## wpiea2020119-print-pdf

## Source details

**Canonical URL:** [wpiea2020119-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020119-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020119-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020119-print-pdf.pdf.json)

---

### Introduction — Gender equality as an economic imperative
- Gender equality framed as a vital economic issue: "Gender equality is more than a moral issue; it is a vital economic issue. For the global economy to reach its potential, we need to create conditions in which all women can reach their potential." — Maurice Obstfeld, March 23, 2017 (IMF 2017).
- Women account for half of the total population but constitute less than a third of the actual workforce (Lagarde 2013).
- UN High-Level Panel: 700 million fewer women than men of working age were in paid employment in 2016; when paid, women tend to work in jobs with relatively low earnings, poor working conditions, and limited career prospects (UN 2016).
- Gender equality and women’s empowerment affect growth, income inequality, demographic shifts, and financial sector stability (Gonzales and others 2015; Kochhar and others 2017; IMF 2018a; IMF and WB 2007; IMF 2017).

### Hypothesis and identification strategy
- Core hypothesis:
  - Higher gender equality supports economic growth by allocating female labor to more productive uses, disproportionately benefiting industries with a higher share of women relative to other industries.
- Mechanisms:
  - Extensive margin: higher gender equality increases female labor force participation, enlarging the talent pool and raising productivity of marginal workers.
  - Intensive margin: higher gender equality allows women to realize fuller labor-market potential (for example, higher career ladders), raising productivity.
- Identification strategy:
  - Adaptation of Rajan and Zingales (1998) DiD approach using cross-industry heterogeneity and cross-country variation in gender inequality.
  - Benchmark industry female shares taken from Sweden (country with lowest GII in period) to estimate industries’ intrinsic gender composition; assumption: ordering of industries by female share in benchmark carries over to other countries.
  - Empirical model: Y_{i,k} = α + β X_{i,k} + γ (Female labor share_i × Gender Inequality_k) + μ_k + ν_i + e_{i,k}; main parameter γ expected to be negative.

### Data, sample, and measures
- Industry-level manufacturing employment and value-added: UNIDO (ISIC 2-digit, Revision 3), manufacturing sectors ISIC 15-37.
- Value-added deflation: Producer Price Index (PPI) for the US.
- Benchmark female employment data: Sweden (main), and alternative benchmarks Australia, Austria, Canada, Ireland, New Zealand.
- Country-level gender inequality: Composite Gender Inequality Index (GII) from IMF (Stotsky and others 2016); range 0–1 where higher = higher gender inequality.
- Sample focus: emerging-market and developing economies in the 1990s.
- Variable definitions (selected):
  - Value-added growth: average growth in real value-added over the 1990s (nominal US$ deflated by US PPI).
  - Labor productivity growth: average growth in value-added per worker over the 1990s.
  - Value-added share: industry value-added as share of total manufacturing in 1990.
  - External finance dependence: industry dependence on external financing (Popov 2014, US firms in 1980s).
  - Alternative gender inequality proxies: adolescent fertility rate (births per 100 women ages 15–19), relative infant mortality (female/male), relative labor participation (male/female), Gender Parity Index (GPI), women’s rights law score.

### Key findings — magnitudes and robustness
- Evidence supports a causal effect of gender inequality on industry-level real economic outcomes; main interaction coefficient (Female share × GII) is negative and statistically significant across specifications.
- Main quantitative magnitudes (Sweden benchmark):
  - An industry at the 75th percentile of the female share in total employment compared to an industry at the 25th percentile grows 1.7 percentage points faster in value-added when located in a country at the 25th percentile of gender inequality rather than in one at the 75th percentile.
  - The same comparison yields 1.2 percentage points faster growth in terms of labor productivity.
- Context for magnitudes (averages):
  - Real annual growth rate of value-added: 2.2 percent per year (mean value-added growth = 0.022).
  - Average growth of labor productivity: 1.2 percent per year (mean labor productivity growth = 0.012).
- Robustness:
  - Results robust to alternative benchmark countries (Canada, Ireland, Austria, New Zealand, Australia) — interaction coefficients remain negative and significant; reported differential growth rates by benchmark: 1.7, 2.2, 1.7, 1.7, 1.9, 2.0 (percent, by benchmark country).
  - Robust to alternative gender inequality measures (Table 7): female share × adolescent fertility rate = -0.024** (0.010); female share × relative infant mortality = -1.538** (0.769); female share × relative labor force participation = -0.037** (0.019); female share × GPI = 0.802** (0.409); female share × women’s rights law score = 0.773** (0.385).
  - Robust to a wide range of omitted-variable interactions and controls: log GDP per capita, log population, log population density, female share in population, exports, imports, openness, FDI inflows, financial development (broad money/GDP), institutional quality (political competition), human capital (teacher-student ratio).
  - Subsample of smaller industries: interaction remains significant and roughly doubles in magnitude (Female share × GII = -1.381** (0.541); differential growth rate = 3.3 percent in that split).
  - External finance dependence checks: external finance dependence interactions with financial development and with GII are generally insignificant; main female share × GII coefficient remains close to baseline.
  - Additional checks: dropping Morocco and Hungary; binary and quartile female-share indicators; averaging or taking maximum across benchmark countries; winsorizing growth — all yield similar negative interaction effects.

