## wpiea2024156-print-pdf

## Source details

**Canonical URL:** [wpiea2024156-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024156-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024156-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024156-print-pdf.pdf.json)

---

### Threshold sensitivity
- Lowering the threshold to one percent leads to an increased share of workers in green jobs.
- Raising the threshold to ten percent results in a decrease in the share of workers in green jobs.
- These threshold adjustments do not alter the overarching finding that the share of green jobs remains similar across countries.

### Data and classification
- Individual-level labor force survey data from five countries: Brazil, Colombia, South Africa, the UK, and the US.
- Occupation classification:
  - International Standard Classification of Occupations for 2008 (ISCO).
  - Jobs classified into 436 groups (4-digit ISCO), with 130 minor (3-digit), 43 sub-major (two-digit) and 10 major groups (1-digit).
- Sample restrictions:
  - Employed individuals aged 16 to 64.
  - Excludes those in military professions.
- Crosswalk and aggregation:
  - Measures of green and polluting that use US 2010 occupation codes are crosswalked to ISCO using US employment weights where there are non-unique matches.
  - Aggregation to three-, two-, and one-digit ISCO groups by simple average of the green task intensity.
  - Minor- and sub-major groups are classified as polluting if at least one unit or sub-major group is polluting.
- Sectoral breakdown uses ISIC: agriculture, construction, manufacturing, market services, non-market services, and mining, energy and water.

### Labor market summary (Table 2, 2023 unless otherwise noted)
- Labour force participation rate (%) by country:
  - Brazil: 70.3
  - Colombia: 68.3
  - South Africa: 62.5
  - United Kingdom: 77.6
  - United States: 72.9
- Female labour force participation rate (%) by country:
  - Brazil: 60.9
  - Colombia: 56.3
  - South Africa: 56.9
  - United Kingdom: 74.2
  - United States: 67.6
- Unemployment rate (%) by country:
  - Brazil: 9.4
  - Colombia: 10.8
  - South Africa: 26.0
  - United Kingdom: 3.8*
  - United States: 3.7
- Average hourly earnings of employees (U.S. dollars) by country:
  - Brazil: 3.12
  - Colombia: 1.88
  - South Africa: 1.38
  - United Kingdom: 23.06
  - United States: 28.72
- Informal employment (%) (available for Brazil, Colombia, South Africa):
  - Brazil: 38.5
  - Colombia: 58.4
  - South Africa: 40.5
- Tertiary school enrollment (% gross) by country:
  - Brazil: 56.8*
  - Colombia: 58.3*
  - South Africa: 24.0
  - United Kingdom: 77.0*
  - United States: 87.9
- Notes:
  - Values marked with an asterisk represent the most recent year.
  - “Labour force participation rate” and “Female labour force participation rate” use ILO modelled estimates for ages 15-64.
  - “Unemployment rate” uses survey data processed by the ILO for ages 15-64.
  - “Average hourly earnings of employee” uses survey data processed by the ILO in current US Dollars.
  - “Informal employment” uses survey data processed by the ILO.

### Aggregates: distribution of green and polluting jobs
- Green jobs share of employment in each country: between 9.4 and 13.2 percent.
- Polluting jobs:
  - AEs: constitute 5 percent of employment.
  - EMs: constitute 9 percent of employment.
- Green intensity:
  - The average greenness score across the working population is 0.5 percentage points higher in AEs than EMs.
- Overarching finding: the share of green jobs is similar across AEs and EMs, though types of green jobs differ.

### Economic structure and occupation distribution
- Occupation grouping: major ISCO groups plotted with skill level in descending order; managers highest skill, elementary occupations lowest.
- Distribution patterns:
  - EMs (Brazil, Colombia, South Africa):
    - Green jobs predominantly cluster among elementary occupations, plant and machine operators, and craft and trade workers.
    - Example: in Brazil these three categories account for 66 percent of green jobs.
  - AEs (UK, US):
    - Green jobs concentrated in managerial, professional, technical positions, and associate professionals.
    - Example: in the UK managers/professionals/associate professionals account for 73 percent of green jobs.
- Green intensity by occupational group:
  - Managers: green intensity close to 10 percent.
  - Professionals, technicians and associate professionals, craft and related trades workers, plant and machine operators and assemblers, and elementary occupations: green intensity around 2 percent.
- Polluting jobs concentration:
  - Predominantly in craft and related trades occupations, with secondary concentration in plant and machine operators and assemblers.
  - Brazil specifics:
    - 63 percent of polluting jobs are within craft and trade occupations.
    - 18 percent in plant and machine operators.
    - Bricklayers and related workers constitute 32 percent of all polluting jobs in Brazil and 47 percent of polluting jobs within crafts and trade category.
    - Agricultural and industrial machinery mechanics: 6 percent of all polluting jobs.
    - Butchers and fishmongers: 6 percent.
    - Sheet-metal workers: 5 percent.
  - UK specifics:
    - Largest share of polluting jobs in craft and trade occupations but only 35 percent (vs 63 percent in Brazil).
    - Professional occupations (e.g., chemical engineer) constitute 14 percent of polluting jobs in the UK versus 1 percent in Brazil.
    - Plant and machinery: 22 percent; elementary occupations: 15 percent (polluting jobs).
- Within major occupation categories (Annex A.3 summary):
  - Managers and professionals: UK and US have almost 35 percent share of green occupations; Colombia has less than 25 percent.
  - Plant and machine operators: US and South Africa have almost 40 percent of these workers in green occupations; Colombia has less than 15 percent.
  - Crafts and trades occupations: more than 40 percent of these workers in Brazil are in polluting occupations, while in the US it is nearly 20 percent.
- Conclusion: most differences in green jobs across countries stem from differences in employment structure rather than differences within major occupation groups.

