## Appendix 1. Country Coverage

## Source details

**Canonical URL:** [Appendix 1. Country Coverage](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16111.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16111.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16111.pdf.json)

---

### I. Introduction: objectives and scope
- Objectives:
  - Investigate the impact high inequality has had on sub-Saharan Africa’s growth performance.
  - Investigate drivers behind income inequality and factors explaining its persistence to better inform policy for inclusive growth and SDG progress.
- Distinctive approach:
  - Tests joint effects of income and gender inequality on growth at different stages of development.
  - Allows the growth-inequality relationship to differ between low-income countries and other country groups (low-income, fragile, and middle-income and oil exporting).
- Sample and data emphasis:
  - Uses newly available data sets to assess joint effects; compares sub-Saharan Africa to other regions such as ASEAN and Latin America and the Caribbean.

### II. Main findings (four-fold)
- System-GMM estimations:
  - Income and gender inequality are found to jointly impede growth mostly in the initial stages of development, resulting in large growth losses in sub-Saharan Africa.
  - Findings are robust to alternative measures of income and gender inequality and to variations in controls in the growth regressions.
- Potential growth dividends from reducing inequalities:
  - Further progress in reducing income and gender inequality could deliver significant sustained growth dividends, particularly for low-income countries.
- Growth decomposition relative to ASEAN:
  - Average annual GDP per capita growth in sub-Saharan African countries could be higher by as much as 0.9 percentage points if income and gender inequality were reduced to the levels observed in the fast-growing Association of South East Asian Nations (ASEAN).
  - The growth shortfall of Latin America and the Caribbean with ASEAN is mainly explained by income inequality.
- Drivers of income inequality:
  - Structural factors and policies drive changes in income inequality over time, with differing relevant factors by region and development level.
  - High income inequality in sub-Saharan Africa is partly driven by:
    - Stage of economic development.
    - Low shares of working age population to total population.
    - Gender inequality.
  - Policies improving access of low-income households and women to opportunities (education, finance) can help reduce inequality.

### III. Data and stylized facts: trends in income and gender inequalities
- Poverty and inequality measures:
  - Poverty data: World Bank’s World Development Indicators database.
  - Income inequality: net Gini (Gini coefficient after taxes and redistributive government programs) from SWIID Version 5.0, October 2014.
  - Gender inequality: United Nations’ Gender Inequality Index (GII).
  - Gender-related legal restrictions: World Bank’s Women, Business and the Law Database.
- Key regional statistics and trends:
  - Sub-Saharan Africa poverty rates: decline from 57 percent in 1990 to less than 43 percent in 2012.
  - Sub-Saharan Africa has the second highest levels of income inequality globally, after Latin America and the Caribbean.
  - Income inequality in sub-Saharan Africa has moved little overall; substantial cross-country variation exists.
  - Gender inequality in sub-Saharan Africa remains among the highest and has declined slower than in other regions despite improvements in education and health outcomes and relatively high female labor force participation rates on average.
  - Sub-Saharan Africa continues to have more legal gender-based restrictions than comparator groups.
- Observed relationships:
  - Relative to the rest of the world, sub-Saharan Africa exhibits markedly higher income inequality at all stages of development.
  - Association between overall gender inequality and income inequality is much weaker in sub-Saharan Africa than elsewhere.
  - Gap between male and female labor force participation rates is on average 15 percentage points lower in sub-Saharan Africa than in the rest of the world (contextual note).

### IV. Inequality and growth: literature and channels
- Income inequality and growth:
  - Evidence that lower net income inequality is associated with faster growth and longer growth spells.
  - Negative channels: imperfect credit markets, socio-political instability, poor governance.
  - Potential positive channels from inequality include higher savings by rich households and incentives for innovation.
  - Empirical heterogeneity: effects may differ by development level and methodological choices.
- Gender inequality and growth:
  - Gender gaps in education, legal rights, finance, employment, and political empowerment affect growth.
  - Female education improves child health, fertility outcomes, and labor participation.
  - Legal restrictions on women’s economic rights correlate with larger gender gaps in labor force participation.
  - Gender gaps in economic participation associated with large GDP losses across income levels.
  - Increased female legislative presence may shift public expenditure toward health and education and raise potential growth.
  - The contribution of gender gaps to growth after controlling for income inequality is less understood.

### V. Empirical methodology and data used in analysis
- Empirical approach:
  - Relates growth in GDP per capita in a sample of 115 countries to indicators of income and gender inequality and standard growth determinants using system-GMM.
- Control variables included:
  - Initial income, initial level of infrastructure, years of schooling, investment to GDP, inflation, indicators of institutional quality, terms-of-trade changes.
- Data handling:
  - Five-year non-overlapping averages for 1995–2014; initial value is first-year observation or latest available observation in previous period.
- Endogeneity and estimation:
  - Two-step system-GMM (Xtabond2 for STATA) used to account for potential endogeneity of inequality and investment variables.
  - Arellano–Bond test and Hansen J statistic used for specification checks; reported p-values indicate no serious second-order serial correlation and no over-identification problems in reported specifications.
- Inequality and gender measures:
  - Initial income inequality: net Gini from SWIID v.5.0 and t20/b40 ratio (WDI and UNU-WIDER).
  - Initial income share of middle class: sum of third and fourth quintiles.
  - Gender inequality: UN GII (0 to 100), averaged over five-year periods and extended 1990–2010 using Gonzales and others (2015b).
  - Female legal inequity index: sum of six binary legal indicators (range 0–6), higher values = less equitable legal rights.

