## _wp15244

## Source details

**Canonical URL:** [_wp15244](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15244.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15244.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15244.pdf.json)

---

### I. Introduction and scope
- Observed trends:
  - The wage (labor) share declined since the 1970s while income inequality rose; on average the wage share declined by 12 percent whereas income inequality increased by 25 percent in some advanced economies over roughly three decades.
- Key question:
  - Whether the declining labor share of income has been a key driver of rising income inequality.
- Main conclusion:
  - The most important determinant of rising income inequality is the growing dispersion of wages, especially at the top of the wage distribution, rather than the decline in the labor share of income.
- Data and coverage:
  - Microeconomic data: Luxembourg Income Study (LIS) household surveys — 231 household surveys covering 43 countries for the period 1978–2010.
  - Macroeconomic data: national accounts for a large set of 93 countries for the period 1970–2013.
- Paper organization:
  - Literature review; decomposition of the Gini index linking factor shares and pseudo-Gini indexes; microeconomic analysis with LIS; macroeconomic exploration; conclusions.

### II. Decomposition methodology (Gini and pseudo-Gini identities)
- Core decomposition identities and marginal effects (as presented):
  - G = Σk sk Ck  (equation (1) in source).
  - Changes approximated by ΔG ≈ Σk s0k ΔCk + Σk C0k Δsk + Σk Δsk ΔCk (equation (2)); the third term is likely close to zero.
  - Marginal impact of a pseudo-Gini index: δG/δCk = sk (equation (3)).
  - Marginal impact of changes in market-income composition (labor vs capital):
    - δGm/δsl = (Cc - Cl) sl? (expressed in source notation as cl l y CC s G  δ δ and more explicitly in Annex as δGm/δsl = (Cc - Cl) s_l?; see equation (23) in Annex).
- Interpretation rules:
  - If pseudo-Gini for capital (Cc) > pseudo-Gini for labor (Cl), then an increase in the labor share reduces inequality; conversely, if Cc < Cl, increasing the labor share increases inequality.
- Treatment of transfers and taxes:
  - Impact of transfers on inequality: ΔG_y - ΔG_m (equation (19) in Annex).
  - Impact of taxation on inequality: ΔG_net - ΔG_y (equation (20) in Annex).

### III. Micro evidence (LIS-based decomposition and sample results)
- Small-sample country decomposition (France, Germany, United Kingdom, United States) — selected exact figures:
  - Change in disposable income Gini (ΔGy_net): US 0.08; UK 0.10; DE 0.03; FR -0.01.
  - Change in gross income Gini (ΔGy): US 0.07; UK 0.10; DE 0.05; FR -0.01.
  - Change in market income Gini (ΔGm): US 0.10; UK 0.13; DE 0.08; FR 0.02.
  - Impact of changes in income shares (labour) on ΔGm (Δsl (C0l - C0c)): US 0.00; UK 0.00; DE 0.01; FR 0.00.
  - Impact of changes in pseudo-Gini indexes:
    - ΔCl (labor pseudo-Gini): US 0.09; UK 0.13; DE 0.06; FR 0.03.
    - ΔCc (capital pseudo-Gini) (=-sl0 ΔCc in table): US 0.01; UK 0.00; DE 0.02; FR 0.00.
  - Gini levels (examples):
    - G0_y_net: US 0.31; UK 0.27; DE 0.26; FR 0.33.
    - G_y_net (final year): US 0.40; UK 0.36; DE 0.29; FR 0.31.
    - G0_m: US 0.41; UK 0.39; DE 0.42; FR 0.44.
    - Gm (final year): US 0.51; UK 0.52; DE 0.49; FR 0.47.
    - G0_l: US 0.44; UK 0.43; DE 0.45; FR 0.46.
    - Gl (final year): US 0.53; UK 0.57; DE 0.54; FR 0.53.
    - G0_c: US 0.92; UK 0.88; DE 0.61; FR 0.97.
    - Gc (final year): US 0.94; UK 0.97; DE 0.87; FR 0.88.
- Extended sample (43 countries; 231 surveys; 1978–2010):
  - Main finding replicated: increase in the pseudo-Gini index of labor income is the dominant driver of rising market-income inequality.
  - Magnitude statements preserved exactly:
    - "A 10 percent increase in the pseudo-Gini index of labor income increases the Gini index for market income by more than 9 percent."
    - "A 10 percent decline in the labor share raises the market-income Gini by about 0.9 percent (effect small but statistically significant; primarily driven by emerging market economies)."
  - Subsample differences:
    - Emerging economies: pseudo-Gini index for capital income is on average higher by 0.16 than in advanced economies; the difference for labor income is 0.07.
- Average marginal impacts (Table 2 exact estimates for the full LIS sample of 43 countries; 231 observations):
  - Impact of a 0.01 change in the share of labor income (δGm/δsl): -0.0004 ** ; St. Dev 0.0012; T -2.2889; P>|t| 0.0272.
  - Impact of a 0.01 increase in the pseudo-Gini index:
    - δGm/δCl: 0.0096 *** ; St. Dev 0.0003; T 250.3138; P>|t| 0.0000.
    - δGm/δCc: 0.0004 *** ; St. Dev 0.0003; T 9.8787; P>|t| 0.0000.
  - Subsample point estimates:
    - Impact of 0.01 change in labor share: Advanced economies -0.0001; Emerging economies -0.0010.
    - Impact of 0.01 increase in labor pseudo-Gini: Advanced 0.0096; Emerging 0.0097.
    - Impact of 0.01 increase in capital pseudo-Gini: Advanced 0.0004; Emerging 0.0003.
- Robustness notes:
  - Results robust to alternative decomposition measures (Annex 1).
  - Inclusion of self-employed earnings in labor income may affect comparability with other studies.
  - The third-order interaction term (Δsk ΔCk) is likely close to zero given slow movement of shares and pseudo-Gini indexes.