### Gender patterns
- Overall pattern: both green and polluting jobs are male-dominated across all countries.
- Share of green jobs by gender:
  - EMs: men account for approximately 83 percent of all green jobs; women account for the rest.
  - AEs: men account for 77 percent of all green jobs; women account for approximately 23 percent.
- Green jobs as share of total employment by gender:
  - EMs:
    - Women: green jobs account for 5 percent of women’s total employment on average.
    - Men: green jobs account for almost 17 percent of men’s total employment on average.
  - AEs:
    - Women: green jobs comprise 6 percent of women’s total employment on average.
    - Men: green jobs comprise 19 percent of men’s total employment on average.
- Polluting jobs by gender:
  - Men hold majority of polluting jobs: EMs 82 percent; AEs 77 percent.
  - Polluting jobs as share of total employment:
    - Women: 4 percent.
    - Men: 13 percent.
    - In AEs: 3 percent for women and 8 percent for men.
- Green intensity by gender:
  - UK, US, and South Africa: significantly higher green intensity across the male-dominated population.
  - Brazil and Colombia: slightly lower green intensity for male-held green jobs relative to AEs, explained by a larger share of green jobs in elementary, trade and craft occupations which have lower green intensity scores.

### Age patterns
- Largest share of workers in green jobs: 31 to 40 years old in both AEs and EMs.
- Polluting jobs share slightly increases with age, peaking around 51 to 60 years old.
- Exception: polluting jobs in South Africa appear more concentrated among young workers.
- No large differences observed between countries or genders in age patterns.

### Education patterns
- Education categories: below high school, high school, some college, college degree and above.
- Green jobs by education:
  - Proportion of workers in green jobs is similar across education levels and countries.
  - Below high school: share in green jobs ranges from 10 percent in Colombia to 12 percent in the US.
  - College degree and above: share in green jobs ranges from 10 percent in Brazil to 16 percent in the US.
- Polluting jobs by education:
  - Polluting jobs are more concentrated at lower levels of education across all sample countries.
  - Brazil and South Africa:
    - Less than high school: Brazil 15 percent; South Africa close to 14 percent work in polluting jobs.
    - High school: Brazil 11 percent; South Africa 14 percent.
  - AEs:
    - Average share of polluting jobs for those with high school education and below is 8 percent.
  - Workers with college education:
    - Share of workers in polluting jobs varies between 2 and 3 percent across both AEs and EMs.

### Interaction: Gender and Education (Section 4.6)
- Distribution of green jobs by gender and education:
  - Men’s distribution in green jobs is relatively uniform across all educational levels.
  - Women’s participation in green jobs is skewed towards higher educational attainment.
  - Country examples:
    - Colombia: 3 percent of women with less than high-school education occupy green jobs, versus 11 percent among women with college education.
    - Men in Colombia: 14 percent (less than high-school) versus 19 percent (college).
  - Interpretation: women with higher education are disproportionately represented in green jobs compared with women with lower education; the gender gap in green job participation narrows across education for men relative to women.
- Concentration of green occupations in STEM:
  - Over 60 percent of occupations classified as green are in STEM fields.
  - Out of the 44 ISCO occupations identified as green, 27 are STEM occupations.
  - Aggregation to 1-digit ISCO groups: Green STEM jobs comprise one managerial occupation, 19 professional occupations, and seven technical occupations.
  - Because identified green STEM occupations require more educational investment, analysis restricts to workers with a college degree or higher.
- Decomposition of the green employment gap (formula as provided):
  - Y_gm − Y_gf = (Y_g,Sm − Y_g,Sf) + (Y_g,Mm − Y_g,Mf) + (Y_g,S,Mm − Y_g,S,Mf) + (Y_g,O m − Y_g,O f)
- Findings from decomposition:
  - STEM occupations explain approximately half of the observed disparity in green employment between men and women in Brazil, Colombia, the UK, and the US (among college-educated workers).
  - In South Africa, STEM occupations account for more than a quarter of the gap.
  - Under-representation of women in managerial positions is a second key factor:
    - South Africa: disparity in managerial roles accounts for 56 percent of the gap among college-educated individuals.
    - United Kingdom: managerial gap contributes 39 percent of the overall discrepancy.
    - Brazil, Colombia, and the US: shortfall in women holding managerial positions contributes 23 percent of the gap among college-educated individuals.
  - Overall implication: both lower female representation in STEM fields and lower female representation in managerial positions are major drivers of the gender employment gap in green jobs among college-educated workers.