### VI. General results — system-GMM baseline (selected coefficients and diagnostics)
- All Countries (selected coefficients from columns (1)–(5)):
  - Initial income per capita (log): -1.420***, -1.728***, -0.998***, -1.921***, -1.209***
  - Initial level of infrastructure: 0.597***, 0.622***, 0.646***, 0.372**, 0.539***
  - Institutional quality (index): 0.046**, 0.054**, 0.022, 0.064***, 0.054**
  - Inequality measures:
    - Initial income inequality (t20/b40 ratio): -0.060
    - Initial income inequality (t20/b40 ratio) x LICs: -0.354***
    - Initial income inequality (net Gini): -0.025
    - Initial income inequality (net Gini) x LICs: -0.048***
    - Initial income share of middle class: 0.158***
    - Gender inequality (lagged): -0.031**
    - Gender inequality x LICs (lagged): -0.039***
    - Female legal inequity (index): -0.246
  - Serial correlation (p-value): 0.103, 0.044, 0.370, 0.121, 0.907
  - Hansen test (p-value): 0.259, 0.230, 0.493, 0.943, 0.553
  - Observations: 381, 429, 269, 430, 334
  - Number of countries: 121, 117, 114, 115, 84
- Emerging and Developing Countries (columns (6)–(10)):
  - Initial income per capita (log): -1.133***, -1.382***, -0.843***, -1.615***, -1.201***
  - Initial level of infrastructure: 0.586***, 0.618***, 0.563***, 0.365*, 0.574***
  - Institutional quality (index): 0.027, 0.035, 0.012, 0.045**, 0.045
  - Inequality measures:
    - Initial income inequality (t20/b40 ratio): -0.116**
    - Initial income inequality (t20/b40 ratio) x LICs: -0.260***
    - Initial income inequality (net Gini): -0.038
    - Initial income inequality (net Gini) x LICs: -0.036**
    - Initial income share of middle class: 0.154***
    - Gender inequality (lagged): -0.057***
    - Gender inequality x LICs (lagged): -0.024**
    - Female legal inequity (index): -0.279
  - Serial correlation (p-value): 0.246, 0.076, 0.400, 0.141, 0.922
  - Hansen test (p-value): 0.677, 0.187, 0.588, 0.904, 0.625
  - Observations: 268, 300, 234, 301, 266
  - Number of countries: 87, 84, 85, 81, 67
- Key empirical signals:
  - Interaction terms with LICs are negative and often highly significant, indicating larger adverse growth associations of initial inequality in LICs.
  - Initial income share of middle class is positive and significant (0.158***; 0.154***).
  - Gender inequality (lagged) is negative and significant in several specifications (-0.031**; -0.057***).
  - Female legal inequity coefficients are negative but not significant in baseline tables (-0.246; -0.279).

### VII. Core magnitudes and extended GMM results (Table 2 highlights)
- Income inequality magnitude:
  - A one-percentage point reduction in the initial Gini coefficient in low-income countries is associated with a 0.15 percentage point cumulative increase in growth over a 5-year period.
- Gender inequality magnitude:
  - A one percentage point reduction in gender inequality in low-income countries is associated with higher cumulative growth over 5 years of 0.2 percentage points.
  - Female legal inequity (index) coefficients reported as negative and significant across extended models (see Table 2 in source).
- Selected control variable coefficients (representative reported values):
  - Initial income per capita (log): -1.234***, -1.347***, -1.081***, -1.746***, -1.148***, -1.608***
  - Fixed capital investment (% GDP): 0.134*, 0.184***, -0.014, 0.093*, 0.132*, 0.028
  - Schooling (years): 0.119, 0.068, 0.159*, 0.045, 0.105, 0.154*
  - Dependent population growth (%): -0.356**, -0.293**, -0.539***, -0.224, -0.306**, -0.286**
  - Infrastructure index: 0.238*, 0.194, 0.270*, 0.294*, 0.237, 0.334**
  - High inflation dummy: -1.583***, -1.627***, -1.621***, -1.228***, -1.472***, -1.552***
  - Terms of trade (percent change): 0.068**, 0.076***, 0.091***, 0.098***, 0.063**, 0.094***
  - Institutional quality (index): 0.047***, 0.063***, 0.040**, 0.080***, 0.064***, 0.054***
- Model diagnostics (selected):
  - Number of instruments: 15, 15, 14, 15, 15, 17
  - Serial correlation (p-value): 0.071, 0.025, 0.202, 0.209, 0.154, 0.274
  - Hansen test (p-value): 0.210, 0.335, 0.319, 0.445, 0.712, 0.700
  - Country fixed effects: Yes
  - Time (period) fixed effects: Yes
  - Observations: 344, 384, 237, 419, 304, 240
  - Number of countries: 110, 106, 104, 115, 78, 78

### VIII. Robustness, channels, and quantile regressions on inequality changes
- Robustness:
  - Income-inequality growth link robust across different inequality measures (net Gini, t20/b40, income share of middle class).
  - Adding gender inequality variables did not diminish the association between income inequality and growth, indicating distinct channels.
- Quantile regression findings (change in net Gini, 1991–2010; 469 five-year changes, 135 countries):
  - Factors associated with reducing net income inequality:
    - GDP per capita growth: negative association (e.g., -0.1010***; -0.0991***).
    - Private capital stock: negative (e.g., -0.0056***).
    - Education spending: negative in several specifications (e.g., -0.1830*).
    - Share of working-age population: negative (e.g., -0.0585*).
    - Fiscal redistribution (market minus net Gini): negative (e.g., -0.0842***).
  - Factors associated with increasing net income inequality:
    - Higher beginning-of-period gender and education inequality.
    - Increased share of natural resource exports (results driven mainly by oil exports).
  - Financial deepening:
    - M2/GDP insignificant in overall sample.
    - Women’s access to financial services (ability to open a bank account) contributed to inequality reductions especially in lower-income country quantiles and SSA 10th quantile regressions.
  - Other structural/macroeconomic correlates (selected exact estimates):
    - Share of agriculture: 0.0515*** (overall), with negative interactions for LIC and SSA in some specifications.
    - Education inequality: 0.0168; 0.0240** in different specifications.
    - Gender inequality index: 0.0301**; 0.0308***.
    - Fiscal redistribution: -0.0842***; interaction for SSA -0.7370*** in one column.
    - Public capital stock: small negative, with SSA interaction -0.0208***.
    - Trade openness: mixed; negative and significant in some SSA specifications (e.g., -0.0356***).

