## _wp1621 — Comparison of the HDI, GDI, and GII

## Source details

**Canonical URL:** [_wp1621 — Comparison of the HDI, GDI, and GII](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1621.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1621.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1621.pdf.json)

---

### Major findings and trends
- General results
  - Trends in individual indicators point toward improvement in education, health, economic opportunity, and political empowerment, but progress is uneven across regions and income groups.
  - South Asia, the Middle East and Central Asia, and sub-Saharan Africa lag other regions despite significant improvement.
  - Low-income developing countries (LIDCs, group consists of 60 countries) tend to lag; gaps between advanced and emerging countries are surprisingly small on some indicators.
- Education
  - Gross secondary enrollment (female to male ratio), 1980-2014:
    - All regions saw an increase in the female to male ratio over this period.
    - Americas and Caribbean ratio is above one for most of the period.
    - Europe is close to one; Asia and Pacific approaches one by the sample end and exceeds Europe’s ratio at sample end.
    - LIDCs: 57 percent have gross secondary enrollment data from 1980-2014; LIDCs continue to lag but made substantial progress narrowing the gap.
  - Robustness: gross secondary enrollment correlates about 0.9 with net enrollment and completion in the sample.
- Health
  - Life expectancy at birth (female to male ratio), 1980-2014:
    - All regions show a ratio above one; the female advantage is decreasing in advanced countries.
    - Number of countries with female to male life expectancy ratio below 1.05 increased, with most of the increase in Africa.
  - Child mortality under age 5 (female to male ratio), 1990-2014:
    - All regions show a ratio below 1 (biological regularity favoring girls).
    - India is the only country in the data set where female child mortality exceeds male; China is an outlier with high relative mortality of girls.
  - Maternal mortality ratio (per 100,000 live births), 1980-2014:
    - All regions achieved declines; sub-Saharan Africa has the highest rate but declined significantly, almost halving.
- Economic opportunity
  - Labor force participation rate (ages 15-64, female to male ratio), 1990-2014:
    - Movement toward greater parity worldwide, with a notable gap in Middle East and Central Asia.
    - Asia and Pacific exhibited a declining participation ratio driven by declining female participation in India and, to a lesser extent, China.
  - Labor force participation rates (percent), 1990-2014:
    - Female labor supply has been rising except in Asia and Pacific; male participation relatively flat or declining in some regions.
  - Mean monthly earnings (female to male ratio), 1995-2011:
    - Using 38 countries with data, wage gaps diminished slowly in advanced and emerging markets.
- Political opportunity
  - Seats held by women in national parliaments (percent), 1990-2014:
    - All regions made substantial progress; women’s representation remains well below 50 percent in every region.
    - Two countries, Rwanda and Bolivia, achieved over half of the legislative body being women in 2015.
    - Rapid improvement in sub-Saharan African representation; by sample end sub-Saharan Africa had higher overall representation than Asia and the Pacific.
- Indices vs indicators
  - Individual indicators are informative for targeted policy; composite indices provide broader perspective but require transparency in choices of indicators, weighting, and aggregation.
  - Indices are highly correlated with income; care is needed to avoid conflating gender equality with standard of living.

### Indices — methodology, time-consistent (TC) extensions, and empirical relationships
- Reconstruction and TC series
  - Authors construct time-consistent (TC) versions of UNDP’s GDI and GII to permit backward extension to 1990 and better trend analysis.
  - Coverage after wage interpolation for the GDI, TC: 146 countries from 1990-2013.
  - GII, TC coverage: 141 countries from 1990-2013.
- GDI (revised UNDP 2014) components and methodological notes
  - Components: life expectancy (with adjustment for average female biological advantage of five years), mean years of schooling, expected years of schooling, and GNI per capita combined with female-to-male wage ratio where available.
  - UNDP substitutes a global average female-to-male wage ratio of 0.8 where wage data are missing; authors find this global average is likely high and instead impute wages by interpolation (for countries with at least 5 years of data) or use regional averages.
  - Alternative TC GDI specification: replace wages with labor force participation rate (LFPR) in standard-of-living sub-index as a robustness check.
  - Key TC result: TC version reduces index values for all countries; in the TC version all countries have index value below one.
- GII construction (summary)
  - Components: reproductive health (MMR, AFR), empowerment (population with at least secondary education, share of parliamentary seats), labor market (LFPR).
  - GII interpretation: lower values when women and men are more equal; higher when more unequal.
  - TC GII very similar to UNDP GII (Spearman’s rho 0.99 for the most recent year), with slight variations due to updated indicators.
- Correlations among indices (Pearson, latest data)
  - GDI — GEI: 0.90***
  - GDI — GGGI: 0.79***
  - GDI — GII: -0.69***
  - GDI — SIGI: -0.75***
  - GDI — WEOI: 0.71***
  - Additional pairwise correlations reported in Table 2 (significance: *** 1% level).
- Correlations among individual indicators (selected, Table 3)
  - Gross secondary enrollment vs Maternal mortality ratio: -0.58*** / -0.55*** (All / LIDCs)
  - Maternal mortality ratio vs Child mortality under-5: 0.45*** / 0.63*** (All / LIDCs)
  - Labor force participation vs Share of female parliamentarians: 0.37*** / 0.31* (All / LIDCs)
  - Gross secondary enrollment vs Life expectancy (at birth): 0.19** / -0.01 (All / LIDCs)
  - Interpretation: correlations between individual indicators are significantly lower than between indices, supporting indices as useful summaries while underscoring indicator-specific analysis for policy.
- Relationship of indices to income (cross-section OLS, Table 4 — coefficients on ln GDP per capita, latest year)
  - GDI: 0.0368*** (0.00403), Constant 0.593*** (0.0396), Observations 146, R-squared 0.391
  - GEI: 0.0599*** (0.00824), Constant 0.194*** (0.0730), Observations 177, R-squared 0.266
  - GGGI: 0.0139*** (0.00519), Constant 0.564*** (0.0484), Observations 127, R-squared 0.077
  - GII: -0.126*** (0.00898), Constant 1.535*** (0.0805), Observations 149, R-squared 0.637
  - SIGI: -0.0729*** (0.0102), Constant 0.814*** (0.0913), Observations 99, R-squared 0.318
  - WEOI: 12.23*** (0.940), Constant -58.44*** (8.451), Observations 126, R-squared 0.643
  - Interpretation: higher income strongly correlated with improved gender-equality measures (sign reversed for indices that increase in inequality such as GII and SIGI).
- Panel results (TC indices, Table 5)
  - Fixed-effects OLS:
    - GDI, TC: ln GDP pc = 0.0676*** (0.00204); Constant = 0.296*** (0.0185); Observations = 2,776; R-squared = 0.295; Number of countries = 151
    - GII, TC: ln GDP pc = -0.157*** (0.00415); Constant = 1.847*** (0.0375); Observations = 3,000; R-squared = 0.333; Number of countries = 138
  - First-differences:
    - ∆ GDI, TC: ∆ ln GDP pc = 0.00831** (0.00420); Constant = 0.00207*** (0.000134); Observations = 2,591; R-squared = 0.009
    - ∆ GII, TC: ∆ ln GDP pc = -0.00501 (0.00611); Constant = -0.00655*** (0.000389); Observations = 2,858; R-squared = 0.000
  - Interpretation: fixed-effects magnify income coefficients; first-difference results show weaker relationships, particularly for GII.