### Descriptive statistics and selected values
- GII (1990, regression sample): mean 0.582; median 0.598; std. dev. 0.129; min 0.308; max 0.807; Obs 48.
- Selected country GII values (Table 2, Panel A): Hungary 0.308; China 0.322; Bulgaria 0.326; Mongolia 0.379; Malta 0.382; Malaysia 0.433; Romania 0.434; Nepal 0.710; India 0.715; Kenya 0.742; Iran 0.747; Niger 0.804.
- Industry female shares (selected, Table 1 sample entries — Sweden shown among benchmark countries):
  - ISIC 15 Food and beverages: Sweden 37.433
  - ISIC 16 Tobacco products: Sweden 52.734
  - ISIC 17 Textiles: Sweden 51.413
  - ISIC 18 Wearing apparel: Sweden 70.639
  - ISIC 20 Wood products (excl. furniture): Sweden 13.995
  - ISIC 33 Medical, precision and optical instr.: Sweden 32.110
  - ISIC 34 Motor vehicles, trailers, semi-trailers: Sweden 16.510
  - ISIC 36 Furniture; manufacturing n.e.c.: Sweden 30.831
- Descriptive stats (Table 3, Panel A, selected):
  - Value-added growth: Mean 0.022; Median 0.026; Std. dev. 0.172; Min -1; Max 1; Obs 692
  - Labor productivity growth: Mean 0.012; Median 0.014; Std. dev. 0.128; Min -0.802; Max 1; Obs 660
  - Value-added share: Mean 0.063; Median 0.035; Std. dev. 0.083; Min 0.000; Max 0.615; Obs 692
  - External fin. dep.: Mean -0.089; Median 0.010; Std. dev. 0.309; Min -0.920; Max 0.280; Obs 17
  - Adolescent fertility rate: Mean 7.918; Median 7.050; Std. dev. 4.660; Min 0.658; Max 22.221; Obs 65
  - Relative infant mortality: Mean 0.828; Median 0.823; Std. dev. 0.049; Min 0.685; Max 0.952; Obs 62
  - GPI: Mean 0.917; Median 0.973; Std. dev. 0.140; Min 0.525; Max 1.080; Obs 57
  - Relative labor participation: Mean 2.225; Median 1.741; Std. dev. 1.527; Min 1.051; Max 8.341; Obs 64
  - Women’s rights law score: Mean 0.643; Median 0.649; Std. dev. 0.138; Min 0.357; Max 0.857; Obs 37

### Main regression coefficients (selected)
- Industry value-added growth (Female share × GII, Table 5 — heteroskedasticity-robust SEs):
  - Sweden (1980s): -0.569** (0.259) → Differential growth rate: 1.7 (percent)
  - Canada (1980s): -0.573*** (0.219) → Differential growth rate: 2.2 (percent)
  - Ireland (1988:95): -0.489** (0.203) → Differential growth rate: 1.7 (percent)
  - Austria (1990s): -0.522** (0.230) → Differential growth rate: 1.7 (percent)
  - New Zealand (1990s): -0.582** (0.238) → Differential growth rate: 1.9 (percent)
  - Australia (1980s): -0.627*** (0.233) → Differential growth rate: 2.0 (percent)
- Labor productivity growth (Female share × GII, Table 6):
  - Sweden (1980s): -0.446** (0.192) → Differential growth rate: 1.2 (percent)
  - Canada (1980s): -0.411*** (0.151) → Differential growth rate: 1.4 (percent)
  - Ireland (1988:95): -0.367*** (0.142) → Differential growth rate: 1.1 (percent)
  - Austria (1990s): -0.412*** (0.155) → Differential growth rate: 1.2 (percent)
  - New Zealand (1990s): -0.436*** (0.165) → Differential growth rate: 1.3 (percent)
  - Australia (1980s): -0.453*** (0.163) → Differential growth rate: 1.3 (percent)

### Policy implications and conclusions
- Empirical conclusion: Gender inequality inhibits growth by constraining the productive use of female labor; high-female-share industries grow relatively faster in countries that are more gender equal.
- Mechanism emphasized: higher gender equality helps allocate female labor to more productive uses via both extensive and intensive margins.
- Policy recommendations (implied by empirical channel):
  - Assign high priority to policies that ensure a level playing field for women.
  - Policy levers include improving the rule of law and women’s legal rights, women’s health, access to education, financial services, and technology.
- Rationale: Policies promoting gender equality are not only matters of human rights, equity, and social justice, but are also relevant levers to boost economic growth—benefiting the economy as a whole.

*Source: wpiea2020119-print-pdf — IMF Working Paper (selected sections: 1. Introduction; 2. Hypothesis and Methodology; 4. Results; Tables and descriptive statistics).*

### 1. Introduction

### 1. Introduction

### Gender equality as an economic imperative
- "Gender equality is more than a moral issue; it is a vital economic issue. For the global economy to reach its potential, we need to create conditions in which all women can reach their potential." — Maurice Obstfeld, March 23, 2017 (IMF 2017).
- Worldwide, productivity growth and the pace of human development are slowing (ILO 2017), and women’s full and effective participation in the workforce and decent work for all are critical to inclusive and sustainable economic growth.
- Women account for half of the total population but constitute less than a third of the actual workforce (Lagarde 2013).
- UN High-Level Panel on Women’s Economic Empowerment: 700 million fewer women than men of working age were in paid employment in 2016; when paid, women tend to work in jobs with relatively low earnings, poor working conditions, and limited career prospects (UN 2016).
- Gender equality and the empowerment of women are central to the development agenda and have implications for growth, income inequality, demographic shifts, and financial sector stability (Gonzales and others 2015; Kochhar and others 2017; IMF 2018a; IMF and WB 2007; IMF 2017).

