## wpiea2024117-print-pdf

## Source details

**Canonical URL:** [wpiea2024117-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024117-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024117-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024117-print-pdf.pdf.json)

---

### III. Introduction — motivation and key summary findings
- Motivation and data:
  - To understand gender gaps in St. Lucia where macro-data show high female participation rates (ECCU female participation rate 71 percent) but anthropological/sociological evidence indicates a matrifocal but non-matriarchal family structure.
  - Data: St. Lucia Labor Force Survey covering 2016Q1-2021Q4 (rich micro dataset).
  - Methods: probit regression with Blinder-Oaxaca decomposition, Mincer earnings regression, analyses across age-groups, education groups, and family characteristics (e.g., presence of young children).
- High-level descriptive context:
  - ECCU female participation rate: 71 percent.
  - World average female participation rate: 53 percent.
  - OECD average female participation rate: 65 percent.
  - Participation gap between men and women in ECCU: 10 percentage points.
  - Participation gaps: OECD average 16 percentage points, LAC average 23 percentage points, world average 26 percentage points.
- Principal econometric findings:
  - Controlling for observables, the participation rate should be 1 percent higher for women than men.
  - Observed overall participation rate gap: 7 percent (women participate 7 percent less than men).
  - Unexplained gap in participation: women are 8 percent less likely to be in the labor market than male workers with the same observable characteristics.
  - Childcare effects:
    - For women: having a young child is associated with lower probability of participation and higher probability of unemployment.
    - For men: having a young child is associated with no difference in probability of participation and lower probability of unemployment.
  - Participation gaps decline with education:
    - Unexplained gap by education level: primary education 13 percentage points; secondary education 6 percentage points; post-secondary 2 percentage points; tertiary 3 percentage points.
  - Labor income gaps:
    - Female workers earn on average 18 percent less than male workers with the same observable characteristics.
    - Unexplained gender gap in labor income by education level: primary education 21 percent; tertiary education 2 percent.
    - Female workers obtain larger gains from education relative to male counterparts.

### II. Data — scope, construction, and key sample facts
- Source and scope:
  - St. Lucia Labor Force Survey (SLLFS) 2016Q1-2021Q4, run by the Central Statistical Office and the Ministry of Finance, Economic Development and the Youth Economy.
  - Working-age population aged 15-64; about 30,000 observations, with about 9,000 reporting labor income.
- Sample construction and weights:
  - Data pooled across years; weights adjusted so each year is represented equally (weight of each observation divided by year-specific sum and multiplied by 100,000 so adjusted weights sum to 100,000 each year).
  - Standard errors clustered at year × quarter × enumeration district level.
- Key variable definitions and measurement notes:
  - Unemployment: broader ILO-recommended definition for small islands.
  - Participation: participates if unemployed, working for pay, or temporarily absent from work during the reference week.
  - Gross monthly income from all employment: available 2016Q1 to 2020Q3; non-missing for about 60 percent of employed people.
    - Among the employed, percentage missing the income variable: 43 percent of men and 39 percent of women.
  - Income reported in 8 brackets (EC$): under 200; 201-399; 400-799; 800-1199; 1200-1999; 2000-3999; 4000-5999; over 6000.
  - Mincer regressions use log of midpoint of income bracket as proxy for log gross monthly income (measurement error caveat).

### III. Descriptive statistics — education, test scores, age profiles, and income distributions
- Education attainment (age 15-64):
  - Women: 21 percent have attained higher education; 79 percent have not.
  - Men: 14 percent have attained higher education; 86 percent have not.
  - Education categories: none or primary; lower or upper secondary; post-secondary (diploma or associate degree); tertiary (university).
- Test scores:
  - Female students on average have higher test scores than male students in St. Lucia and other ECCU countries (Harmonized Test Scores; scale 300 to 625).
- Age profiles and gender gaps by education:
  - Workers with higher education:
    - Gender gap in participation: 2 percentage points in early thirties, rising to 13 percentage points in early sixties.
    - Little gender gap in unemployment across age groups except under 25.
  - Workers without higher education:
    - Gender gap in participation: 9 percentage points in early thirties, rising to 24 percentage points in early sixties.
    - Example age 30-34:
      - Unemployment rate: 1 percentage point higher for women with higher education than men with similar education.
      - Unemployment rate: 15 percentage points higher for women without higher education than men without higher education.
- Labor income distributions by education:
  - With higher education:
    - Women are less likely to be in the lower (under 2,000 EC$) and upper (6,000 EC$ and over) ends and more likely in middle brackets (2,000-5,999 EC$) relative to men.
  - Without higher education:
    - Women more likely in lower brackets (under 1,200 EC$) and less likely in upper brackets (1,200 EC$ and over) compared to men.

### IV. Methodology — econometric models and covariates
- Models and decompositions:
  - Probit regressions for participation and unemployment; Blinder-Oaxaca decompositions applied to first-order approximation of probit (Yun 2004).
  - Mincer earnings regression (linear) of log earnings on education, experience, experience squared, tenure, occupation, industry, log hours, sector indicators; Blinder-Oaxaca decomposition applied with pooled reference coefficients.
  - Robustness: Logit-based decompositions reported in Annex III; ordered logistic regression for income brackets reported in Annex III.
- Key covariates:
  - Education categories, household structure indicators (spouse; child 0-4; child 5-14; single parent; person 65+), district indicators (10 districts), age groups (5-year intervals), time fixed effects.
  - Mincer Z vector includes education; quadratic potential experience; tenure categories; occupation (9); industry (21); log usual hours; central government indicator; self-employment indicators; time fixed effects.
  - Potential experience defined as years of work experience if worked every year after graduation from highest education.