### Regression analysis: Gender and Job’s Greenness (Section 5.3)
- Mincerian regression specification:
  - ln(w_{i,t,c}) = α_c + β_f_c · I[Female_{i,t,c} = 1] + β_g_c · I[Green_{i,t,c} = 1] + β_fg_c · I[Female_{i,t,c} = 1]·I[Green_{i,t,c} = 1] + γ_c · X_{i,t,c} + ε_{i,t,c}
  - I[Green_{i,t,c} = 1]: green job indicator (Green is measured by having a green job index at least of 5%).
  - X_{i,t,c}: controls for age, education, male and female marriage status, sectors, and informality (for emerging countries).
- Main regression findings (Table 5):
  - Brazil:
    - Female (β_f_c): -0.22 ∗∗∗ (0.004)
    - Green (β_g_c): 0.04 ∗∗∗ (0.003)
    - Female×Green (β_fg_c): 0.02 ∗∗ (0.008)
    - Observations: 565,452
    - R-squared: 0.393
  - Colombia:
    - Female (β_f_c): -0.14 ∗∗∗ (0.006)
    - Green (β_g_c): 0.09 ∗∗∗ (0.006)
    - Female×Green (β_fg_c): 0.07 ∗∗∗ (0.011)
    - Observations: 119,801
    - R-squared: 0.539
  - United Kingdom:
    - Female (β_f_c): -0.11 ∗∗∗ (0.007)
    - Green (β_g_c): 0.04 ∗∗∗ (0.010)
    - Female×Green (β_fg_c): 0.07 ∗∗∗ (0.018)
    - Observations: 30,028
    - R-squared: 0.393
  - United States:
    - Female (β_f_c): -0.15 ∗∗∗ (0.001)
    - Green (β_g_c): 0.11 ∗∗∗ (0.001)
    - Female×Green (β_fg_c): 0.05 ∗∗∗ (0.002)
    - Observations: 5,563,305
    - R-squared: 0.291
- Interpretation:
  - Gender wage gap in non-green jobs (β_f_c): women earn lower hourly wages than men across all countries (examples reported: 11 percent in the UK, 14 percent in Colombia).
  - Green wage premium for men (β_g_c): positive across countries (examples reported: 4 percent in Brazil, 11 percent in the US).
  - Interaction effect (β_fg_c): positive and statistically significant across countries; supplementary effects for women in green jobs:
    - Brazil: 2 percentage points
    - United States: 5 percentage points
    - Colombia: 7 percentage points
    - United Kingdom: 7 percentage points
  - Interpretation: green jobs disproportionately increase the wages of women, suggesting a smaller gender pay gap in green employment after controls.
  - Robustness: Annex C reports further robustness exercises controlling for NAICS 2-digit sector level; results remain broadly unchanged.

### AI exposure, complementarity, and greenness (Section 6.1–6.3)
- AI exposure index:
  - Felten et al. (2021)’s AI Occupational Exposure (AIOE), normalized between zero and one; correlates 10 AI applications with 52 occupational abilities.
- AI complementarity:
  - Pizzinelli et al. (2023) approach using O*NET work contexts and job zones to capture social, ethical, legal, and technical factors that determine complementarity beyond exposure.
  - Complementarity index constructed from six components (arithmetic mean), normalized to 0–1.
- Six components:
  a. Communication: 1) Face-to-Face, 2) Public Speaking.
  b. Responsibility: 3) Responsibility for outcomes, 4) Responsibility for others’ health.
  c. Physical Conditions: 5) Outdoors Exposed, 6) Physical Proximity.
  d. Criticality: 7) Consequence of Error, 8) Freedom of Decisions, 9) Frequency of Decisions.
  e. Routine: 10) Degree of Automation (inverted scale), 11) Unstructured vs Structured Work.
  f. Skills: 12) Job Zones.
- AI and greenness distributions:
  - U-shaped relationship: jobs with lowest and highest AI exposure have the highest greenness scores (examples: environmental engineers vs refuse sorters).
  - Employment distribution by AI exposure and greenness:
    - AEs:
      - Approximately 75 percent of green jobs fall into the high AI exposure category.
      - Around 60 percent of non-green jobs are in the high AI exposure category.
    - EMs:
      - Both green and non-green jobs have lower AI exposure, each around 40 percent in the high exposure category.
  - Complementarity patterns:
    - UK and US: roughly 60 percent of green jobs have high complementarity with AI, versus only 30 percent of non-green jobs.
    - In AEs: non-green jobs are highly exposed but lack the same level of complementarity as green jobs.
    - In EMs: green jobs demonstrate higher complementarity (around 35 percent) compared to non-green jobs (below 20 percent).
- Polluting jobs and AI:
  - Polluting jobs generally show significantly lower AI exposure across examined countries.
  - UK: only 21 percent of polluting job workers are exposed to AI (compared with 75 percent in green jobs and 66 percent in neutral jobs).
  - EMs: exposure to AI among polluting jobs is notably minimal, with less than 5 percent of such jobs being AI-exposed.
  - In both AEs and EMs, AI exposure in polluting jobs is primarily in jobs with high complementarity rather than substitution.