### Literature, measurement, and empirical challenges
- Extensive literature documents gender inequality in opportunities (education, health, finance) and outcomes (employment, earnings), with a rich literature on the determinants of the gender wage gap.
- Theoretical contributions propose that gender inequality may hamper economic development (Galor and Weil 1996; Lagerlöf 2003) through effects on human capital and fertility.
- Cross-country empirical approaches typically regress per capita income growth on proxies of gender inequality while controlling for standard growth covariates; these raise endogeneity concerns, including reverse causality and omitted variables.
- Instrumental variable strategies are an avenue but finding a plausible instrument that affects growth only through gender inequality is challenging.
- This paper uses a broader composite gender inequality index (GII) that evaluates both equality of opportunities and outcomes—including women’s empowerment, female reproductive health, and labor market variables (Gaye and others 2010; UNDP 2014; Stotsky and others 2016).

### Identification strategy and empirical approach
- Focus: a channel through which higher gender equality may support economic growth — by allocating female labor to its more productive use, enabling firms to better use available labor resources.
- Hypothesis: higher gender equality disproportionately benefits industries that typically have a greater share of women in employment relative to other industries.
- Mechanisms:
  - Extensive margin: higher gender equality increases female labor force participation, enlarging the talent pool and raising the productivity of marginal workers.
  - Intensive margin: higher gender equality allows women to fully develop labor-market potential (for example, higher career ladders), raising productivity.
  - Effects are expected to be more evident for high-female-share industries (defined relative to other industries).
- Methodology: adapt the difference-in-differences (DiD) application by Rajan and Zingales (1998).
  - Identify an industry’s intrinsic gender composition using the share of female labor in total labor in Sweden (the country with by far the lowest GII in the relevant period), assuming Sweden’s labor market is relatively frictionless regarding women’s access and attitudes to jobs across industries.
  - Assume the benchmark country’s estimated industries’ “clean” gender compositions carry over to other countries, enabling within-country, across-industry comparisons: high-female-share industries should grow relatively faster in more gender-equal countries.
  - This exploits heterogeneity across manufacturing industries and addresses many endogeneity concerns in aggregate cross-country studies.
- Clarifications:
  - "High-female-share" denotes an industry’s share of female employment is high relative to other industries (not necessarily >50 percent).
  - The approach does not imply the benchmark country’s industry gender composition is optimal or that horizontal segregation is desirable; it is used as an exogenous source of variation.

### Data and scope
- Industry-level manufacturing employment data from the UN Industrial Development Organization (UNIDO) database.
- Country-level composite gender inequality index (GII) from IMF releases (Stotsky and others 2016).
- Sample: a large sample of emerging-market and developing economies.
- The GII measures a broader set of gender inequalities than measures focusing only on education.

### Key findings and magnitudes
- Evidence supports a causal effect of gender inequality on real economic outcomes at the industry level.
- Quantitative estimates:
  - An industry at the 75th percentile of the female share in total employment, compared to an industry at the 25th percentile, grows 1.7 percentage points faster in terms of value-added when it is located in a country at the 25th percentile of gender inequality rather than in one at the 75th percentile.
  - The same comparison yields 1.2 percentage points faster growth in terms of labor productivity.
- Context for magnitudes:
  - The real annual growth rate of value-added is, on average, 2.2 percent per year.
  - The average growth of labor productivity is 1.2 percent.
- Robustness: results are robust to different measures of gender inequality and to a wide range of alternative explanations including outliers, measurement error, omitted variables, and reverse causality.

*Source: wpiea2020119-print-pdf - 1. Introduction*

### 2. Hypothesis and Methodology

### 2. Hypothesis and Methodology

### Hypothesis
- Industries with a typically greater share of women in their employment compared to other industries grow relatively slower in countries that, a priori, have higher gender inequality.
- Mechanisms:
  - Extensive margin: Higher gender equality → more and better educated women enter the labor force → larger talent pool → higher productivity of the marginal worker → higher industry productivity and growth.
  - Intensive margin: Higher gender equality → women fulfill fuller potential (e.g., promotion to managerial positions) → growth-promoting effects.
- Effects are predicted to be stronger for high-female-share industries:
  - Extensive margin: newly hired women disproportionately join industries that typically hire more women.
  - Intensive margin: unlocking women’s potential is more beneficial where female labor share is greater.

### Identification strategy and empirical specification
- Follow the DiD identification strategy of RZ (Rajan and Zingales) using cross-industry variation in gender compositions and cross-country variation in gender inequality.
- Estimated model:
  - Y_{i,k} = α + β X_{i,k} + γ (Female labor share_i × Gender Inequality_k) + μ_k + ν_i + e_{i,k}
    - Dependent variable: average real growth rate of value-added in industry i in country k over the period of the 1990s.
    - μ_k and ν_i: country and industry fixed effects.
    - X_{i,k}: industry i’s value-added share of manufacturing in country k in 1990 (to account for possible convergence effects).
    - Main coefficient of interest: γ on the interaction (Female labor share_i × Gender Inequality_k). If the hypothesis is correct, γ should be negative.
- Timing and exogeneity:
  - Use all country-level variables from the first year of the analysis (following RZ). Test analyzes how ex-ante gender inequality affects ex-post industry growth.
  - Industry female share is defined using a benchmark country with the least gender frictions in labor markets—Sweden (lowest GII). Shares are averaged across years for corresponding periods (see Table 1 in source).