### IV. Macro panel evidence for advanced economies (empirical framework and main regression findings)
- Estimating equation (symbolic, preserved as in source):
  - y_it = α + β s_lit + γ I_lit + θ r_it + x_it'ρ + ε_it  (equation (5) in source).
    - s_l = wage (labor) share.
    - I_l = inequality (dispersion) of labor income (proxy: ratio top 10 percent salaries to bottom 90 percent salaries; robustness: 10-to-90 labor income for OECD).
    - r = redistributive impact of tax and welfare system (proxied by ratio of public revenues to GDP, and social protection and health spending).
    - x_it = other control variables.
- Data, coverage, and limitations (advanced economies subsection):
  - Micro data: LIS (panel 1: 26 countries; 174 observations; panel 2: 17 countries; 57 observations).
  - Macro panel: unbalanced panel covering a large sample, period from the 1970s to 2013.
  - Sample size remark: "The sample includes about 350 observations for the Gini equation."
  - Caveats: income surveys underreport capital income and poorly capture the extreme top; Gini more sensitive to middle of distribution; recent work shows top 1 percent shares rising faster.
- Estimation methodology:
  - Panel fixed effects preferred per Breusch and Pagan Lagrange multiplier test.
  - Endogeneity treated with panel instrumental variables and full information maximum likelihood estimators.
  - Instruments for wage share in IVs: lagged labor share, lagged unemployment rate, employment in services sector, proxy for wage-bargaining framework.
  - Instrument diagnostics reported:
    - R2 of the first-step wage share equation = 0.86.
    - R2 of regressing the Gini index on instruments = 0.01.
- Core regression results (fixed effects, summary):
  - Wage share:
    - In simple specifications, inequality declines when the wage share increases (negative coefficient).
    - The wage-share coefficient becomes not significant when dispersion of labor income is included.
  - Dispersion of labor income:
    - Robust, large, and positive association with the Gini index; reported coefficients (examples from Table 3): 0.0242 ***, 0.0203 ***, 0.0174 ***, 0.0173 ***, 0.0173 ***, 0.0161 *** (absolute t statistics in parentheses in source).
  - Public policy / redistributive proxies (examples preserved from tables):
    - Public Revenues: negative and significant in many specifications, e.g., -0.0011 ***, -0.0008 **, -0.0007 **, -0.0008 ** (Table 3); Table 4: -0.0008 **, 0.0000, -0.0014 ***.
    - Public Social Protection Spending: coefficients not consistently significant; examples: -0.0011, -0.0006, -0.0009, -0.0007 (Table 3); Table 4: -0.0009, -0.0010, -0.0039 *.
    - Public Health Spending: negative and in some specifications statistically significant, e.g., -0.0046 *, -0.0055 **, -0.0070 *** (Table 3); Table 4: -0.0055 **, -0.0021, -0.0039 *.
  - Globalization proxies:
    - Economic Globalization: positive and significant in some specifications, e.g., 0.0007 *** (Table 3); Table 4 reports 0.0007 ***, 0.0007, 0.0003 *.
    - Financial Globalization: positive and significant in at least one specification in Table 3, e.g., 0.0094 **.
- Endogeneity and IV results:
  - Robustness via panel IVs and full information ML: labor share negative but small effect; inequality of labor income remains large and significantly positive.