### AI, green jobs, and gender (Section 6.3)
- Gender differences:
  - Women in green jobs face higher levels of AI exposure than men across the sampled countries; disparity less pronounced in AEs than in EMs.
  - Drivers:
    - In EMs, lower AI exposure among men in green jobs is due to prevalence of manual occupations for men.
    - In AEs, green jobs are more professional, technical, or managerial, with greater AI exposure.
- Potential for women to benefit from AI:
  - Women hold a larger proportion of employment in professional and technical occupations, which present substantial potential for complementarity with AI.
  - Managerial positions (highest complementarity) are typically male-dominated but fewer in number than professional and technical occupations.
- Gender patterns in polluting jobs:
  - Within polluting jobs women exhibit higher exposure and greater complementarity with AI than men.
  - Overall: women occupy more cognitive roles (higher exposure and complementarity), men occupy more manual and non-cognitive roles (lower exposure).
- Education and AI exposure in green jobs:
  - College-educated workers in green jobs face higher AI exposure than those with lower educational levels.
  - Around 90 percent of college-educated workers, across most countries, are engaged in occupations with high AI exposure.
  - Workers without a high school diploma are largely in elementary occupations and have significantly lower AI exposure; in Brazil, Colombia, and the US, fewer than 20 percent of these workers are in high-exposure occupations.
  - UK and South Africa: over 40 and 30 percent, respectively, of workers with only a middle school education or less are in high-exposure occupations.
  - Complementarity potential by education:
    - Across all countries, over 70 percent of college-educated workers in green jobs could benefit from high AI complementarity.

### Green Jobs, AI, and Labor Market Returns (Section 6.4)
- Expanded regression specification includes AI exposure/complementarity indicators:
  - I[HELC_{i,t,c} = 1]: high-exposure and low-complementarity to AI indicator.
  - I[HEHC_{i,t,c} = 1]: high-exposure and high-complementarity to AI indicator.
  - Interaction terms: Female×Green, Female×HELC, Female×HEHC, Green×HELC, Green×HEHC, Female×Green×HELC, Female×Green×HEHC.
- Table 6 — Coefficient highlights:
  - Brazil (Observations: 565,452; R-squared: 0.395):
    - Female: -0.15 ∗∗∗ (0.004)
    - Green: 0.04 ∗∗∗ (0.004)
    - HELC: 0.07 ∗∗∗ (0.004)
    - HEHC: 0.10 ∗∗∗ (0.004)
    - Female×Green: -0.09 ∗∗∗ (0.018)
    - Female×Green×HEHC: 0.22 ∗∗∗ (0.021)
  - Colombia (Observations: 119,801; R-squared: 0.543):
    - Female: -0.12 ∗∗∗ (0.007)
    - Green: 0.07 ∗∗∗ (0.007)
    - HELC: 0.17 ∗∗∗ (0.007)
    - HEHC: 0.16 ∗∗∗ (0.008)
    - Green×HELC: -0.29 ∗∗∗ (0.021)
    - Female×Green×HELC: 0.43 ∗∗∗ (0.050)
  - United Kingdom (Observations: 30,028; R-squared: 0.399):
    - Female: -0.07 ∗∗∗ (0.011)
    - Green: 0.06 ∗∗∗ (0.019)
    - HELC: 0.11 ∗∗∗ (0.012)
    - HEHC: 0.18 ∗∗∗ (0.013)
    - Note: some table formatting shows inconsistent concatenations in reported cells; numbers preserved as presented.
  - United States (Observations: 5,563,305; R-squared: 0.296):
    - Female: -0.16 ∗∗∗ (0.001)
    - Green: 0.11 ∗∗∗ (0.002)
    - HELC: 0.11 ∗∗∗ (0.001)
    - HEHC: 0.18 ∗∗∗ (0.001)
    - Female×Green: 0.06 ∗∗∗ (0.006)
    - Green×HELC: -0.11 ∗∗∗ (0.003)
- Key empirical findings:
  - Green wage premium for men: positive and significant across countries for men, ranging between 4 and 11 percent for workers in low-exposure to AI jobs.
  - For men in HEHC occupations the green premium is generally still positive across most countries.
  - For men in HELC occupations the green premium does not exist or becomes negative in Colombia.
  - Gender pay gap and AI complementarity in green jobs:
    - Overall: gender pay gap is lower in green jobs that have more potential to complement AI.
    - Low-exposure occupations:
      - Brazil and Colombia: gender pay gap in green jobs is larger than in non-green counterparts by 9 and 7 percentage points, respectively.
      - United Kingdom and United States: gender pay gap in low-exposed green jobs is smaller than in non-green counterparts by between 7 and 10 percentage points.
    - HELC occupations:
      - Colombia and the US show statistically significant differences: green jobs have a lower gender pay gap than non-green counterparts by 36 and 5 percentage points, respectively.
    - HEHC occupations:
      - Overall gender pay gap in green jobs is lower than in non-green counterparts.
      - Brazil and the United States exhibit statistically significant differences at 12 and 8 percentage points, respectively.