### Key identification assumptions
- Intrinsic industry differences: Industries intrinsically differ in their tendency to employ women due to industry-specific relative marginal product of labor (MPL) between women and men (horizontal segregation driven by relative MPL).
- Cross-country ordering: The ordering of industries by female share in the benchmark country carries over to other countries (e.g., if industry A employs more women than industry B in Sweden, the same ordering holds in Turkey). Full equality of shares across countries is not required—only consistent ordering.
- Stability of industry gender compositions:
  - Reported low variation of industries’ female share across time and benchmark countries: e.g., within Sweden over the 1980s the variation of the female share in an industry across time is around 4 percent of the mean.
  - High correlations of industries’ gender compositions between Sweden and alternative benchmark countries (Panel B of Table 3).
- Rationale for benchmark-country approach:
  - Actual country-specific industry female shares are equilibrium outcomes potentially contaminated by gender-based frictions (e.g., discrimination).
  - Using Sweden in the 1980s (high gender equality) provides a cleaner proxy for industries’ intrinsic gender composition linked to relative MPLs.
  - Aggregate technological changes shaping relative MPLs are assumed worldwide; Sweden in the 1980s is treated as a suitable proxy for manufacturing-sector industry compositions in the sample of emerging-market and developing economies in the 1990s.
- Robustness: Results are robust to using alternative dates and countries for benchmarking. Benchmark countries should have relatively low gender inequality compared to the regression sample; analysis focuses on emerging-market and developing economies rather than developed economies.

### Data construction and sample
- Industry data:
  - Source: ISIC 2-digit industry-level dataset from UNIDO (Revision 3).
  - Coverage restricted to manufacturing sectors (ISIC 15-37).
  - Value-added: nominal value-added converted to real value-added using the Producer Price Index (PPI) from the US.
  - Female employment data for benchmark countries from UNIDO: Australia, Austria, Canada, Ireland, New Zealand, and Sweden (Sweden is main benchmark).
  - Observations dropped if value-added share of manufacturing < 0 or > 1.
  - Industries dropped if female employment data missing for majority of benchmark countries or most years in the relevant period.
  - Final sample: 17 industries.
  - Value-added share of an industry in a country’s total manufacturing value-added measured in 1990; if missing in 1990, use 1991, 1992, or 1993; otherwise observation omitted.
  - Labor productivity = real value-added / number of employees.
  - Use country-industry observations where average growth rate for a sector (over the 1990s) is calculated as the average of at least three data points on growth rates.
  - External finance dependence (robustness check): measures from Popov (2014) for US firms in the 1980s mapped to industries.

### Country-level measures and covariates
- Main gender inequality measure:
  - Composite Gender Inequality Index (GII) from IMF (based on Stotsky and others 2016), range 0 to 1 where higher GII indicates higher gender inequality.
  - GII in 1990 within regression sample varies from approximately 0.3 in Hungary to approximately 0.8 in Morocco.
  - GII in benchmark countries in 1990 ranges from 0.093 in Sweden to 0.284 in New Zealand.
- Alternative gender inequality proxies (used in robustness checks):
  - Adolescent fertility rate: births per 100 women ages 15–19 years.
  - Relative infant mortality: ratio of infant female mortality to infant male mortality.
  - Relative labor participation ratio: male labor participation rate relative to female rate.
  - Gender Parity Index (GPI): ratio of female students to male students.
  - Women’s rights law score: constructed from Women’s Legal Rights data as (number of negative answers) / (number of total replied questions) in 1990.
  - Pairwise correlations among these measures and with GII reported in Panel C of Table 3 indicate consistency.
- Covariates:
  - Institutional quality: political competition index from Polity IV (range 0 to 10; higher = greater political competition).
  - Financial development proxy: broad money from World Bank WDI.
  - Human capital proxy: teacher-student ratio in secondary education (Barro 1991).
  - Other controls from WDI: real GDP (log), population (log), FDI inflows (scaled by GDP), exports and imports (scaled by GDP), trade over GDP (openness), population density (people per km^2, in logs).
- Missing data rules:
  - If a country-level variable is missing in 1990, use earliest available year in the 1990s; if not available until 1993, observation not used in corresponding robustness check.

*Source: IMF — "2. Hypothesis and Methodology" (wpiea2020119-print-pdf).*

### 4. Results

### 4. Results

### 4.1 Baseline results: interaction of industry female share and country GII
- Specification: industry value-added growth as dependent variable; controls include country and industry fixed effects; main coefficient is the interaction between an industry’s share of female labor and a country’s GII, plus industry value-added share in total manufacturing.
- Benchmark for industries’ gender composition: Sweden (preferred).
- Main empirical finding (Sweden benchmark, Table 5, column 1):
  - The coefficient on the interaction term is negative and highly statistically significant.
  - Predicted differential: an industry at the 75th percentile of female shares (rubber and plastics products) compared to an industry at the 25th percentile of female shares (non-metallic mineral products) grows 1.7 percentage points faster in value-added when located in Costa Rica (25th percentile of GII) than in Cameroon (75th percentile of GII).
  - Context: the real value-added growth rate is, on average, 2.2 percent per year; thus the 1.7 percentage points differential is substantial.
- Robustness to alternative benchmark countries (Australia, Austria, Canada, Ireland, New Zealand):
  - Statistical significance remains; magnitude increases in some cases.
- Concern addressed: labor supply versus productivity
  - Using real labor productivity growth (value-added based) as dependent variable (Table 6) yields a similar result:
    - Predicted differential: 1.2 percentage points faster value-added labor productivity growth for the 75th vs 25th percentile female-share industry in Costa Rica relative to Cameroon.
    - Context: the value-added labor productivity growth rate is, on average, 1.2 percent per year; thus the 1.2 percentage points differential is large.