### IX. Implications for sub-Saharan Africa (SSA)
- Growth decomposition vs ASEAN-5 (Indonesia, Malaysia, Philippines, Thailand, Vietnam):
  - SSA’s average annual real GDP per capita growth has been about 1½ percentage points lower than ASEAN-5 over the last decade.
  - Reducing three inequality indicators to ASEAN levels could boost SSA’s annual per capita GDP growth by on average about 0.9 of a percentage point.
  - Closing the infrastructure gap has a roughly similar impact (~0.9 percentage point).
  - Compared with Latin America and the Caribbean, reducing gender inequality and legal gender-related restrictions yields sizeable growth effects for SSA.
- Subgroup-specific SSA findings:
  - Low-income countries (excluding fragile states):
    - Catch-up effect from low initial income relative to ASEAN contributes about 2½ percentage points of real GDP per capita growth.
    - This catch-up is more than undone by weak infrastructure and lower human capital accumulation.
    - Gender inequality accounts for ¾ percentage points of the growth differential.
  - Fragile states:
    - Lower quality of infrastructure and political institutions explains the largest fraction of the growth differential.
    - Reducing gender inequality could boost annual per capita GDP growth by 2/3 percentage points.
  - Middle-income and oil-exporting countries:
    - Reducing income inequality to ASEAN levels is important to raise growth.
    - Removing legal gender-related restrictions shows particularly strong growth payoffs for oil-exporting SSA countries (caveat on correlation vs causation).

### X. Policy-relevant implications and recommendations
- Core conclusions:
  - Income and gender inequality, including legal gender-related restrictions, impede growth mainly in countries at earlier stages of development, causing large growth losses in SSA.
  - Addressing both income and gender inequalities can deliver significant sustained growth dividends, particularly for low-income countries.
  - Gender inequality affects growth through channels distinct from those of income inequality; policies addressing both are complementary.
- Policy priorities for SSA (empirically suggested):
  - Reduce income inequality (through redistribution and targeted policies) and gender inequality (including removal of legal barriers).
  - Close infrastructure and human capital gaps.
  - Accelerate the demographic transition by raising investment in human capital.
  - Improve progressivity of tax systems and expenditures; move away from blunt instruments like universal subsidy programs.
  - Expand access to finance and opportunities for women and low-income households.
- Design considerations:
  - Policies must be tailored to country-specific situations, administrative capacity, and potential tradeoffs.
  - Ensure redistribution policies are carefully designed to avoid adverse effects on growth.
  - Removal of legal restrictions on women and increased accessibility of education and healthcare may catalyze social-norm changes.

### XI. Country coverage (appendix listing)
- High income countries (World Bank definition) include: United States, United Kingdom, Austria, Belgium, Denmark, France, Germany, Italy, Luxembourg, Netherlands, Norway, Sweden, Switzerland, Canada, Japan, Finland, Greece, Iceland, Ireland, Malta, Portugal, Spain, Australia, New Zealand, Chile, Uruguay, Barbados, Trinidad and Tobago, Cyprus, Israel, Hong Kong SAR, China, Korea Rep., Singapore, Russian Federation, Czech Republic, Slovak Republic, Estonia, Latvia, Lithuania, Croatia, Slovenia, and Poland.
- Middle income (Upper Middle Income) countries include: Turkey, South Africa, Argentina, Brazil, Colombia, Costa Rica, Dominican Republic, Ecuador, Mexico, Panama, Peru, Venezuela RB, Belize, Jamaica, St. Lucia, Suriname, Iran Islamic Rep., Iraq, Jordan, Malaysia, Maldives, Thailand, Algeria, Angola, Botswana, Gabon, Mauritius, Seychelles, Namibia, Tunisia, Fiji, Marshall Islands, Azerbaijan, Belarus, Albania, Kazakhstan, Bulgaria, China, Turkmenistan, Serbia, Montenegro, Hungary, Macedonia FYR, Bosnia and Herzegovina, and Romania.
- Low income (Lower Middle Income or Low Income) countries include: Bolivia, El Salvador, Guatemala, Haiti, Honduras, Nicaragua, Paraguay, Guyana, Syrian Arab Republic, Egypt Arab Rep., Yemen Rep., Afghanistan, Bangladesh, Bhutan, Cambodia, Sri Lanka, India, Indonesia, Lao PDR, Nepal, Pakistan, Philippines, Vietnam, Djibouti, Burundi, Cameroon, Cabo Verde, Central African Republic, Chad, Comoros, Congo Rep., Congo Dem. Rep., Benin, Ethiopia, The Gambia, Ghana, Guinea-Bissau, Guinea, Côte d'Ivoire, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Morocco, Mozambique, Niger, Nigeria, Zimbabwe, Rwanda, São Tomé and Príncipe, Senegal, Sierra Leone, Sudan, Swaziland, Tanzania, Togo, Uganda, Burkina Faso, Zambia, Papua New Guinea, Micronesia Fed. Sts., Armenia, Georgia, Kyrgyz Republic, Moldova, Tajikistan, Ukraine, Uzbekistan, and Mongolia.

*Source: _wp16111 - Appendix 1. Country Coverage (PDF).*

### Appendix 1. Country Coverage ...........................................................................................