### Implications, unresolved questions, and research needs
- Main conclusions:
  - Similar proportions of workers in green jobs across five economies but large disparities in occupational distribution: in AEs green jobs concentrate among high-skilled and cognitive occupations; in EMs many green jobs are manual, especially in construction.
  - Green jobs are disproportionately held by men in both AEs and EMs, driven by underrepresentation of women in STEM fields and managerial roles.
  - There is a green wage premium and narrower gender pay gaps in green jobs.
  - Many green jobs are well-positioned to harness AI benefits; green jobs with greater capacity to leverage AI exhibit reduced gender pay gaps, potentially attracting more women to green occupations.
- Identified gaps and priorities for further work:
  - Need for further studies examining the resilience and expansion potential of green jobs over time.
  - Need to identify factors that could enhance or eliminate the current green jobs’ wage premium, particularly for women.
  - Need for in-depth analysis of the evolving impact of AI on the green economy, wage structures, and the gender gap to inform policies promoting an inclusive, equitable green job market.

*Source: wpiea2024156-print-pdf - Annex A and Figure A.1 show the importance of threshold to determine the share of workers in green jobs*

### Annex A and Figure A.1 show the importance of threshold to determine the share of workers in green

### Annex A and Figure A.1 show the importance of threshold to determine the share of workers in green jobs

### Threshold sensitivity
- Lowering the threshold to one percent leads to an increased share of workers in green jobs.
- Raising the threshold to ten percent results in a decrease in the share of workers in green jobs.
- These threshold adjustments do not alter the overarching finding that the share of green jobs remains similar across countries.

### Data
- Individual-level data from labor force surveys covering five countries: Brazil, Colombia, South Africa, the UK, and the US.
- Occupation classification:
  - International Standard Classification of Occupations for 2008 (ISCO).
  - Jobs classified into 436 groups (4-digit ISCO), with 130 minor (3-digit), 43 sub-major (two-digit) and 10 major groups (1-digit).
- Sample restrictions:
  - Employed individuals aged 16 to 64.
  - Excludes those in military professions.
- Crosswalk and aggregation:
  - Measures of green and polluting that use US 2010 occupation codes are crosswalked to ISCO using US employment weights where there are non-unique matches.
  - Aggregation to three-, two-, and one-digit ISCO groups by simple average of the green task intensity.
  - Minor- and sub-major groups are classified as polluting if at least one unit or sub-major group is polluting.
- Sectoral breakdown uses ISIC: agriculture, construction, manufacturing, market services, non-market services, and mining, energy and water.

### Labor market summary (Table 2, 2023 unless otherwise noted)
- Labour force participation rate (%) by country:
  - Brazil: 70.3
  - Colombia: 68.3
  - South Africa: 62.5
  - United Kingdom: 77.6
  - United States: 72.9
- Female labour force participation rate (%) by country:
  - Brazil: 60.9
  - Colombia: 56.3
  - South Africa: 56.9
  - United Kingdom: 74.2
  - United States: 67.6
- Unemployment rate (%) by country:
  - Brazil: 9.4
  - Colombia: 10.8
  - South Africa: 26.0
  - United Kingdom: 3.8*
  - United States: 3.7
- Average hourly earnings of employees (U.S. dollars) by country:
  - Brazil: 3.12
  - Colombia: 1.88
  - South Africa: 1.38
  - United Kingdom: 23.06
  - United States: 28.72
- Informal employment (%) (available for Brazil, Colombia, South Africa):
  - Brazil: 38.5
  - Colombia: 58.4
  - South Africa: 40.5
- Tertiary school enrollment (% gross) by country:
  - Brazil: 56.8*
  - Colombia: 58.3*
  - South Africa: 24.0
  - United Kingdom: 77.0*
  - United States: 87.9
- Notes:
  - Values marked with an asterisk represent the most recent year.
  - “Labour force participation rate” and “Female labour force participation rate” use ILO modelled estimates for ages 15-64.
  - “Unemployment rate” uses survey data processed by the ILO for ages 15-64.
  - “Average hourly earnings of employee” uses survey data processed by the ILO in current US Dollars.
  - “Informal employment” uses survey data processed by the ILO.

### Aggregates: cross-country distribution of green and polluting jobs
- Green jobs share of employment in each country: between 9.4 and 13.2 percent.
- Polluting jobs:
  - AEs: constitute 5 percent of employment.
  - EMs: constitute 9 percent of employment.
- Green intensity:
  - The average greenness score across the working population is 0.5 percentage points higher in AEs than EMs.
- Overarching finding: the share of green jobs is similar across AEs and EMs, though types of green jobs differ.