### Data availability, quality issues, and measurement caveats
- Coverage and availability rules
  - Coverage sample: 188 countries grouped by income; a country is “available” for an indicator if it has at least 50 percent of the data for the years 1980-2014.
- Key coverage statistics (preserve exact figures)
  - Economic opportunity:
    - Of 21 identified economic opportunity indicators, only labor force participation rate and youth unemployment have significant global coverage.
    - Over 93 percent and 90 percent of countries have data on labor force participation rate and youth unemployment rate, respectively.
    - Of the 60 LIDCs, 58 countries have data for labor force participation rate and 56 countries have data for youth unemployment rate.
  - Education:
    - Gross secondary enrollment coverage: 130 countries (70 percent of total).
    - Of the LIDCs, 57 percent have gross secondary enrollment data from 1980-2014.
  - Health:
    - Crude birth rate, adolescent fertility rate, life expectancy, and under-5 mortality rate are available in 97 percent of countries.
    - HIV prevalence is available for 56 out of 60 LIDCs.
    - Maternal mortality data available only every five years.
  - Political opportunity:
    - Share of female seats in parliament: 127 out of 188 countries have reported data.
- Appendix D wage-data specifics and critique
  - Ratio of female to male wage data: poorest coverage; maximum-country-year coverage available for only 68 countries in the year with highest coverage.
  - Regional bias in wage observations: 44 percent of observations cover Europe and Central Asia; 0.6 percent for sub-Saharan Africa.
  - Authors’ wage statistics (Table D1):
    - Ratio of Female to Male Wage: Range 0.41-1.5; Mean (s.d.) 0.77 (.12); Countries 27; Years 1995-2011; Source: ILO
  - Authors’ imputation approach: interpolate countries with at least 5 years of wage data; use regional averages for countries without sufficient data (instead of UNDP’s global average 0.8).
  - Wage coverage by region (Table D2 — Percent of Observations / Share of Population):
    - East Asia and Pacific: 15.9 / 33.7
    - Europe and Central Asia: 43.9 / 14.1
    - Latin America and Caribbean: 24.0 / 8.5
    - Middle East and North Africa: 10.4 / 5.3
    - North America: 1.8 / 5.4
    - South Asia: 3.2 / 22.6
    - Sub-Saharan Africa: 0.6 / 10.1
- Indicator-specific cautions
  - Life expectancy at birth can mask selective reduction of females via sex-selective abortion and differential child treatment; female natural advantage about four to five years should inform equality targets.
  - Maternal mortality ratio is regression-model estimated and can be measured with error; civil registration and vital statistics need improvement.
  - Mean nominal monthly earnings mix hours and pay-per-hour effects and have limited coverage; authors note the UNDP assumed global wage ratio 0.8 may be high.
  - Political representation measured by parliamentary seats does not capture ministerial roles or subnational representation.

### Policy relevance and recommendations
- Data policy and measurement
  - Improve civil registration and vital statistics to enhance maternal mortality measurement.
  - Expand sex-disaggregated reporting across household surveys and administrative sources.
  - Prioritize time-use data and intra-household allocation data to capture unpaid work burdens and resource distribution.
  - Support international initiatives listed in Appendix A and Table A2 (EDGE, Data2X, Gender Data Portal, Women, Business and the Law, etc.) to improve coverage and comparability.
- Use of indicators and indices for policy
  - Individual indicators: most useful for targeted policy objectives and monitoring specific outcomes (education, health, labor participation).
  - Indices: useful as summary statistics but require transparency about indicator choice, weighting, aggregation, and consistent time-series construction.
  - A good index should be simple, transparent, avoid mixing well-being with empowerment, and be robust to minor changes in included variables and weights.
  - With gender equality embedded in the SDGs, governments should use indicators and indices to guide budget and policy decisions to close gender gaps and advance women’s well being.

### Key numeric benchmarks and empirical statistics (preserved exactly)
- LIDC grouping: 60 countries.
- GDI world average (UNDP, 2013 data, 148 countries): 0.920 (indicating a gap of 8 percent in the HDI for females from that for males).
- Regional GDI averages (UNDP 2013 data):
  - OECD countries: 0.964
  - Latin American and the Caribbean: 0.963
  - South Asia: 0.860
  - Arab countries: 0.866
  - Sub-Saharan African countries: 0.867
- GDI original vs TC Spearman correlations (Table D3):
  - GDI Original vs TC version: 0.96***
  - GDI Original vs TC version with LFPR: 0.75***
  - TC version vs TC version with LFPR: 0.81***
- GDI, TC coverage after wage interpolation: 146 countries from 1990-2013.
- GII, TC coverage: 141 countries from 1990-2013.
- Expected years of schooling growth rates (authors’ findings):
  - Average yearly growth rate: females 2.3 percent; males 1.7 percent.
  - Average ten year growth rate: females 16.8 percent; males 11.4 percent.
- GDI (TC) panel fixed-effects coefficient on ln GDP pc: 0.0676*** (0.00204); Observations = 2,776; Number of countries = 151.
- GII (TC) panel fixed-effects coefficient on ln GDP pc: -0.157*** (0.00415); Observations = 3,000; Number of countries = 138.

*Source: IMF working paper section "1. Comparison of the HDI, GDI, and GII" (content unit provided)._

### References .............................................................................................................

### References

### Tables
- 1.      Definitions of Selected Gender-Related Indicators ............................................................. 8
- 2.      Correlations Between the Gender Equality Indices ............................................................34
- 3.      Correlations Between the Indicators ...................................................................................35
- 4.      Regression of Gender Equality Indices on Per Capita Income ...........................................40
- 5.      Regression of Time Consistent Gender Equality Indices on Per Capita Income ...............41

### Figures
- 1.      Gross Secondary Enrollment  .............................................................................................12
- 2.      Life Expectancy at Birth (Female to male ratio) ................................................................14
- 3.      Life Expectancy at Birth (Years)  .......................................................................................15
- 4.      Number of Countries with Female to Male Life Expectancy Ratio Below 1.05 ................16
- 5.      Child Mortality, Under the Age of 5 (Female to male ratio)  .............................................17
- 6.      Maternal Mortality Ratio (Modeled estimate, per 100,000 live births) ..............................18
- 7.      Labor Force Participation Rate, Ages 15-64 (Female to male ratio) ..................................20
- 8.      Labor Force Participation Rate, Ages 15-64 (Percent) .......................................................21
- 9.      Mean Monthly Earnings of Employees (Female to male ratio) .........................................22
- 10.      Seats Held by Women in National Parliaments (Percent of total) ......................................23
- 11.      GDI, Time Consistent Version ...........................................................................................37
- 12.      GDI, Time Consistent Version, with Labor Force Participation Rate ................................38
- 13.      GII, Time Consistent Version .............................................................................................39

### Box
- Box

*Source: _wp1621 - References .............................................................................................................*

### 1. Comparison of the HDI, GDI, and GII ...............................................................................25

### 1. Comparison of the HDI, GDI, and GII

### Introduction
- The paper examines trends in indicators of gender equality and advancement of women using individual indicators and gender equality indices.
- Introduces reconstructed, backward-consistent versions of the UNDP’s Gender Development Index (GDI) and Gender Inequality Index (GII).
- Time horizon and scope:
  - Trends analyzed over several decades, with many indicators presented for 1980-2014.
  - The Sustainable Development Goals adopted in 2015 set targets over the next 15 years; Goal 5 explicitly calls for gender equality and the empowerment of women and girls.
- Key high-level findings:
  - Trends in individual indicators point toward improvement in education, health, economic opportunity, and political empowerment, but progress is uneven across regions and income groups.
  - South Asia, the Middle East and Central Asia, and sub-Saharan Africa lag other regions despite significant improvement.
  - Low-income developing countries (LIDCs) tend to lag; gaps between advanced and emerging countries are surprisingly small on some indicators.