### Appendix 1. Country Coverage

### I. Introduction: objectives and scope
- Objectives:
  - Investigate the impact high inequality has had on sub-Saharan Africa’s growth performance.
  - Investigate drivers behind income inequality and factors explaining its persistence to better inform policy for inclusive growth and SDG progress.
- Distinctive approach:
  - Tests joint effects of income and gender inequality on growth at different stages of development.
  - Allows the growth-inequality relationship to differ between low-income countries and other country groups (low-income, fragile, and middle-income and oil exporting).
- Sample and data emphasis:
  - Uses newly available data sets to assess joint effects; compares sub-Saharan Africa to other regions such as ASEAN and Latin America and the Caribbean.

### II. Main findings (four-fold)
- Using system-GMM estimations:
  - Income and gender inequality are found to jointly impede growth mostly in the initial stages of development, resulting in large growth losses in sub-Saharan Africa.
  - Findings are robust to alternative measures of income and gender inequality and to variations in controls in the growth regressions.
- Potential growth dividends from reducing inequalities:
  - Further progress in reducing income and gender inequality could deliver significant sustained growth dividends, particularly for low-income countries.
- Growth decomposition relative to ASEAN:
  - Average annual GDP per capita growth in sub-Saharan African countries could be higher by as much as 0.9 percentage points if income and gender inequality were reduced to the levels observed in the fast-growing Association of South East Asian Nations (ASEAN).
  - The growth shortfall of Latin America and the Caribbean with ASEAN is mainly explained by income inequality.
- Drivers of income inequality:
  - Both structural factors and policies drive changes in income inequality over time, with differing relevant factors by region and development level.
  - High income inequality in sub-Saharan Africa is partly driven by:
    - Stage of economic development.
    - Low shares of working age population to total population.
    - Gender inequality.
  - Policies improving access of low-income households and women to opportunities (education, finance) can help reduce inequality.

### III. Data and stylized facts: trends in income and gender inequalities
- Poverty and inequality measures:
  - Poverty data from the World Bank’s World Development Indicators database.
  - Income inequality measured by the net Gini (Gini coefficient after taxes and redistributive government programs).
  - Net Gini data source: Standardized World Income Inequality Database (SWIID, Solt 2014); SWIID Version 5.0, October 2014.
  - Gender inequality measured by the United Nations’ Gender Inequality Index (GII).
  - Gender-related legal restrictions measured by the World Bank’s Women, Business and the Law Database.
- Key regional statistics and trends:
  - Sub-Saharan Africa poverty rates: decline from 57 percent in 1990 to less than 43 percent in 2012.
  - Sub-Saharan Africa has the second highest levels of income inequality globally, after Latin America and the Caribbean.
  - Income inequality in sub-Saharan Africa has moved little overall; substantial cross-country variation exists (e.g., Lesotho, Niger, Sierra Leone experienced significant declines; in one third of countries with data, cumulative growth over 1995–2011 associated with increases in income inequality).
  - Gender inequality in sub-Saharan Africa remains among the highest and has declined slower than in other regions despite improvements in education and health outcomes and relatively high female labor force participation rates on average.
  - Sub-Saharan Africa continues to have more legal gender-based restrictions than comparator groups.
- Observed relationships:
  - Kuznets-style relationship: relative to the rest of the world, sub-Saharan Africa exhibits markedly higher income inequality at all stages of development (Figure 4).
  - Association between overall gender inequality and income inequality is much weaker in sub-Saharan Africa than elsewhere (Figure 5), suggesting other factors contribute substantially to income inequality in the region.
  - Note on labor force gaps: gap between male and female labor force participation rates is on average 15 percentage points lower in sub-Saharan Africa than in the rest of the world (contextual explanation provided in footnote).

### IV. Inequality and growth: literature and channels
- Income inequality and growth:
  - Growing evidence that lower net income inequality is associated with faster growth and longer growth spells (references summarized).
  - Negative association channels include:
    - Imperfect credit markets limiting low-income households’ investment in education and physical capital (Galor and Zeira 1993; Corak 2013).
    - Socio-political instability and poor governance reducing private investment.
  - Possible growth-promoting channels for inequality noted (higher savings by rich households, incentives for innovation, returns to education).
  - Empirical heterogeneity:
    - Effect of inequality on growth may change with development level; may be more binding at low income levels (Barro 2000).
    - Lack of consensus reflects nonlinearity, composite nature of inequality measures, different empirical methods, and omitted dimensions (e.g., gender inequality).
- Gender inequality and growth:
  - Gender inequality assessed via gaps in opportunities (education, legal system, finance) and outcomes (employment, political empowerment).
  - Education gaps reduce human capital quality and growth (Klasen 1999); female education improves child health, fertility, labor participation (Mitra, Bang, and Biswas 2015).
  - Legal restrictions on women’s economic rights correlate with larger gender gaps in labor force participation (Gonzales and others 2015a).
  - Gender gaps in economic participation associated with large GDP losses across income levels (Elborgh-Woytek and others 2013; Stotsky 2006).
  - Increased female legislative presence may shift public expenditure toward health and education and raise potential growth (Mitra, Bang, and Biswas 2015).
  - Empirical gaps: the contribution of gender gaps to growth after controlling for income inequality is less understood; many studies examine gender dimensions separately.

### V. Empirical methodology and data used in analysis
- Empirical approach:
  - Relates growth in GDP per capita in a sample of 115 countries to indicators of income and gender inequality and standard growth determinants using econometric analysis (system-GMM referenced earlier).
- Control variables included:
  - Initial income (catch-up effect).
  - Initial level of infrastructure.
  - Years of schooling.
  - Investment to GDP (capital and labor quality).
  - Inflation.
  - Indicators of institutional quality (macroeconomic and political management).
  - Terms-of-trade changes (external shocks).
- Appendix information:
  - Appendix Table 1 provides exact definitions of variables and data sources used in regressions (referenced).

