## wp18105

## Source details

**Canonical URL:** [wp18105](https://www.imf.org/-/media/files/publications/wp/2018/wp18105.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18105.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18105.pdf.json)

---

### I. Scope, data and methodology
- Study sample: 100 countries from 1980 to 2014.
- Focus: impact of fiscal and structural policies on gender inequality:
  - Gender gaps in labor force participation.
  - Gender gaps in education (years of schooling and years in tertiary education).
- Methodological novelty: application of Bayesian Model Averaging (BMA) to identify robust determinants of gender inequality and address model uncertainty.
- Key caveat: results are presented as robust associations rather than definitive causal relationships.
- Data: comprehensive cross-country dataset (1980–2014) drawing on World Development Indicators, Women, Business and the Law, IMF Fiscal Affairs, IMF WEO, United Nations, ILO, Polity, ICRG, Barro-Lee (2013), Botero and others (2004), Campos and Nugent (2012).
- BMA implementation notes:
  - Posterior probability interpretation rule of thumb: < 50% = no evidence; 50%–75% = weak evidence; > 95% = strong evidence.
  - Level of development and its square included in all BMA regressions as basic controls.

### II. Main empirical findings (associations)
- Education gaps — policies and correlates associated with narrower gender gaps:
  - Improved Sanitation Facilities (posterior inclusion probability = 1.00; coefficient -0.017).
  - Public Education Expenditure (posterior inclusion probability = 1.00; coefficient -0.130).
  - Marriage Age Gap (posterior inclusion probability = 1.00; coefficient 0.095) — smaller marriage age gap associated with lower education gaps.
  - log GDP per capita and log GDP per capita squared enter with Posterior Inclusion Probability 1.00 (coefficients -0.504 and 0.038 respectively), indicating a nonlinear relationship with education gaps.
  - Adolescent Fertility Rate appears (posterior inclusion probability = 0.36; coefficient 0.001) and is strongly associated in panel regressions.
  - Other associations in panel regressions: Political Stability and Public Safety, Financial Institutions Access, Public Health Expenditure, Maternal Mortality Ratio, Improved Water Source.
- Labor force participation gaps — policies and correlates associated with narrower gender gaps:
  - Demographics: Fertility Rate (posterior inclusion probability = 1.00; coefficient 4.25) and Marriage Age Gap (posterior inclusion probability = 0.96; coefficient 0.89) — higher fertility and larger marriage age gaps associated with wider labor gaps.
  - Education: Female Avg. Years of Total Schooling (posterior inclusion probability = 0.97; coefficient -1.26) — higher female education associated with lower labor gaps.
  - Legal rights and institutions: Daughter Inheritance Rights (posterior inclusion probability = 1.00; coefficient -4.37) and Female Head of Household (%) (posterior inclusion probability = 1.00; coefficient -9.13); Control of Corruption (posterior inclusion probability = 0.74; coefficient -1.41).
  - Infrastructure: Improved Sanitation Facilities (posterior inclusion probability = 0.99; coefficient -0.68) with an interaction Improved Sanitation Facilities x log GDP per capita (posterior inclusion probability = 0.99; coefficient 0.09) — larger impact at lower GDP per capita. Telephone Subscription Rate and Telephone Subscription Rate x Female Education (posterior inclusion probability = 1.00 for the interaction; coefficient -0.04) show complementarities with female education.
  - Labor Market Protection: Labor Market Protection Index (posterior inclusion probability = 1.00; coefficient -16.93) and Labor Market Protection Index (squared) (posterior inclusion probability = 0.96; coefficient 3.47) — evidence of a nonlinear relationship (initial increases in protection narrow gaps; at higher protection levels further increases can widen gaps).
  - Fiscal: Public Education Expenditure appears with posterior inclusion probability = 0.17 (coefficient -0.07) and in panel regressions public education expenditure is sometimes strongly negative (e.g., -1.257*** in one specification).
- Stylized patterns from panel regressions:
  - Nonlinear development effects on labor gaps: at low development female LFP high; at middle incomes female LFP decreases (gap widens); at higher incomes female LFP rises again (gap narrows).
  - Fertility coefficients in Table 2 examples: 3.923***; 4.149***; 2.279***; 3.231***; 3.142***; 4.620*** (statistically significant and positive).
  - Years of total schooling (female) coefficients in Table 2 examples: -0.758***; -0.541***; -1.658***; -0.779***; -2.438***; -3.926***.

