## wp18132 — Box 1 and Section 2: The Role of Corporate Income Taxes; Corporate taxation mitigates arbitrage in response to taxation of entrepreneurial income

## Source details

**Canonical URL:** [wp18132 — Box 1 and Section 2: The Role of Corporate Income Taxes; Corporate taxation mitigates arbitrage in response to taxation of entrepreneurial income](https://www.imf.org/-/media/files/publications/wp/2018/wp18132.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18132.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18132.pdf.json)

---

### Box 1 — Main point and enumerated finding
- Main point:
  - The corporate income tax plays an important role in enforcing the taxation of labor income.
- Enumerated finding:
  - 1. While dividends can easily be taxed at the shareholder level, taxing reinvested earnings would be difficult without a tax at the corporate level.
- Fragment/continuation marker:
  - a

### Corporate taxation and income-shifting (theory and implications)
- Key conceptual points:
  - Distinguishing labor income from capital income can be difficult (or impossible) when individuals can freely choose the form through which they declare their income (IMF, 2014).
  - When the PIT base can be shifted to some alternative tax base that is taxed at a lower rate (such as corporate income), the optimal tax theory implies that the optimal tax rate on personal income rises with the tax rate on the alternative base.
- Formal expression (notation preserved from source):
  - Optimal top income tax rate allowing for income shifting is given as:
    - 푡푡∗ = (1 + 푠푠∙휏휏∙푎푎푎푎) / (1 + 푎푎푎푎),
      - where 푠푠 is the share of marginal income shifted from the individual base, 휏휏 is the tax rate on the alternative tax base (for example, corporate or capital income), and all other parameters are as previously defined, with the marginal welfare weight set to zero.
- External pressures:
  - International tax competition and capital mobility have driven a steady downward trend in statutory corporate income tax rates in recent decades (Figure 2 in source), which may also exert downward pressure on PIT rates.
- Withholding and retained earnings:
  - Note (footnote a): distributed earnings can, in principle, be taxed through withholding taxes, but many economies—especially developing economies—have tax treaties restricting withholding taxes on foreign shareholders. In economies with a corporate-level tax payable only on distributed profits, the corporate income tax cannot fulfill the withholding function on retained earnings.

### Empirical estimates of tax elasticities (ETIs) — literature and meta-evidence
- Summary of literature findings:
  - Neisser (2017) meta-analysis: most estimates range from 0 to 1, with a peak at around 0.3.
  - Saez, Slemrod, and Giertz (2012) review: best available estimates range from 0.12 to 0.4, with a midpoint of 0.25.
  - Some papers report much higher elasticities (e.g., Feldstein (1995) finds ETIs above 1) and some report zero or negative elasticities.
  - Negative ETIs are deemed unlikely to be genuine economic responses (likely due to omitted variables).
- Time-trend evidence (mixed):
  - Saez, Slemrod, and Giertz (2012): lower ETIs in the 1990s in the United States compared to the 1980s.
  - Brewer, Saez and Shepard (2010): varying UK elasticities across reforms: 0.4 (1988) vs 0.1–0.3 (1979) by method; long-term estimates 0.6–0.9 for both 1962-1978 and 1978-2003, with difference-in-difference dropping from 0.9 to 0.6.
  - Kleven and Schultz (2014) for Denmark: declining elasticities over time, attributed to stronger enforcement and compliance.
  - Piketty, Saez, and Stantcheva (2014) using 18 OECD countries: increase in average ETIs from 0–0.2 (1960–80) to 0.6–0.8 (1981–2010).
- Overall assessment:
  - Literature yields ambiguous evidence on ETI trends over recent decades.

