party trade.
  - Guvenen et al. (2019): outbound profit shifting from US multinationals affects measured aggregate productivity in the United States by magnitudes ranging from 0.09 to 0.24 percent.

### Theory — channels and BoP accounting implications
- Transfer price manipulation
  - Arm’s length principle often hard to apply for unique or intangible intra-firm transactions, giving multinationals leeway to allocate profits to low-tax jurisdictions.
  - Theoretical neutrality: transfer price manipulation changes composition between trade and income balances but not aggregate current account, ignoring taxes.
  - When taxes are included, profit shifting to a foreign jurisdiction that taxes the profit causes a net weakening of the current account of the sending (high-tax) country by the tax paid to the foreign government.
  - Table summaries (conceptual):
    - Example 1 (HQ → LT profit shift): aggregate current account unchanged ignoring tax; with tax, current account changes by xtLT or -xtLT depending on country.
    - Example 2 and 3 (affiliate-to-affiliate shifts): aggregate current accounts generally unchanged except for effects via taxes or tax savings x(tHT-tLT).
- International debt shifting
  - Use of intercompany debt or third-party debt allocation can shift profits via deductible interest.
  - Related-party loans: interest payments affect income accounts bilaterally but offsetting retained earnings/dividend effects leave net income unchanged absent taxes.
  - Third-party borrowing bundled in high-tax jurisdictions can change bilateral current accounts while leaving aggregate current accounts unchanged, with real effects again arising via taxes paid on shifted profits.
- Location of intangible assets (IP)
  - Legal ownership of patents, trademarks in low-tax jurisdictions generates royalties recorded as trade in services; shifting the asset changes trade and income account composition.
  - Sales of patents/copyrights enter trade in R&D services (affect trade account and capital account offsets); non-produced non-financial assets (e.g., brand names) enter capital account without affecting balances.
  - Pricing such transfers is difficult due to private information about future returns.
- Treaty shopping
  - Use of conduit/intermediate entities can exploit bilateral withholding tax differences.
  - Example: if LT imposes no withholding tax, a $1 dividend outflow is fully distributed; if LT imposes withholding tax rate t, the dividend outflow reduces the income balance by $1 – t.
  - Treaty shopping that accesses zero-withholding treaty networks neutralizes withholding tax effects from perspective of sender and ultimate recipient; intermediate countries collect any nonzero withholding taxes, improving their income balance by the amount collected.

### Data collection and measurement issues
- Trade data (goods) are generally based on administrative records (IMTS) and typically more reliable; trade in services often relies on ITRS or surveys with variable coverage.
- Income account data are more likely survey-based; ITRS may not capture ultimate counterpart or distinguish portfolio vs direct income; reinvested earnings often collected via surveys with variable quality.
- Given data differences, profit shifting may affect measured but not true current accounts if income-account offsets (especially reinvested earnings) are missed.
- GDP ratio sensitivity
  - Presentation of current account as share of GDP amplifies sensitivity to profit shifting because GDP includes net exports but not foreign income; hence transfer-price-induced trade changes affect GDP while offsetting income-account entries do not.
- Merchanting
  - Merchanting is service trade (buying and selling without processing). If undertaken by residents abroad, it is an export service.
  - Merchanting has been identified as important in explaining current account balances in some countries and may be underreported.
  - Merchanting among related parties can enable profit shifting similar to transfer-price manipulation; merchanting among third parties generates genuine profits without profit-shifting scope.

*Source: wpiea2021041-print-pdf - 6. Regression Results – Tax Base Gainers, Breakdown of Income Balance*

### 2. Example B is slightly  more complicated in that  bilateral current accounts are affected by the

### wpiea2021041-print-pdf - 2. Example B is slightly  more complicated in that  bilateral current accounts are affected by the

### D. Portfolio Investment
- Retained earnings related to portfolio investment are not counted in the income account.
- Fischer et al. (2019) find retained earnings related to portfolio investment range from 1.2 to 7.8 percent of GDP in some financial centers.
- Theoretical point: portfolio investors cannot shift profits; profit shifting requires jointly controlled multinational affiliates.
- Practical caveat: misclassification errors can cause some direct investment to be recorded as portfolio investment (e.g., missing data on controlling stockholders), linking profit shifting to reported national current accounts.

### IV. EMPIRICAL ANALYSIS
#### A. Graphical Evidence
- Testable hypotheses from theory:
  - Transfer price manipulation → stronger trade balances and weaker income balances.
  - Profit shifting through debt → does not affect trade balance or net income balance.
  - Current accounts should not differ by attractiveness for profit shifting, except for any real tax effect that strengthens income balance of beneficiaries and measurement errors (likely in income balance).
- Empirical pattern: countries likely to benefit from profit shifting (tax base gainers) tend to have stronger trade balances and weaker income balances.
- The documented pattern holds across definitions of tax base gainers, and for averages over 2010-2018, the last year, the last five years, and since 1990.