### III. Quantitative robustness (selected coefficients, model fit, samples)
- Education panel regression highlights (dependent variable: Gender Gap in Years of Schooling, Male Minus Female):
  - Selected coefficient examples: Log (GDP per capita) values include -0.327; -1.635***; -1.173***; 7.956**; -0.635*; 13.477***.
  - Log (GDP per capita) squared: 0.010; 0.091***; 0.070***; -0.601***; 0.047**; -0.971***.
  - Adolescent fertility rate: 0.006***; 0.006***; 0.000; 0.056***; 0.003**; 0.057***.
  - Marriage age gap: 0.045**; 0.157***; 0.158***; 0.496***; 0.084***; 0.210.
  - Improved sanitation facilities: -0.016***; -0.014; -0.017***; 0.010.
  - Observations across columns: 1773; 1688; 1656; 1076; 1448; 928.
  - Adjusted R-squared across columns: 0.432; 0.377; 0.429; 0.516; 0.469; 0.561.
- Labor force participation panel regression highlights (Table 2 examples):
  - Labor Market Protection Index: -8.107***; -9.173***; -9.336***; -15.908***.
  - Labor Market Protection Index Sq.: 1.471*; 1.800**; 1.770**; 3.554***.
  - Telephone Subscription Rate: 0.409***; 0.693***; 0.709***.
  - Interaction Telephone Subscription Rate x Female Education: -0.046***; -0.078***; -0.077***.
  - Public education expenditure: -1.257*** (in one specification); Public education expenditure x Emerging Economy: 0.935***.
  - Observations across specifications: 1842; 1769; 1387; 1360; 1360; 1172.
  - Adjusted R-squared across specifications: 0.679; 0.664; 0.746; 0.762; 0.762; 0.781.
- Labor market protection — male/female disaggregation:
  - Labor Market Protection Index effects on female LFP/employment-to-pop ratios: 15.365*** (s.e. 3.282); 20.448*** (s.e. 3.781).
  - Effects on male LFP/employment-to-pop ratios: -1.841 (not significant); 4.484** (s.e. 1.899).
  - Labor Market Protection Index (squared) coefficients: -2.961***; 0.752*; -4.763***; -1.216** (columns (1)–(4)).
- Sub-indices (Annex II, Table A2) — selected regression coefficients on gender gap in labor force participation:
  - Maximum number of working days per week: 8.01*** (s.e. 2.73).
  - Can the workweek extend to 50 hours per week: -5.56 (s.e. 4.78).
  - Are there restrictions on night work: 3.60 (s.e. 2.78).
  - Notice period for redundancy dismissal after 20 yrs of employment: -0.54*** (s.e. 0.18).
  - Priority rules that apply to redundancies: -3.87* (s.e. 2.30).
  - Severance pay for redundancy dismissal after 20 yrs: 0.045 (s.e. 0.031).
  - Observations in these models: 120–121; Adjusted R-squared range: 0.433–0.464.

### IV. Regional decompositions and magnitudes (2005–2010 averages)
- Decomposition uses BMA coefficients and five-year regional averages to split male–female LFP gap into contributions from policies, demographics/health, region-specific factors, and level of development.
- Demographics and health outcomes widened the labor force participation gap in all regions:
  - Widening by as much as 33 percentage points in sub-Saharan Africa.
  - Much smaller widening in North America and Europe and Central Asia (about 10 percentage points).
  - Within demographics, fertility contributed the largest impact, in particular in sub-Saharan Africa (+24½ percentage points).
  - Marriage age gap widened the gap: about 2½ percentage points in North America to about 6 percentage points in sub-Saharan Africa and South Asia.
- Policies narrowed the gap in all regions (magnitudes vary):
  - Policies accounted for about 45–49 percentage points lower labor force participation gaps in sub-Saharan Africa, Middle East and North Africa (MENA) and South Asia.
  - Policies lower the gap by 66 percentage points in North America.
- Regional effects (relative to North America) widened the gender gap in MENA, Latin America and the Caribbean, East Asia and Pacific, and South Asia, but lowered the gap in sub-Saharan Africa.

### V. Policy-relevant conclusions and recommendations
- Overarching methodological point:
  - BMA identifies a smaller set of variables that are robustly related to gender gaps while addressing model uncertainty — useful for policy prioritization.
- Policy takeaways:
  - Prioritize female education across regions; large catch-up potential in sub-Saharan Africa and South Asia.
    - Education-related impacts from decompositions: e.g., narrowing by as much as 47 percentage points in North America; more than 41 percentage points in Europe and Central Asia; more than 35 percentage points in East Asia and the Pacific.
  - Invest in infrastructure, notably improved sanitation facilities in low-income countries (stronger effects at lower GDP per capita).
  - Strengthen legal rights for women (equal inheritance rights; women’s right to head a household) to reduce labor force participation gaps.
  - Implement social policies and campaigns to reduce adolescent fertility and narrow the marriage age gap.
  - Design labor market policies carefully:
    - Strengthening job protection can help where protection is low.
    - Excessive labor market rigidity at high protection levels can widen gender gaps.
  - Increase public spending on education (and health) where feasible; public education expenditure is robustly linked to narrower education gaps and associated with narrower labor gaps in some specifications and regions.
  - Improve institutional environment (control of corruption, lower economic risk) as complementary measures to boost female labor force participation.
- Operational notes for policymakers:
  - Emphasize policies that are robust across many specifications (as identified by high posterior inclusion probabilities in BMA).
  - Consider interactions and nonlinearities (e.g., sanitation x log GDP per capita; labor protection squared) when designing reforms.

*Italic: Source — IMF Working Paper wp18105 (excerpted references, empirical results, Annex I and Annex II).*

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

### wp18105 - References

### I. Introduction and scope
- Study sample: 100 countries from 1980 to 2014.
- Focus: impact of fiscal and structural policies on gender inequality, specifically:
  - Gender gaps in labor force participation.
  - Gender gaps in education (years of schooling and years in tertiary education).
- Methodological novelty: application of Bayesian Model Averaging (BMA) to identify robust determinants of gender inequality and address model uncertainty arising from many potential regressors.
- Key caveat: results are presented as robust associations rather than definitive causal relationships.