### Annex III — Blinder-Oaxaca decomposition (labor force participation, unemployment, and labor income)
- Decomposition identity and interpretation:
  - ȳm − ȳf = θ̂′(Z̄m − Z̄f) + (−η̂).
    - Explained gap: θ̂′(Z̄m − Z̄f).
    - Unexplained gap: −η̂.
  - Explained gap reflects observable differences (may reflect social constraints); unexplained gap captures discrimination and unobserved differences.
- Labor force participation and unemployment (sample sizes and key decomposition results):
  - Participation regression sample: 15-64, 29,169 observations.
  - Unemployment regression sample: 15-64 in labor force, 23,280 observations.
  - Male labor force participation coefficient: 83.4% (S.E. 0.3%, P-Value 0.00).
  - Female labor force participation coefficient: 76.5% (S.E. 0.4%, P-Value 0.00).
  - Difference (male − female) participation: 6.9% (S.E. 0.5%, P-Value 0.00).
  - Explained component of participation gap: -0.9% (S.E. 0.2%, P-Value 0.00).
    - Education contribution to explained: -0.5% (S.E. 0.1%, P-Value 0.00).
    - Household contribution to explained: -0.1% (S.E. 0.1%, P-Value 0.34).
    - Others: -0.3% (S.E. 0.2%, P-Value 0.07).
  - Unexplained component of participation gap: 7.8% (S.E. 0.5%, P-Value 0.00).
  - Male unemployment coefficient: 20.8% (S.E. 0.4%, P-Value 0.00).
  - Female unemployment coefficient: 26.9% (S.E. 0.5%, P-Value 0.00).
  - Difference (male − female) unemployment: -6.1% (S.E. 0.6%, P-Value 0.00).
  - Explained component of unemployment gap: 1.7% (S.E. 0.2%, P-Value 0.00).
    - Education contribution: 1.5% (S.E. 0.1%, P-Value 0.00).
    - Household contribution: -0.1% (S.E. 0.1%, P-Value 0.35).
    - Others: 0.3% (S.E. 0.2%, P-Value 0.08).
  - Unexplained component of unemployment gap: -7.8% (S.E. 0.6%, P-Value 0.00).
- Interpretation:
  - Accounting for observables widens the participation gap from 6.9 to 7.8 percentage points (explained component negative).
  - Unexplained gap implies female workers are around 8 percentage points less likely to be in the labor market than male workers with same observables.
  - For unemployment, accounting for characteristics widens the gap from -6.1 to -7.8 percentage points; education is a key explanatory factor.

### Childcare and interactions with young children
- Definition: Young child = child younger than 5 in the household.
- Participation gaps by education group (average predicted probabilities):
  - Primary: 13 percentage points gap.
  - Secondary: 6 percentage points gap.
  - Post-secondary: 2 percentage points gap.
  - Tertiary: 3 percentage points gap.
- Unemployment gaps by education group (average predicted probabilities):
  - Primary: -14 percentage points gap.
  - Secondary: -7 percentage points gap.
  - Post-secondary: -2 percentage points gap.
  - Tertiary: 1 percentage point gap.
- Average predicted probabilities by gender and young children:
  - Men:
    - Having a young child: not associated with lower participation.
    - Having a young child: associated with 2 percentage points lower probability of unemployment.
  - Women:
    - Having a young child: associated with 3 percentage points lower probability of participation.
    - Having a young child: associated with 4 percentage points higher probability of unemployment.
- Heterogeneity by female education:
  - Women with higher education and young children:
    - Participation down by 5 percentage points on average.
    - Unemployment does not change on average.
  - Women without higher education and young children:
    - Participation down by 2 percentage points on average.
    - Unemployment up by 5 percentage points on average.
- Interpretation:
  - Childcare responsibilities affect women’s employment more than men’s, with channels differing by education level.

### Labor income — Mincer regression and decompositions
- Main Mincer result:
  - Female coefficient: -0.18 (S.E. 0.02) — interpreted as female workers earn on average 18 percent less than male workers with same observables.
- Key Mincer coefficients (selected):
  - Potential experience: 0.02 (S.E. 0.00).
  - Tenure:
    - 6 months - 1 year: 0.06 (S.E. 0.05).
    - 1-5 years: 0.19 (S.E. 0.04).
    - 5-10 years: 0.40 (S.E. 0.04).
    - more than 10 years: 0.55 (S.E. 0.05).
  - Education:
    - secondary: 0.27 (S.E. 0.02).
    - post-secondary: 0.70 (S.E. 0.04).
    - tertiary: 0.91 (S.E. 0.04).
  - Log hours: 0.65 (S.E. 0.05).
  - Government indicator: 0.18 (S.E. 0.03).
  - Self-employed with employees: 0.21 (S.E. 0.06).
  - Self-employed without employees: -0.29 (S.E. 0.03).
  - Industry highlights: mining 0.82 (0.12); electricity 0.65 (0.11); water 0.57 (0.09); construction 0.54 (0.06); accommodation 0.52 (0.05); financial & insurance 0.67 (0.07); extraterritorial 0.61 (0.13).
- Decomposition (mean log income):
  - Male mean log income: 7.49 (S.E. 0.02, P-Value 0.00).
  - Female mean log income: 7.35 (S.E. 0.02, P-Value 0.00).
  - Difference (male − female): 0.14 (S.E. 0.02, P-Value 0.00) — raw gap of 14 percent (log difference measure).
  - Explained component: -0.04 (S.E. 0.02, P-Value 0.01).
    - Potential Experience: 0.00 (S.E. 0.00, P-Value 0.57).
    - Tenure: 0.02 (S.E. 0.00, P-Value 0.00).
    - Education: -0.09 (S.E. 0.01, P-Value 0.00).
    - Occupation: 0.04 (S.E. 0.01, P-Value 0.00).
    - Industry: 0.01 (S.E. 0.01, P-Value 0.36).
    - Hours Worked: 0.02 (S.E. 0.00, P-Value 0.00).
    - Government: -0.02 (S.E. 0.00, P-Value 0.00).
    - Self-Employed with Employees: 0.01 (S.E. 0.00, P-Value 0.00).
    - Self-Employed w/o Employees: -0.02 (S.E. 0.00, P-Value 0.00).
  - Unexplained component: 0.18 (S.E. 0.02, P-Value 0.00).
  - Sample counts: N male 4,754; N female 4,333.
- Interpretation:
  - Raw 14 percent log gap decomposes into -4 percent explained and 18 percent unexplained.
  - After accounting for characteristics, the gender gap in labor income increases from 14 to 18 percent.
  - Within explained components, positive contributors: tenure (2 percent), occupation (4 percent), log hours (2 percent), self-employed with employees (1 percent). Negative contributors: education (-9 percent), government (-2 percent), self-employed without employees (-2 percent).
  - Similar results when applying Heckman selection correction.

### Gender gap by education and education premia
- Unexplained gender gaps (Mincer allowing gap to vary by education):
  - primary: 21 percent (significant).
  - secondary: 21 percent (significant).
  - post-secondary: 8 percent (not significant).
  - tertiary: 2 percent (not significant).
- Education premia conditional on occupation and industry (relative to primary or less):
  - Men:
    - secondary: 27 percent.
    - post-secondary: 63 percent.
    - tertiary: 80 percent.
  - Women:
    - secondary: 27 percent.
    - post-secondary: 76 percent.
    - tertiary: 99 percent.
- Education premia excluding occupation and industry:
  - Men:
    - secondary: 37 percent.
    - post-secondary: 84 percent.
    - tertiary: 104 percent.
  - Women:
    - secondary: 39 percent.
    - post-secondary: 107 percent.
    - tertiary: 135 percent.
- Marginal gains from education:
  - Women with post-secondary education earn 68 percent more than women with secondary education.
  - Men with post-secondary education earn 47 percent more than men with secondary education.
- Interpretation:
  - Unexplained gender gap in labor income decreases with higher education.
  - Gains from education are larger for women than men; higher education is associated with decreased unexplained gap.
  - Causal interpretation requires assumption that residual in Mincer regression is uncorrelated with education.