### 4.2 Alternative measures of gender inequality
- Alternative narrow measures tested (Table 7): adolescent fertility rate, relative infant mortality, relative labor participation, GPI, women’s rights law score.
- Interpretation of measures:
  - Higher adolescent fertility rate, relative infant mortality, relative labor participation indicate higher gender inequality.
  - Lower GPI and lower women’s rights law score indicate higher gender inequality.
- Results:
  - Conclusions are robust to using these narrower measures.
  - Inequality in both opportunities (women’s rights law, education, health) and outcomes (employment) plays a role for growth.
  - Results reported using Swedish industry gender compositions; similar results obtained with alternative benchmark countries (results not reported).

### 4.3 Omitted variable bias checks (Table 8)
- Controlled-interaction checks (female share interacted with):
  - (1) log of real GDP per capita (reverse causality concern: development lowering GII).
  - (2) log of total population (bias from very low population affecting female labor participation).
  - (3) log of population density (urbanization, commuting, social norms).
  - (4) female share in population (ratio of number of females to number of males).
  - (5)–(7) export, import, openness (export+import/GDP), and FDI (net inflows as share of GDP).
  - (9) financial development (broad money as share of GDP).
  - (10) institutional quality (political competition from Polity IV).
  - (11) human capital (teacher-to-student ratio in secondary education).
- Result: None of these additional controls significantly alters the main coefficient; conclude results unlikely driven by omitted variables.

### 4.4 Reverse causality concerns and subsample of small industries
- Issue: economic growth lowers gender inequality and GII persistence may induce reverse causality despite using GII from 1990 and growth in the 1990s.
- Mitigation: estimate model on subsample of smaller industries (industries with less than a median of value-added shares in their respective countries).
  - Rationale: small industries are less likely to drive country-level growth and less likely to generate macro feedback to gender equality.
- Result:
  - Interaction coefficient between female share and GII remains statistically significant and roughly doubles in magnitude.
  - Interpretation: increase in magnitude likely due to higher growth potential of smaller industries.

### 4.5 External finance dependence (Table 9)
- Concern: finance-growth literature (Rajan and Zingales) may confound results if external finance dependence and financial development correlate with gender composition and GII.
- Data: industries’ external finance dependence from Popov (2014) for US firms in the 1980s.
- Findings:
  - Interaction term between external finance dependence and financial development is positive but insignificant (consistent with RZ directionally).
  - External finance dependence is insignificant when interacted with GII alone or alongside interaction with female share.
  - Financial development does not appear to be a relevant omitted factor.
  - Main coefficient of interest remains close to baseline estimates in Table 5.

### 4.6 Additional robustness checks (Table 10)
- Tests performed:
  - (1) Drop two countries with highest and lowest GII (Morocco and Hungary) to test outlier sensitivity.
  - (2) Use a binary dummy for industries above/below median female share in Sweden (1 if above median, 0 otherwise).
  - (3) Use categorical variable with four quartile-based categories of female share (1 = below 25th percentile, 2 = 25th–50th, etc.).
  - (4) For each industry, assign the average female share across six benchmark countries.
  - (5) For each industry, assign the maximum female share across six benchmark countries.
  - (6) Winsorize mean growth rates at 1 and 99 percent (alternative to restricting growth between -1 and +1).
- Result: All alternative specifications yield results similar to the baseline.

### Key substantive findings and magnitudes
- Differential growth in value-added (Sweden benchmark):
  - 1.7 percentage points faster growth (75th vs 25th percentile female-share industry) in Costa Rica vs Cameroon.
  - Average real value-added growth rate: 2.2 percent per year.
- Differential growth in value-added labor productivity:
  - 1.2 percentage points faster growth (75th vs 25th percentile female-share industry) in Costa Rica vs Cameroon.
  - Average value-added labor productivity growth rate: 1.2 percent per year.
- Robustness: results hold across alternative benchmarks, alternative GII measures, controls for many potential confounders (GDP, population, density, demographics, trade measures, FDI, financial development, institutions, human capital), subsamples, and alternative constructions of industry female shares.

### Policy implications and conclusions (Section 5 summary)
- Evidence indicates gender inequality inhibits economic growth by constraining the productive use of female labor: high-female-share industries grow relatively faster in countries that are more gender equal.
- Methodology: difference-in-differences adaptation using within-country, across-industry variation in emerging market and developing economies in the 1990s (UNIDO data) and country GII (Stotsky and others 2016).
- Robustness: extensive checks address endogeneity, measurement error, reverse causality, omitted variables, and outliers.
- Mechanism emphasized: higher gender equality helps allocate female labor to more productive uses.
- Policy recommendations (implied by empirical channel):
  - Assign high priority to policies that ensure a level playing field for women.
  - Specific policy levers include improving the rule of law and women’s legal rights, women’s health, access to education, financial services, and technology.
- Rationale: such policies are not only matters of human rights, equity, and social justice, but are also relevant levers to boost economic growth—benefiting the economy as a whole.

*Source: 4. Results, wpiea2020119-print-pdf*

### References

### References

### Primary literature and themes
- Core topics covered by the cited literature:
  - Gender inequality, women’s economic empowerment, and gender gaps in education and labor markets.
  - Links between gender inequality and economic growth, productivity, and diversification.
  - Finance, financial inclusion, and the gendered effects of technology and automation on employment.
  - Measurement and indices: Gender Inequality Index (GII), Human Development Indices.
  - Policy-oriented IMF outputs: Staff Discussion Notes and Working Papers on gender and macroeconomic policy.

- Frequently cited institutions and publication types:
  - International Monetary Fund: IMF Working Papers and IMF Staff Discussion Notes (multiple entries, e.g., IMF Working Paper 19/91; IMF Staff Discussion Note 13/10; IMF Staff Discussion Note 18/05; IMF Staff Discussion Note 18/06; IMF Staff Discussion Note 18/07; IMF Staff Discussion Note 20/01).
  - World Bank and UN publications (World Development Report 2012; UNDP Technical Notes).
  - Academic journals: Quarterly Journal of Economics; Journal of Monetary Economics; American Economic Review; Journal of Economic Growth; Journal of Development Economics; Journal of Economic Literature; Feminist Economics; World Development; Journal of Macroeconomics; Journal of Economic Perspectives.