### II. Main empirical findings (associations)
- Policies linked to narrower gender gaps in education:
  - Good sanitation facilities are strongly associated with a narrower education gap between males and females in emerging markets and low-income economies.
  - Low adolescent fertility.
  - A narrower marriage age gap (used as a proxy for cultural norms and under-age marriage).
  - Public spending on education.
- Policies linked to narrower gender gaps in labor force participation:
  - Improvements in infrastructure (electricity, better access to water, improved sanitation facilities).
  - Stronger legal rights for females.
  - Low adolescent fertility.
  - A narrower marriage age gap.
  - A stronger institutional environment (measured by the corruption index).
- Labor market protection:
  - Non-linear relationship with labor force participation gaps.
  - At lower levels of labor market protection, stronger protection is associated with narrower labor force participation gaps.
  - At higher levels of protection, excessive labor market rigidity diminishes that effect (a “plateau” exists where changes have minimal effect).
  - Findings align with World Bank (2013): tighter “employment protection legislation and minimum wage can shift employment away from young people, women, the less skilled”.

### III. Literature syntheses and quantitative evidence cited
- Time-allocation models of female labor supply (Becker 1965) underpin much of the theoretical framework.
- Education and wages are important determinants of female employment:
  - Eckstein and Lifshitz (2011): education accounts for one-third of the increase in female employment; wages explain about 20 percent.
- Fertility and marriage effects:
  - Bloom and others (2009): each birth on average decreases women’s labor supply by almost two years during a woman’s reproductive life (panel of 97 countries).
  - Mishra and Smyth (2010): a 1 percent increase in the fertility rate results in a 0.4 percent decrease in female labor force participation rates in G7 countries.
  - De Laat and Sevilla-Sanz (2011): cross-country positive relationship between fertility and female labor force participation when men’s contribution to home production is higher.
  - Field and Ambrus (2008): each additional year that marriage is delayed is associated with about one-fifth additional year of schooling and 5.6 percent higher literacy for women (rural Bangladesh).
  - Stimple and Stadelmann (2016): an additional age difference between husband and wife of one year reduces female secondary schooling completion rates by 14 percentage points.
- Fiscal and expenditure policy impacts:
  - Tax credits or benefits for low-wage earners can stimulate labor force participation.
  - Switching from family income taxation to individual income taxation can support female labor force participation (Elborgh-Woytek and others 2013).
  - Child care price elasticity of female labor supply estimated to range from –0.13 to –0.2:
    - Reducing the price of childcare by 50 percent could be associated with an increase of 6.5 to 10 percent in the labor supply of young mothers.
  - Infrastructure and education spending:
    - Das et al. (2015): female labor force participation in India would rise by 2 percentage points if Indian states increased education spending by 1 percent of GDP.
    - Norando (2010): differences in basic infrastructure explain much of the cross-country variation in female labor force participation.
    - Lamb (2015): collecting water is estimated to take as much as 26 percent of women’s time in rural Africa.
- Legal restrictions:
  - Gonzales and others (2015) and World Bank (2013): restrictions on women’s rights to inheritance and property and legal impediments to economic activities are strongly associated with larger gender gaps in labor force participation.

### IV. Methodological remarks and motivation for BMA
- Large number of proposed determinants in the literature creates:
  - Difficulty identifying the most strongly related variables.
  - Large model uncertainty given constrained sample sizes in macro studies.
  - Need for prioritization of policy actions that provide the best “bang for the buck”.
- Bayesian Model Averaging (BMA) is used to address model uncertainty and identify the most robust determinants of gender gaps.
- BMA has precedent in macroeconomic literature (Sala-i-Martin et al. 2004; Jones and Schneider 2006; Moral-Benito 2012; Du et al. 2013) and is applied here for the first time to study determinants of gender inequality.

### V. Stylized facts from preliminary data analysis
- Aggregate observations across country groups:
  - Some gender gaps are small in the global sample, but disaggregation shows variation:
    - Education gaps are closed in advanced markets but persist in EMDEs (emerging markets and developing economies).
- Correlations highlighted:
  - Higher public spending on education is associated with a narrower gender gap in years of schooling in both developed and developing countries; relationship appears stronger in developing countries.
  - Improved access to sanitation facilities is negatively correlated with the gender gap in years of schooling in emerging markets (effect not evident in advanced economies).
  - Improved access to finance is associated with a narrower gender gap in tertiary enrollment, but not in secondary or primary enrollment.

*Source: wp18105 - References (IMF working paper content provided).*

### 17. Similarly, we find evidence for a range of possible determinants of gender gaps

### 17. Similarly, we find evidence for a range of possible determinants of gender gaps

### Determinants of labor force participation gaps (stylized facts)
- More equal legal rights are associated with narrower labor force participation gaps (Figure 2, panel 1).
- Higher public spending on education is associated with lower labor force participation gaps; effect mainly driven by advanced economies (Figure 2, panel 2).
- The impact of labor market regulation on labor force participation gaps is ambiguous: some relationship in advanced economies, less so in emerging and developing economies (Figure 2, panel 3).

### Empirical specification (fixed-effect panel model)
- Estimated relationship:
  - Gap_{k,i,t} = α + δPolicy_{k,i,t} + βX_{k,i,t} + μ_k + ν_t + ε_{k,i,t}
- Definitions:
  - Gap_{k,i,t}: gender gap in either labor force participation or education in country i in region k in year t.
  - Policy_{k,i,t}: includes
    - Fiscal policy: public spending on education and on health (tax rate on secondary earners and income vs. family taxation not available for most EMDEs).
    - Infrastructure: access to sanitation facilities, access to electricity, access to improved water source, telephone subscription rates.
    - Structural policies: labor market protection index, trade openness, political risk rating, control of corruption.
    - Legal variables from World Bank’s Women, Business and the Law: equal inheritance rights for daughters and sons, women’s right to head a household, guaranteed equality before the law.
  - X_{k,i,t}: controls including Log(GDP per capita) and its square, fertility and neonatal mortality rates, marriage age gap and absolute age at marriage.
  - μ_k: region fixed effects; ν_t: year fixed effects.
- Estimation approaches: ordinary least squares and Bayesian Model Averaging (BMA). Specification yields associations not causal statements.