### Probit regressions — labor force participation and unemployment (Table A1 highlights)
- Samples:
  - Participation: 15-64, Obs = 29,169.
  - Unemployment: 15-64 in labor force, Obs = 23,280.
- Selected marginal effects/probit coefficients (standard errors in parentheses):
  - Female (participation): -0.51 (0.036), -0.34 (0.025), -0.37 (0.027) across specifications.
  - Female (unemployment): 0.46 (0.038), 0.26 (0.026), 0.28 (0.027).
  - Education indicators (example participation column): Secondary 0.16; Post-secondary 0.035; Tertiary 0.074.
  - Household controls examples:
    - Spouse: participation ~ -0.040 to -0.046; unemployment ~ -0.13 to -0.12.
    - Child under 5: participation examples -0.074, 0.0010, -0.045; unemployment examples 0.051, -0.089, -0.069.
  - Interaction terms examples:
    - Female x Secondary: 0.21 and -0.21 (S.E. 0.048, 0.050).
    - Female x Post-Secondary: 0.41 and -0.35 (S.E. 0.089, 0.096).
    - Female x Tertiary: 0.39 and -0.50 (S.E. 0.086, 0.10).
- Standard errors clustered at enumeration district × year × quarter.

### Mincer regressions — log gross monthly income (Table A2 highlights)
- Sample sizes:
  - Employed 15-64: around Obs = 9,087 (e.g., 9,088 in column (1)).
  - Central government sample: Obs = 1,723 in specification (6).
- Female coefficients across specifications: -0.21, -0.19, -0.18, -0.21, -0.32, -0.32, -0.17 (standard errors ~0.017–0.050).
- Education returns (examples across specifications):
  - Secondary: 0.28, 0.27, 0.27, 0.27, 0.37, 0.45, 0.24.
  - Post-secondary: 0.72, 0.70, 0.70, 0.63, 0.83, 0.58, 0.61.
  - University: 0.95, 0.93, 0.91, 0.80, 1.04, 0.81, 0.73.
- Log hours worked: 0.67, 0.65, 0.65, 0.78, 0.70, 0.65.
- R-squared range across columns: 0.379, 0.408, 0.424, 0.425, 0.353, 0.575, 0.429.

### Conditional quantile regressions — premia on unobserved factors (Table A3)
- Sample: employed 15-64, Obs = 9,087.
- Female coefficients by quantile (robust SEs):
  - 25th percentile: -0.17 (0.018).
  - 50th percentile: -0.20 (0.017).
  - 75th percentile: -0.22 (0.017).
- Gender gap in conditional labor income distribution:
  - 25th percentile: 17 percent.
  - 50th percentile: 20 percent.
  - 75th percentile: 22 percent.
- Other highlights:
  - Tenure >10 years: ~0.51–0.55 across quantiles (S.E. ~0.029–0.047).
  - Post-secondary premia: 0.65, 0.76, 0.81 across 25th/50th/75th.
  - Tertiary premia: 0.90, 0.98, 0.94 across 25th/50th/75th.
  - Median interpretation: median female conditional on observables earns 20 percent less than male counterpart; tenure >10 years associated with 55 percent higher income; post-secondary associated with 76 percent higher income; 10 percent more hours associated with 6 percent higher income.

### Robustness checks — Blinder-Oaxaca (Logit) and Heckman selection
- Logit-based Blinder-Oaxaca (selected):
  - Participation predicted: Male 83.3% (S.E. 0.3%); Female 76.5% (S.E. 0.4%); Difference 6.9% (S.E. 0.5%); Explained -0.9% (S.E. 0.2%); Unexplained 7.7% (S.E. 0.5%).
  - Unemployment predicted: Male 20.9% (S.E. 0.4%); Female 26.9% (S.E. 0.5%); Difference -6.0% (S.E. 0.6%); Explained 1.7% (S.E. 0.2%); Unexplained -7.8% (S.E. 0.6%).
- Blinder-Oaxaca with Heckman selection (log income; selected):
  - Male coefficient: 8.08 (S.E. 0.06, P-Value 0.00).
  - Female coefficient: 7.94 (S.E. 0.07, P-Value 0.00).
  - Difference: 0.14 (S.E. 0.08, P-Value 0.08).
  - Explained: -0.04 (S.E. 0.02, P-Value 0.01).
  - Unexplained: 0.19 (S.E. 0.08, P-Value 0.02).
  - Sample counts: N male 12,331; selected male 4,754; N female 12,472; selected female 4,333.

### Year-by-year regressions (Tables A6–A8 highlights)
- Participation probit female coefficients by year (15-64):
  - 2016: -0.43 (0.052)
  - 2017: -0.39 (0.054)
  - 2018: -0.44 (0.056)
  - 2019: -0.27 (0.053)
  - 2020: -0.25 (0.051)
  - 2021: -0.44 (0.066)
  - Observations by year: 2016 5,688; 2017 5,489; 2018 5,002; 2019 4,873; 2020 4,867; 2021 3,250.
- Unemployment probit female coefficients by year (labor force sample):
  - 2016: 0.33 (0.046)
  - 2017: 0.30 (0.052)
  - 2018: 0.27 (0.060)
  - 2019: 0.25 (0.053)
  - 2020: 0.36 (0.063)
  - 2021: 0.24 (0.072)
  - Observations by year: 2016 4,719; 2017 4,484; 2018 4,125; 2019 3,951; 2020 3,430; 2021 2,571.
- Mincer female coefficients by year (employed sample):
  - 2016: -0.11 (0.029)
  - 2017: -0.21 (0.037)
  - 2018: -0.20 (0.034)
  - 2019: -0.18 (0.040)
  - 2020: -0.28 (0.062)
  - Observations by year (employed): 2016 2,506; 2017 2,080; 2018 1,944; 2019 1,812; 2020 745.
  - R-squared by year: 0.443, 0.435, 0.473, 0.440, 0.415 (2016–2020).

### Ordered logistic regression and predicted income-category probabilities (Table A9, Figure A1)
- Sample: employed 15-64, Obs = 9,087.
- Selected ordered logit coefficients (standard errors):
  - Female: -0.59 (0.05).
  - Potential experience: 0.06 (0.01).
  - Tenure: 6 months - 1 year 0.17 (0.16); 1-5 years 0.58 (0.12); 5-10 years 1.13 (0.12); more than 10 years 1.61 (0.14).
  - Education: secondary 0.76 (0.07); post-secondary 2.06 (0.12); tertiary 2.76 (0.13).
  - Occupation examples: managers 1.31 (0.15); professionals 1.62 (0.12); clerical support 0.86 (0.10); service and sales 0.29 (0.07).
  - Industry examples: mining 2.60 (0.37); manufacturing 0.93 (0.17); accommodation 1.82 (0.15).
  - Government: 0.56 (0.10).
  - Self w/ employee: 0.67 (0.19); Self w/o employee: -0.86 (0.10).
  - Log hours: 2.02 (0.13).
  - Cutpoints reported (Cut1–Cut7) with standard errors (e.g., Cut1 5.28 (0.49), Cut7 15.82 (0.53)).
- Predicted probabilities by income bracket shown in Figure A1 (income brackets in Eastern Caribbean Dollars):
  - Under 200; 201 - 399; 400 - 799; 800 - 1199; 1200 - 1999; 2000 - 3999; 4000 - 5999; Over 6000.
  - Figure A1 reports average predicted probability of being in each income category by gender with 95 percent confidence intervals.