### Data sources, figures, and notes
- Figure 1:
  - Title: Gender inequality and real GDP per capita growth.
  - Source: World Bank WDI database for GDP growth rates in real terms, IMF for GII in 1990.
  - Notes: "This graph plots the average growth rate of real GDP per capita for emerging-market and developing economies over the 1990s and GII in 1990."

- Figure 2:
  - Title: Mean and standard deviation of the share of female labor in total labor within each industry over the 1980s in Sweden.

- Figure 3:
  - Title: Mean and standard deviation of gender inequality index (GII) over the 1990s.
  - Axis/legend text shown: "Gender Inequality Index in the sample of countries 1990-2000" and labels "Standard Deviation" and "Mean".

- Tables:
  - Table 1: "Industries’ female shares (%) in countries with the lowest gender inequality index (GII)". Notes: Reports the average ratio of female employees to total employees in industries in benchmark countries over the corresponding period; periods restricted by availability of UNIDO data.
  - Table 2: "Gender inequality index (GII) by countries in 1990" (Panel A shown).

### Selected industry female shares (Table 1 — sample entries)
- ISIC 15 Food and beverages: Sweden 37.433; Canada 28.686; Ireland 23.738; Austria 39.917; New Zealand 28.340; Australia 29.651.
- ISIC 16 Tobacco products: Sweden 52.734; Canada 36.911; Ireland 40.203; Austria 37.888; New Zealand 42.248; Australia 40.086.
- ISIC 17 Textiles: Sweden 51.413; Canada 44.788; Ireland 44.743; Austria 51.768; New Zealand 45.429; Australia 47.219.
- ISIC 18 Wearing apparel: Sweden 70.639; Canada 75.387; Ireland 78.421; Austria 85.502; New Zealand 77.378; Australia 74.598.
- ISIC 20 Wood products (excl. furniture): Sweden 13.995; Canada 8.098; Ireland 10.894; Austria 17.628; New Zealand 11.640; Australia 12.662.
- ISIC 33 Medical, precision and optical instr.: Sweden 32.110; Canada 36.169; Ireland 56.064; Austria 41.513; New Zealand 35.798; Australia 46.588.
- ISIC 34 Motor vehicles, trailers, semi-trailers: Sweden 16.510; Canada 14.930; Ireland 21.036; Austria 13.803; New Zealand 19.674; Australia 10.834.
- ISIC 36 Furniture; manufacturing n.e.c.: Sweden 30.831; Canada 28.852; Ireland 35.713; Austria 31.422; New Zealand 24.261; Australia 25.319.

### Selected country GII values (Table 2, Panel A — sample entries as listed)
- Hungary 0.308
- China 0.322
- Bulgaria 0.326
- Mongolia 0.379
- Malta 0.382
- Malaysia 0.433
- Romania 0.434
- Sri Lanka 0.462
- Trinidad and Tobago 0.466
- Uruguay 0.468
- Philippines 0.493
- Mauritius 0.506
- Argentina 0.516
- Barbados 0.531
- Mexico 0.529
- Venezuela 0.540
- Peru 0.598
- South Africa 0.611
- Bolivia 0.629
- Turkey 0.627
- Honduras 0.632
- Egypt 0.644
- Senegal 0.651
- Bangladesh 0.669
- Cameroon 0.674
- Nepal 0.710
- India 0.715
- Kenya 0.742
- Iran 0.747
- Belize 0.755
- Niger 0.804

*References section and tables/figures as presented in the source PDF.*

### 0.488 Colombia 0.583 Iraq 0.660 Jordan 0.806

### wpiea2020119-print-pdf - 0.488 Colombia 0.583 Iraq 0.660 Jordan 0.806

### Descriptive statistics (Table 3 — Panel A)
- Value-added growth: Mean 0.022; Median 0.026; Std. dev. 0.172; Min -1; Max 1; Obs 692
- Labor productivity growth: Mean 0.012; Median 0.014; Std. dev. 0.128; Min -0.802; Max 1; Obs 660
- Value-added share: Mean 0.063; Median 0.035; Std. dev. 0.083; Min 0.000; Max 0.615; Obs 692
- Log GDP per cap: Mean 8.193; Median 8.419; Std. dev. 0.779; Min 6.372; Max 9.442; Obs 46
- Log population: Mean 16.268; Median 16.338; Std. dev. 1.910; Min 12.142; Max 20.850; Obs 47
- Log pop. density: Mean 3.841; Median 3.791; Std. dev. 1.485; Min 0.341; Max 7.009; Obs 47
- Female share in popul.: Mean 0.502; Median 0.502; Std. dev. 0.009; Min 0.476; Max 0.539; Obs 47
- Fin. Development: Mean 0.418; Median 0.343; Std. dev. 0.229; Min 0.115; Max 1.270; Obs 45
- Export: Mean 0.302; Median 0.254; Std. dev. 0.181; Min 0.059; Max 0.758; Obs 47
- Import: Mean 0.326; Median 0.282; Std. dev. 0.203; Min 0.046; Max 0.896; Obs 47
- Trade: Mean 0.628; Median 0.563; Std. dev. 0.375; Min 0.150; Max 1.645; Obs 47
- FDI inflows: Mean 0.011; Median 0.009; Std. dev. 0.011; Min -0.010; Max 0.053; Obs 47
- Teacher-stud. ratio: Mean 0.052; Median 0.051; Std. dev. 0.017; Min 0.027; Max 0.104; Obs 39
- Political competition: Mean 5.711; Median 7; Std. dev. 3.455; Min 0; Max 10; Obs 45
- External fin. dep.: Mean -0.089; Median 0.010; Std. dev. 0.309; Min -0.920; Max 0.280; Obs 17
- GII: Mean 0.582; Median 0.598; Std. dev. 0.129; Min 0.308; Max 0.807; Obs 48
- Adolescent fertility rate: Mean 7.918; Median 7.050; Std. dev. 4.660; Min 0.658; Max 22.221; Obs 65
- Relative infant mortality: Mean 0.828; Median 0.823; Std. dev. 0.049; Min 0.685; Max 0.952; Obs 62
- GPI: Mean 0.917; Median 0.973; Std. dev. 0.140; Min 0.525; Max 1.080; Obs 57
- Relative labor part.: Mean 2.225; Median 1.741; Std. dev. 1.527; Min 1.051; Max 8.341; Obs 64
- Women’s rights law score: Mean 0.643; Median 0.649; Std. dev. 0.138; Min 0.357; Max 0.857; Obs 37