### Results — Gender gaps in education (high-level findings)
- Development and demographics explain variation in education gaps; scope for policy action is large.
- Nonlinear relationship between gender gaps and GDP per capita:
  - At higher income levels, increases in GDP per capita are associated with narrower gender gaps in tertiary education (columns (4) and (6) of Table 1).
- Fertility and marriage gaps:
  - Higher adolescent fertility rates are strongly associated with wider gender gaps in education.
  - Larger marriage age gap is associated with larger gender gaps in education.
- Institutions and infrastructure:
  - Political stability and public safety indices link to narrower gender gaps in education.
  - Improved sanitation facilities are associated with narrower gaps in years of schooling.
- Financial and fiscal:
  - Higher financial development associated with lower gender gaps in tertiary education.
  - Higher public spending on education and health associated with lower gender gaps in tertiary education.

- Selected coefficients and statistics from Table 1 (dependent variable: Gender Gap in Years of Schooling, Male Minus Female):
  - Log (GDP per capita): values include -0.327; -1.635***; -1.173***; 7.956**; -0.635*; 13.477*** (standard errors reported in parentheses).
  - Log (GDP per capita) squared: 0.010; 0.091***; 0.070***; -0.601***; 0.047**; -0.971***.
  - Maternal mortality ratio: 0.001***; 0.000**; 0.000; -0.010***; 0.000; -0.001.
  - Adolescent fertility rate: 0.006***; 0.006***; 0.000; 0.056***; 0.003**; 0.057***.
  - Marriage age gap: 0.045**; 0.157***; 0.158***; 0.496***; 0.084***; 0.210.
  - Political Stability and Public Safety: -0.123***; -0.125***; -3.440***; -0.058**; -2.426***.
  - Improved sanitation facilities: -0.016***; -0.014; -0.017***; 0.010.
  - Financial institutions access: -8.489***; 0.092; -5.430**.
  - Public education expenditure: -0.130***; -0.202*.
  - Public health expenditure: 0.003; -1.519***.
  - Observations: 1773; 1688; 1656; 1076; 1448; 928.
  - Adjusted R-squared: 0.432; 0.377; 0.429; 0.516; 0.469; 0.561.
  - Note: Columns (4) and (6) dependent variable is gender gap in tertiary school enrollment rate (male – female).

### Results — Gender gaps in labor force participation (high-level findings)
- Development stage effects:
  - Nonlinear relationship: at low development female LFP high; at middle incomes female LFP decreases (gender gap widens); at higher incomes female LFP rises again (gender gap narrows).
- Health, demographics, and social factors:
  - Neonatal mortality rate, fertility rate, and marriage age gap are significantly related to labor force participation gaps.
  - Fertility rate coefficients strongly positive: e.g., 3.923***; 4.149***; 2.279***; 3.231***; 3.142***; 4.620*** (Table 2).
  - Marriage age gap coefficients positive: 2.306***; 2.498***; 1.036***; 0.572*; 0.526; 1.165***.
- Education:
  - Years of total schooling female associated with lower gender gaps in LFP: -0.758***; -0.541***; -1.658***; -0.779***; -2.438***; -3.926***.
- Institutions and legal equality:
  - Equal inheritance rights for daughters strongly related to lower labor force participation gaps (e.g., -5.734***; -9.528***; -8.446***; -8.848***; -4.702***).
  - Women’s right to head of household coefficients: -8.926***; -9.846***; -10.338***; -9.681***; -7.238***.
  - Control of Corruption: -0.923**; -1.767***; -1.463***; -1.333***; -0.071.
- Structural policies:
  - Economic Risk Rating: -0.214***; -0.152***; -0.127**; -0.101.
  - Labor Market Protection Index: negative linear term and positive squared term indicating nonlinearity:
    - Labor Market Protection Index: -8.107***; -9.173***; -9.336***; -15.908***.
    - Labor Market Protection Index Sq.: 1.471*; 1.800**; 1.770**; 3.554***.
  - Interpretation: at lower protection levels, increases narrow gender gaps; at higher protection levels, increases widen gender gaps.
- Infrastructure:
  - Telephone Subscription Rate: 0.409***; 0.693***; 0.709***.
  - Interaction Telephone Subscription Rate x Female Education: -0.046***; -0.078***; -0.077***.
  - Improved sanitation facilities: -0.375*; -0.043; -0.286***.
  - Interaction Improved sanitation facilities x log GDP per capita: 0.064***; 0.028***; 0.044***.
- Financial access:
  - Financial institutions access: 1.268; -5.850 (coefficients in some specifications).
  - Financial institutions access x education: 0.196; 0.851.
- Fiscal policies:
  - Public education expenditure: -1.257*** (in one specification).
  - Public education expenditure x Emerging Economy: 0.935***.
  - Public health expenditure: 0.292 (in that specification).
- Model fit and sample:
  - Observations across specifications: 1842; 1769; 1387; 1360; 1360; 1172.
  - Adjusted R-squared: 0.679; 0.664; 0.746; 0.762; 0.762; 0.781.