### Conclusion and policy-relevant findings
- Key takeaways:
  - High female labor force participation is important but not sufficient as a benchmark; a broader set of indicators and granular analysis are warranted.
  - Econometric analysis should be complemented with holistic understanding of historical and social context shaping gender traits.
- Policy implications and good practices (tailored suggestions):
  - Emerging market and developing economies: invest in infrastructure, support female entrepreneurs by increasing access to finance, promote equal rights for women.
  - Advanced economies: increase parity between maternity and paternity leave, promote access to affordable, high-quality childcare.
  - St. Lucia-specific:
    - Results reinforce the importance of education in bridging gender gaps.
    - Women without higher education face disproportionate disadvantage, highlighting need for targeted policy initiatives.

*IMF Working Paper — Annex I and Annex III tables and figures from "Inequality in a More Equal World—Labor Market Gender Gaps in St. Lucia".*

### Annex I. Additional Tables .............................................................................................

### Annex I. Additional Tables

### III. Introduction — motivation and key summary findings
- Motivation:
  - To understand gender gaps in St. Lucia where macro-data show high female participation rates (ECCU female participation rate 71 percent) but anthropological/sociological evidence indicates a matrifocal but non-matriarchal family structure.
  - Data: St. Lucia Labor Force Survey covering 2016Q1-2021Q4 (rich micro dataset).
  - Methods: probit regression with Blinder-Oaxaca decomposition, Mincer earnings regression, analyses across age-groups, education groups, and family characteristics (e.g., presence of young children).
- High-level descriptive context:
  - ECCU female participation rate: 71 percent.
  - World average female participation rate: 53 percent.
  - OECD average female participation rate: 65 percent.
  - Participation gap between men and women in ECCU: 10 percentage points.
  - Participation gaps: OECD average 16 percentage points, LAC average 23 percentage points, world average 26 percentage points.
- Principal econometric findings (preserving exact reported magnitudes):
  - Controlling for observables, the participation rate should be 1 percent higher for women than men (reflecting women’s higher education, etc.).
  - Observed overall participation rate gap: 7 percent (women participate 7 percent less than men).
  - Unexplained gap in participation: women are 8 percent less likely to be in the labor market than male workers with the same observable characteristics.
  - Young children disproportionately affect female workers, particularly those without higher education:
    - For women: having a young child is associated with lower probability of participation and higher probability of unemployment.
    - For men: having a young child is associated with no difference in probability of participation and lower probability of unemployment.
  - Participation gaps decline with education:
    - Unexplained gap in labor force participation by education level: primary education 13 percentage points; secondary education 6 percentage points; post-secondary 2 percentage points; tertiary 3 percentage points.
  - Labor income gaps:
    - Female workers earn on average 18 percent less than male workers with the same observable characteristics.
    - Unexplained gender gap in labor income by education level: primary education 21 percent; tertiary education 2 percent.
    - Female workers obtain larger gains from education relative to male counterparts (higher education premia for women).

### II. Data — scope, construction, and key sample facts
- Data source and scope:
  - St. Lucia Labor Force Survey (SLLFS) during 2016Q1-2021Q4, run by the Central Statistical Office and the Ministry of Finance, Economic Development and the Youth Economy.
  - Focus on working-age population aged 15-64; yields about 30,000 observations, with about 9,000 reporting labor income.
- Sample and weighting:
  - Data pooled across available years; weights adjusted so each year is represented equally (weight of each observation divided by year-specific sum and multiplied by 100,000 so adjusted weights sum to 100,000 each year).
  - Standard errors clustered at year × quarter × enumeration district level.
- Dependent variables and definitions:
  - Unemployment: broader definition—individual did not work for pay, was not temporarily absent from work, and wanted to work during the reference week (ILO recommendation noted for small islands).
  - Participation: participates if unemployed, working for pay, or temporarily absent from work during the reference week.
  - Gross monthly income from all employment: available 2016Q1 to 2020Q3; non-missing for about 60 percent of employed people in sample.
    - Among the employed, percentage missing the income variable: 43 percent of men and 39 percent of women.
  - Income reporting: 8 brackets—under 200; 201-399; 400-799; 800-1199; 1200-1999; 2000-3999; 4000-5999; over 6000 Eastern Caribbean Dollars (EC$).
  - For Mincer regressions, log of midpoint of income bracket used as proxy for individual log gross monthly income (acknowledged measurement error caveat).
- Notes on robustness:
  - Results from an ordered logistic regression using income brackets reported in Annex III.

### III. Descriptive statistics — education, test scores, age profiles, and income distributions
- Education attainment (working-age population, age 15-64):
  - Women: 21 percent have attained higher education; 79 percent have not.
  - Men: 14 percent have attained higher education; 86 percent have not.
  - Education categories used: “none or primary”, “lower or upper secondary”, “post-secondary (diploma or associate degree)”, and “tertiary (university)”.
- Test scores:
  - Female students on average have higher test scores than male students in St. Lucia and other ECCU countries (Harmonized Test Scores referenced; scale 300 minimal attainment to 625 advanced attainment).
- Age profiles and gender gaps by education (highlights and exact comparative points):
  - For workers with higher education:
    - Gender gap in labor force participation rate: 2 percentage points in early thirties, rising to 13 percentage points in early sixties.
    - Little gender gap in unemployment across all age groups except under 25.
  - For workers without higher education:
    - Gender gap in participation: 9 percentage points in early thirties, rising to 24 percentage points in early sixties.
    - Example for age group 30-34:
      - Unemployment rate: 1 percentage point higher for women with higher education than men with similar education.
      - Unemployment rate: 15 percentage points higher for women without higher education than men without higher education.
- Labor income distribution by education:
  - Workers with higher education:
    - Women are less likely to be in the lower (under 2,000 EC$) and upper (6,000 EC$ and over) ends and more likely to be in middle brackets (2,000-5,999 EC$) relative to men.
  - Workers without higher education:
    - Clear pattern of men earning more than women: women more likely in lower income brackets (under 1,200 EC$) and less likely in upper brackets (1,200 EC$ and over) compared to men.