### Female shares in benchmark countries (Table 3 — Panel A excerpt)
- Sweden: Mean 31.552; Median 32.110; Std. dev. 15.359; Min 13.995; Max 70.639; Obs 17
- Canada: Mean 28.259; Median 28.686; Std. dev. 17.436; Min 7.473; Max 75.387; Obs 17
- Ireland: Mean 31.711; Median 26.799; Std. dev. 18.062; Min 9.321; Max 78.421; Obs 17
- Austria: Mean 31.081; Median 29.574; Std. dev. 17.976; Min 11.749; Max 85.502; Obs 17
- New Zealand: Mean 28.970; Median 24.261; Std. dev. 16.688; Min 10.232; Max 77.378; Obs 17
- Australia: Mean 28.551; Median 28.017; Std. dev. 17.093; Min 7.605; Max 74.598; Obs 17

### Correlations (Table 3 — Panel B & C)
- Correlations of female shares across benchmark countries (selected):
  - Sweden–Canada: 0.923
  - Sweden–Ireland: 0.820
  - Sweden–Austria: 0.937
  - Sweden–New Zealand: 0.935
  - Sweden–Australia: 0.934
- Pairwise correlations between gender inequality measures (Panel C):
  - GII with Adolescent fertility rate: 0.541***  
  - GII with Relative infant mortality: 0.568***  
  - GII with Relative labor force participation: 0.427***  
  - GII with GPI: -0.572***  
  - GII with Law score: -0.555***  
  - Adolescent fertility with Relative infant mortality: 0.228
  - Adolescent fertility with GPI: -0.408***  
  - Relative infant mortality with GPI: -0.670***  
  - Relative labor force participation with GPI: -0.182***  
  - Law score with GPI: 0.339**

  (Statistical significance: * p<0.10 ** p<0.05 *** p<0.01)

### Industry growth patterns by gender inequality (Table 4)
- Average residual growth of high-female-share industries (1990s; industries above 75th percentile of female share using Sweden benchmark):
  - In high gender inequality countries: 0.4
  - In low gender inequality countries: 1.4
  - Growth differential reported in table: 1.8
- Average residual growth of low-female-share industries (1990s; industries below 25th percentile of female share):
  - In high gender inequality countries: -0.4
  - In low gender inequality countries: -0.4
  - Low-female-share industry growth difference reported: 0.0
- Notes: Residual growth rates obtained after regressing average annual real value-added growth in the 1990s on industry and country fixed effects. Two countries with highest and lowest GII (Morocco and Hungary) dropped to reduce outlier bias.

### Main regression results: Industry value-added growth (Table 5)
- Specification: Dependent variable = average real value-added growth (1990s); interaction = Female share x GII; female shares from benchmark countries.
- Female share x GII coefficients (heteroskedasticity-robust SEs in parentheses):
  - Sweden (1980s): -0.569** (0.259)
  - Canada (1980s): -0.573*** (0.219)
  - Ireland (1988:95): -0.489** (0.203)
  - Austria (1990s): -0.522** (0.230)
  - New Zealand (1990s): -0.582** (0.238)
  - Australia (1980s): -0.627*** (0.233)
- Initial value-added share coefficients: all -0.005 with SE (0.012)
- R square range: 0.409–0.411; Adjusted R square range: 0.348–0.350
- No. of observations: 692 (all specifications)
- Differential growth rate (percent): 1.7, 2.2, 1.7, 1.7, 1.9, 2.0 (by benchmark country)

### Main regression results: Labor productivity growth (Table 6)
- Specification: Dependent variable = real labor productivity growth (value-added per worker) over the 1990s.
- Female share x GII coefficients:
  - Sweden (1980s): -0.446** (0.192)
  - Canada (1980s): -0.411*** (0.151)
  - Ireland (1988:95): -0.367*** (0.142)
  - Austria (1990s): -0.412*** (0.155)
  - New Zealand (1990s): -0.436*** (0.165)
  - Australia (1980s): -0.453*** (0.163)
- Initial value-added share coefficients: 0.009–0.010 with SE (0.007)
- R square range: 0.374–0.376; Adjusted R square range: 0.309–0.311
- No. of observations: 660 (all specifications)
- Differential growth rate (percent): 1.2, 1.4, 1.1, 1.2, 1.3, 1.3