### Indicators versus Indices and Data Issues
- Two approaches:
  - Sex-disaggregated individual indicators: informative on specific aspects, useful for policy targeting, but may not move together and can miss broader relationships.
  - Composite indices (e.g., HDI-derived GDI, UNDP’s revised stand-alone GDI and the GII): provide broader perspective but involve choice of indicators, weighting, and aggregation that introduce arbitrariness.
- Methodological choices in this paper:
  - Present trends (not levels) to avoid prescriptive judgments about “correct” gender gaps.
  - Use relative measures, mainly female-to-male ratios.
- Data availability and quality issues:
  - Sex-disaggregated data suffer from spotty coverage across time and countries.
  - Good indicators should have broad country coverage and long time-series availability; many do not.
  - Most microeconomic data come from household surveys; many key variables available only at household level (e.g., consumption, assets).
  - Lack of data on intra-household resource allocation and time-use limits analysis of unpaid work burdens.
- International efforts to improve data:
  - Evidence and Data for Gender Equality (EDGE) project: three-year effort to supply data for the UN Statistical Commission’s Minimum Set of Gender Indicators (covers economic and political participation, access to resources, education, health, human rights).
  - Data2X as a supporting international gender data initiative.

### Trends in Selected Indicators (definitions and measurement choices)
- Selected indicator areas: education, health, economic opportunity, political opportunity.
- Education:
  - Indicator used: gross secondary enrollment rate (chosen for availability and high correlation with net enrollment and completion).
  - Definition preserved: “Gross enrollment rate is the ratio of total enrollment, regardless of age, to the population of the age group that officially corresponds to the level of education shown.”
- Health:
  - Primary indicator: life expectancy at birth (wide coverage, annual frequency).
  - Supplementary indicators for developing-country focus: child mortality (under-5) and maternal mortality ratio.
  - Notes on interpretation:
    - Life expectancy at birth can mask selective reduction of female population via sex-selective abortion and differential child treatment.
    - Female natural life expectancy is about four to five years longer than men’s; equality targets in health might aim at this biological norm rather than exact parity.
    - Maternal mortality ratio: number of women who die from pregnancy-related causes per 100,000 live births; data are estimated with regression models and can be measured with error.
- Economic opportunity:
  - Primary indicator: labor force participation rate, ages 15-64 (proportion of population aged 15-64 that is economically active).
  - Wages: mean nominal monthly earnings of employees by sex are presented but suffer from limited coverage and mix hours and pay-per-hour effects.
- Political opportunity:
  - Indicator used: seats held by women in national parliaments (percentage of parliamentary seats in a single or lower chamber held by women).
  - Caveats: parliamentary share alone does not fully capture political empowerment; does not account for ministerial or other government levels and parliaments’ differing roles across countries.

### Key empirical and methodological points
- Comparison strategy:
  - Constructed time series for UNDP’s GDI and GII to examine trends consistent backward in time; choice motivated by prominence of UNDP indices and similarity of individual component variables.
  - Parallel presentation of individual indicators and indices to attribute index trends to underlying indicators.
- Indicator properties and limitations emphasized:
  - For education, gross secondary enrollment rate chosen because of data availability and correlation (about 0.9) with net enrollment and completion rates in the sample.
  - For life expectancy, life expectancy at age 60 is an alternative but has less frequent availability, so life expectancy at birth was preferred.
  - For labor force participation, alternative measures that subdivide the variable (e.g., informal vs formal) are not widely available for the cross-country sample.
  - Maternal mortality: data quality issues due to scarce country-level collection; improvement in civil registration and vital statistics is an important goal.

### World Trends by Region and Income Group
- Geographic grouping: IMF classification used — sub-Saharan Africa, the Americas and Caribbean, Asia and the Pacific, Europe, and the Middle East and Central Asia.
- Treatment of newly-formed countries: enter the sample in the year they came into being (examples in the sample include Timor-Leste and South Sudan).
- LIDC grouping:
  - LIDCs group consists of 60 countries.
  - LIDC eligibility: designated as eligible for IMF-supported Poverty Reduction and Growth Trust (PRGT) and had per capita gross national income less than an unadjusted US$2,390 in 2013 (IMF, 2014).
  - India, Pakistan, and the Philippines are excluded from the LIDC grouping despite per capita income below US$2,390 because they are not PRGT-eligible and are viewed as resembling emerging markets.
  - Europe has only one LIDC (Moldova); Europe is excluded from LIDC analyses because of this.
- Empirical patterns summarized:
  - When grouped by region, South Asia, the Middle East and Central Asia, and sub-Saharan Africa lag other regions though all have recorded significant improvement.
  - When grouped by economic development level, countries at all levels generally made progress; LIDCs tend to lag; some indicators show small gaps between advanced and emerging countries.

### Use of Indicators and Indices for Policy Analysis
- Indicators:
  - Useful for setting targeted policy objectives and monitoring specific outcomes (education, health, labor participation).
  - Recommendation implicit: improve sex-disaggregation and frequency of key indicators to better guide policy.
- Indices:
  - Provide a broader picture but require transparency about indicator choice, weighting, and aggregation; different formulations explain differing country rankings.
- Data policy relevance:
  - Improve civil registration and vital statistics to enhance maternal mortality measurement.
  - Expand sex-disaggregated reporting across household surveys and administrative sources.
  - Prioritize time-use data and intra-household allocation data to capture unpaid work burdens and resource distribution.

*Source: IMF working paper section "1. Comparison of the HDI, GDI, and GII" (content unit provided).*

### 2014. Some countries may have switched income classification over time but were held in

### _wp1621 - 2014. Some countries may have switched income classification over time but were held in

### Education
- Data construction:
  - Non-overlapping five-year averages used to smooth trends and account for data gaps.
  - Each averaged country observation weighted by the five-year average population share in the region.
  - Female to male ratios examined for all indicators except maternal mortality and proportion of seats held by women in national parliaments.
- Gross secondary enrollment (female to male ratio), 1980-2014:
  - All regions saw an increase in the female to male ratio over this period.
  - Persistent gaps remain, particularly in sub-Saharan Africa and the Middle East and Central Asia.
  - The Americas and Caribbean ratio is above one for most of the period.
  - Europe is close to one, though dropping a bit toward the end of the sample.
  - Asia and Pacific ratio approaches one by the end of the sample and exceeds Europe’s ratio at sample end.
- Trends by income level:
  - All groups saw convergence toward one.
  - LIDCs continue to lag but made substantial progress narrowing the gap.
  - Regional trends among LIDCs show upward movement in sub-Saharan Africa, Asia and the Pacific, and the Middle East and Central Asia.
  - LIDCs in sub-Saharan Africa and the Middle East and Central Asia still lag LIDCs elsewhere.
- Notes and robustness:
  - Few countries change broad income classification over time; from sample beginning to end only 10 countries move from developing and emerging to advanced (all but Korea are small countries).
  - If data available for only portion of five years, average generated from available years; robustness check using countries with at least half the annual observations found no significant difference.