### Data and methodology used in the empirical analysis
- Data sources:
  - World Wealth & Income Database (WID) for income distribution data, combining surveys with other sources to better represent the richest individuals.
  - OECD tax database (including historical versions) for central government tax rates and thresholds.
- Limitations noted:
  - WID does not separate taxable from non-taxable income (cannot exclude tax-exempt items like pensions where relevant).
  - Personal income tax bracket detail for combined central plus sub-central rates not available; estimated ETIs are based on central rates only (but combined actual rates are used for comparisons when available).
- ETI definition and estimation approach (notation preserved):
  - ETI defined as percentage change in taxable income (Y) relative to percentage change in net-of-tax rate (1 − t) for top earners; alternatively using income share (s) to control for general real-income changes.
  - Elasticity formula (Equation (3) in source):
    - 푎푎푖 = ∆푙푙푙푙(푌푌푖) / ∆푙푙푙푙(1−푡푡푖)  표표표표 ∆푙푙푙푙(푠푠푖) / ∆푙푙푙푙(1−푡푡푖)
  - When top percentiles face multiple tax brackets, average marginal tax rate for the top group is calculated assuming a Pareto distribution; example formula for two tax rates:
    - 푡푡̅ = (1−푤푤) 푡푡푙푙푙푙푙푙 + 푤푤푡푡ℎ푖푖푖푖ℎ,
      - where w is the share of top percentile income that exceeds the top tax bracket.
  - Difference-in-difference (DD) ETI identified from differences between top percentile and next four percentiles:
    - 푎푎DD = [∆ln(푌푌1) − ∆ln(푌푌2−5)] / [∆ln(1−푡푡1) − ∆ln(1−푡푡2−5)] (Equation (4) in source)
  - Pareto assumptions to recover real incomes from income shares:
    - 푌푌푖 = 푎푎/(푎푎−1) 푚푚푖 (Equation (5) in source)
    - Pareto index a estimated via survival function rearrangement (Equation (6) in source).
- Sample and calculation rules:
  - At least one ETI obtained for about half of 35 OECD economies over 1981–2016.
  - Twelve ETI definitions calculated across up to 16 economies; the approach yielding the most observations was ETI using real incomes under the short calculation window, producing 89 elasticities (years 1982–2013, except 2008).
  - To avoid spurious large elasticities from very small tax changes, tax reforms with statutory rate changes less than 1 percentage point were disregarded (robustness checked with 2 and 3 percentage point cutoffs).

### Empirical results — estimated ETIs and regression findings
- Descriptive and robustness highlights:
  - Large outliers observed in ETI estimates (e.g., maximum of 12,235 percent for some DD estimates); medians are more informative than means.
  - Dropping elasticities exceeding 100 in absolute value reduces mean elasticities (real income and income share) to 0.2, aligning with the median.
- Central empirical finding on typical ETI size:
  - An ETI of around 0.2 appears to be a reasonable figure based on multiple estimation methods and robustness checks.
  - Using the short calculation window and the DD approach, ETI ~0.2; long calculation window DD yields ETIs of 0 to 0.3 (0.2 close to average).
  - About a quarter of estimated ETIs are negative (likely due to omitted variables), but if independent of reform timing should not bias average ETI.
- Regression specifications (notation preserved):
  - Cross-sectional/time-series regression (Equation (7) in source):
    - ln Y_i = α + a_i ln(1−t_i) + x′β + f_c + ε
  - Difference-in-difference regression (Equation (8) in source):
    - ln Y_i = α + D_1 + a_5 ln(1−t_i) + a_DD ln(1−t_i) × D_1 + x′β + f_c + ε
- Regression findings:
  - Regression-based average elasticities broadly corroborate ETI ~0.2 when control variables (output gap and capital account openness) are included.
  - Most coefficients on ln(net-of-tax rate) are significant at least at 10 percent.
- ETI trends over time:
  - Visual inspection of Figures 4 and 5 shows no clear time trend.
  - Formal regressions of ETIs on a time trend (36 regressions with various sample cuts) yield only 5 significant coefficients (4 of 5 positive); three of these significant regressions have very few observations.
  - Robustness checks controlling for output gap and splitting samples by social spending or welfare-state classification reveal no consistent trend.
  - Conclusion: no strong evidence of a systematic increase in ETIs over the last three decades.

### Implied optimal top tax rates and comparisons to actual rates
- Approach and assumptions:
  - Use average ETI ≈ 0.2 and each economy’s Pareto index to calculate revenue-maximizing top tax rates (marginal welfare weight on rich = 0), then compare to actual combined rates.
- Key empirical comparison:
  - Figure 3 in source: revenue-maximizing tax rates on the horizontal axis; actual combined top PIT rates (including local taxes) on the vertical axis.
  - Finding: all economies in the sample are below the revenue-maximizing optimal tax rates under the assumptions used.
- Caveats and interpretation:
  - Actual tax rates being below revenue-maximizing rates does not prove suboptimality: the actual welfare weight for rich individuals may exceed zero, or a particular economy’s ETI might be higher than the cross-country average.
  - Adding consumption taxes (e.g., VAT) to the comparison could bring many economies closer to the implied optimum, but this is problematic because consumption taxes have different incidence and elasticities, and may not be simply additive when targeting top earners.
- Pareto index and income distribution trend:
  - Pareto index for top ventile shows a clear downward trend over ~35 years (Figure 6), implying a larger share of income in the top tail and, all else equal, calling for higher optimal top tax rates.