### IV. Methodology — econometric models and covariates
- Econometric approach:
  - Probit regressions for labor force participation and unemployment; Blinder-Oaxaca decompositions applied to those probit models.
  - Mincer earnings regression (linear) of log earnings on education, experience, experience squared, and other covariates; Blinder-Oaxaca decomposition applied to earnings regression.
  - Robustness: Blinder-Oaxaca decompositions using Logit regressions reported in Annex III Table A4; ordered logistic regression for income brackets reported in Annex III.
- Probit model specification (participation example):
  - Participation modeled as indicator where α + βF_i + γ′X_i + ε_i^P > 0.
  - F_i: indicator for being female.
  - X_i: covariates including education, household structure, district, age group, and time fixed effects (annual and quarter indicators).
  - Error term ε_i^P assumed standard normal conditional on observables.
  - Unemployment modeled similarly but sample restricted to workers participating in the labor market.
- Construction of covariates:
  - Education categorical: none or primary; lower or upper secondary; post-secondary (diploma or associate degree); tertiary (university).
  - Household structure indicators: having a spouse; at least one child age 0-4; at least one child age 5-14; single parent (no spouse, at least one child age 0-14, and household head); at least one person age 65 or over in household.
  - District indicators: the 10 districts of St. Lucia (Castries divided into Castries City and Castries Rural; Anse la Raye and Canaries combined).
  - Age group indicators: 5-year intervals 15-19, 20-24, ..., 60-64.
- Mincer earnings regression specification:
  - y_i = δ + ηF_i + θ′Z_i + ε_i^M, where y_i is natural log of gross monthly income from all employment, F_i indicator for female, Z_i vector of covariates, ε_i^M mean-zero residual.
  - Z_i includes: education; quadratic function of potential experience; tenure at current job; occupation (9 categories); industry (21 categories); log usual hours worked per week; indicator for working in central government or statutory board; indicator for self-employed with employees; indicator for self-employed without employees; time fixed effects.
  - Potential experience defined as years of work experience if worked every year after graduation from highest level of education.
  - Tenure categories: less than 6 months; 6 months or more but less than 1 year; 1 year or more but less than 5 years; 5 years or more but less than 10 years; 10 years or more.
- Complementary analyses:
  - Conditional quantile regressions of log labor income on gender and observables reported in Annex II.
  - Ordered logistic regression of labor income brackets on gender and observables reported in Annex III.

*International Monetary Fund — IMF Working Paper excerpt (Annex I. Additional Tables).*

### Annex III (Table A9, Figure A1).

### Annex III (Table A9, Figure A1). Blinder-Oaxaca Decomposition

### Methodology
- The Blinder-Oaxaca decomposition is applied to both the probit regressions and the Mincer earnings regression to split the raw gender gap into an explained part and an unexplained part.
- For the probit regressions, the decomposition is implemented on the first-order approximation of the probit model (Yun 2004).
- For the Mincer earnings regression, the standard Blinder-Oaxaca decomposition is used with reference coefficients taken from the pooled estimation (both women and men in the sample).
- Notation: ȳm and ȳf denote sample mean of log monthly labor income for men and women; Z̄m and Z̄f denote sample mean of covariates for men and women.
- Decomposition identity used: ȳm − ȳf = θ̂′(Z̄m − Z̄f) + (−η̂).
  - Explained gap: θ̂′(Z̄m − Z̄f)
  - Unexplained gap: −η̂
- Interpretation notes:
  - Explained gap reflects differences in observable characteristics in Z; some of this may reflect discrimination if observable differences are driven by social constraints.
  - Unexplained gap captures effects of discrimination and unobserved group differences in productivity and taste.

### Labor Force Participation and Unemployment (Blinder-Oaxaca results)
- Sample definitions:
  - Labor force participation regression: 15-64 years old individuals, with 29,169 observations.
  - Unemployment regression: 15-64 years old individuals who are in the labor force, with 23,280 observations.
  - Coefficients used for decompositions come from pooled regressions. Standard errors clustered at enumeration district x year x quarter level.
- Key summary results from Table 1 (labor force participation and unemployment):
  - Male labor force participation coefficient: 83.4% (S.E. 0.3%, P-Value 0.00)
  - Female labor force participation coefficient: 76.5% (S.E. 0.4%, P-Value 0.00)
  - Difference (male − female) in participation: 6.9% (S.E. 0.5%, P-Value 0.00)
  - Explained component of participation gap: -0.9% (S.E. 0.2%, P-Value 0.00)
    - Education contribution to explained: -0.5% (S.E. 0.1%, P-Value 0.00)
    - Household contribution to explained: -0.1% (S.E. 0.1%, P-Value 0.34)
    - Others: -0.3% (S.E. 0.2%, P-Value 0.07)
  - Unexplained component of participation gap: 7.8% (S.E. 0.5%, P-Value 0.00)
  - Male unemployment coefficient: 20.8% (S.E. 0.4%, P-Value 0.00)
  - Female unemployment coefficient: 26.9% (S.E. 0.5%, P-Value 0.00)
  - Difference (male − female) in unemployment: -6.1% (S.E. 0.6%, P-Value 0.00)
  - Explained component of unemployment gap: 1.7% (S.E. 0.2%, P-Value 0.00)
    - Education contribution: 1.5% (S.E. 0.1%, P-Value 0.00)
    - Household contribution: -0.1% (S.E. 0.1%, P-Value 0.35)
    - Others: 0.3% (S.E. 0.2%, P-Value 0.08)
  - Unexplained component of unemployment gap: -7.8% (S.E. 0.6%, P-Value 0.00)
- Interpretation:
  - After accounting for individual characteristics, the labor force participation gap widens from 6.9 to 7.8 percentage points (explained component negative at -0.9 percentage point).
  - The unexplained gap implies female workers are around 8 percentage points less likely to be in the labor market than male workers with the same observable characteristics.
  - For unemployment, accounting for characteristics widens the gender gap from -6.1 to -7.8 percentage points — i.e., female workers in the labor force are 8 percentage points more likely to be unemployed than males with the same observables.
  - Education is a key driver: female workers are more educated on average and education has sizable explanatory power over unemployment rate.

### Childcare and Young Children (Interactions with female indicator)
- Definition: Young child = child younger than 5 in the household.
- From probit regressions interacting female indicator with young child indicator:
  - Gender gaps in labor force participation by education group (average predicted probabilities, Figure 5):
    - Primary: 13 percentage points gap
    - Secondary: 6 percentage points gap
    - Post-secondary: 2 percentage points gap
    - Tertiary: 3 percentage points gap
  - Gender gaps in unemployment by education group (average predicted probabilities, Figure 5):
    - Primary: -14 percentage points gap
    - Secondary: -7 percentage points gap
    - Post-secondary: -2 percentage points gap
    - Tertiary: 1 percentage point gap
  - Average predicted probabilities by gender and young children (Figure 6):
    - For men:
      - Having a young child: not associated with lower participation
      - Having a young child: associated with 2 percentage points lower probability of unemployment
    - For women:
      - Having a young child: associated with 3 percentage points lower probability of participation
      - Having a young child: associated with 4 percentage points higher probability of unemployment
  - Heterogeneity by female education (Figure 7):
    - Women with higher education and young children:
      - Participation goes down by 5 percentage points on average
      - Unemployment does not change on average
    - Women without higher education and young children:
      - Participation goes down by 2 percentage points on average
      - Unemployment goes up by 5 percentage points on average
- Interpretation:
  - Women’s employment is more negatively affected by childcare responsibilities than men’s, with different channels by education level:
    - Women with higher education: negative effect mainly via participation channel.
    - Women without higher education: negative effects via both participation and unemployment channels.