### Health
- Life expectancy at birth (female to male ratio), 1980-2014:
  - All regions show a ratio above one (female life expectancy exceeds male).
  - The female to male ratio is markedly decreasing in advanced countries.
  - Americas and Caribbean (dominated by developing countries) show same declining ratio.
  - Remaining regions and developing countries generally show flat or rising ratio in recent years (after earlier marked declines in Africa and the Middle East).
  - Asia and Pacific shows a consistent increase in the gender ratio over time.
- Life expectancy (years), 1980-2014:
  - Gender gaps narrowed because male life expectancy rose faster than female life expectancy, not because female life expectancy worsened.
  - In sub-Saharan Africa, female life expectancy was basically flat over a substantial period, only rising toward the end, while male life expectancy gradually rose and almost closed the gap.
  - The flat female life expectancy pattern in sub-Saharan Africa may reflect disproportionate impact of HIV/AIDS on African women.
- Number of countries with female to male life expectancy ratio below 1.05, 1980-2014:
  - Increasing number of countries saw their ratio fall below 1.05, with most of the increase in Africa.
- Child mortality under age 5 (female to male ratio), 1990-2014:
  - All regions show a ratio below 1, reflecting biological regularity in favor of girls.
  - India is the only country in the data set where child mortality of females exceeds that of males.
  - China is an outlier with high relative mortality of girls.
  - All regions show decline in female and male child mortality over time; changing gender gap reflects differences in rate of decline.
  - By region, Africa and the Middle East and Central Asia have higher relative female mortality than other regions except Asia (ratio raised by China and India).
- Maternal mortality ratio (modeled estimate, per 100,000 live births), 1980-2014:
  - All regions achieved declines in maternal mortality.
  - Sub-Saharan Africa has the highest rate but declined significantly, almost halving.
  - Asia and the Pacific and the Middle East and Central Asia achieved significant declines.
  - Decline in the Americas and Caribbean was gradual; ratio remains slightly higher than Europe.
  - Emerging markets and LIDCs achieved large reductions, with progress spread across LIDCs.

### Economic opportunity
- Labor force participation rate, ages 15-64 (female to male ratio), 1990-2014:
  - General movement toward greater gender parity in labor force participation worldwide.
  - Notable gap between the Middle East and Central Asian region and other regions.
  - Asia and the Pacific exhibited a declining participation ratio in recent years, driven by declining female participation in India and, to a lesser extent, China.
  - Excluding India, participation ratio is much higher in the Asia and Pacific region; excluding China and India, participation ratio has risen slightly over the sample period.
- Trends by income level:
  - Advanced economies and LIDCs moved significantly toward parity in the past two decades.
  - Striking increase in female participation in the Americas and Caribbean.
  - Emerging and developing economies have lagged, reflecting the weight of China and India.
- LIDCs by region:
  - Middle East and Central Asia lags; Asia and the Pacific trend is relatively flat.
  - Sub-Saharan Africa has the highest ratio, even while women remain concentrated in informal sector and subsistence agriculture.
- Labor force participation rates (percent), 1990-2014:
  - Except in the Asia and Pacific region, female labor supply has been rising; male participation has been relatively flat and is even declining in some regions.
- Mean monthly earnings of employees (female to male ratio), 1995-2011:
  - Using 38 countries with relevant data, wage gaps between women and men diminished slowly in this period in advanced and emerging markets.

### Political opportunity
- Seats held by women in national parliaments (percent of total), 1990-2014:
  - All regions made substantial progress in improving women’s legislative representation.
  - Women’s representation remains well below 50 percent in every region.
  - Two countries, Rwanda and Bolivia, achieved over half of the legislative body being women in 2015.
  - Middle East and Central Asia made gains commensurate with other regions in increasing political representation despite relative lack of economic progress for women.
  - Rapid improvement in sub-Saharan African women’s political representation; by sample end sub-Saharan Africa had a higher overall representation of women than Asia and the Pacific.
- Trends by income level:
  - Countries at all income levels made progress in increasing women’s political representation, with LIDCs making more rapid progress.
- LIDCs by region and notable country examples:
  - In sub-Saharan African LIDCs, Rwanda and Senegal show significant increases in female parliamentarians in the late 2000s (may reflect quotas).
  - In the Asia and Pacific region, Nepal shows a sizeable increase.
  - In the Middle East and Central Asia, all LIDCs except Yemen contribute to regional improvement.
  - In the Americas and Caribbean, all LIDCs except Haiti contribute to the increase.

### Overview and Comparison of Gender Equality Indices
- Indices examined:
  - UNDP’s GDI and GII.
  - Global Gender Gap Index (GGGI) of the World Economic Forum (WEF).
  - Social Institutions and Gender Index (SIGI) of the OECD.
  - Women’s Economic Opportunity Index (WEOI) of the Economist Intelligence Unit (EIU).
  - Gender Equity Index (GEI).
  - Relative Status of Women (RSW) index.
  - Gender Gap Measure (GGM) index.
- Gender Development Index (GDI), old and revised:
  - Old GDI introduced in 1995 alongside the Gender Empowerment Measure (GEM).
  - Old GDI criticized for being a measure of how the HDI score was lowered by gender inequalities rather than a measure of gender equality.
  - UNDP suspended the old GDI in 2010 and introduced a revised GDI in 2014.
  - Aside from a change in computation, UNDP did not change the GDI components: education, health, and income variables of the HDI.

*Sources: World Bank, World Development Indicators database; ILO, ILOSTAT; Mithra and Farid (2013); Liu, Arai, Kanda, Lee, Glasser, and Tamashiro (2012); Thorslund, Wastesson, Agahi, Lagergren, and Parker (2013); World Bank (2011); ILO (2013); Das, Jain-Chandra, Kochhar, and Kumar (2015); Klasen and Pieters (2015); UN Women (2015); UNDP (2014a, 2014b); IMF staff estimates.*

### Box 1 lays out the methodology of the calculation of the old and revised GDIs and

### Box 1 — Methodology and Evolution of the HDI, GDI, and GII

### Methodology overview: old GDI (original) and key methodological elements
- The ‘old’ GDI employed three sub-indices with variables reflecting educational attainment, health status, and income, mirroring the HDI structure.
- Health and education indicators were scaled on the ratio of the difference between actual and assumed minimum value of the indicator and the difference between the assumed maximum and minimum values of the indicator.
- The income indicator used a different construction based on the Atkinson formulation for the utility of income (see UNDP (1995) for details).
- Each sub-index was calculated separately for females and males.
- A gender-equity-sensitive indicator (GESI) was calculated for each sub-index using the formula involving PFP (proportion of female population), PMP (proportion of male population), and an inequality aversion coefficient Ɛ.
  - UNDP assumes Ɛ = 2.
- For the standard of living index, the international average real adjusted GDP per capita was multiplied by its GESI to derive a gender-inequality adjusted GDP per capita.
- Aggregation for the original GDI: simple average of the three sub-indices (arithmetic mean).

### Revisions and methodological changes over time (1995 → 2014)
- 1995 HDI
  - Sub-indices: Life expectancy index; Educational attainment index (adult literacy and combined enrollment); Standard of living index (Real GDP per capita PPP$, min. PPP$100 and max. PPP$40,000), with GDP index based on Atkinson utility approach.
  - Aggregation: simple average of the three sub-indices.
- Original GDI (based on 1995 HDI)
  - Same sub-indices as 1995 HDI, calculated for females and males, with gender-equity-sensitive adjustments and GESI applied; GDP per capita index same as 1995 HDI.
- 1999 HDI
  - Same as 1995 HDI except: GDP per capita uses logarithms; minimum set to PPP$200.
- 1999 GDI
  - Same as original GDI but incorporates 1999 HDI changes and changes in computation of the adjusted income index.
- 2000 HDI
  - Same as 1999 HDI except: adult literacy taken from age 15 and above; combined enrollment ratio uncapped.
- 2000 GDI
  - Based on 2000 HDI.
- 2010 HDI
  - Sub-indices: Life expectancy index (LE) same as 2000 HDI; Education index (ED) built from mean years of schooling and expected years of schooling; GNI index uses GNI per capita (PPP$).
  - Aggregation: geometric mean of sub-indices (geometric averaging applied to dimension indices).
- 2014 GDI (revised GDI)
  - Direct measure of the gender gap using components of the revised HDI with sex-disaggregated data.
  - Health variable: life expectancy with an adjustment for an average female biological advantage of five years.
  - Education variables: mean years of schooling and expected years of schooling.
  - Income methodology replaced by GNI per capita (PPP$ at 2011 constant prices) combined with the ratio of female to male wages in all sectors (not just non-agricultural sectors).
    - Where countries lack sex-disaggregated wage data, UNDP uses a presumed global average female-to-male wage ratio of 0.8.
      - (The source notes: "Our calculation suggests this estimate is high.")
  - Aggregation: geometric mean used to aggregate sub-indices; comparison between females and males relies on the simple ratio of female HDI to male HDI.
  - Index interpretation: typically ranges from 0 to 1 but may exceed 1 when females are more advantaged; numbers closer to 1 imply more equal gender relations.
  - Countries ranked by absolute deviation of the GDI from 1 (gender parity).
  - GDI remains a measure of gender equality (not directly female disadvantage), though ratios could be capped at gender equality to convert to female disadvantage.