### Other drivers, welfare weights, political economy, and conclusions
- Welfare-weight interpretation:
  - Using ETI = 0.2 and Pareto index = 2.2, Figure 7 maps implied optimal top PIT rates to the social welfare weight g for top earners.
  - Reconciling the substantial decline in observed top marginal PIT rates with optimal-tax theory would require an implied rise in the social welfare weight on high-income earners from close to 0 to more than 0.5 over ~35 years—an implication at odds with survey evidence.
- Public attitudes:
  - World Values Survey evidence (Figure 8) shows societal preferences in favor of redistribution in OECD countries have become stronger since the 1980s, which implies a reduction, not an increase, in social welfare weight on high-income earners.
- Political economy explanation:
  - Declines in top PIT rates are likely driven by political economy factors: better-off individuals may have greater political influence via lobbying, media access, and higher political engagement.
  - Literature cited: Ardanaz and Scartascini (2011) on legislative malapportionment; Rodriguez (2004) on rent-seeking and redistribution.
- Alternative possibility:
  - Past top tax rates (e.g., in the 1980s) might have been higher than optimal; current declines could be movement toward optimal rates. The source notes this is theoretically possible but not very plausible given concerns about rising inequality and lack of high-income tax buoyancy.
- Overall conclusions drawn in the source:
  - No systematic rise in ETIs detected that would explain the widespread decline in top personal income tax rates.
  - Changes in income distribution (Pareto index) and public support for redistribution do not explain declining top PIT rates.
  - Political economy forces and non-optimal past tax settings are more plausible drivers of observed declines in top personal income tax rates.

*Source: wp18132 — IMF staff estimates and analysis as presented in “2. Corporate taxation mitigates arbitrage in response to taxation of entrepreneurial income,” and Box 1: The Role of Corporate Income Taxes.*

### Box 1. The Role of Corporate Income Taxes

### Box 1. The Role of Corporate Income Taxes

### Main point
- The corporate income tax plays an important role in enforcing the taxation of labor income:

### Enumerated finding
- 1. While dividends can easily be taxed at the shareholder level, taxing reinvested earnings would be difficult without a tax at the corporate level.

### Fragment/continuation marker
- a

*Source: wp18132 - Box 1. The Role of Corporate Income Taxes*

### 2. Corporate taxation mitigates arbitrage in response to taxation of entrepreneurial income,

### 2. Corporate taxation mitigates arbitrage in response to taxation of entrepreneurial income

### Corporate taxation and income shifting
- Distinguishing labor income from capital income can be difficult (or impossible) when individuals can freely choose the form through which they declare their income (IMF, 2014).
- When the PIT base can be shifted to some alternative tax base that is taxed at a lower rate (such as corporate income), the optimal tax theory implies that the optimal tax rate on personal income rises with the tax rate on the alternative base.
- Optimal top income tax rate allowing for income shifting is given in the source as:
  - 푡푡
    ∗
    =
    (
    1 +푠푠∙휏휏∙푎푎푎푎
    )
    (
    1 +푎푎푎푎
    )
    , 
    - where 푠푠 is the share of marginal income shifted from the individual base, 휏휏 is the tax rate on the alternative tax base (for example, corporate or capital income), and all other parameters are as previously defined, with the marginal welfare weight set to zero.
- International tax competition and capital mobility have driven a steady downward trend in statutory corporate income tax rates in recent decades (Figure 2 in source), which may also exert downward pressure on PIT rates.
- Note from source (footnote a): distributed earnings can, in principle, be taxed through withholding taxes, but many economies—especially developing economies—have tax treaties restricting withholding taxes on foreign shareholders. In economies with a corporate-level tax payable only on distributed profits, the corporate income tax cannot fulfill the withholding function on retained earnings.

### Empirical estimates of tax elasticities (ETIs) — literature and meta-evidence
- Existing empirical literature yields a wide range of ETI estimates.
  - Neisser (2017) meta-analysis: most estimates range from 0 to 1, with a peak at around 0.3.
  - Saez, Slemrod, and Giertz (2012) review: best available estimates range from 0.12 to 0.4, with a midpoint of 0.25.
  - Some papers report much higher elasticities (e.g., Feldstein (1995) finds ETIs above 1) and some report zero or negative elasticities.
- Negative ETIs are deemed unlikely to be genuine economic responses (likely due to omitted variables).
- Time-trend findings in literature are mixed:
  - Saez, Slemrod, and Giertz (2012) find lower ETIs in the 1990s in the United States compared to the 1980s.
  - Brewer, Saez and Shepard (2010) find varying elasticities across UK reforms: 0.4 (1988) vs 0.1–0.3 (1979) by method; long-term estimates 0.6–0.9 for both 1962-1978 and 1978-2003, with difference-in-difference dropping from 0.9 to 0.6.
  - Kleven and Schultz (2014) for Denmark find declining elasticities over time, attributed to stronger enforcement and compliance.
  - Piketty, Saez, and Stantcheva (2014) using 18 OECD countries find an increase in average ETIs from 0–0.2 (1960–80) to 0.6–0.8 (1981–2010).
- Overall literature yields ambiguous evidence on ETI trends over recent decades.