### V. Determinants of wage (labor) income dispersion (Table 5 summaries)
- Main correlates (selected coefficients preserved exactly as reported):
  - Financial Globalization: positive association with wage dispersion; example coefficients: 0.0719 **, 0.0701 *, 0.037 *, 0.1531 *, 0.0788 *** (absolute z statistics in parentheses in the source).
  - Unemployment: positive association; example coefficients: 0.0082 *, 0.0066 *, 0.0231 **, 0.0075 **.
  - Industry Unionization: negative association; example coefficients: -0.0118 ***, -0.024 ***, -0.01 ***.
  - Tertiary Education (share): negative association; example coefficients: -0.018 *** , -0.009 ***.
  - Government Spending (primary government spending as proxy): negative association; example coefficient: -0.009 ***.
- Text interpretation:
  - Higher financial globalization and higher unemployment levels are associated with higher dispersion of wages.
  - Higher unionization in industry, a higher share of educated workers, and higher primary government spending help reduce wage dispersion.

### VI. Policy implications and conclusions (as stated in source)
- Core conclusion:
  - The most important determinant of income inequality is the dispersion of labor income rather than the aggregate wage (labor) share.
- Factors associated with rising wage dispersion:
  - Growing financial globalization, decline in industry unionization, decline in the size of the state.
- Policy recommendations (preserved wording and emphasis):
  - Policymakers should pay attention to labor market outcomes and wage dispersion to avoid unfavorable distributional consequences.
  - Public policies that support inclusive growth—e.g., promoting participation in the labor market and strengthening human capital of low-income groups—may prevent the rise in economic disparities.
- Research caveat:
  - Deeper analysis of the drivers of wage dispersion is beyond the paper's scope and could be taken up in further research.

### 1. Income Inequality and Wage Share in Group of Seven Countries ____________________5

### 1. Income Inequality and Wage Share in Group of Seven Countries ____________________5

### Main Sections
- 1. Income Inequality and Wage Share in Group of Seven Countries ____________________5
- 2. Average Marginal Impact on the Gini Index for Market Income of Changes in the Labor Share and Pseudo-Gini Indexes for Labor and Capital ______________________________12
- 3. Advanced and Emerging Economies: Marginal Impact on the Gini Index for Market Income of Changes in the Labor Share and Pseudo-Gini Indexes for Labor and Capital  ___13

### Tables
- 1. Decomposition of Changes in Inequality (Measured by the Gini Index) ______________11
- 2. Average effects on the Gini index for Market Income ____________________________13
- 3. Determinants of Income Inequality, Fixed Effects _______________________________15
- 4. Determinants of Income Inequality—Robustness Checks _________________________17
- 5. Determinants of Wage Dispersion  ___________________________________________18

### Annexes
- 1. Gini Coefficients, Pseudo-Gini (or Concentration) Indexes, and Gini Correlations  _____20
- 2. Inequality Decomposition Using the LIS Data Set _______________________________23
- 3. Description of the Database  ________________________________________________27

### Annex Tables
- 2.1. Countries Included in the Analysis  _________________________________________24
- 3.1. Countries Considered in the Estimation and Descriptive Statistics for Inequality  _____28

*Source: _wp15244 - 1. Income Inequality and Wage Share in Group of Seven Countries ____________________5*

### References ________________________________________________________________31

### References ________________________________________________________________31

### I. INTRODUCTION
- Observed trends: the wage (labor) share declined since the 1970s while income inequality rose; on average the wage share declined by 12 percent whereas income inequality increased by 25 percent in some advanced economies over roughly three decades.
- Public and policy interest highlighted by statements in 2006–2007 (Ben Bernanke; Germany’s finance minister) and heightened after the global financial crisis.
- Key question examined: whether the declining labor share of income has been a key driver of rising income inequality.
- Main conclusion summarized: the most important determinant of rising income inequality is the growing dispersion of wages, especially at the top of the wage distribution, rather than the decline in the labor share of income.
- Data and scope overview:
  - Microeconomic data: Luxembourg Income Study (LIS) household surveys — 231 household surveys covering 43 countries for the period 1978–2010.
  - Macroeconomic data: national accounts for a large set of 93 countries for the period 1970–2013.
- Paper organization: literature review; decomposition of the Gini index linking factor shares and pseudo-Gini indices; microeconomics analysis with LIS; macroeconomic exploration; conclusions.