*Source: _wp16111 - Appendix 1. Country Coverage (PDF).*

### Appendix 1 also provides information on the coverage of the countries. Due to the paucity of data

### _wp16111 - Appendix 1 also provides information on the coverage of the countries. Due to the paucity of data

### Data coverage and inequality measures
- Country coverage varies between models due to paucity of data across inequality measures and growth determinants.
- Five-year periods are used for all variables; the initial value is the observation in the first year or, if unavailable, the latest available observation in the previous period.
- Income inequality measures used:
  - Initial income inequality: net Gini taken from the SWIID v.5.0. (Empirical results broadly similar if the traditional (market) Gini is used.)
  - Initial ratio of the top 20 to the bottom 40 percentiles of the income distribution (t20/b40 ratio). Source: World Bank’s World Development Indicators (WDI) database, supplemented by the UNU-WIDER database.
  - Initial income share of middle class: sum of income shares of the third and fourth quintiles. Data sources same as for the t20/b40 ratio.
- The analysis consistently indicates that inequality remains high in sub-Saharan Africa and Latin America.

### Gender inequality and legal restrictions
- Primary gender inequality measure: UN’s Gender Inequality Index (GII), extended for 1990–2010 using GII methodology from Gonzales and others (2015b). The GII:
  - Ranges between 0 (equal) and 100 (unequal).
  - Is averaged over each five-year period.
- Female legal inequity index:
  - Constructed as the sum of six legal indicators (in 0-1 format) representing women’s legal rights to earn and hold income and wealth.
  - Ranges between 0 and 6; higher values correspond to less equitable legal rights for women.
  - Six indicators included:
    - (a) unmarried women have equal property rights for immovable property;
    - (b) married women have equal inheritance rights;
    - (c) joint titling of property is default for married couples;
    - (d) married women can get a job or pursue a profession;
    - (e) adult married woman can open a bank account;
    - (f) married woman can sign contracts (without permission from another family member).
- Note: income inequality measures capture inequality at the household level; the gender inequality index captures a combination of intra- and inter-household inequality to the extent that women are household members or heads.

### Empirical strategy and estimation
- Sample period: 1995–2014.
- Non-overlapping five-year averages used to abstract from business cycle fluctuations and deal with data gaps.
- Endogeneity concerns:
  - Two-step system generalized methods of moments (system-GMM) and initial levels of inequality for each 5-year period are used to account for possible endogeneity of inequality and investment variables.
  - The Xtabond2 package for STATA (Roodman 2009) is used to estimate the system-GMM regressions.
- Specification tests:
  - Arellano–Bond test statistic used to test for no second-order serial correlation in errors; the test fails to reject the null hypothesis of zero correlation.
  - Hansen J statistic used to test for over-identification due to large numbers of instruments from system-GMM; the statistic passes criteria for no over-identification problem in the model specifications.

### General results (reduced-form growth models with inequality)
- Modeling approach: start with a simple Solow growth model with added inequality indicators. Simplified models control for initial income, initial level of infrastructure, and institutional quality, plus individual inequality indicators.
- Inequality indicators included: net Gini, initial income share of middle class, t20/b40 ratio; gender measures included GII (lagged) and female legal inequity index.

- Table 1. Growth, Inequality and Gender Inequality: System GMM Results from Baseline Model — Selected coefficients and test statistics

  - All Countries (columns (1)–(5)):
    - Initial income per capita (log): -1.420***, -1.728***, -0.998***, -1.921***, -1.209***
    - Initial level of infrastructure: 0.597***, 0.622***, 0.646***, 0.372**, 0.539***
    - Institutional quality (index): 0.046**, 0.054**, 0.022, 0.064***, 0.054**
    - Measures of Inequality:
      - Initial income inequality (t20/b40 ratio): -0.060
      - Initial income inequality (t20/b40 ratio) x LICs: -0.354***
      - Initial income inequality (net Gini): -0.025
      - Initial income inequality (net Gini) x LICs: -0.048***
      - Initial income share of middle class: 0.158***
      - Gender inequality (lagged): -0.031**
      - Gender inequality x LICs (lagged): -0.039***
      - Female legal inequity (index): -0.246
    - Constant: 6.470***, 9.310***, -1.652, 12.879***, 2.664*
    - Serial correlation (p-value): 0.103, 0.044, 0.370, 0.121, 0.907
    - Hansen test (p-value): 0.259, 0.230, 0.493, 0.943, 0.553
    - Observations: 381, 429, 269, 430, 334
    - Number of countries: 121, 117, 114, 115, 84

  - Emerging and Developing Countries (columns (6)–(10)):
    - Initial income per capita (log): -1.133***, -1.382***, -0.843***, -1.615***, -1.201***
    - Initial level of infrastructure: 0.586***, 0.618***, 0.563***, 0.365*, 0.574***
    - Institutional quality (index): 0.027, 0.035, 0.012, 0.045**, 0.045
    - Measures of Inequality:
      - Initial income inequality (t20/b40 ratio): -0.116**
      - Initial income inequality (t20/b40 ratio) x LICs: -0.260***
      - Initial income inequality (net Gini): -0.038
      - Initial income inequality (net Gini) x LICs: -0.036**
      - Initial income share of middle class: 0.154***
      - Gender inequality (lagged): -0.057***
      - Gender inequality x LICs (lagged): -0.024**
      - Female legal inequity (index): -0.279
    - Constant: 5.677***, 8.293***, -1.484, 12.667***, 2.770
    - Serial correlation (p-value): 0.246, 0.076, 0.400, 0.141, 0.922
    - Hansen test (p-value): 0.677, 0.187, 0.588, 0.904, 0.625
    - Observations: 268, 300, 234, 301, 266
    - Number of countries: 87, 84, 85, 81, 67