### Economic structure and occupation distribution
- Occupation grouping: major ISCO groups plotted with skill level in descending order; managers highest skill, elementary occupations lowest.
- Distribution patterns:
  - EMs (Brazil, Colombia, South Africa):
    - Green jobs predominantly cluster among elementary occupations, plant and machine operators, and craft and trade workers.
    - Example: in Brazil these three categories account for 66 percent of green jobs (examples: refuse workers, building construction laborers, heavy truck drivers).
  - AEs (UK, US):
    - Green jobs concentrated in managerial, professional, technical positions, and associate professionals.
    - Example: in the UK managers/professionals/associate professionals account for 73 percent of green jobs (examples: engineering professionals, financial and investment advisors).
- Green intensity by occupational group:
  - Managers: green intensity close to 10 percent.
  - Professionals, technicians and associate professionals, craft and related trades workers, plant and machine operators and assemblers, and elementary occupations: green intensity around 2 percent.
- Polluting jobs concentration:
  - Predominantly in craft and related trades occupations, with secondary concentration in plant and machine operators and assemblers.
  - Brazil specifics:
    - 63 percent of polluting jobs are within craft and trade occupations.
    - 18 percent in plant and machine operators.
    - Bricklayers and related workers constitute 32 percent of all polluting jobs in Brazil and 47 percent of polluting jobs within crafts and trade category.
    - Agricultural and industrial machinery mechanics: 6 percent of all polluting jobs.
    - Butchers and fishmongers: 6 percent.
    - Sheet-metal workers: 5 percent.
  - UK specifics:
    - Largest share of polluting jobs in craft and trade occupations but only 35 percent (vs 63 percent in Brazil).
    - Professional occupations (e.g., chemical engineer) constitute 14 percent of polluting jobs in the UK versus 1 percent in Brazil.
    - Plant and machinery: 22 percent; elementary occupations: 15 percent (polluting jobs).
- Within major occupation categories (Annex A.3 summary):
  - Managers and professionals: UK and US have almost 35 percent share of green occupations; Colombia has less than 25 percent.
  - Plant and machine operators: US and South Africa have almost 40 percent of these workers in green occupations; Colombia has less than 15 percent.
  - Crafts and trades occupations: more than 40 percent of these workers in Brazil are in polluting occupations, while in the US it is nearly 20 percent.
- Conclusion: most differences in green jobs across countries stem from differences in employment structure rather than differences within major occupation groups.

### Gender
- Overall pattern: both green and polluting jobs are male-dominated across all countries.
- Share of green jobs by gender:
  - EMs: men account for approximately 83 percent of all green jobs; women account for the rest.
  - AEs: men account for 77 percent of all green jobs; women account for approximately 23 percent.
- Green jobs as share of total employment by gender:
  - EMs:
    - Women: green jobs account for 5 percent of women’s total employment on average.
    - Men: green jobs account for almost 17 percent of men’s total employment on average.
  - AEs:
    - Women: green jobs comprise 6 percent of women’s total employment on average.
    - Men: green jobs comprise 19 percent of men’s total employment on average.
- Polluting jobs by gender:
  - Men hold majority of polluting jobs: EMs 82 percent; AEs 77 percent.
  - Polluting jobs as share of total employment:
    - Women: 4 percent.
    - Men: 13 percent.
    - In AEs: 3 percent for women and 8 percent for men.
- Green intensity by gender:
  - UK, US, and South Africa: significantly higher green intensity across the male-dominated population.
  - Brazil and Colombia: slightly lower green intensity for male-held green jobs relative to AEs, explained by a larger share of green jobs in elementary, trade and craft occupations which have lower green intensity scores.

### Age
- Distribution by age cohort:
  - Largest share of workers in green jobs: 31 to 40 years old in both AEs and EMs.
  - Polluting jobs share slightly increases with age, peaking around 51 to 60 years old.
  - Exception: polluting jobs in South Africa appear more concentrated among young workers.
- No large differences observed between countries or genders in age patterns.

### Education
- Education categories: below high school, high school, some college, college degree and above.
- Green jobs by education:
  - Proportion of workers in green jobs is similar across education levels and countries.
  - Below high school: share in green jobs ranges from 10 percent in Colombia to 12 percent in the US.
  - College degree and above: share in green jobs ranges from 10 percent in Brazil to 16 percent in the US.
- Polluting jobs by education:
  - Polluting jobs are more concentrated at lower levels of education across all sample countries.
  - Brazil and South Africa:
    - Less than high school: Brazil 15 percent; South Africa close to 14 percent work in polluting jobs.
    - High school: Brazil 11 percent; South Africa 14 percent.
  - AEs:
    - Average share of polluting jobs for those with high school education and below is 8 percent.
  - Workers with college education:
    - Share of workers in polluting jobs varies between 2 and 3 percent across both AEs and EMs.

*Source: wpiea2024156-print-pdf - Annex A and Figure A.1 show the importance of threshold to determine the share of workers in green jobs*

### 4.6    Interaction Gender and Education

### 4.6    Interaction Gender and Education

### Distribution of green jobs by gender and education
- Figure 6 presents the employment share of workers in green, polluting, green and polluting, and neutral jobs by education and gender.
- Key observed patterns:
  - Men’s distribution in green jobs is relatively uniform across all educational levels.
  - Women’s participation in green jobs is skewed towards higher educational attainment.
  - Country examples (exact values from the source):
    - Colombia: 3 percent of women with less than high-school education occupy green jobs, versus 11 percent among women with college education.
    - Men in Colombia: 14 percent (less than high-school) versus 19 percent (college).
- Interpretation: women with higher education are disproportionately represented in green jobs compared with women with lower education; the gender gap in green job participation narrows across education for men relative to women.

### Concentration of green occupations in STEM
- Over 60 percent of occupations classified as green are in STEM fields.
- 2018 evidence: the average STEM worker earned two-thirds more than those employed in other fields (Funk and Parker (2018)).
- Gender gaps in STEM enrollment:
  - Males are 15 to 17 percentage points more likely than females to enroll in tertiary STEM education in upper-middle-income and high-income economies.
  - In low-income countries the male–female enrollment difference is 7-percentage points.
- Given the concentration of green jobs in STEM and the low share of female STEM graduates, the role of STEM education is examined as a driver of the gender employment gap in green jobs.