### II. REVIEW OF THE LITERATURE
- Historical context:
  - Factor income shares were central in classical and mid-20th century political economy (Ricardo; Kaldor 1961).
  - From the 1970s the topic receded due to perceived constancy of factor shares and broader ownership of capital by households.
- Renewed interest since early 2000s, with three motivations cited by Atkinson (2009):
  1. Link between macro (national accounts) and micro (household) incomes.
  2. Potential explanatory role of factor shares for personal income inequality.
  3. Social-justice concerns about fairness across income sources.
- Proposed drivers of the declining labor share in literature:
  - Globalization, technological change, international capital mobility, wage bargaining, financial liberalization, decline in relative price of investment goods.
  - Karabarbounis and Neiman (2014): the lower price of investment goods explains roughly half of the observed decline in the labor share.
- Two strands of empirical work:
  1. Documenting and explaining shifts in the labor share (country studies, macro cross-country studies).
  2. Examining consequences for personal income inequality (household survey decompositions).
- Prior findings:
  - Some studies (e.g., Jacobson and Occhino) find the decline in labor share associated with increased Gini in specific country cases (United States: a 1 percent decrease in labor share increases Gini by 0.15–0.33 percent).
  - ILO (2015) conclusion: “inequality starts in the labor market.”

### III. INCOME SHARES OR THE DISTRIBUTION OF INCOME? A LOOK AT HOUSEHOLD DATA
- Methodology:
  - Inequality measure: Gini index.
  - Decomposition approach follows Lerman and Yitzhaki (1985) and CBO (2011): total Gini as weighted average of pseudo-Gini (concentration) indices of income components where weights are income shares.
  - Key identities and marginal effects:
    - G = Σk sk Ck  (equation (1))
    - Changes approximated by ΔG ≈ Σk s0k ΔCk + Σk C0k Δsk + Σk Δsk ΔCk (equation (2)), with the third addend likely close to zero.
    - Marginal impact of pseudo-Gini: δG/δCk = sk (equation (3)).
    - Marginal impact of changes in income shares (labor vs capital): δG/δsl = (Cl - Cc) sl,y (?) as expressed in the paper: cl l y CC s G  δ δ  (equation (4) in source notation).
- Small-sample country decomposition (France, Germany, United Kingdom, United States):
  - Periods: US 1979–2013; UK 1979–2010; DE 1978–2010; FR 1978–2010.
  - Change in disposable income Gini (ΔGy_net): US 0.08; UK 0.10; DE 0.03; FR -0.01.
  - Change in gross income Gini (ΔGy): US 0.07; UK 0.10; DE 0.05; FR -0.01.
  - Change in market income Gini (ΔGm): US 0.10; UK 0.13; DE 0.08; FR 0.02.
  - Impact of changes in income shares (labour) on ΔGm (Δsl (C0l - C0c)): US 0.00; UK 0.00; DE 0.01; FR 0.00.
  - Impact of changes in pseudo-Gini indexes:
    - ΔCl (labor pseudo-Gini): US 0.09; UK 0.13; DE 0.06; FR 0.03.
    - ΔCc (capital pseudo-Gini) (=-sl0 ΔCc in table): US 0.01; UK 0.00; DE 0.02; FR 0.00.
  - Gini levels reported (initial and final years) for y_net, y, m, l, c are included in the table (examples):
    - G0_y_net: US 0.31; UK 0.27; DE 0.26; FR 0.33.
    - G_y_net in the final year: US 0.40; UK 0.36; DE 0.29; FR 0.31.
    - G0_m: US 0.41; UK 0.39; DE 0.42; FR 0.44.
    - Gm in the final year: US 0.51; UK 0.52; DE 0.49; FR 0.47.
    - G0_l: US 0.44; UK 0.43; DE 0.45; FR 0.46.
    - Gl in the final year: US 0.53; UK 0.57; DE 0.54; FR 0.53.
    - G0_c: US 0.92; UK 0.88; DE 0.61; FR 0.97.
    - Gc in the final year: US 0.94; UK 0.97; DE 0.87; FR 0.88.
- Extended sample results (43 countries; 231 income surveys; 1978–2010 start years vary by country):
  - Main finding replicated: the dominant driver of rising market-income inequality is the increase in the pseudo-Gini index of labor income; increased concentration of capital income also raises inequality but less so because labor is the largest component of market income.
  - Magnitudes and marginal effects (sample averages):
    - A 10 percent increase in the pseudo-Gini index of labor income increases the Gini index for market income by more than 9 percent.
    - A 10 percent decline in the labor share raises the market-income Gini by about 0.9 percent (effect small but statistically significant; primarily driven by emerging market economies).
  - Differences by subsample:
    - Emerging economies: pseudo-Gini index for capital income is on average higher by 0.16 than in advanced economies; the difference for labor income is 0.07.
- Average marginal impacts illustrated (Figure 2 summary):
  - Labor share effect on Gini small and negative.
  - Labor income pseudo-Gini has the largest positive marginal impact on the market-income Gini.
  - Capital income pseudo-Gini has a small positive marginal impact.
- Table 2 — Average effects on the Gini index for market income (sample of 43 countries; 231 observations):
  - Impact of a 0.01 change in the share of labor income (δGm/δsl): -0.0004 ** ; St. Dev 0.0012; T -2.2889; P>|t| 0.0272.
  - Impact of a 0.01 increase in the pseudo-Gini index:
    - δGm/δCl: 0.0096 *** ; St. Dev 0.0003; T 250.3138; P>|t| 0.0000.
    - δGm/δCc: 0.0004 *** ; St. Dev 0.0003; T 9.8787; P>|t| 0.0000.
  - Significance levels: * 10%, ** 5%, *** 1%.
  - Subsample point estimates (reported):
    - Impact of 0.01 change in labor share: Advanced economies -0.0001; Emerging economies -0.0010.
    - Impact of 0.01 increase in labor pseudo-Gini: Advanced 0.0096; Emerging 0.0097.
    - Impact of 0.01 increase in capital pseudo-Gini: Advanced 0.0004; Emerging 0.0003.
- Robustness and interpretation:
  - Results robust to alternative decomposition measures (Annex 1).
  - Inclusion of self-employed earnings in labor income in LIS definitions may affect direct comparability with other studies (e.g., CBO).
  - The third-order decomposition term (interaction of share changes and pseudo-Gini changes) is likely close to zero given slow movement of both shares and pseudo-Gini indices.
- Empirical implication emphasized: rising inequality of labor income (dispersion within wages) — especially at the top — is the primary driver of rising market-income inequality across a broad set of countries; changes in the labor share play a smaller but statistically significant role, with larger marginal effects observed in emerging markets.