### Labor Income (Mincer regression and Blinder-Oaxaca decomposition)
- Main Mincer regression result:
  - Female coefficient: -0.18 (S.E. 0.02) — interpreted as female workers earn on average 18 percent less than male workers with the same observable characteristics.
- Other Mincer regression findings (Table 2, coefficients preserved):
  - Potential experience coefficient: 0.02 (S.E. 0.00)
  - Tenure effects:
    - 6 months - 1 year: 0.06 (S.E. 0.05)
    - 1-5 years: 0.19 (S.E. 0.04)
    - 5-10 years: 0.40 (S.E. 0.04)
    - more than 10 years: 0.55 (S.E. 0.05)
    - Interpretation: having more than ten years of tenure at the current job is associated with 55 percent higher income compared to less than 6 months of tenure.
  - Education coefficients:
    - secondary: 0.27 (S.E. 0.02)
    - post-secondary: 0.70 (S.E. 0.04)
    - tertiary: 0.91 (S.E. 0.04)
    - Interpretation: post-secondary (non-tertiary) education associated with 70 percent higher income compared to primary or less.
  - Log hours coefficient: 0.65 (S.E. 0.05) — working 10 percent more hours is associated with 6.5 percent higher income.
  - Government indicator: 0.18 (S.E. 0.03) — government employees earn 18 percent higher income compared to private employees.
  - Self-employed with employees: 0.21 (S.E. 0.06) — 21 percent higher income.
  - Self-employed without employees: -0.29 (S.E. 0.03) — 28 percent lower income.
  - Industry highlights (coefficients and S.E.):
    - mining: 0.82 (0.12)
    - electricity: 0.65 (0.11)
    - water: 0.57 (0.09)
    - construction: 0.54 (0.06)
    - accommodation: 0.52 (0.05)
    - financial & insurance: 0.67 (0.07)
    - extraterritorial: 0.61 (0.13)
- Decomposition results for labor income (Table 3):
  - Mean log income:
    - Male: 7.49 (S.E. 0.02, P-Value 0.00)
    - Female: 7.35 (S.E. 0.02, P-Value 0.00)
    - Difference (male − female): 0.14 (S.E. 0.02, P-Value 0.00) — raw gap of 14 percent (log difference measure)
  - Explained component: -0.04 (S.E. 0.02, P-Value 0.01)
    - Potential Experience: 0.00 (S.E. 0.00, P-Value 0.57)
    - Tenure: 0.02 (S.E. 0.00, P-Value 0.00)
    - Education: -0.09 (S.E. 0.01, P-Value 0.00)
    - Occupation: 0.04 (S.E. 0.01, P-Value 0.00)
    - Industry: 0.01 (S.E. 0.01, P-Value 0.36)
    - Hours Worked: 0.02 (S.E. 0.00, P-Value 0.00)
    - Government: -0.02 (S.E. 0.00, P-Value 0.00)
    - Self-Employed with Employees: 0.01 (S.E. 0.00, P-Value 0.00)
    - Self-Employed w/o Employees: -0.02 (S.E. 0.00, P-Value 0.00)
  - Unexplained component: 0.18 (S.E. 0.02, P-Value 0.00)
  - Sample sizes:
    - N of observations male: 4,754
    - N of observations female: 4,333
  - Interpretation:
    - Out of the 14 percent raw gap in log labor income, -4 percent is explained and 18 percent is unexplained.
    - After accounting for individual characteristics, the gender gap in labor income increases from 14 to 18 percent (unexplained component).
    - Within the explained components, positive contributors to the gap: tenure (2 percent), occupation (4 percent), log hours worked (2 percent), self-employed with employees indicator (1 percent).
    - Negative contributors to the gap: education (-9 percent), government indicator (-2 percent), self-employed without employees indicator (-2 percent).
    - Results are similar when Heckman correction is applied to address selection bias (Annex III Table A5).
  - Note on measures:
    - The 14 percent raw gender gap calculated from the Mincer regression is the difference in average log male income and average log female income. It differs from the gap computed as (average male income − average female income)/average male income (12 percent).

### Gender Gap by Education Group and Education Premia
- Unexplained gender gap by education (from Mincer regression allowing gender gap to vary by education):
  - Unexplained gender gaps:
    - primary: 21 percent (significant)
    - secondary: 21 percent (significant)
    - post-secondary: 8 percent (not significant)
    - tertiary: 2 percent (not significant)
  - Pattern holds when focusing on government employees: large and significant unexplained gaps for primary and secondary, small and insignificant for post-secondary and tertiary.
- Education premia (conditional on occupation and industry):
  - Men relative to primary or less:
    - secondary: 27 percent
    - post-secondary: 63 percent
    - tertiary: 80 percent
  - Women relative to primary or less:
    - secondary: 27 percent
    - post-secondary: 76 percent
    - tertiary: 99 percent
- Education premia (excluding occupation and industry — including gains through occupation & industry):
  - Men relative to primary or less:
    - secondary: 37 percent
    - post-secondary: 84 percent
    - tertiary: 104 percent
  - Women relative to primary or less:
    - secondary: 39 percent
    - post-secondary: 107 percent
    - tertiary: 135 percent
- Marginal gains from education (largest gender difference at post-secondary level):
  - Women with post-secondary education earn 68 percent more than women with secondary education.
  - Men with post-secondary education earn 47 percent more than men with secondary education.
- Interpretation:
  - The unexplained gender gap in labor income decreases with higher education.
  - Gains from education are larger for women than for men both because of direct education returns and because higher education is associated with a decrease in the unexplained gender gap.
  - Causal interpretation requires the assumption that residual εi in the Mincer regression is uncorrelated with education.

### Conclusion and Policy-relevant Findings
- Two key aspects highlighted:
  - High female labor force participation is useful but not sufficient as a benchmark; a broader set of indicators and granular analysis are warranted.
  - Econometric analysis should be complemented with holistic understanding of historical and social context shaping gender traits.
- Policy implications and good practices (tailored to country context):
  - For emerging market and developing economies: invest in infrastructure, support female entrepreneurs by increasing access to finance, promote equal rights for women.
  - For advanced economies: increase parity between maternity and paternity leave, promote access to affordable, high-quality childcare.
  - For St. Lucia specifically:
    - Results reinforce the importance of education in bridging gender gaps.
    - Women without higher education face disproportionate disadvantage in the labor force, highlighting the need for policy initiatives targeting this segment.