### Alternative gender inequality measures (Table 7)
- Replacement of GII with alternative proxies (using Sweden female shares):
  - Female share x Adolescent fertility rate: -0.024** (0.010); Differential growth rate 2.5; No. obs. 911
  - Female share x Relative infant mortality: -1.538** (0.769); Differential growth rate 1.4; No. obs. 878
  - Female share x Relative labor force participation: -0.037** (0.019); Differential growth rate 0.7; No. obs. 911
  - Female share x Gender parity index (GPI): 0.802** (0.409); Differential growth rate 2.6; No. obs. 807
  - Female share x Women’s rights law score: 0.773** (0.385); Differential growth rate 3.2; No. obs. 537
- Initial value-added share coefficients in these specifications: -0.014 (SE 0.009–0.010) except law score where 0.004 (0.012)
- R square values: 0.365–0.385; Adjusted R square: 0.303–0.323

### Robustness checks (Table 8)
- Baseline: GII constructed by Stotsky and others (2016); female shares from Sweden.
- Main interaction (Female share x GII) remains negative and significant across specifications:
  - Examples: -0.581*** (0.234); -0.598** (0.271); -0.535** (0.270); -0.597** (0.276); -0.546** (0.259); -0.552** (0.259); -0.541** (0.258); -0.573** (0.276)
  - In the split-sample focusing on smaller industries (column 12): Female share x GII = -1.381** (0.541)
- Initial value-added share coefficients remain near -0.004 to -0.007 (SEs ~0.011–0.013)
- No. of observations varies by specification: e.g., 666, 675, 675, 675, 675, 675, 675, 675, 651, 653, 551, 322
- Differential growth rate (percent) across robustness checks: 1.7, 1.7, 1.6, 1.7, 1.6, 1.6, 1.6, 1.7, 1.4, 1.5, 1.7, 3.3

### External finance dependence and financial development (Table 9)
- Tests incorporating external finance dependence (Popov (2014) industry data) and financial development:
  - Female share x GII: -0.623** (0.287) in column (1); -0.575** (0.292) in column (2)
  - Ext. fin. Dep. x GII: 0.150 (0.133) in col (1); -0.045 (0.148) in col (2); -0.051 (0.161) in later cols
  - Female share x fin. dev.: 0.261 (0.202); 0.299 (0.196); 0.281 (0.191) in columns reported
  - Ext fin. Dep. x fin. dev.: 0.031 (0.146); 0.029 (0.145)
- Initial value-added share: -0.004 to -0.007 (SE ~0.012–0.013)
- R square: 0.338–0.409; Adjusted R square: 0.268–0.347
- No. of observations: 692 for some specifications; 651 for others
- Differential growth rate reported: 1.8 (column 2) and 1.5 (column 5)

### Additional robustness checks (Table 10)
- Variants include: dropping Morocco and Hungary; binary high-female-share dummy; quartile categorical female-share variable; mean across six benchmark countries; maximum across five benchmark countries; winsorizing mean growth rates at 1 and 99 percentile.
- Female share x GII coefficients (examples):
  - Column (1): -0.774** (0.303)
  - Column (2) (high-female-share dummy): -0.215*** (0.071)
  - Column (3) (quartiles): -0.063** (0.030)
  - Column (4) (average across six benchmarks): -0.683*** (0.152)
  - Column (5) (maximum across five benchmarks): -0.602** (0.214)
  - Column (6) (winsorized growth): -0.507** (0.233)
- Initial value-added share coefficients: -0.004 to -0.006 (SE ~0.009–0.012)
- R square range: 0.405–0.444; Adjusted R square range: 0.344–0.386
- No. of observations: 658, 692, 692, 692, 692, 692

### Variable definitions (Table A1 — selected)
- Female shares: The average ratio of female employees to total employees in industries in a benchmark country over the corresponding period. Data source: UNIDO, Rev.3
- Value-added growth: The average growth in value-added in an industry over the 1990s. Nominal values in US dollars deflated by PPI in the US. Data source: UNIDO, Rev.3; PPI for the US from IMF’s IFS (PPI for the US, 2010=100, FRED referenced in notes)
- Labor productivity growth: Average growth in value-added–based labor productivity in an industry over the 1990s. Data source: UNIDO, Rev.3
- Value-added share: Industry value-added as share of total manufacturing in 1990. Data source: UNIDO, Rev.3
- External finance dependence: Industry dependence on external financing calculated using US firms in the 1980s. Data source: Popov (2014)
- Log GDP per cap.: Log of GDP per capita (constant US dollars) in 1990. Data source: WB
- Log pop.: Log of population in 1990. Data source: WB
- Log pop. density: Log of population per km2 area in 1990. Data source: WB
- Financial development: Broad money as a share of GDP in 1990. Data source: WB
- Export / Import / Trade: Export and import as shares of GDP in 1990; trade = export + import. Data source: WB
- FDI inflows: Net inflows as a share of GDP in 1990. Data source: WB
- Teacher-student ratio: Teachers per student in secondary education in 1990. Data source: WB
- Political competition: Index coded 0 to 10 (higher = more competition). Data source: Polity IV
- GII: Gender Inequality Index in 1990, constructed by Stotsky and others (2016). Data source: IMF
- Adolescent fertility rate: Births per 100 females ages 15–19 in 1990. Data source: WB
- GPI: Ratio of female students to male students in 1990. Data source: WB
- Relative infant mortality: Ratio of female infant mortality to male infant mortality in 1990. Data source: WB
- Relative labor participation: Ratio of male labor force participation to female labor force participation in 1990. Data source: WB
- Women’s rights law score: Ratio of affirmative legal-rights responses to total codified questions in 1990. Data source: Women, Business and the Law (WBL)

*Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020119-print-pdf.pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020119-print-pdf.pdf_