### Labor market protection — disentangling effects on men and women
- Regressions of male and female labor force participation rates on determinants show:
  - Labor Market Protection Index increases female LFP rate: 15.365*** (standard error 3.282) in column (1).
  - Labor Market Protection Index effect on male LFP rate: -1.841 (not statistically significant) in column (2).
  - Labor Market Protection Index on female employment-to-pop ratio: 20.448*** (standard error 3.781) in column (3).
  - Labor Market Protection Index on male employment-to-pop ratio: 4.484** (standard error 1.899) in column (4).
  - Labor Market Protection Index (squared) coefficients: -2.961***; 0.752*; -4.763***; -1.216** (columns (1)–(4) respectively).
- Interpretation:
  - Stronger labor protection significantly increases female labor force participation and employment-to-population ratios.
  - Some (weaker) evidence that stronger labor protection lowers male labor force participation.
  - Sub-index evidence (Annex II): larger maximum number of working days per week associated with a wider gender gap; longer notice period for redundancy dismissal narrows the gender gap.

### Robustness and model uncertainty — Bayesian Model Averaging (BMA)
- Motivation:
  - Many potential determinants tested; policymakers need clarity on which are most fundamental and robust.
  - Classical approaches struggle with model uncertainty; BMA addresses this by averaging across models.
- BMA approach:
  - Averages across a large set of models given priors; each model receives a weight based on the Bayesian information criterion (BIC/SBIC) and choice of priors.
  - Posterior probability interpretation rule of thumb: < 50% = no evidence; 50%–75% = weak evidence; > 95% = strong evidence.
  - Level of development and its square are included in all BMA regressions as basic controls.

*Italic: Source — IMF Working Paper wp18105 (excerpt: section 17 and surrounding material).*

### 26. The results highlight a smaller set of variables that is robustly related with lower

### 26. The results highlight a smaller set of variables that is robustly related with lower

### Robust determinants of gender gaps in education
- Key variables robustly associated with lower education gaps (posterior inclusion probability = 100 percent):
  - Improved Sanitation Facilities
  - Public Education Expenditure
  - Marriage Age Gap (lower age-gap between men and women at marriage associated with lower gaps)
- Other associations reported in the BMA results (Gender Gap in Years of Schooling, in Years, Male Minus Female):
  - log GDP per capita: Coefficient -0.504; Posterior Inclusion Probability 1.00
  - log GDP per capita squared: Coefficient 0.038; Posterior Inclusion Probability 1.00
  - Improved Sanitation Facilities: Coefficient -0.017; Posterior Inclusion Probability 1.00
  - Public Education Expenditure: Coefficient -0.130; Posterior Inclusion Probability 1.00
  - Marriage Age Gap: Coefficient 0.095; Posterior Inclusion Probability 1.00
  - Adolescent Fertility Rate: Coefficient 0.001; Posterior Inclusion Probability 0.36
  - Female Head of Household (%): Coefficient 0.003; Posterior Inclusion Probability 0.05
  - Maternal Mortality Ratio: Coefficient 0.000; Posterior Inclusion Probability 0.05
  - Improved Water Source: Coefficient 0.000; Posterior Inclusion Probability 0.04
  - Financial Institutions Access: Coefficient 0.005; Posterior Inclusion Probability 0.04
  - Political Risk Rating: Coefficient 0.000; Posterior Inclusion Probability 0.03
  - Trade Openness: Coefficient 0.000; Posterior Inclusion Probability 0.03
  - Daughter Inheritance Rights: Coefficient 0.001; Posterior Inclusion Probability 0.03
  - Public Health Expenditure: Coefficient 0.000; Posterior Inclusion Probability 0.03
- Sample size for education BMA:
  - Number of countries = 120
  - Number of observations = 1468
- Note: posterior probabilities of < 50%, 50% - 75% and > 95% are interpreted as no evidence, weak evidence and strong evidence of an effect, respectively.