### Decomposition of the green employment gap
- The decomposition formula (as given in the source) for green jobs (g) gap between men (m) and women (f) in their employment share (Y):
  - Y_gm − Y_gf = (Y_g,Sm − Y_g,Sf) + (Y_g,Mm − Y_g,Mf) + (Y_g,S,Mm − Y_g,S,Mf) + (Y_g,O m − Y_g,O f)
  - Interpretation: the total gap is decomposed into contributions from non-managerial STEM occupations (S), managerial non-STEM occupations (M), managers in STEM occupations (S,M), and other green occupations (O).
- Identification of STEM occupations:
  - Based on O*NET (2018) categorization, STEM occupations fall into four domains: 1. life and physical science, engineering, mathematics, and information technology occupations; 2. social science occupations; 3. architecture occupations; 4. health occupations.
  - US occupational classifications were crosswalked to 4-digit ISCO codes; duplicated matches averaged using US employment weights to enable international comparability.
  - Out of the 44 ISCO occupations identified as green, 27 are STEM occupations.
- Aggregation to 1-digit ISCO groups:
  - Green STEM jobs comprise: one managerial occupation, 19 professional occupations, and seven technical occupations.
  - According to ISCO, managers, professionals, and technical roles require higher skill levels (3 and 4).
  - Because identified green STEM occupations require more educational investment, the analysis is restricted to workers with a college degree or higher.

### Findings from the decomposition (Figure 7)
- STEM education’s role:
  - In Brazil, Colombia, the UK, and the US, STEM occupations are responsible for approximately half of the observed disparity in green employment between men and women (among college-educated workers).
  - In South Africa, the contribution of STEM occupations to this gender gap exceeds a quarter.
- Managerial representation’s role:
  - Under-representation of women in managerial positions is a second key factor.
  - South Africa: disparity in managerial roles accounts for 56 percent of the gap in green employment among college-educated individuals.
  - United Kingdom: managerial gap contributes 39 percent of the overall discrepancy.
  - Brazil, Colombia, and the US: shortfall in women holding managerial positions contributes 23 percent of the gap in green employment for college-educated individuals.
- Overall implication: both lower female representation in STEM fields and lower female representation in managerial positions are major drivers of the gender employment gap in green jobs among college-educated workers.

*Source: wpiea2024156-print-pdf - 4.6    Interaction Gender and Education*

### 5.3    Regression analysis

### 5.3    Regression analysis

### Mincerian regression specification
- Regression estimated:
  ln(w_{i,t,c}) = α_c + β_f_c · I[Female_{i,t,c} = 1] + β_g_c · I[Green_{i,t,c} = 1] + β_fg_c · I[Female_{i,t,c} = 1]·I[Green_{i,t,c} = 1] + γ_c · X_{i,t,c} + ε_{i,t,c}
- Variables:
  - w_{i,t,c}: wage-per-hour remuneration.
  - I[Female_{i,t,c} = 1]: woman indicator.
  - I[Green_{i,t,c} = 1]: green job indicator (Green is measured by having a green job index at least of 5%).
  - X_{i,t,c}: controls for age, education, male and female marriage status, sectors, and informality (for emerging countries).
- Reported outcomes: percentage change in wage-per-hour remuneration from being a woman (β_f_c), from working in a green job (β_g_c), and from the interaction of being a woman in a green job (β_fg_c).

### Main regression findings (Table 5: Mincerian Regression Analysis: Gender and Job’s Greenness)
- Coefficients (with standard errors in parentheses) and significance:
  - Brazil
    - Female (β_f_c): -0.22 ∗∗∗ (0.004)
    - Green (β_g_c): 0.04 ∗∗∗ (0.003)
    - Female×Green (β_fg_c): 0.02 ∗∗ (0.008)
    - Observations: 565,452
    - R-squared: 0.393
  - Colombia
    - Female (β_f_c): -0.14 ∗∗∗ (0.006)
    - Green (β_g_c): 0.09 ∗∗∗ (0.006)
    - Female×Green (β_fg_c): 0.07 ∗∗∗ (0.011)
    - Observations: 119,801
    - R-squared: 0.539
  - United Kingdom
    - Female (β_f_c): -0.11 ∗∗∗ (0.007)
    - Green (β_g_c): 0.04 ∗∗∗ (0.010)
    - Female×Green (β_fg_c): 0.07 ∗∗∗ (0.018)
    - Observations: 30,028
    - R-squared: 0.393
  - United States
    - Female (β_f_c): -0.15 ∗∗∗ (0.001)
    - Green (β_g_c): 0.11 ∗∗∗ (0.001)
    - Female×Green (β_fg_c): 0.05 ∗∗∗ (0.002)
    - Observations: 5,563,305
    - R-squared: 0.291
- Controls included in all regressions: Age, Age-Squared, Age-Cohort (cohorts of five years from 15 to 64 years old), Education (less than elementary finished, elementary finished, high school finished, college finished or more), Married, Married-Female, Sector (main ISIC aggregates). Informality included for Brazil and Colombia; not included for UK and US.
- Significance notation: ∗ p <0.10, ∗∗ p <0.05, ∗∗∗ p <0.01.