*Italic: Source: Authors’ calculations on Luxembourg Income Study (LIS) data and macroeconomic national accounts as presented in the source content.*

### 1. Advanced Economies

### 1. Advanced Economies

### Data, coverage, and limitations
- Micro data: Luxembourg Income Study data used; average values across countries (panel 1: 26 countries; 174 observations or income surveys; panel 2: 17 countries; 57 observations or income surveys).
- Macro panel: unbalanced panel covering a large sample of countries, period from the 1970s to 2013 (country coverage varies).
- Sample size remark: "The sample includes about 350 observations for the Gini equation."
- Survey and measurement caveats:
  - Income surveys underreport capital income and poorly capture the extreme top of the distribution; analysis likely underestimates developments at the top.
  - Gini index is more sensitive to the middle of the distribution than to tails.
  - Recent work suggests top 1 percent shares have been increasing even more rapidly than other percentiles (Alvaredo and others 2013).

### Empirical framework
- Goal: confirm, using macro data and standard regression analysis, whether the wage share affects inequality and whether distribution of labor income has a larger impact on the Gini index than the wage share.
- Estimating equation (symbolically): y_it = α + β s_lit + γ I_lit + θ r_it + x_it'ρ + ε_it (equation (5) in source). Definitions preserved as in source:
  - s_l = wage (labor) share
  - I_l = inequality (dispersion) of labor income (proxy: ratio top 10 percent salaries to bottom 90 percent salaries; for robustness also 10-to-90 labor income available for OECD)
  - r = redistributive impact of tax and welfare system (proxied by ratio of public revenues to GDP, and social protection and health spending)
  - x_it = other control variables
- Methodology: panel techniques (fixed effects preferred per Breusch and Pagan Lagrange multiplier test); also panel instrumental variables and full information maximum likelihood estimators used to treat the wage share as endogenous.