### Robust determinants of gender gaps in labor force participation
- Numerous country characteristics and policies are robustly related to gender gaps in labor force participation:
  - Demographics
    - Higher fertility rates and wider age gaps at marriage are associated with wider gender gaps in the labor market (posterior inclusion probability > 95 percent).
    - Equal minimum ages by law for men and women could help decrease the gap.
  - Education
    - Higher levels of female education are strongly associated with lower labor force participation gaps.
  - Legal rights
    - Equal legal rights are associated with narrower gaps in labor force participation.
    - Equal inheritance rights for daughters and sons, and women’s right to be head of a household, enter with a posterior inclusion probability of 100 percent and are economically significant.
  - Infrastructure
    - Improved sanitation facilities decrease the gap, particularly at early stages of development; the impact moderates as GDP per capita increases.
    - Telephone subscription (landline) rates are related to narrower gaps in countries where female education is on average higher (interaction effects).
    - Empirical example: access to a cellphone for all women, and access to a computer for married women, is significantly related to higher female labor force participation in Costa Rica (Ivanova, Makioka, and Wong, 2017).
  - Labor market protection
    - Stronger labor market protection is associated with lower labor force participation gaps at lower levels of protection; the marginal impact declines as protection strengthens.
    - At higher levels of protection, increases in protection are associated with higher gender gaps in labor force participation.
  - Institutional environment
    - Better control of corruption and lower economic risk are weakly associated with narrower labor force participation gaps.
- BMA results (Determinant of Gender Gap in Labor Force Participation, in Percent, Male Minus Female):
  - log GDP per capita: Coefficient 58.62; Posterior Inclusion Probability 1.00
  - log GDP per capita squared: Coefficient -3.36; Posterior Inclusion Probability 1.00
  - Fertility Rate: Coefficient 4.25; Posterior Inclusion Probability 1.00
  - Daughter Inheritance Rights: Coefficient -4.37; Posterior Inclusion Probability 1.00
  - Female Head of Household (%): Coefficient -9.13; Posterior Inclusion Probability 1.00
  - Labor Market Protection Index: Coefficient -16.93; Posterior Inclusion Probability 1.00
  - Telephone Subscription Rate x Female Education: Coefficient -0.04; Posterior Inclusion Probability 1.00
  - Improved Sanitation Facilities: Coefficient -0.68; Posterior Inclusion Probability 0.99
  - Improved Sanitation Facilities x log GDP per capita: Coefficient 0.09; Posterior Inclusion Probability 0.99
  - Female Avg. Years of Total Schooling: Coefficient -1.26; Posterior Inclusion Probability 0.97
  - Labor Market Protection Index (squared): Coefficient 3.47; Posterior Inclusion Probability 0.96
  - Marriage Age Gap: Coefficient 0.89; Posterior Inclusion Probability 0.96
  - Telephone Subscription Rate: Coefficient 0.31; Posterior Inclusion Probability 0.88
  - Control of Corruption: Coefficient -1.41; Posterior Inclusion Probability 0.74
  - Economic Risk Rating (higher means lower risks): Coefficient -0.10; Posterior Inclusion Probability 0.56
  - Personal Income Tax Rate: Coefficient 0.03; Posterior Inclusion Probability 0.48
  - Neonatal Mortality Rate: Coefficient 0.05; Posterior Inclusion Probability 0.31
  - Political Risk Rating (higher means lower risks): Coefficient 0.02; Posterior Inclusion Probability 0.23
  - Public Education Expenditure: Coefficient -0.07; Posterior Inclusion Probability 0.17
  - Financial Institutions Access x female Education: Coefficient 0.04; Posterior Inclusion Probability 0.05
  - Financial Institutions Access: Coefficient -0.40; Posterior Inclusion Probability 0.05
- Sample size for labor force BMA:
  - Number of countries = 96
  - Number of observations = 1126
- Note: posterior probabilities of < 50%, 50% - 75% and > 95% are interpreted as no evidence, weak evidence and strong evidence of an effect, respectively.

### Quantifying policy and structural effects across regions (2005-2010)
- Decomposition approach:
  - Uses BMA-derived coefficients and five-year regional averages to decompose the male-to-female labor force participation gap into impacts from:
    - policies (institutions, structural policies, infrastructure, fiscal expenditure)
    - demographics and health outcomes (neonatal mortality rate, fertility rate, marriage age gap)
    - region-specific factors (region FE) and level of development
- Main regional findings:
  - Demographics and health outcomes have widened the labor force participation gap in all regions.
    - Widening by as much as 33 percentage points in sub-Saharan Africa.
    - Much smaller effects in North America and Europe and Central Asia (about 10 percentage points).
    - Within demographics, fertility contributed the largest impact, in particular in sub-Saharan Africa (+24½ percentage points).
    - The marriage age gap widened the gap from about 2½ percentage points in North America to about 6 percentage points in sub-Saharan Africa and South Asia.
  - Policies have narrowed the gap in all regions, with variation in magnitude:
    - Policies accounted for about 45-49 percentage points lower labor force participation gaps in sub-Saharan Africa, Middle East and North Africa (MENA) and South Asia.
    - Policies lower the gap by 66 percentage points in North America.
  - Regional effects (compared to North America as the base region) widened the gender gap in MENA, Latin America and the Caribbean, East Asia and Pacific, and South Asia, but lowered the gap in sub-Saharan Africa.

### Which policies deliver the largest effects by region
- Education
  - Educating girls has narrowed the gap substantially in all regions:
    - Narrowed by as much as 47 percentage points in North America.
    - More than 41 percentage points in Europe and Central Asia.
    - More than 35 percentage points in East Asia and the Pacific.
    - High potential to lower gaps in South Asia and sub-Saharan Africa.
- Institutions (including legal rights)
  - Contributed more than 10 percentage points to narrowing the gap in all regions except MENA.
- Infrastructure
  - Enhancing sanitation facilities can play a large role at lower levels of development, with strong potential in South Asia and sub-Saharan Africa.
- Public expenditure
  - Expenditures on health and education have a direct negative impact on the labor force participation gap and influence the gap indirectly through better educational outcomes.

### Conclusions and policy implications
- Methodological contribution:
  - Application of Bayesian Model Averaging (BMA) to gender gaps in education and labor force participation highlights determinants that are robust across many model specifications and addresses model uncertainty.
- Policy-relevant findings:
  - Education gaps:
    - Adequate sanitation facilities and narrower marriage age gaps are robustly associated with lower gender gaps in education.
    - Low adolescent fertility, stable and safe environments (political stability and public safety), higher financial development, and higher public spending on education and health are associated with narrower gender gaps in tertiary education.
  - Labor force participation gaps:
    - Adequate infrastructure—especially improved sanitation facilities—strong legal rights for females, low adolescent fertility, and a small marriage age gap are robustly associated with narrower labor force participation gaps.
    - Labor market protection has a non-linear relationship: increases in protection narrow gaps when initial protection is low, but further increases at high protection levels may widen gaps.
    - Higher public expenditure on education is associated with narrower labor force participation gaps in advanced economies.
    - Stronger institutional environments (better control of corruption and lower economic risk) provide additional evidence of boosting female labor force participation.
- Policy recommendations:
  - Prioritize female education across regions, with particular focus on sub-Saharan Africa and South Asia for catch-up potential.
  - Invest in infrastructure, notably sanitation facilities in low-income countries.
  - Implement social policies and campaigns to reduce adolescent fertility and narrow the marriage age gap.
  - Design labor market policies carefully: strengthening job protection can help where protection is low, but excessive rigidity can widen gender gaps.