*IMF Working Paper — Annex III (Table A9, Figure A1) from "Inequality in a More Equal World—Labor Market Gender Gaps in St. Lucia"*

### Annex I. Additional Tables

### Annex I. Additional Tables

### Probit regressions — labor force participation and unemployment (Table A1)
- Samples:
  - Labor force participation: 15-64 years old individuals. Observations: 29,169 (columns (1)-(3)).
  - Unemployment: 15-64 years old individuals who are in the labor force. Observations: 23,280 (columns (4)-(6)).
- Key coefficient estimates (selected, marginal effects/probit coefficients):
  - Female:
    - Labor force participation: -0.51, -0.34, -0.37 (standard errors (0.036), (0.025), (0.027)).
    - Unemployment: 0.46, 0.26, 0.28 (standard errors (0.038), (0.026), (0.027)).
  - Education indicators (effects vary by specification):
    - Secondary: 0.16; Post-secondary: 0.035; Tertiary: 0.074 (first set of columns; standard errors reported).
    - On unemployment columns, coefficients for education are negative (e.g., Post-secondary -0.50, -0.68; Tertiary -0.37, -0.65).
  - Household and demographic controls (examples):
    - Spouse: around -0.040 to -0.046 for participation; -0.13 to -0.12 for unemployment.
    - Child under 5: participation coefficients vary (e.g., -0.074, 0.0010, -0.045); unemployment coefficients vary (e.g., 0.051, -0.089, -0.069).
    - Elder: participation around -0.19 (first columns) and unemployment around 0.11 (later columns).
  - Interaction terms (selected):
    - Female x Secondary: 0.21 and -0.21 (standard errors (0.048), (0.050)).
    - Female x Post-Secondary: 0.41 and -0.35 (standard errors (0.089), (0.096)).
    - Female x Tertiary: 0.39 and -0.50 (standard errors (0.086), (0.10)).
    - High-Skilled x Child under 5: 0.34 and -0.069 (standard errors (0.16), (0.18)).
- Notes:
  - Coefficients on age group, district, year, and quarter indicators are omitted.
  - Standard errors are clustered at enumeration district x year x quarter level.

### Mincer regressions — log gross monthly income (Table A2)
- Samples:
  - Specifications (1)-(5) and (7): employed individuals aged 15-64. Observations around 9,087; example: Obs = 9,088 for column (1).
  - Specification (6): employed by the central government. Observations = 1,723.
- Key coefficient estimates (log income regressions; standard errors in parentheses):
  - Female:
    - Columns show -0.21, -0.19, -0.18, -0.21, -0.32, -0.32, -0.17 (standard errors (0.017) to (0.050) across specifications).
  - Experience: 0.021, 0.020, 0.021, 0.021, 0.026, 0.0063, 0.020 (standard errors ~0.0023–0.0048).
  - Education returns relative to base:
    - Secondary: 0.28, 0.27, 0.27, 0.27, 0.37, 0.45, 0.24 (standard errors ~0.024–0.031).
    - Post-secondary: 0.72, 0.70, 0.70, 0.63, 0.83, 0.58, 0.61 (standard errors ~0.040–0.096).
    - University (tertiary): 0.95, 0.93, 0.91, 0.80, 1.04, 0.81, 0.73 (standard errors ~0.040–0.081).
  - Log hours worked: 0.67, 0.65, 0.65, 0.78, 0.70, 0.65 (standard errors ~0.052–0.076).
  - Government employment: positive premium (e.g., 0.18, 0.18, 0.20, 0.11 in some columns; standard errors reported).
  - Interaction terms (selected):
    - Female # Secondary: -0.0025, 0.027, -0.14, -0.0024 (standard errors (0.035)–(0.040)).
    - Female # Post-secondary: 0.13, 0.24, 0.34, 0.042 (standard errors (0.063)–(0.098)).
    - Government # Female: -0.11 (standard error (0.056)) in one specification; other triple interactions reported with standard errors.
- Goodness of fit:
  - R-squared ranges: 0.379, 0.408, 0.424, 0.425, 0.353, 0.575, 0.429 across columns.

### Premia on unobserved factors — conditional quantile (Table A3 and Annex II)
- Sample: employed individuals aged 15-64. Observations = 9,087 for quantile regressions.
- Conditional quantile regression estimates (selected coefficients; robust standard errors in parentheses):
  - Female at conditional quantiles:
    - 25th percentile: -0.17 (0.018)
    - 50th percentile (median): -0.20 (0.017)
    - 75th percentile: -0.22 (0.017)
  - Experience:
    - 25th: 0.021 (0.0022)
    - 50th: 0.023 (0.0023)
    - 75th: 0.027 (0.0024)
  - Tenure (examples):
    - more than 10 years: 0.51, 0.55, 0.52 (standard errors (0.029), (0.046), (0.047)).
  - Education premia (relative to primary or less):
    - Secondary: 0.18, 0.26, 0.30 (standard errors (0.024), (0.022), (0.025)).
    - Post-secondary: 0.65, 0.76, 0.81 (standard errors (0.042), (0.037), (0.032)).
    - Tertiary: 0.90, 0.98, 0.94 (standard errors (0.035), (0.041), (0.033)).
  - Log hours: 0.59, 0.60, 0.65 (standard errors (0.020), (0.021), (0.037)).
  - Government: 0.21, 0.16, 0.16 (standard errors (0.033), (0.035), (0.024)).
  - Self-employed:
    - Self w/ employees: 0.031, 0.16, 0.42 (standard errors (0.070), (0.10), (0.039)).
    - Self w/o employees: -0.41, -0.29, -0.16 (standard errors (0.032), (0.034), (0.040)).
- Main interpretation and empirical findings:
  - The gender gap in labor income is increasing across conditional quantiles: 17, 20 and 22 percent for workers at the 25th, 50th and 75th percentiles of the conditional labor income distribution, respectively.
  - If the conditional ranking τ reflects unobservable skills or effort, the unexplained gender gap appears to increase with unobservable skills or effort.
  - Median regression (τ = 50) broadly corroborates OLS estimates:
    - Median female workers conditional on observables earn 20 percent less than their male counterparts.
    - Having more than 10 years of tenure is associated with 55 percent higher income.
    - Post-secondary (non-tertiary) education is associated with 76 percent higher income compared to primary education or less.
    - Working 10 percent more hours is associated with 6 percent higher income.
    - Government employees and self-employed with employees earn 16 percent higher income while self-employed without employees earn 29 percent lower income compared to private employees with otherwise the same observables.