- Key empirical signals from the results presented:
  - Negative and sometimes statistically significant interactions of inequality measures with LICs (low-income countries) indicate larger adverse growth associations of initial inequality in LICs in several specifications (e.g., t20/b40 ratio x LICs: -0.354*** for All Countries; -0.260*** for Emerging and Developing Countries).
  - Initial income share of middle class is positively and significantly associated with growth (0.158*** for All Countries; 0.154*** for Emerging and Developing Countries).
  - Gender inequality (lagged) shows negative and statistically significant associations with growth in several specifications (e.g., -0.031** for All Countries; -0.057*** for Emerging and Developing Countries).
  - Female legal inequity (index) coefficients are negative but not statistically significant in the reported specifications (-0.246 for All Countries; -0.279 for Emerging and Developing Countries).
  - Specification tests reported (Arellano–Bond and Hansen J) indicate no serious second-order serial correlation and no over-identification problem in the reported model specifications based on the provided p-values.

*Source: Authors' calculations.*

### 1. The dependent variable is real GDP per capita growth, averaged over non-overlapping 5-year periods, for 1995-2014. Se

### _wp16111 - 1. The dependent variable is real GDP per capita growth, averaged over non-overlapping 5-year periods, for 1995-2014. Se

### Inequality and growth — core findings
- A robust negative association exists between average real GDP per capita growth and initial income distribution; this negative association holds primarily for countries at early stages of development.
- High income inequality impedes growth where credit constraints and credit market imperfections are more common.
- The negative association between countries’ average growth rates and their income distribution at the beginning of 5-year periods is robust to different measures of income inequality and to dropping developed countries from the sample.
- A clear negative relationship is present between real GDP per capita growth and gender inequality.
- The negative association between growth and inequality in low-income countries is robust to inclusion of other covariates (inflation, percent change in terms of trade, dependent population growth, schooling, etc.).
- The models capture non-linearity using interaction terms for low-income countries while estimating on the whole sample.

### Key estimated magnitudes (from extended GMM models, Table 2)
- Income inequality:
  - A one-percentage point reduction in the initial Gini coefficient in low-income countries is associated with a 0.15 percentage point cumulative increase in growth over a 5-year period.
- Gender inequality:
  - A one percentage point reduction in gender inequality in low-income countries is associated with higher cumulative growth over 5 years of 0.2 percentage points in low-income countries.
  - Female legal inequity (index) coefficients are negative and significant across models (see Table 2 model columns).
- Selected control variable coefficients (representative, exact values as reported in Table 2):
  - Initial income per capita (log): -1.234 ***, -1.347 ***, -1.081 ***, -1.746 ***, -1.148 ***, -1.608 ***
  - Fixed capital investment (% GDP): 0.134 *, 0.184 ***, -0.014, 0.093 *, 0.132 *, 0.028
  - Schooling (years): 0.119, 0.068, 0.159 *, 0.045, 0.105, 0.154 *
  - Dependent population growth (%): -0.356 **, -0.293 **, -0.539 ***, -0.224, -0.306 **, -0.286 **
  - Infrastructure index: 0.238 *, 0.194, 0.270 *, 0.294 *, 0.237, 0.334 **
  - High inflation dummy: -1.583 ***, -1.627 ***, -1.621 ***, -1.228 ***, -1.472 ***, -1.552 ***
  - Terms of trade (percent change): 0.068 **, 0.076 ***, 0.091 ***, 0.098 ***, 0.063 **, 0.094 ***
  - Institutional quality (index): 0.047 ***, 0.063 ***, 0.040 **, 0.080 ***, 0.064 ***, 0.054 ***
- Model diagnostics and sample info (as reported):
  - Number of instruments: 15, 15, 14, 15, 15, 17
  - Serial correlation (p-value): 0.071, 0.025, 0.202, 0.209, 0.154, 0.274
  - Hansen test (p-value): 0.210, 0.335, 0.319, 0.445, 0.712, 0.700
  - Country fixed effects: Yes (all models)
  - Time (period) fixed effects: Yes (all models)
  - Observations: 344, 384, 237, 419, 304, 240
  - Number of countries: 110, 106, 104, 115, 78, 78

### Robustness and channels
- The income-inequality growth link is robust to different inequality measures (Gini, top 20 to bottom 40 ratio, income share of middle class).
- Gender inequality affects growth through channels other than income inequality: adding gender inequality variables did not diminish the association between income inequality and growth.
- Possible alternative channels include gender gaps in educational attainment, rural-urban income inequality, and inequality arising from dependence on natural resource exports.

### Implications for sub-Saharan Africa (SSA)
- Growth decomposition comparing SSA with ASEAN-5 (Indonesia, Malaysia, Philippines, Thailand, Vietnam):
  - SSA’s average annual real GDP per capita growth has been about 1½ percentage points lower than ASEAN-5 over the last decade.
  - Reducing three inequality indicators to ASEAN levels could boost SSA’s annual per capita GDP growth by on average about 0.9 of a percentage point.
  - The impact from closing the infrastructure gap is of roughly the same order of magnitude (~0.9 percentage point).
  - Compared with Latin America and the Caribbean, reducing gender inequality and legal gender-related restrictions yields sizeable growth effects for SSA.
- Subgroup-specific findings within SSA (Figure 7 summary):
  - Low-income countries (excluding fragile states):
    - Catch-up effect from low initial income relative to ASEAN contributes about 2½ percentage points of real GDP per capita growth.
    - This catch-up is more than undone by weak infrastructure and lower human capital accumulation.
    - Gender inequality accounts for ¾ percentage points of the growth differential.
  - Fragile states:
    - Lower quality of infrastructure and political institutions explains the largest fraction of the growth differential.
    - Reducing gender inequality could boost annual per capita GDP growth by 2/3 percentage points.
  - Middle-income and oil-exporting countries:
    - Reducing income inequality to ASEAN levels is an important factor to raise growth.
    - Removing legal gender-related restrictions shows particularly strong growth payoffs for oil-exporting SSA countries (caveat noted on potential correlation rather than causation).