*Source: IMF Working Paper (content unit: "26. The results highlight a smaller set of variables that is robustly related with lower")*

### ANNEX I. DATA

### ANNEX I. DATA

### Data construction
- Comprehensive cross-country dataset of economic, political, institutional, and social variables for the period 1980-2014.
- Data sources include: World Development Indicators, Women, Business and the Law, IMF Fiscal Affairs, IMF WEO, United Nations, International Labor Organization, Polity, International Country Risk Guide, Barro-Lee (2013), Botero and others (2004), and Campos and Nugent (2012).
- Selection criterion: choose the most standard and widely accepted data source for each policy variable to maximize comparability across countries.
- Table A1 provides summary statistics and data sources for the policy variables used.

### Table A1. Summary statistics (selected variables and sources)
- Gender Gaps
  - Labor Force Participation Gap (M - F): Advanced Markets — No. of Obs 792; Mean 17.0; Std. Dev 9.0. Emerging and Developing Markets — No. of Obs 3408; Mean 25.7; Std. Dev 18.2. Source: World Bank, WDI
  - Literacy Gap (M - F): Advanced Markets — No. of Obs 248; Mean 3.5; Std. Dev 4.1. Emerging and Developing Markets — No. of Obs 2924; Mean 10.4; Std. Dev 10.3. Source: World Bank, WDI
  - Primary School Enrollment Gap (M - F): Advanced Markets — No. of Obs 1256; Mean 0.5; Std. Dev 1.8. Emerging and Developing Markets — No. of Obs 5513; Mean 8.8; Std. Dev 12.4. Source: World Bank, WDI
  - Secondary School Enrollment Gap (M - F): Advanced Markets — No. of Obs 1218; Mean -0.6; Std. Dev 6.8. Emerging and Developing Markets — No. of Obs 5138; Mean 3.1; Std. Dev 9.8. Source: World Bank, WDI
  - Tertiary School Enrollment Gap (M - F): Advanced Markets — No. of Obs 1238; Mean -2.8; Std. Dev 13.3. Emerging and Developing Markets — No. of Obs 4323; Mean -1.0; Std. Dev 8.0. Source: World Bank, WDI
  - Years of Schooling Gap: Advanced Markets — No. of Obs 1683; Mean -3.3; Std. Dev 6.9. Emerging and Developing Markets — No. of Obs 5559; Mean 5.2; Std. Dev 9.0. Source: Barro and Lee (2013)
- Access to Finance
  - Financial Institutions Access: Advanced Markets — No. of Obs 1122; Mean 0.4; Std. Dev 0.4. Emerging and Developing Markets — No. of Obs 4828; Mean 0.1; Std. Dev 0.2. Source: World Bank, FINDEX
- Institutions
  - Daughter Inheritance Rights: Advanced Markets — No. of Obs 1872; Mean 1.0; Std. Dev 0.1. Emerging and Developing Markets — No. of Obs 8117; Mean 0.7; Std. Dev 0.5. Source: World Bank, WBL
  - Female Head of Household: Advanced Markets — No. of Obs 1872; Mean 0.9; Std. Dev 0.2. Emerging and Developing Markets — No. of Obs 8117; Mean 0.7; Std. Dev 0.5. Source: World Bank, WBL
  - Guaranteed Equality by Law: No. of Obs 3079; Mean 0.9; Std. Dev 0.3. Source: World Bank, WBL
  - Political Risk Rating: Advanced Markets — No. of Obs 984; Mean 80.7; Std. Dev 8.9. Emerging and Developing Markets — No. of Obs 3896; Mean 57.3; Std. Dev 11.6. Source: ICRG
  - Government Effectiveness: Advanced Markets — No. of Obs 627; Mean 1.5; Std. Dev 0.5. Emerging and Developing Markets — No. of Obs 2837; Mean -0.4; Std. Dev 0.7. Source: World Bank, WGI
  - Regulatory Quality: Advanced Markets — No. of Obs 627; Mean 1.4; Std. Dev 0.4. Emerging and Developing Markets — No. of Obs 2839; Mean -0.4; Std. Dev 0.7. Source: World Bank, WGI
  - Rule of Law: Advanced Markets — No. of Obs 646; Mean 1.4; Std. Dev 0.5. Emerging and Developing Markets — No. of Obs 2877; Mean -0.4; Std. Dev 0.7. Source: World Bank, WGI
  - Control of Corruption: Advanced Markets — No. of Obs 627; Mean 1.5; Std. Dev 0.7. Emerging and Developing Markets — No. of Obs 2839; Mean -0.4; Std. Dev 0.7. Source: World Bank, WGI
- Demographics
  - Female Avg. Years of Total Schooling: No. of Obs 4402; Mean 6.5; Std. Dev 3.2. Source: World Bank, DHS
  - Contraceptive Prevalence: Advanced Markets — No. of Obs 628; Mean 72.9; Std. Dev 9.0. Emerging and Developing Markets — No. of Obs 3284; Mean 39.9; Std. Dev 21.5. Source: World Bank, WDI
  - Maternal Mortality Ratio: Advanced Markets — No. of Obs 768; Mean 9.7; Std. Dev 5.8. Emerging and Developing Markets — No. of Obs 3480; Mean 297.1; Std. Dev 357.8. Source: World Bank, WDI
  - Adolescent Fertility Rate: Advanced Markets — No. of Obs 637; Mean 16.8; Std. Dev 10.5. Emerging and Developing Markets — No. of Obs 1996; Mean 78.9; Std. Dev 29.5. Source: United Nations
  - Marriage Age Gap: Advanced Markets — No. of Obs 1215; Mean 2.7; Std. Dev 0.7. Emerging and Developing Markets — No. of Obs 4368; Mean 4.2; Std. Dev 1.8. Source: United Nations
- Infrastructure
  - Telephone Subscription Rate: Advanced Markets — No. of Obs 1783; Mean 34.7; Std. Dev 18.1. Emerging and Developing Markets — No. of Obs 7141; Mean 6.6; Std. Dev 9.2. Source: World Bank, WDI
  - Access to Electricity: Advanced Markets — No. of Obs 714; Mean 98.7; Std. Dev 5.4. Emerging and Developing Markets — No. of Obs 3192; Mean 65.3; Std. Dev 33.4. Source: World Bank, WDI
  - Improved Water Source: Advanced Markets — No. of Obs 735; Mean 99.6; Std. Dev 1.2. Emerging and Developing Markets — No. of Obs 3328; Mean 79.6; Std. Dev 18.8. Source: World Bank, WDI
  - Improved Sanitation Facilities: Advanced Markets — No. of Obs 690; Mean 99.6; Std. Dev 1.1. Emerging and Developing Markets — No. of Obs 3276; Mean 60.8; Std. Dev 30.6. Source: World Bank, WDI
- Fiscal Policies
  - Public Education Expenditure: Advanced Markets — No. of Obs 961; Mean 5.0; Std. Dev 1.3. Emerging and Developing Markets — No. of Obs 2829; Mean 4.4; Std. Dev 2.7. Source: World Bank, WDI
  - Public Health Expenditure: Advanced Markets — No. of Obs 594; Mean 6.1; Std. Dev 1.7. Emerging and Developing Markets — No. of Obs 2635; Mean 3.2; Std. Dev 2.1. Source: World Bank, WDI
- Structural Policies
  - Employment Protection Index: Advanced Markets — No. of Obs 1062; Mean 1.5; Std. Dev 0.7. Emerging and Developing Markets — No. of Obs 3571; Mean 1.6; Std. Dev 0.4. Source: Campos and Nugent (2012)
  - Trade Openness: Advanced Markets — No. of Obs 1075; Mean 95.0; Std. Dev 76.2. Emerging and Developing Markets — No. of Obs 4718; Mean 79.3; Std. Dev 45.7. Source: IMF, WEO