### Main regression findings (fixed effects and baseline results)
- Core empirical conclusion: inequality is driven more by dispersion of labor income than by the wage share itself.
- Wage (labor) share:
  - In simple specifications, inequality declines when the wage share increases (negative coefficient in column 1).
  - The estimated coefficient on the wage share is significant only when dispersion of labor income is not included; when dispersion is included, the wage-share coefficient becomes not significant.
- Dispersion of labor income:
  - Robust, large, and positive association with the Gini index (statistically significant across specifications).
  - Examples of estimated coefficients in source tables (exact values reported in the chapter):
    - Table 3: "Dispersion of Labor income" coefficients reported such as 0.0242 ***, 0.0203 ***, 0.0174 ***, 0.0173 ***, 0.0173 ***, 0.0161 *** (absolute value of t statistics in parentheses).
    - Table 4 (robustness): Dispersion of Labor income 0.0173 ***, 0.3226 ***, 0.6036 *** (absolute value of z statistics in parentheses).
- Public policy / redistributive proxies:
  - Public Revenues: negative and significant relationship with Gini (e.g., -0.0011 ***, -0.0008 **, -0.0007 **, -0.0008 ** in Table 3; Table 4 reports public revenues -0.0008 **, 0.0000, -0.0014 *** in different specifications).
  - Public Social Protection Spending: coefficients not consistently significant (examples: -0.0011, -0.0006, -0.0009, -0.0007 in Table 3; Table 4 reports -0.0009, -0.0010, -0.0039 * in some specifications).
  - Public Health Spending: negative and in some specifications statistically significant (e.g., -0.0046 *, -0.0055 **, -0.0070 *** in Table 3; Table 4 reports -0.0055 **, -0.0021, -0.0039 *).
- Globalization proxies:
  - Economic Globalization: positive and significant in some panel specifications (e.g., 0.0007 *** in Table 3; Table 4 reports 0.0007 ***, 0.0007, 0.0003 *).
  - Financial Globalization: positive and significant in at least one specification in Table 3 (0.0094 **).

### Endogeneity checks and instruments
- Concern: simultaneity between income inequality and wage share.
- Robustness approaches:
  - Panel instrumental variables estimation and full information maximum likelihood (ML) estimator treating labor share as endogenous.
  - Instrument list (used in panel IVs): lagged labor share, lagged unemployment rate, employment in the services sector, and a proxy for the wage-bargaining framework.
  - Instrument validity diagnostics reported in text:
    - R2 of the first-step wage share equation = 0.86.
    - R2 of regressing the Gini index on the set of instruments = 0.01.
  - Robustness results (Table 4): labor share negative but small effect on inequality; inequality of labor income remains large and significantly positive.

### Determinants of wage (labor) income dispersion (simple regression, Table 5)
- Findings (selected coefficients preserved exactly as reported in the chapter):
  - Financial Globalization: positive association with wage dispersion; example coefficients: 0.0719 **, 0.0701 *, 0.037 *, 0.1531 *, 0.0788 *** (absolute values of z statistics in parentheses).
  - Unemployment: positive association with dispersion; example coefficients: 0.0082 *, 0.0066 *, 0.0231 **, 0.0075 **.
  - Industry Unionization: negative association with dispersion; example coefficients: -0.0118 ***, -0.024 ***, -0.01 ***.
  - Tertiary Education (share): negative association with dispersion; example coefficients: -0.018 *** , -0.009 ***.
  - Government Spending (primary government spending as proxy for size of the state): negative association; example coefficient: -0.009 ***.
- Interpretation in text: higher financial globalization and higher unemployment levels are associated with higher dispersion of wages; higher unionization in industry, a higher share of educated workers, and higher primary government spending help reduce wage dispersion.

### Policy implications and conclusions
- Core conclusion: the most important determinant of income inequality is the dispersion of labor income rather than the aggregate wage (labor) share.
- Explanation: household income is largely labor earnings, and its distribution has become more unequal.
- Factors associated with rising wage dispersion: growing financial globalization, decline in industry unionization, and decline in the size of the state.
- Policy recommendations (as stated in source):
  - Policymakers should pay attention to labor market outcomes and wage dispersion to avoid unfavorable distributional consequences.
  - Public policies that support inclusive growth—e.g., promoting participation in the labor market and strengthening human capital of low-income groups—may prevent the rise in economic disparities.
- The paper notes that deeper analysis of the drivers of wage dispersion is beyond its scope and could be taken up in further research.