#### B. Regression Analysis
- Benchmark panel specification (expressing external balances relative to GNI):
  - B/GNI_it = β τ_it + γ′ x_it + c_i + λ_t + ε_it
  - τ = statutory corporate income tax rate; x = control variables; c_i = country fixed effects; λ_t = year fixed effects.
- Group-fixed effect (GFE) extension:
  - B/GNI_it = β τ_it + γ′ x_it + c_i + η_gi + ε_it
  - η_gi = group-period fixed effects; number of groups set to 2.
- Data: panel of 81 countries from 1990-2018 based on IMF Balance of Payments, WEO, and updated Chinn and Ito (2006) data.

Expected signs (Table 3):
- Transfer price manipulation:
  - Current account β = 0
  - Trade account β < 0
  - Income account β > 0
  - Net equity income β > 0
  - Net interest income β = 0
- Debt shifting:
  - Current account β = 0
  - Trade account β = 0
  - Income account β = 0
  - Net equity income β > 0
  - Net interest income β < 0

Key regression findings (Table 4 and extensions):
- Statutory corporate tax rate results (Table 4):
  - Column (2): Corporate tax rate coefficient on Trade Balance = -0.103*** (robust SE reported in table).
  - Column (3): Corporate tax rate coefficient on Income Balance = 0.109***.
  - Interpretation: a 10-percentage point higher tax rate weakens the trade balance to GNI-ratio by about 1 percentage point and improves the income balance by about 1 percentage point, on average (magnitudes in columns 2 and 3 roughly offset each other).
- Robustness checks:
  - Omitting terms of trade or using distance of own tax rate to world average does not change results.
  - Considering net equity and net interest income as dependent variables in Table 4 yielded insignificant results.
- GFE specification (columns 4–6 of Table 4):
  - Similar findings but indicate the effect of corporate income tax on the income balance does not fully offset the effect on the trade balance, producing a higher effect on the current account—consistent with a real tax effect and misattribution of retained earnings.
- Tax-base-gainer dummy analysis (Table 5):
  - Using group dummies instead of tax rate yields a net positive effect on the current account for base gainers ranging from 4.8 to 7.2 percent of GNI across different groupings (Tørsløv, Clausing, Cobham groupings).
  - This indicates non-linearity: effects concentrated in major hubs with large profits for nonresidents.
- Income balance breakdown (Table 6):
  - Tax-base-gainer groups show significantly higher net interest income; net equity income is significantly weaker for one grouping and insignificant for the others.

Quantification of components of the current account boost for tax base gainers:
- Average estimated boost of current account for tax base gainers ≈ 6 percentage points of GNI (average of coefficients in specified columns).
- Decomposition (equation (3) in source):
  - ΔCA ≈ real tax effect + measurement error
  - Empirical decomposition:
    - Real tax effect ≈ 2.1% of GNI (computed from Tørsløv et al. (2018) revenue estimates divided by GNI).
    - Measurement error ≈ 3.9% of GNI.
  - Aggregate statement in source:
    - ΔCA ≈ 6% of GNI = 2.1% of GNI (real tax effect) + 3.9% of GNI (measurement error).

Additional empirical notes:
- If the current account is overestimated due to these issues, estimated national savings would also be affected because of the identity between current account and savings minus investment.
- Data sample sizes and model fit indicators (selected):
  - Observations in Table 4: 1,535 (columns 1–3) and 911 (columns 4–6).
  - R^2 reported range in Table 4: 0.668 to 0.845 (first three columns) and 0.735 to 0.836 (last three columns).
  - Observations in Table 5: 1,588; R^2 range: 0.285 to 0.406.
  - Observations in Table 6: reported per subtable (1,477 and 1,232) with R^2 values reported per regression.

### V. CONCLUSION
- Theoretical conclusion:
  - Profit shifting does not distort aggregate current account balances in theory; it affects components (trade and income balances) with offsetting accounting entries when shifting occurs via transfer price manipulation, and affects debit/credit sides of income balance for intragroup lending.
  - Bilateral balances between particular country pairs can be significantly affected.
  - Measured current account ratio to GDP can be affected by tax-motivated arrangements because net exports enter GDP while payments to nonresidents are not deducted.
  - Real tax effect: profits taxed and retained locally do not repatriate through the income account.
  - Measurement errors are more likely in the income balance than trade balance, so theoretical offsets may not appear in data.
- Empirical summary:
  - On average, profit shifting affects trade and income balances symmetrically, with no significant impact on current account balances.
  - Exception: countries that attract significant mobile profits (tax base gainers) show current account to GNI ratio raised by about 6 percentage points.
  - Of this 6 percentage points: approximately one third (2.1% of GNI) due to the real tax effect; approximately two thirds (3.9% of GNI) due to measurement errors.
- Policy implications:
  - In countries with significant tax-related external flows, macroeconomists should focus on aggregate current account balance and exercise caution interpreting its components and bilateral current accounts.
  - The aggregate current account should be treated cautiously when there are doubts about accurately monitoring items on the income account, notably reinvested foreign earnings.

*Source: Authors’ analysis in the IMF working paper content provided.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021041-print-pdf.pdf_