### Gender Inequality Index (GII) methodology summary
- GII components: reproductive health, empowerment, and labor market variables.
  - Reproductive health (RH): maternal mortality ratio (MMR) and adolescent fertility rate (AFR; ages 15–19).
  - Empowerment (EM): population with at least secondary education (SE); shares of parliamentary seats (PR).
  - Labor market (LM): labor force participation rates (LFPR).
- Construction:
  - Each sub-index is calculated for each gender using geometric means across dimensions.
  - Female and male means are aggregated by the harmonic mean to create the equally distributed gender index (EDGI).
  - GII is based on the gap between 1 and the ratio of the calculated harmonic mean (actual data) to a reference standard that assumes females and males are equal.
  - GII interpretation: lower value when women and men are more equal; higher value when more unequal.
- Rationale for indicators:
  - MMR chosen because its improvement is a global policy priority and indicates access to health.
  - AFR included to capture risks and consequences of early childbearing on standards of living and human capital.

### Empirical summary statistics (UNDP 2014b, 2013 data)
- World average GDI value: 0.920 (indicating a gap of 8 percent in the HDI for females from that for males), based on a calculation for 148 countries in 2013.
- Regional averages:
  - OECD countries: 0.964
  - Latin American and the Caribbean: 0.963
  - South Asia: 0.860
  - Arab countries: 0.866
  - Sub-Saharan African countries: 0.867
- Note: A number of countries have GDI scores higher than 1 because female education achievement and female life expectancy (exceeding males by more than five years) can produce ratios above parity.

### Critiques, limitations, and suggested improvements (literature highlights)
- Broad critiques of the old GDI:
  - Income variable is problematic in measurement (Klasen 2006a, 2006b).
  - Paid income ignores women’s unpaid/home activities (Folbre 2006).
  - Earned income may not translate into differences in household spending or reflect women’s control of earnings (Chant 2006).
  - Suggestions: replace non-agricultural wages with economy-wide wages and improve earned income data coverage (Klasen 2006b); replace earned income with labor force participation (Dijkstra 2002).
  - Old GDI measured gender equality rather than female disadvantage by compounding gender gaps (Klasen 2006b).
- Critiques of revised indices:
  - GII mixes well-being and empowerment measures and combines female-only progress with gender-gap variables, which biases against poor countries (Klasen and Schüler 2011; Klasen 2014; Permanyer 2013).
  - Complexity and limited transparency in construction reduce intuition and interpretability.
  - Maternal mortality is an important driver of GII but suffers from poor data quality (Klasen 2014).
  - Klasen (2014) argues revised GDI is a better measure of gender parity but notes continued weakness in use of earnings.
  - Permanyer (2013) notes that variables unique to women prevent the index from taking a value of 0 when men and women are equal in other dimensions.
- Specific methodological concerns for other indices:
  - GEI caps female-to-male ratios at equality to prevent female advantage offsetting female disadvantage.
  - GGGI truncates results at equality benchmarks and uses complex weighting based on standard deviations, raising comparability concerns over time (Klasen and Schüler 2011; Hawken and Munck 2013).
  - SIGI uses principal components and squared sub-indices (incorporating inequality aversion) but may overweight large gaps and faces challenges capturing context-specific social institution biases (Van Staveren 2013; Branisa et al. 2014).
  - WEOI focuses on economic advancement and uses an unweighted average across 29 indicators; it allows compensation of female advantage with female disadvantage (EIU, 2012).

### Alternative and complementary indices (brief summaries)
- Gender Equality Index (GEI)
  - 21 indicators from six sources including attitudinal indicators (World Values Surveys); uses matching percentiles method; values generally range 0 to 1 with higher = more equal; caps female-to-male ratios at equality.
- Global Gender Gap Index (GGGI)
  - 14 indicators across educational attainment, health and well-being, economic participation and opportunity, and political empowerment; truncated at equality benchmarks (life expectancy benchmark 1.06; sex-ratio at birth benchmark 0.944); unweighted average of sub-indices; increasing in gender equality.
- Social Institutions and Gender Index (SIGI)
  - Focuses on social institutions as sources of gender inequality across five categories: discriminatory family codes; restricted physical integrity; bias toward sons; laws on restricted resources and assets; restricted civil liberties; sub-indices aggregated using principal components analysis and squared values to incorporate inequality aversion; values 0 to 1 with higher = more inequality.
- Women’s Economic Opportunity Index (WEOI)
  - 29 indicators across labor policy and practice, access to finance, education and training, women’s legal and social status, and general business environment; values range 0 to 100 with higher = more equal; allows compensation across indicators (measure of women’s opportunity relative to men).

*Source: Box 1, comparison and methodology of the HDI, GDI, and GII (excerpts as provided).*

### 3. A value less than 1 implies bias against women and a value greater than 1 implies bias

### 3. A value less than 1 implies bias against women and a value greater than 1 implies bias

### Diversity and construction of gender indices
- Indices differ in focus, choice of indicators, and aggregation methodology; they are summary measures of gender inequality rather than measures solely of bias against women.
- The income variable has the highest variation and thus “drives the results,” creating potential confounding between gender equality and standard of living; this can be addressed by introducing a weighting scheme.
- Klasen and Schüler (2011) propose the Gender Gap Measure (GGM) which:
  - replaces earnings with labor force participation, and
  - uses the geometric rather than arithmetic mean.
- The revised GDI adopts the female-to-male ratio idea but continues to rely on income rather than labor force participation.

### Quantitative correlations among indices (Table 2)
- Pearson correlations between indices (latest data available for each index):
  - GDI — GEI: 0.90***
  - GDI — GGGI: 0.79***
  - GDI — GII: -0.69***
  - GDI — SIGI: -0.75***
  - GDI — WEOI: 0.71***
  - GEI — GGGI: 0.76***
  - GEI — GII: -0.74***
  - GEI — SIGI: -0.81***
  - GEI — WEOI: 0.76***
  - GGGI — GII: -0.53***
  - GGGI — SIGI: -0.69***
  - GGGI — WEOI: 0.62***
  - GII — SIGI: 0.69***
  - GII — WEOI: -0.88***
  - SIGI — WEOI: -0.74***
- Note: Negative correlations for the GII and SIGI reflect that these indices are increasing in more inequality, while the others increase in more equality.
- Significance notation: *** Indicates significance at the 1% level, ** at the 5% level, and * at the 10% level.

### Correlations among individual indicators (Table 3) — selected entries
- Correlations (All countries / LIDCs where given):
  - Gross secondary enrollment vs Life expectancy (at birth): 0.19** / -0.01
  - Gross secondary enrollment vs Maternal mortality ratio: -0.58*** / -0.55***
  - Gross secondary enrollment vs Labor force participation (ages 15-64): 0.02 / 0.15
  - Maternal mortality ratio vs Child mortality, under the age of 5: 0.45*** / 0.63***
  - Labor force participation vs Share of female parliamentarians: 0.37*** / 0.31*
  - Child mortality vs Gross secondary enrollment: -0.41*** / -0.41**
- Interpretation: Correlations between individual indicators are significantly lower than between indices, supporting the role of indices as useful summary measures while highlighting the relevance of individual indicators for focused analysis.
- Sources for indicators: World Bank, World Development Indicators database; and IMF staff estimates.
- Significance notation: *** Indicates significance at the 1% level, ** at the 5% level, and * at the 10% level.