*Source: _wp15244 - 1. Advanced Economies (chapter text and tables as supplied).*

### Annex 2. Inequality Decomposition Using the LIS Data Set

### Annex 2. Inequality Decomposition Using the LIS Data Set

### Definitions and income components
- Reference unit: household; income definition: per capita equivalent income computed using the LIS equivalence scale.
  - Footnote: The LIS equivalence scale is defined as the square root of the number of individuals in the household.
- Total gross income y is defined as market income m plus transfers g:
  - gmy
- Transfer income g comprises private transfers (alimony, remittances, transfers from nonprofit institutions) and public transfers (pensions, unemployment benefits, disability benefits). Public transfers make up the bulk of transfer income.
- Gross market income m is the sum of labor l and capital income c:
  - clm
  - Note: The labor income definition used includes both wages from paid employment and income from self employment.
- Net (or disposable) household income is total income minus taxes t:
  - tyy
    net
    

### Inequality decomposition framework and key equations
- Breakdown of changes in inequality in market income over a certain period (empirical counterpart of equation (9)) is expressed as:
  - 
    
    
    0impact  indicesion    t concentra
    00
    impact   shares   income
    00
    
    
    ccllccllccllm
    CsCsCsCsCsCsG
    (17)
  - In equation (17), sl and sc, and Cl and Cc are, respectively, the income shares and pseudo-Gini indices for l and c, and 0 is the base year (or the initial year in the analysis, which varies depending on the country).
- Given that income shares add up to 1 (lc
  ss), equation (17) can be rewritten as:
  - 
    
    
    0impact  indices  tion   concentra
    00
    impact  shares   income
    00
    
    
    cllccllcllm
    CCsCsCsCCsG
    (18)
  - The observed impact of changes in income composition on inequality depends on the initial values of the pseudo-Gini indices for labor and capital.

### Impact of transfers and taxation
- Impact of transfers on inequality:
  - my
    GG
    (19)
- Impact of taxation on inequality:
  - y
    y
    GG
    net
    
    (20)

### Marginal effects of market-income composition
- Gini index for gross market income:
  - ccllm
    sCsCG
    (21)
- Since lc
  ss1 (22), the marginal impact from a variation in market-income composition is:
  - cl
    l
    m
    CC
    s
    G
    
    δ
    δ
    (23)
- Interpretation:
  - If the pseudo-Gini index for capital is higher than that for labor, then an increase (reduction) in the labor share reduces (raises) inequality.
  - This requires:
    - l
      Gini
      c
      Gini
      l
      c
      GG
      ρ
      ρ
      
      (24)
    - Equivalently, in terms of average incomes:
      - 
        
        c
        mFc
        mFl
        l
        ,cov
        ,cov
        
        (25)
      - Which requires average labor income to be “sufficiently” higher than average capital income.

### Countries included in the LIS-based analysis (Annex Table 2.1)
- List of countries considered (and indication if income components are recorded gross or net of taxes; definition may vary by year of survey, in this case both gross, net or mixed are listed). Years shown follow each country exactly as in the source.
  - Austral i a ( gross)1981; 1985; 1989; 1995; 2001; 2003; 2008; 2010
  - Austri a ( ne t; gross)1994; 1997; 2000; 2004
  - Be l gi um (ne t; gross)1985; 1988; 1992; 1995; 1997; 2000
  - Brazi l  ( gross)2006; 2009; 2011
  - Canada (gross)1981; 1987; 1991; 1994; 1997; 1998; 2000; 2004; 2007; 2010
  - China (gross)2002
  - Col ombi a ( gross)2004; 2007; 2010
  - Cze ch Re publ i c (gross)1992; 1996; 2004
  - De nmark ( gross)1987; 1992; 1995; 2000; 2004; 2007; 2010
  - Egypt ( ne t)2012
  - Estoni a (mi xe d, gross)2000; 2004; 2007; 2010
  - Fi nl and ( mi x e d, gross)1987; 1991; 1995; 2000; 2004; 2007; 2010
  - France  (mi xe d; gross)1978; 1984; 1989; 1994; 2000; 2005; 2010
  - Ge rmany ( gross)1978; 1981; 1983; 1984; 1989; 1994; 2000; 2004; 2007; 2010
  - Gre e ce  ( ne t; gross)1995; 2000; 2004; 2007; 2010
  - Guatemal a (gross)2006
  - Hungary (ne t)1991; 1994; 1999; 2005; 2007; 2009; 2012
  - Ice l and (gross)2004; 2007; 2010
  - India (net)2004
  - Ire l and ( gross; ne t)1987; 1994; 1995; 1996; 2000; 2004; 2007; 2010
  - Israe l  ( gross)1979; 1986; 1992; 1997; 2001; 2005; 2007; 2010
  - Ital y ( ne t; mi x e d)1986; 1987; 1989; 1991; 1993; 1995; 1998; 2000; 2004; 2008; 2010
  - Japan (gross)2008
  - Lux e mbourg ( ne t; gross)1985; 1991; 1994; 1997; 2000; 2004; 2007; 2010
  - Me x i co (ne t)1984; 1989; 1992; 1994; 1996; 1998; 2000; 2002; 2004; 2008; 2010
  - Ne the rl ands ( gross)1983; 1987; 1990; 1993; 1999; 2004; 2007; 2010
  - Norway ( gross)1979; 1986; 1991; 1995; 2000; 2004; 2007; 2010
  - Pe ru ( ne t)2004
  - Pol and (ne t; mi x e d; gross)1992; 1995; 1999; 2004; 2007; 2010
  - Romani a ( gross)1995; 1997
  - Russi a (ne t)2000; 2004; 2007; 2010
  - Se rbi a (ne t)2006; 2010; 2013
  - Sl ovak Re publ i c ( gross; ne t)1992; 1996; 2004; 2007; 2010
  - Sl ove ni a (ne t)1997; 1999; 2004; 2007; 2010
  - South Afri ca ( gross)2008; 2010
  - South Korea (Gross)2006
  - Spai n ( ne t; gross)1980; 1985; 1990; 1995; 2000; 2004; 2007; 2010
  - Swe de n ( gross)1981; 1987; 1992; 1995; 2000; 2005
  - Swi tze rl and ( gross)1982; 1992; 2000; 2002; 2004
  - Tai wan ( gross)1981; 1986; 1991; 1995; 1997; 2000; 2005; 2007; 2010
  - Uni te d Ki ngdom ( gross)1979; 1986; 1991; 1995; 1994; 1999; 2004; 2007; 2010
  - Uni te d State s (gross)1979; 1986; 1991; 1994; 1997; 2000; 2004; 2007; 2010; 2013
  - Uruguay (net)2004
- Source and cut-off date (as stated in the annex table): Source: LIS database. Cut off date for data February 24, 2015.
- Note in original table: for a detailed definition of the recording method (gross, net or mixed) of taxes see http://www.lisdatacenter.org/