### Data and methodology used in the empirical analysis
- Data sources used:
  - World Wealth & Income Database (WID) for income distribution data, combining surveys with other sources to better represent the richest individuals.
  - OECD tax database (including historical versions) for central government tax rates and thresholds.
- Limitations:
  - WID does not separate taxable from non-taxable income (cannot exclude tax-exempt items like pensions where relevant).
  - Personal income tax bracket detail for combined central plus sub-central rates not available; estimated ETIs are based on central rates only (but combined actual rates are used for comparisons when available).
- ETI definitions and calculations:
  - ETI defined as percentage change in taxable income (Y) relative to percentage change in net-of-tax rate (1 − t) for top earners, alternatively using income share (s) to control for general real-income changes.
  - Equation provided in source for elasticity (notation preserved as in source):
    - 푎푎
      푖푖
      =
      ∆푙푙푙푙
      (
      푌푌
      푖푖
      )
      ∆푙푙푙푙
      (
      1−푡푡
      푖푖
      )
       표표표표 
      ∆푙푙푙푙
      (
      푠푠
      푖푖
      )
      ∆푙푙푙푙
      (
      1−푡푡
      푖푖
      )
      (Equation (3) in source)
  - When top percentiles face multiple tax brackets, average marginal tax rate for the top group is calculated assuming a Pareto distribution; example formula in source for two tax rates:
    - 푡푡
      ̅
      =
      (
      1−푤푤
      )
      푡푡
      푙푙푙푙푙푙
      +푤푤푡푡
      ℎ푖푖푖푖 ℎ
      , where w is the share of top percentile income that exceeds the top tax bracket.
  - Difference-in-difference (DD) ETI identified from differences between top percentile and next four percentiles (top ventile without top percentile) to control for non-tax developments affecting entire top ventile:
    - 푎푎
      퐷퐷퐷퐷
      =
      ∆ln
      (
      푌푌
      1
      )
      −∆ln
      (
      푌푌
      2−5
      )
      ∆ln
      (
      1−푡푡
      1
      )
      −∆ln
      (
      1−푡푡
      2−5
      )
      (Equation (4) in source)
  - Pareto distribution assumptions used to estimate real incomes from income shares:
    - 푌푌
      푖푖
      =
      푎푎
      푎푎−1
      푚푚
      푖푖
      (Equation (5) in source)
    - Pareto index a estimated via survival function rearrangement:
      - 푎푎=
        푙푙푙푙
        (
        0.2
        )
        푙푙푙푙 (푚푚
        5
        /푚푚
        1
        )
        =
        푙푙푙푙
        (
        0.2
        )
        푙푙푙푙 (푠푠
        5
        /5푠푠
        1
        )
      (Equation (6) in source)
- Sample and approach:
  - At least one ETI obtained for about half of 35 OECD economies over 1981–2016.
  - Twelve ETI definitions calculated across up to 16 economies; the approach yielding the most observations was ETI using real incomes under the short calculation window, producing 89 elasticities (years 1982–2013, except 2008).
  - To avoid spurious large elasticities from very small tax changes, tax reforms with statutory rate changes less than 1 percentage point were disregarded (robustness checked with 2 and 3 percentage point cutoffs).