### Time-consistent (TC) extensions of UNDP indices
- Because the GDI and GII are not constructed consistently over time, the authors construct time-consistent (TC) versions of the UNDP’s GDI and GII to permit backward extension and more accurate trend analysis.
- Appendix D (not reproduced here) provides construction details and differences versus UNDP calculations.
- Regional and income-level aggregations use population-weighted averages.

### Trends over time (figures summary)
- GDI, TC version (Figure 11):
  - All regions show a trend toward greater gender equality.
  - Middle East and Central Asia lags considerably behind other regions.
  - Sub-Saharan Africa improves only modestly; Asia and Pacific surpasses Sub-Saharan Africa by the sample end.
  - Emerging market countries make the most significant improvement starting from a relatively weak position.
- GDI, TC with labor force participation replacing wages (Figure 12):
  - Sub-Saharan Africa performs relatively better and Middle East and Central Asia relatively worse compared with the wage-based GDI because of high female labor participation in Sub-Saharan Africa and low participation in Middle East and Central Asia.
- GII, TC version (Figure 13):
  - Y axis inverted (increasing GII = higher gender inequality).
  - Similar trend of improving gender equity across regions.
  - Contrasts with GDI: Middle East and Central Asia and Sub-Saharan Africa show roughly the same degree of gender equality, with Middle East and Central Asia starting worse but ending better than Sub-Saharan Africa.
  - Differences reflect GII’s inclusion of maternal mortality and adolescent fertility (which disadvantage Sub-Saharan Africa) and the higher female labor force participation and political representation in Sub-Saharan Africa.

### Relationship of indices to income — cross-section results (Table 4)
- OLS regressions of each index on ln GDP per capita (latest year of data for each index):
  - Coefficients on ln GDP pc:
    - GDI: 0.0368*** (0.00403)
    - GEI: 0.0599*** (0.00824)
    - GGGI: 0.0139*** (0.00519)
    - GII: -0.126*** (0.00898)
    - SIGI: -0.0729*** (0.0102)
    - WEOI: 12.23*** (0.940)
  - Constants:
    - GDI: 0.593*** (0.0396)
    - GEI: 0.194*** (0.0730)
    - GGGI: 0.564*** (0.0484)
    - GII: 1.535*** (0.0805)
    - SIGI: 0.814*** (0.0913)
    - WEOI: -58.44*** (8.451)
  - Observations: 146 (GDI), 177 (GEI), 127 (GGGI), 149 (GII), 99 (SIGI), 126 (WEOI)
  - R-squared: 0.391 (GDI), 0.266 (GEI), 0.077 (GGGI), 0.637 (GII), 0.318 (SIGI), 0.643 (WEOI)
- Interpretation:
  - Higher income is strongly correlated with improved measures of gender equality (sign reversed for GII and SIGI).
  - The GII regression shows particularly high explanatory power of income; GDI somewhat lower but still substantial.
  - Indices remain highly dependent on variables that vary with country income level; they can confound gender equality and standard of living.

### Relationship of TC indices to income — panel results (Table 5)
- OLS with cross-section fixed effects:
  - GDI, TC: ln GDP pc = 0.0676*** (0.00204); Constant = 0.296*** (0.0185); Observations = 2,776; R-squared = 0.295; Number of countries = 151
  - GII, TC: ln GDP pc = -0.157*** (0.00415); Constant = 1.847*** (0.0375); Observations = 3,000; R-squared = 0.333; Number of countries = 138
- OLS in first differences:
  - ∆ GDI, TC: ∆ ln GDP pc = 0.00831** (0.00420); Constant = 0.00207*** (0.000134); Observations = 2,591; R-squared = 0.009
  - ∆ GII, TC: ∆ ln GDP pc = -0.00501 (0.00611); Constant = -0.00655*** (0.000389); Observations = 2,858; R-squared = 0.000
- Interpretation:
  - With fixed effects, coefficients on income increase in absolute value.
  - In first-difference specifications, income remains positive and significant for the GDI but loses significance for the GII, underscoring sensitivity to specification.

### Uses for policy analysis and decision making
- Individual indicators:
  - Most useful for focused analysis tied to specific policy areas.
  - Advantages: straightforward to understand and interpret.
  - Limitation: data deficiencies in some indicators; international initiatives are improving coverage.
- Indices:
  - Useful as summary statistics of a country’s status on gender equity.
  - Limitations stem from choice of variables, weighting, aggregation methodology, and inconsistent time-series construction.
  - A good index should be simple, transparent, avoid mixing well-being with empowerment, and be robust to minor changes in included variables and weights.
- Policy relevance:
  - With gender equality embedded in the SDGs, governments should use indicators and indices to guide budget and policy decisions aimed at closing gender gaps and advancing women’s well being.
  - Both indicators and indices provide objective input for evaluating fiscal and other policies.

### Conclusion — key findings
- The paper constructs TC versions of the UNDP’s GDI and GII to permit consistent backward time-series analysis.
- Trends in individual indicators (education, health, economic and political opportunity) disaggregated by region and development level show overall progress in gender equality, with persistent regional and dimensional heterogeneity.
- Women have made significant progress in closing education gaps and some health indicators (including maternal mortality), and in economic and political gaps, though significant gaps remain.
- Advanced countries generally do better, but emerging and some LIDCs have made substantial improvements in specific areas; progress can occur even with significant income differences among countries.
- Indices are highly correlated with income; care is needed in aggregate analysis to avoid conflating gender equality with standard of living.
- Both indicators and indices are valuable to policymakers when used appropriately and with an understanding of their construction and limitations.

*Sources: See appendix C, table 1; Barro and Lee; ILO, ILOSTAT; World Bank, World Development Indicators database; and IMF staff estimates.*

### References

### _wp1621 - References

### Major bibliographic sources
- Wide-ranging references on gender equality, measurement indices, and gender-disaggregated data drawn from international organizations, academic journals, and policy institutions, including:
  - African Development Bank, Agenor & Canuto, Bardhan & Klasen, Barro & Lee, Bauer, Bertrand, Branisa et al., Buvinic et al., Casarico & Profeta, Chant, Clinton Foundation and Gates Foundation, Das et al., Dijkstra, Dilli et al., Economist Intelligence Unit, European Institute for Gender Equality, Fernandez & Fogli, Foa & Tanner, Folbre, Gaye et al., Gonzales et al., Grown, Hawken & Munck, IFPRI, ILO, IMF, Klasen and coauthors, Liu et al., Mithra & Farid, Morrison et al., OECD, Permanyer, Schüler, Thorslund et al., UNDP, UN, UN Women, Van Staveren et al., World Bank, World Economic Forum.
- Data sources heavily referenced include World Bank databank, World Development Indicators, ILOSTAT Database, UN and UN agencies’ databases, and IMF staff estimates.

### Appendix A. Gender-Related Data — scope and structure
- Appendix purpose: information on data with specific disaggregation by sex drawn from World Bank databases.
- Data series classified under five broad categories:
  - Economic opportunity
  - Education
  - Health
  - Political opportunity
  - Violence against women
- Table A1 lists indicators under each category with start year and sex-disaggregation status (yes / not applicable / collected but not yet available indicated by –).
- Notes:
  - Footnote: "This appendix was prepared by Carla Intal."
  - Footnote: "Data indicated with – are collected but not yet available."