### What drives income inequality? (empirical strategy and main correlates)
- Empirical approach:
  - Quantile regressions on a sample of 135 countries between 1991 and 2010; 469 observations of non-overlapping 5-yearly changes in inequality.
  - Interaction terms used for low-income countries (LIC) and sub-Saharan Africa (SSA).
  - Regressions relate five-year changes in inequality to beginning-of-period values of candidate drivers; GDP per capita growth and share of exports in natural resources cover the contemporaneous five-year period.
- Factors associated with reducing net income inequality (Table 3 results):
  - GDP per capita growth (negative association with change in net Gini in overall sample; coefficients reported include -0.1010*** and -0.0991*** in different specifications).
  - Private capital stock (negative, e.g., -0.0056***).
  - Education spending (negative in several specifications, e.g., -0.1830*).
  - Share of working-age population (negative, e.g., -0.0585*).
  - Fiscal redistribution (market minus net Gini) (negative, e.g., -0.0842***).
- Factors associated with increasing net income inequality:
  - Higher beginning-of-period gender and education inequality (positive coefficients reported).
  - Increased share of natural resource exports is associated with increases in inequality (results driven mainly by oil exports).
- Financial deepening:
  - No clear effect in overall sample quantile regressions (M2/GDP insignificant in overall sample).
  - For the 10th quantile (large reductions in inequality), enhancing women’s access to financial services (women’s right to open bank account dummy) played a role in reducing inequality mainly for lower-income countries (large positive coefficient in some specifications for the overall sample and strong negative effects in SSA 10th quantile regressions).
- Other structural/macroeconomic correlates (selected exact estimates from Table 3):
  - Share of agriculture: 0.0515*** (overall sample), with interaction terms negative for LIC and SSA in some specifications.
  - Education inequality: coefficients reported include 0.0168 and 0.0240** depending on specification.
  - Gender inequality index: 0.0301** and 0.0308*** in different specifications.
  - Fiscal redistribution: -0.0842*** (overall), with stronger negative interactions for SSA in some specifications (-0.7370*** in one column).
  - Public capital stock: small negative coefficients with significant negative interaction for SSA (e.g., -0.0208***).
  - Trade openness: mixed; negative and significant in some SSA specifications (e.g., -0.0356***).

### Policy-relevant implications and conclusions
- Income and gender inequality, including legal gender-related restrictions, impede growth mainly in countries at earlier stages of development, with large growth losses in sub-Saharan Africa.
- Further progress in reducing income and gender inequality could deliver significant sustained growth dividends, particularly for low-income countries.
- Gender inequality affects growth via channels distinct from those of income inequality; policies addressing both are complementary.
- For SSA, policy priorities suggested by the empirical findings include:
  - Reducing income inequality (through redistribution and targeted policies) and gender inequality (including removal of legal barriers).
  - Closing infrastructure and human capital gaps.
  - Accelerating the demographic transition by raising investment in human capital.
  - Improving the progressivity of tax systems and expenditures and moving away from blunt instruments like universal subsidy programs.
- The analysis underscores that addressing inequality can be as important as addressing traditional growth gaps (infrastructure, investment, institutions) for raising per capita growth in SSA.

*Source: Authors' calculations and regression results as reported in the supplied content.*

### 0.9 percentage points if inequality of gender and income was reduced to the levels observed in the

### _wp16111 - 0.9 percentage points if inequality of gender and income was reduced to the levels observed in the

### Key findings on growth and inequality
- Reduction of inequality of gender and income to the levels observed in the fast-growing ASEAN region would raise growth by 0.9 percentage points, with variations across country groups.
- In Latin America and the Caribbean, gender inequality does not appear to be a main contributor to the region’s growth differentials with ASEAN.
- High levels of income inequality in sub-Saharan Africa appear partly driven by structural features of its countries; fiscal and financial sector policies that influence access of low-income households and women to opportunities also matter.
- GDP per capita growth, the private capital stock, education spending, gender inequality, the share of the working age population, and fiscal redistribution are generally related to changes in income inequality in the global sample.

### Drivers of changes in inequality (regional and cross-country)
- Structural factors and policies have been driving changes in inequality over time, but relevant factors differ by region and level of development.
- The evidence indicates that:
  - Structural country features are important drivers of high income inequality in sub-Saharan Africa.
  - Fiscal policy and financial sector policies that affect access for low-income households and women can influence inequality.
  - Country-specific variations imply that multiple, interacting factors drive inequality.

### Policy implications and recommendations
- Policies, if carefully designed, could help reduce inequality; however:
  - Policies must be tailored to country-specific situations and take into account administrative capacity and potential tradeoffs.
  - There is a need to ensure that policies aimed at reducing income inequality are carefully designed so that they themselves do not adversely affect economic growth.
- Specific channels for policy action highlighted:
  - Removal of legal restrictions on women and increasing accessibility of education and healthcare may provide an impetus for social norms to change.
  - Fiscal redistribution and measures expanding access to finance and opportunities for women and low-income households are relevant policy levers.