### Empirical results — estimated ETIs and regression findings
- Descriptive and robustness highlights:
  - Large outliers observed in ETI estimates (e.g., maximum of 12,235 percent for some DD estimates); medians are more informative than means.
  - Dropping elasticities exceeding 100 in absolute value reduces mean elasticities (real income and income share) to 0.2, aligning with the median.
- Central empirical finding on typical ETI size:
  - An ETI of around 0.2 appears to be a reasonable figure based on multiple estimation methods and robustness checks.
  - Using the short calculation window and the DD approach, ETI ~0.2; long calculation window DD yields ETIs of 0 to 0.3 (0.2 close to average).
  - About a quarter of estimated ETIs are negative (likely due to omitted variables), but if independent of reform timing should not bias average ETI.
- Regression-based average elasticities:
  - Regression specification for elasticity (notation preserved):
    - 푙푙푙푙 푌푌
      푖푖
      =훼훼+푎푎
      푖푖
      푙푙푙푙
      (
      1−푡푡
      푖푖
      )
      +푥푥
      ′
      훽훽+푓푓
      푐푐
      +휀휀 
      (Equation (7) in source)
  - Difference-in-difference regression specification:
    - 푙푙푙푙 푌푌
      푖푖
      =훼훼+퐷퐷
      1
      +푎푎
      5
      푙푙푙푙
      (
      1−푡푡
      푖푖
      )
      +푎푎
      퐷퐷퐷퐷
      푙푙푙푙
      (
      1−푡푡
      푖푖
      )
      ×퐷퐷
      1
      +푥푥
      ′
      훽훽+푓푓
      푐푐
      +휀휀 
      (Equation (8) in source)
  - Regression estimates (Table 2 and Table 3 in source) broadly corroborate ETI ~0.2 when control variables (output gap and capital account openness) are included; most coefficients on ln(net-of-tax rate) significant at least at 10 percent.
- ETI trends over time:
  - Visual inspection of Figures 4 and 5 shows no clear time trend.
  - Formal regressions of ETIs on a time trend (36 regressions with various sample cuts) yield only 5 significant coefficients (4 of 5 positive); three of these significant regressions have very few observations and thus limited weight.
  - Robustness checks controlling for output gap and splitting samples by social spending or welfare-state classification reveal no consistent trend.
  - Conclusion: no strong evidence of a systematic increase in ETIs over the last three decades.

### Implied optimal top tax rates and comparisons to actual rates
- Using an average ETI of around 0.2 and each economy’s Pareto index, revenue-maximizing top tax rates (marginal welfare weight on rich = 0) are calculated and compared to actual combined rates.
- Key empirical comparison:
  - Figure 3 in source shows revenue-maximizing tax rates on the horizontal axis and actual combined top PIT rates (including local taxes) on the vertical axis.
  - Finding: all economies in the sample are below the revenue-maximizing optimal tax rates under the assumptions used.
- Caveats and interpretation:
  - Actual tax rates being below revenue-maximizing rates does not prove they are suboptimal: actual welfare weight for rich individuals may exceed zero, or a particular economy’s ETI might be higher than the cross-country average.
  - Adding consumption taxes (e.g., VAT) to the comparison could bring many economies closer to the implied optimum, but this is problematic because consumption taxes have different incidence and elasticities, and may not be simply additive when targeting top earners.
- Pareto index and income distribution trends:
  - Pareto index for top ventile shows a clear downward trend over ~35 years (Figure 6), implying a larger share of income in the top tail and, all else equal, calling for higher optimal top tax rates.

### Other drivers and interpretation: welfare weights, political economy, and conclusions
- Welfare weight on top earners:
  - Using ETI = 0.2 and Pareto index = 2.2, Figure 7 maps implied optimal top PIT rates to the social welfare weight g for top earners.
  - To reconcile the substantial decline in observed top marginal PIT rates with optimal-tax theory would require an implied rise in the social welfare weight on high-income earners from close to 0 to more than 0.5 over ~35 years—an implication at odds with survey evidence.
- Public attitudes:
  - World Values Survey evidence (Figure 8) shows societal preferences in favor of redistribution in OECD countries have become stronger since the 1980s, which implies a reduction, not an increase, in social welfare weight on high-income earners.
- Political economy explanation:
  - Declines in top PIT rates are likely driven by political economy factors: better-off individuals may have greater political influence via lobbying, media access, and higher political engagement.
  - Literature cited in source: Ardanaz and Scartascini (2011) on legislative malapportionment, Rodriguez (2004) on rent-seeking and redistribution.
- Alternative possibility:
  - Past top tax rates (e.g., in the 1980s) might have been higher than optimal; current declines could be movement toward optimal rates. Source notes this is theoretically possible but not very plausible given concerns about rising inequality and lack of high-income tax buoyancy.
- Overall conclusion from the source:
  - No systematic rise in ETIs detected that would explain the widespread decline in top personal income tax rates.
  - Changes in income distribution (Pareto index) and public support for redistribution do not explain declining top PIT rates.
  - Political economy forces and non-optimal past tax settings are more plausible drivers of observed declines in top personal income tax rates.

*Source: IMF staff estimates and analysis as presented in “2. Corporate taxation mitigates arbitrage in response to taxation of entrepreneurial income,” from wp18132.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18132.pdf_