### Appendix A — Key indicators and metadata (preserve original labels and years)
- Economic opportunity indicators (selection with start year and sex-disaggregated status as listed):
  - Access to credit - Yes
  - Access to ICT, Internet, Mobile Banking 2008 Yes
  - Average number of hours spent on:
    - unpaid child care - Yes
    - unpaid domestic work 1990 Yes
    - unpaid housework - Yes
  - Employees by category:
    - Agriculture 1980 Yes
    - Industry 1980 Yes
    - Services 1980 Yes
  - Female-headed households 1990 Not applicable
  - Female professional and technical workers 2006 Not applicable
  - Firms with:
    - Female top manager 2007 Not applicable
    - Female participation in ownership 2003 Not applicable
    - Women in managerial posts - Not applicable
  - Labor force participation rate (ages 15+) 1990 Yes
  - Labor force participation rate (ages 15-64) 1990 Yes
  - Land owners - Yes
  - Proportion of employed:
    - who are employers 1990 Yes
    - who are own account workers 1990 Yes
  - Wage equality ratio - Not applicable
  - Wage workers (percent) 1980 Yes
  - Youth unemployment 1991 Yes

- Education indicators (selection):
  - Completion rates:
    - Primary 1980 Yes
    - Secondary 1980 Yes
  - Educational attainment:
    - Primary 2000 Yes
    - Lower secondary 2000 Yes
    - Post secondary 2000 Yes
    - Upper secondary 2000 Yes
    - Tertiary 2000 Yes
  - Gross enrollment rates:
    - Primary 1980 Yes
    - Secondary 1980 Yes
    - Tertiary 1980 Yes
  - Net primary enrollment rate 1980 Yes
  - Female graduates in:
    - Sciences 1998 Not applicable
    - Social science, business and law 1998 Not applicable
    - Engineering, manufacturing, and construction (start year not listed)
  - Survival rate to grade 5 1980 Yes
  - Youth literacy rate 1980 Yes

- Health indicators (selection):
  - Adolescent fertility rate 1980 Not applicable
  - Births by skilled health worker 1984 Not applicable
  - Contraceptive prevalence 1980 Not applicable
  - Crude birth rate 1980 Not applicable
  - HIV prevalence 1990 Yes
  - Life expectancy 1980 Yes
  - Maternal mortality ratio 1990 Not applicable
  - Mean age at marriage 1980 Yes
  - Mortality rate, under-5 1980 Yes

- Political opportunity indicators (selection):
  - Share of female judges - Not applicable
  - Female legislators, senior officials and managers 1987 Not applicable
  - Female police officers - Not applicable
  - Female seats in parliament and ministerial bodies 1990 Not applicable

- Violence against women indicators (selection):
  - Proportion of women aged 15-49 subjected to physical or sexual violence in the last 12 months — Not applicable
    - by an intimate partner (percent) - Not applicable
    - by persons other than intimate partner (percent) - Not applicable
  - Women married at age 18 1986 Not applicable
  - Women who believe a husband is justified in beating his wife (various reasons) 1999–2005 Not applicable
    - when she argues with him (percent) 2000 Not applicable
    - when she burns the food (percent) 2000 Not applicable
    - when she goes out without telling him (percent) 2000 Not applicable
    - when she neglects the children (percent) 2000 Not applicable
    - when she refuses sex with him (percent) 1999 Not applicable
    - any of the five reasons (percent) 2005 Not applicable

### Data availability — coverage snapshots and thresholds
- Coverage sample: 188 countries, grouped by level of income.
- Availability rule: a country is treated as “available” for an indicator if it has at least 50 percent of the data for the years 1980-2014.
- Rationale: Preference to encompass the period in which reforms in gender equality became a focus of international efforts.
- Visual convention: darker shade implies greater data availability; a full dark bar indicates complete data coverage and a full light bar indicates no data coverage.
- Ranking: Indicators are ranked from left to right by number of countries with 50 percent of the data available.

### Data availability — key empirical coverage points (preserve exact figures)
- Economic opportunity:
  - Of the 21 identified gender indicators for economic opportunity, only labor force participation rate and youth unemployment have significant global data coverage.
  - Over 93 percent and 90 percent of countries have data on labor force participation rate and youth unemployment rate, respectively.
  - Data coverage on low-income countries is particularly high for these two variables: out of 60 LIDCs, 58 countries have data for labor force participation rate and 56 countries have data for youth unemployment rate.
  - Other economic indicators (wage work, employment by category, female employers, own account workers) are highly skewed towards high-income countries.

- Education:
  - Gross secondary enrollment has the highest coverage: 130 countries have data, accounting for 70 percent of the total.
  - Of the LIDCs, 57 percent have data from 1980-2014.
  - Other indicators with reasonable coverage: primary completion rates, gross tertiary enrollment, net primary enrollment, and survival rate in school to grade 5.

- Health:
  - Four indicators—crude birth rate, adolescent fertility rate, life expectancy, and under-5 mortality rate—are available in 97 percent of countries.
  - HIV prevalence is available for 56 out of 60 LIDCs.
  - Maternal mortality data are available only every five years.

- Political opportunity:
  - Share of female seats in parliament: 127 out of 188 countries have reported data.
  - Other indicators (share of female legislators, senior officials and managers; share of female judges; share of female police officers) have scarce data availability.

### Table A2 — Data initiatives (organization and focus; URLs preserved as listed)
- Initiatives and organizations with exclusive or significant focus on improving gender-related statistics include:
  - Women, Business and the Law — IFC, World Bank — Data collection, dissemination — http://wbl.worldbank.org/
  - Gender Equality Data Portal — World Bank — Data collection, organization, dissemination — http://datatopics.worldbank.org/gender/
  - Data2X — UN Foundation — Data and advocacy consultancy, dissemination — http://data2x.org/
  - Hunger Report — Bread for the World Institute — Advocacy, data visualization — http://hungerreport.org/missingdata/
  - No Ceiling — Clinton and Gates Foundations — Advocacy, data visualization — http://noceilings.org/about/
  - EDGE Initiative — UN Statistics Division — Data collection, dissemination — http://genderstats.org/EDGE
  - Gender Data Portal — OECD — Data collection, organization, dissemination — http://www.oecd.org/gender/data/
  - Gender and Land Rights Database — FAO — Data collection, dissemination — http://www.fao.org/gender-landrights-database/en/
  - World Policy Analysis Center — UCLA — Data collection, dissemination — http://worldpolicycenter.org/

### Appendix B. List of countries by region and LIDC classification (preserve presentation)
- Table B1 lists countries included in the sample and their regional classification; bolded countries indicate LIDCs (presentation preserved in original table).
- Regions included: Africa, Asia Pacific, Europe, Middle East & Central Asia, The Americas & Caribbean.
- Countries enumerated under each region exactly as presented in the source (sample includes 188 countries).