### Robustness checks — Blinder-Oaxaca decompositions and selection (Annex III, Tables A4–A5)
- Data: St. Lucia Labor Force Survey 2016-21 for selected decompositions; other samples as noted in tables.
- Blinder-Oaxaca (Logit) decomposition results (selected):
  - Labor Force Participation:
    - Male predicted: 83.3% (S.E. 0.3%, P-Value 0.0%).
    - Female predicted: 76.5% (S.E. 0.4%, P-Value 0.0%).
    - Difference: 6.9% (S.E. 0.5%, P-Value 0.0%).
    - Explained: -0.9% (S.E. 0.2%, P-Value 0.0%).
    - Unexplained: 7.7% (S.E. 0.5%, P-Value 0.0%).
  - Unemployment:
    - Male predicted: 20.9% (S.E. 0.4%, P-Value 0.0%).
    - Female predicted: 26.9% (S.E. 0.5%, P-Value 0.0%).
    - Difference: -6.0% (S.E. 0.6%, P-Value 0.0%).
    - Explained: 1.7% (S.E. 0.2%, P-Value 0.0%).
    - Unexplained: -7.8% (S.E. 0.6%, P-Value 0.0%).
- Blinder-Oaxaca with Heckman selection correction (log income; selected coefficients):
  - Male coefficient: 8.08 (S.E. 0.06, P-Value 0.00).
  - Female coefficient: 7.94 (S.E. 0.07, P-Value 0.00).
  - Difference: 0.14 (S.E. 0.08, P-Value 0.08).
  - Explained: -0.04 (S.E. 0.02, P-Value 0.01), with component contributions:
    - Potential Experience: 0.00 (S.E. 0.00, P-Value 0.57).
    - Tenure: 0.02 (S.E. 0.00, P-Value 0.00).
    - Education: -0.09 (S.E. 0.01, P-Value 0.00).
    - Occupation: 0.04 (S.E. 0.01, P-Value 0.00).
    - Hours Worked: 0.02 (S.E. 0.00, P-Value 0.00).
    - Government: -0.02 (S.E. 0.00, P-Value 0.00).
    - Self-employed with employees: 0.01 (S.E. 0.00, P-Value 0.00).
    - Self-employed w/o employees: -0.02 (S.E. 0.00, P-Value 0.00).
  - Unexplained: 0.19 (S.E. 0.08, P-Value 0.02).
  - Sample counts reported: N of observations male 12,331; selected male 4,754; N female 12,472; selected female 4,333.

### Year-by-year probit and Mincer regressions (Tables A6–A8)
- Participation probit coefficients by year (15-64 sample; base education primary or less; base age 15-19; standard errors clustered):
  - Female coefficients by year:
    - 2016: -0.43 (0.052)
    - 2017: -0.39 (0.054)
    - 2018: -0.44 (0.056)
    - 2019: -0.27 (0.053)
    - 2020: -0.25 (0.051)
    - 2021: -0.44 (0.066)
  - Education coefficients (Secondary examples):
    - 2016: 0.34 (0.072); 2017: 0.25 (0.075); 2018: 0.24 (0.073); 2019: 0.31 (0.070); 2020: 0.24 (0.069); 2021: 0.32 (0.090).
  - Observations by year: 2016: 5,688; 2017: 5,489; 2018: 5,002; 2019: 4,873; 2020: 4,867; 2021: 3,250.
- Unemployment probit coefficients by year (labor force sample; selected highlights):
  - Female coefficients by year:
    - 2016: 0.33 (0.046)
    - 2017: 0.30 (0.052)
    - 2018: 0.27 (0.060)
    - 2019: 0.25 (0.053)
    - 2020: 0.36 (0.063)
    - 2021: 0.24 (0.072)
  - Education coefficients (Post-secondary examples show strong negative association with unemployment, e.g., 2016: -0.76 (0.11), 2018: -1.20 (0.12)).
  - Observations by year: 2016: 4,719; 2017: 4,484; 2018: 4,125; 2019: 3,951; 2020: 3,430; 2021: 2,571.
- Mincer regressions by year (employed sample; selected coefficients):
  - Female coefficients by year (log gross monthly income):
    - 2016: -0.11 (0.029)
    - 2017: -0.21 (0.037)
    - 2018: -0.20 (0.034)
    - 2019: -0.18 (0.040)
    - 2020: -0.28 (0.062)
  - Education premia by year (examples):
    - Post-secondary: 2016 0.60 (0.070); 2017 0.87 (0.095); 2018 0.66 (0.076); 2019 0.76 (0.090); 2020 0.54 (0.13).
    - Tertiary: 2016 0.79 (0.066); 2017 0.95 (0.094); 2018 0.98 (0.075); 2019 1.08 (0.083); 2020 0.91 (0.16).
  - Observations by year: 2016: 2,506; 2017: 2,080; 2018: 1,944; 2019: 1,812; 2020: 745.
  - R-squared by year: 0.443, 0.435, 0.473, 0.440, 0.415 (2016–2020).

### Ordered logistic regression and predicted income-category probabilities (Tables A9 and Figure A1)
- Sample: employed individuals aged 15-64. Observations: 9,087.
- Ordered logistic estimated coefficients (selected; standard errors in parentheses):
  - Female: -0.59 (0.05).
  - Potential experience: 0.06 (0.01).
  - Tenure (selected): 6 months - 1 year: 0.17 (0.16); 1-5 years: 0.58 (0.12); 5-10 years: 1.13 (0.12); more than 10 years: 1.61 (0.14).
  - Education: secondary 0.76 (0.07); post-secondary 2.06 (0.12); tertiary 2.76 (0.13).
  - Occupation (selected): managers 1.31 (0.15); professionals 1.62 (0.12); clerical support 0.86 (0.10); service and sales 0.29 (0.07).
  - Industry (selected): mining 2.60 (0.37); manufacturing 0.93 (0.17); accommodation 1.82 (0.15).
  - Government: 0.56 (0.10).
  - Self w/ employee: 0.67 (0.19); Self w/o employee: -0.86 (0.10).
  - Log hours: 2.02 (0.13).
  - Cutpoints reported (Cut1–Cut7) with standard errors (e.g., Cut1 5.28 (0.49), Cut7 15.82 (0.53)).
- Predicted probabilities by income bracket (Figure A1 note):
  - Income brackets are in Eastern Caribbean Dollars: Under 200; 201 - 399; 400 - 799; 800 - 1199; 1200 - 1999; 2000 - 3999; 4000 - 5999; Over 6000.
  - Figure A1 shows average predicted probability of being in each income category by gender, with 95 percent confidence intervals; estimated coefficients are in Table A9.

*Annex tables and figures are sourced from the St. Lucia Labor Force Survey 2016-21 and authors’ calculations as presented in the IMF Working Paper "Inequality in a More Equal World—Labor Market Gender Gaps in St. Lucia".*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024117-print-pdf.pdf_