### Development goals and historical context
- The paper emphasizes importance for countries, particularly in sub-Saharan Africa, to reduce income and gender inequalities to achieve sustained growth.
- The recently launched SDGs explicitly recognize income and gender inequality as separate goals, which is encouraging and indicates global recognition of these priorities.
- Historical precedent: Once the Millennium Development Goals were announced, countries started efforts that led to considerable progress in improving primary education attainment; the paper urges building on that impetus to reduce income and gender inequalities.

### Country coverage (appendix summary)
- High income countries (World Bank definition) listed include: United States, United Kingdom, Austria, Belgium, Denmark, France, Germany, Italy, Luxembourg, Netherlands, Norway, Sweden, Switzerland, Canada, Japan, Finland, Greece, Iceland, Ireland, Malta, Portugal, Spain, Australia, New Zealand, Chile, Uruguay, Barbados, Trinidad and Tobago, Cyprus, Israel, Hong Kong SAR, China, Korea Rep., Singapore, Russian Federation, Czech Republic, Slovak Republic, Estonia, Latvia, Lithuania, Croatia, Slovenia, and Poland.
- Middle income (Upper Middle Income) countries listed include: Turkey, South Africa, Argentina, Brazil, Colombia, Costa Rica, Dominican Republic, Ecuador, Mexico, Panama, Peru, Venezuela RB, Belize, Jamaica, St. Lucia, Suriname, Iran Islamic Rep., Iraq, Jordan, Malaysia, Maldives, Thailand, Algeria, Angola, Botswana, Gabon, Mauritius, Seychelles, Namibia, Tunisia, Fiji, Marshall Islands, Azerbaijan, Belarus, Albania, Kazakhstan, Bulgaria, China, Turkmenistan, Serbia, Montenegro, Hungary, Macedonia FYR, Bosnia and Herzegovina, and Romania.
- Low income (Lower Middle Income or Low Income) countries listed include: Bolivia, El Salvador, Guatemala, Haiti, Honduras, Nicaragua, Paraguay, Guyana, Syrian Arab Republic, Egypt Arab Rep., Yemen Rep., Afghanistan, Bangladesh, Bhutan, Cambodia, Sri Lanka, India, Indonesia, Lao PDR, Nepal, Pakistan, Philippines, Vietnam, Djibouti, Burundi, Cameroon, Cabo Verde, Central African Republic, Chad, Comoros, Congo Rep., Congo Dem. Rep., Benin, Ethiopia, The Gambia, Ghana, Guinea-Bissau, Guinea, Côte d'Ivoire, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Morocco, Mozambique, Niger, Nigeria, Zimbabwe, Rwanda, São Tomé and Príncipe, Senegal, Sierra Leone, Sudan, Swaziland, Tanzania, Togo, Uganda, Burkina Faso, Zambia, Papua New Guinea, Micronesia Fed. Sts., Armenia, Georgia, Kyrgyz Republic, Moldova, Tajikistan, Ukraine, Uzbekistan, and Mongolia.

### Variables used in empirical analyses (selected descriptions)
- Initial income inequality (t20/b40): Ratio of income distribution at the top 20 relative to that of the bottom 40 percent of population. Source: Primarily World Development Indicators (WDI), augmented by UNU-WIDER.
- Initial income inequality (Gini): Traditional Gini measure (net Gini used in paper). Source: Standardized World Income Inequality Database (SWIID v.5.0).
- Initial income share of middle class: Sum of income shares of the third and fourth quintiles. Source: Primarily WDI, augmented by UNU-WIDER.
- Lagged gender inequality: Index calculated using the UN methodology covering 1990–2010 periods. Source: IMF staff estimates (details reported in Gonzales et al., 2015b).
- Female legal inequity: Sum of six binary indicators representing restrictions on selected women’s legal rights (range 0–6). Source: World Bank Women, Business and the Law (WBL) database.
- Initial income per capita: Logged real GDP per capita in the first year in each five-year period. Source: Penn World Tables (PWT v.8.0).
- Fixed capital investment: Gross fixed capital formation in percent of GDP, averaged over five-year periods. Source: Primarily PWT, some augmentation from WDI.
- Schooling: Average years of schooling (population aged 15+). Source: Barro-Lee database.
- Share of working-age population: Percentage of population aged 15–64 (value taken from initial year of each period). Source: UN Population database.
- Education inequality: Education Gini from Castelló-Climent and Doménech (2014), constructed from Barro and Lee database.
- Fiscal redistribution: Difference between market Gini and net Gini (value taken from initial year). Source: SWIID v.5.0.
- Education Spending: Percentage of public spending allocated to education (value from initial year). Source: World Bank WDI database.
- Financial depth: Ratio of broad money (M2) to GDP (value from initial year). Source: World Bank WDI database.
- Women’s financial inclusion: Dummy variable equal to 1 if women have the ability to open bank accounts. Source: World Bank WBL database.
- Public capital stock: Estimate of general government capital stock constructed from government investment flows (in 2005 dollar prices). Source: IMF Investment and Capital Stock database.
- Private capital stock: Estimate constructed from private investment flows (in 2005 dollar prices) (value from initial year). Source: IMF World Economic Outlook database.
- Trade openness: Sum of total exports and imports as a percentage of GDP. Source: IMF World Economic Outlook database.
- Change in net Gini: Difference of net Gini in levels over each 5-year interval. Source: SWIID v.5.0.
- Dependent population growth: Average annual percentage change in the non-working age population (under 15 or above 64). Source: UN Population database.
- High inflation: Dummy variable equal to 1 if average annual inflation in consumer prices over a five-year period is more than 15 percent. Source: IMF World Economic Outlook database.
- Infrastructure index: Composite index based on electricity consumption, access to water, and access to any type of phones. Electricity consumption from IEA; access to water and telephones from WDI.

*Source: _wp16111 - 0.9 percentage points if inequality of gender and income was reduced to the levels observed in the (IMF working paper content provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16111.pdf_