### Data, estimation sample, and descriptive statistics (Annex 3 excerpt)
- Annex Table 3.1 reports earliest and latest values for the Gini index for countries included in the estimation sample. Selected entries (Country — Earliest Observation Gini Year — Latest Observation Gini Year) appear exactly as in the source; examples include:
  - Argentina EME 0.461995 0.442007
  - Armenia EME 0.342003 0.312008
  - Australia ADV 0.281981 0.342008
  - Austria ADV 0.231987 0.272011
  - Belgium ADV 0.231985 0.242011
  - Brazil EME 0.552004 0.522008
  - Canada ADV 0.321971 0.322008
  - China EME 0.361996 0.422005
  - France ADV 0.291979 0.312012
  - Germany ADV 0.271973 0.282012
  - Mexico EME 0.521996 0.452010
  - Netherlands ADV 0.251983 0.222012
  - Norway ADV 0.221979 0.232012
  - South Africa EME 0.571995 0.632009
  - Spain ADV 0.321980 0.342012
  - United Kingdom ADV 0.271969 0.362011
  - United States ADV 0.321974 0.372010
- Note: ADV = advanced economy; EME = emerging market economy; LIDC = low-income and developing countries.
- Data sources used in the estimation analysis (as listed):
  - Disposable Gini index: OECD, Eurostat, World Bank's World Development Indicators, LIS, Socio-Economic Database for Latin America and the Caribbean.
  - Wage share: ILO database; adjusted wage share when available; European Commission’s Annual Macroeconomic Database (AMECO) used for longer series where applicable.
  - Unemployment rate: IMF World Economic Outlook.
  - Employment rate in the services sector: ILO.
  - Wage-setting institutional variables: Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts data set, 1960–2011 (ictwss_Coord and ictwss_Type).
  - Dispersion of labor income: ratio of total income of the top 10 percent to the bottom 10 percent, from World Bank’s World Development Indicators.
  - Ratios of public revenue, social protection spending, and health expenditures to GDP: IMF World Economic Outlook; Eurostat; OECD; World Health Organization; UNESCO; CEPALSTAT; Asian Development Bank; World Bank; IMF International Financial Statistics.
  - Economic globalization: KOF Index of Globalization (Dreher, Gaston, and Martens 2008).
  - Financial globalization: log of total foreign assets and liabilities divided by GDP, computed from updated and extended versions of the Lane and Milesi-Ferretti (2007) dataset.

*Source: Annex 2 and associated tables and notes in the supplied IMF document.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15244.pdf_