*Italic: Source — wp18105 - ANNEX I. DATA (PDF).*

### ANNEX II. LABOR MARKET PROTECTION INDEX

### Labor market protection indices used
- Baseline index: Employment Protection Legislation Index from Campos and Nugent (2012), consisting of four components:
  1. Conditions of work (hours of work, weekly rest and paid leave)
  2. Employment security termination of employment
  3. Conditions of employment (labor contracts, wages and personnel management)
  4. General provisions (labor codes, general labor and employment acts)
- Robustness checks use:
  - Employment Laws Index from Botero, Djankov, La Porta, Lopez de Silanes and Shleifer (QJE 2004)
  - World Bank Doing Business Index and its sub-indices (adopts BDLLS methodology and extends coverage)

### World Bank Doing Business sub-components (index consists of)
  1. Cost of increasing hours worked
  2. Cost of firing workers
  3. Dismissal procedures
  4. Alternative employment contracts

### Table A2. Sub-Indices of Labor Market Protection — selected regression results
- Dependent variable: gender gap in labor force participation
- Coefficients (standard errors in parentheses). Models (1) through (8):

  - Can the workweek for a single worker extend to 50 hours per week: -5.56 (4.78)
  - Are there restrictions on night work: 3.60 (2.78)
  - Are there restrictions on "weekly holiday" work: -2.92 (2.86)
  - What is the maximum number of working days per week: 8.01*** (2.73)
  - Any retraining or reassignment obligation before making a worker redundant: -4.04 (3.12)
  - Any priority rules that apply to redundancy dismissals or lay-offs: -3.87* (2.30)
  - Notice period for redundancy dismissal after 20 yrs of employment: -0.54*** (0.18)
  - Severance pay for redundancy dismissal after 20 yrs of employment: 0.045 (0.031)

- No. Observations:
  - Models (1)–(3): 121
  - Model (4): 120
  - Models (5)–(8): 120
- Adjusted R-squared:
  - (1) 0.437
  - (2) 0.435
  - (3) 0.433
  - (4) 0.458
  - (5) 0.439
  - (6) 0.441
  - (7) 0.464
  - (8) 0.441

- Additional model information:
  - All models include GDP per capita, region and time fixed effects.
  - Robust standard errors in parentheses.
  - Significance notation: * p < 0.10     ** p < 0.05     *** p < 0.01

*Italic: Source — wp18105 - ANNEX I. DATA (PDF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18105.pdf_