*Sources: References and appendices as provided in the IMF working paper content unit.*

### Appendix C.  Data on Gender Indices

### Appendix C.  Data on Gender Indices

### Index coverage and metadata (Table C1)
- Old Gender-Related Development Index (GDI)
  - Developer: UNDP
  - Publication date: 1995, 2000-2007/2008, 2009
  - Web link: http://hdr.undp.org/en/global-reports
  - Data year: 1992, 1998-2005, 2007
  - Country coverage: 194
  - Note: The old GDI was discontinued after the 2007 index, which was reported in the 2009 Human Development Report.
- Revised GDI
  - Developer: UNDP
  - Publication date: 2014
  - Web link: http://hdr.undp.org/en/data
  - Data year: 2013
  - Country coverage: 148
  - Note: The revised GDI was introduced in the 2014 Human Development Report, using 2013 data.
- Gender Inequality Index (GII)
  - Developer: UNDP
  - Publication date: 2014
  - Web link: http://hdr.undp.org/en/data
  - Data year: 2008, 2011-2013
  - Country coverage: 195
- Gender Equality Index (GEI)
  - Developer: Institute of Social Studies, Erasmus University
  - Publication date: Regularly updated
  - Web link: http://www.indsocdev.org/
  - Data year: Every 5 years, 1990 onward
  - Country coverage: 209
- Social Institutions and Gender Index (SIGI)
  - Developer: OECD
  - Publication date: 2009, 2012, 2014
  - Web link: http://genderindex.org/
  - Data year: 2009, 2012, 2014
  - Country coverage: 108
- Global Gender Gap Index (GGGI)
  - Developer: World Economic Forum
  - Publication date: 2006-2014
  - Web link: http://reports.weforum.org/global-gender-gap-report-2014/
  - Data year: 2006-2014
  - Country coverage: 142
- Women's Economic Opportunity Index (WEOI)
  - Developer: Economist Intelligence Unit
  - Publication date: 2010, 2012
  - Web link: http://graphics.eiu.com/upload/WEO_2012_v0.4.4_FINAL_FOR_PUBLIC_RELEASE.xls
  - Data year: 2010, 2012
  - Country coverage: 128

*The number of countries reflects the sample size for the latest listed year of data.*

---

### Appendix D. Replication, Extension, and Revision of the GDI and GII

### Objective and approach
- Purpose: Extend the UNDP’s new GDI and the GII back to 1990 and assess sensitivity of the GDI to replacement or re-estimation of some variables.
- Term used: Time consistent (TC) versions for replicated and extended series.
- Extension limit: Series extended back only to 1990 (despite some indicators available back to 1950) because of limited data for some indicators and decision not to impute extensively beyond wages.

### Construction of the GDI, TC version
- Approach: Followed the steps of computing the GDI as described in HDR 2014 technical notes; data updated since UNDP original calculations.
- Wage data handling:
  - Ratio of female to male wage data availability: poorest coverage.
  - Maximum-country-year coverage: data available for only 68 countries in the year with highest coverage.
  - Regional bias: 44 percent of observations cover Europe and Central Asia; 0.6 percent for sub-Saharan Africa.
  - UNDP method: uses global weighted wage ratio average of 0.8 to substitute missing wages.
  - Authors’ method: impute missing wage data using interpolation for countries with at least 5 years of data; use a regional average for countries without sufficient data (instead of UN global average).
- Coverage after wage interpolation: 146 countries from 1990-2013.
- Alternative specification: substituted labor force participation rate (LFPR) for wages in the “standard of living” sub-index as an alternative robustness check.
- Key finding on index values: TC version reduces index values for all countries; unlike UNDP’s GDI which showed some countries with gender disparity in favor of women, all countries in the TC version have an index value below one.

### Indicators included in the GDI (Table D1)
- Life Expectancy (F)
  - Range: 22.7-86.7
  - Mean (s.d.): 67.33 (11.25)
  - Countries: 230
  - Years: 1960-2013
  - Source: World Development Indicators, World Bank
- Mean Years of Schooling (F)
  - Range: 0-13.64
  - Mean (s.d.): 5.67 (3.46)
  - Countries: 156
  - Years: 1950-2012
  - Source: Barro and Lee (2014) and UNESCO Institute for Statistics
- Expected Years of Schooling (F)
  - Range: 0.36-20.84
  - Mean (s.d.): 10.77 (4.13)
  - Countries: 101
  - Years: 1970-2013
  - Source: UNESCO Institute for Statistics
- Ratio of Female to Male Wage
  - Range: 0.41-1.5
  - Mean (s.d.): 0.77 (.12)
  - Countries: 27
  - Years: 1995-2011
  - Source: ILO
- Female Share of Economically Active Population
  - Range: 0.096-0.56
  - Mean (s.d.): 0.40 (.095)
  - Countries: 184
  - Years: 1990-2014
  - Source: ILO
- GNI per capita PPP (2011 $)
  - Range: 307-156,408
  - Mean (s.d.): 15,172.78 (19,408.85)
  - Countries: 189
  - Years: 1980-2013
  - Source: ILO
- Population (F)
  - Range: 25.98-544,386.9
  - Mean (s.d.): 11,053.89 (41,981.62)
  - Countries: 184
  - Years: 1980-2013
  - Source: ILO

### Expected years of schooling: data availability and growth rates
- Availability: Expected years of schooling has limited availability across time; for many countries only one observation between 1970-2013.
- UNDP practice: uses observations from as far back as 2002 to calculate the 2013 GDI.
- Authors’ growth rate findings:
  - Average yearly growth rate in expected years of schooling: females 2.3 percent; males 1.7 percent.
  - Average ten year growth rate: females 16.8 percent; males 11.4 percent.
- Implication: Using older observations (e.g., 1990) to represent a country’s gap in later years (e.g., 2000) may misrepresent the gap because the gap narrowed substantially, converging around 2000.

### Wage data coverage by region (Table D2)
- East Asia and Pacific: Percent of Observations 15.9; Share of Population 33.7
- Europe and Central Asia: Percent of Observations 43.9; Share of Population 14.1
- Latin America and Caribbean: Percent of Observations 24.0; Share of Population 8.5
- Middle East and North Africa: Percent of Observations 10.4; Share of Population 5.3
- North America: Percent of Observations 1.8; Share of Population 5.4
- South Asia: Percent of Observations 3.2; Share of Population 22.6
- Sub-Saharan Africa: Percent of Observations 0.6; Share of Population 10.1

### Critique of UNDP wage imputation
- UNDP global average wage ratio used: 0.8.
- Two main issues identified:
  - Over-representation of Europe and Central Asia in wage observations relative to population.
  - Global average assumes missing-data countries have same wage ratio as observed countries, potentially penalizing countries with true ratios below the global average.

### Correlations: GDI original vs. TC versions (Table D3)
- Spearman rank correlations (significance noted):
  - GDI Original vs. TC version: 0.96*** (significant at the 1 percent level)
  - GDI Original vs. TC version with LFPR: 0.75*** (significant at the 1 percent level)
  - TC version vs. TC version with LFPR: 0.81*** (significant at the 1 percent level)
- Note: The TC versions are strongly correlated with the UNDP’s GDI but show measurable differences.

### Construction of the GII, TC version
- Coverage: 141 countries from 1990-2013.
- Indicators included (five):
  - Maternal mortality ratio (MMR)
  - Adolescent fertility rate
  - Share of female seats in national parliaments
  - Educational attainment at secondary and tertiary levels
  - Labor force participation rate
- Interpolation and extrapolation methods:
  - MMR: available in five-year intervals beginning 1990; linear interpolation used for in-between years.
  - Adolescent fertility rate: data from United Nations Department of Economic and Social Affairs via WDI; generally good coverage and consistent time series.
  - Female seats in parliament: data from Inter-Parliamentary Union beginning 1990; linear interpolation used for missing years, including gaps around eight years in some countries.
  - Educational attainment (population over age 25 with at least secondary education):
    - Sources: Barro and Lee (2014) and UNESCO Institute for Statistics.
    - Method: Add percent secondary to percent tertiary from Barro and Lee; supplement with UNESCO “population with at least secondary education (+25)”; linear interpolation when missing data are between two points; use most recent year when interpolation not possible (some countries use data as far back as 2010 for the most recent index year).
  - Labor force participation rate: ILO data via World Bank; generally complete time series 1990-2013; use most recent available year when recent years are missing.
- Reconstruction method: Followed steps described in UNDP Human Development Report 2014 technical notes.
- Comparison with UNDP GII:
  - Spearman’s rho between TC GII and UNDP GII for the most recent year: 0.99 (significant at the 1 percent level).
  - Slight variation attributed to updates in indicators used.

*Sources: Tables and text as provided in the Appendix C and Appendix D material.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1621.pdf_