### Interpretation of regression coefficients and key quantitative results
- Gender wage gap in non-green jobs (β_f_c):
  - Women earn lower hourly wages than men across all countries in the sample.
  - Reported gaps: 11 percent in the UK, 14 percent in Colombia (ranges reported in text).
- Green wage premium for men (β_g_c):
  - Men experience a positive wage premium from transitioning to green jobs across all countries.
  - Reported green wage premia: 4 percent in Brazil, 11 percent in the US.
- Interaction effect (β_fg_c):
  - Statistically significant positive interaction between green jobs and women across countries.
  - Supplementary effects for women in green jobs:
    - Brazil: 2 percentage points
    - United States: 5 percentage points
    - Colombia: 7 percentage points
    - United Kingdom: 7 percentage points
  - Interpretation: green jobs disproportionately increase the wages of women, suggesting a smaller gender pay gap in green employment after controlling for other variables.
- Robustness: Annex C reports further robustness exercises controlling for NAICS 2-digit sector level; results remain broadly unchanged.

---

### 6    The Future of Work (selected sections relevant to AI, greenness, and gender)

### 6.1 Data and methodology for AI exposure and complementarity
- AI exposure index used: Felten et al. (2021)’s AI Occupational Exposure (AIOE), normalized between zero and one; correlates 10 AI applications with 52 occupational abilities.
- AI complementarity approach: Pizzinelli et al. (2023) using O*NET work contexts and job zones to capture social, ethical, legal, and technical factors that determine complementarity beyond exposure.
- O*NET inputs:
  - Work contexts: 11 contexts identified as most relevant to AI’s likelihood to replace or be implemented under supervision; aggregated into 5 groups and then into 6 components for this study.
  - Job zones: clusters by education, training, and experience requirements; occupations with longer professional development are expected to be better positioned to incorporate AI complementarity.
- Construction of complementarity index:
  - Six components (arithmetic mean of context scores), then complementary index = arithmetic mean of the six components, normalized to 0–1.

- Six components (as organized in the source):
  a. Communication: 1) Face-to-Face, 2) Public Speaking.
  b. Responsibility: 3) Responsibility for outcomes, 4) Responsibility for others’ health.
  c. Physical Conditions: 5) Outdoors Exposed, 6) Physical Proximity.
  d. Criticality: 7) Consequence of Error, 8) Freedom of Decisions, 9) Frequency of Decisions.
  e. Routine: 10) Degree of Automation (inverted scale), 11) Unstructured vs Structured Work.
  f. Skills: 12) Job Zones.

### 6.2 AI and greenness: distributions and quantitative patterns
- Occupations mapped by greenness score and AIOE exposure: U-shaped relationship—jobs with lowest and highest AI exposure have the highest greenness scores. Example occupations:
  - Environmental engineers: green score close to one and highly exposed to AI.
  - Refuse sorters: green score close to one and much lower exposure.
- Employment distribution by AI exposure and greenness (selected quantitative findings):
  - Advanced economies (AEs):
    - Approximately 75 percent of green jobs fall into the high AI exposure category.
    - Around 60 percent of non-green jobs are in the high AI exposure category.
  - Emerging markets (EMs):
    - Both green and non-green jobs have lower AI exposure, each around 40 percent in the high exposure category.
  - Complementarity patterns:
    - UK and US: roughly 60 percent of green jobs have high complementarity with AI, versus only 30 percent of non-green jobs.
    - In AEs: non-green jobs are highly exposed but lack the same level of complementarity as green jobs.
    - In EMs: green jobs demonstrate higher complementarity (around 35 percent) compared to non-green jobs (below 20 percent).
- Polluting jobs and AI:
  - Polluting jobs generally show significantly lower AI exposure across examined countries.
  - UK: only 21 percent of polluting job workers are exposed to AI (compared with 75 percent in green jobs and 66 percent in neutral jobs).
  - EMs: exposure to AI among polluting jobs is notably minimal, with less than 5 percent of such jobs being AI-exposed.
  - In both AEs and EMs, AI exposure in polluting jobs is primarily in jobs with high complementarity rather than substitution.
  - Interpretation: limited AI exposure in polluting jobs is attributable to their predominantly manual and non-cognitive nature.

### 6.3 AI, green jobs, and gender (key patterns)
- Gender differences in AI exposure and complementarity within green jobs:
  - Women in green jobs face higher levels of AI exposure than men across the sampled countries; this disparity is less pronounced in AEs than in EMs.
  - Drivers:
    - In EMs, the lower AI exposure among men in green jobs is due to the prevalence of manual occupations for men.
    - In AEs, green jobs are more often professional, technical, or managerial, which involves greater AI exposure.
- Potential for women to benefit from AI:
  - Despite higher exposure, women hold a larger proportion of employment in professional and technical occupations, which present substantial potential for complementarity with AI.
  - Managerial positions (highest complementarity with AI) are typically male-dominated but fewer in number than professional and technical occupations.
- Gender patterns in polluting jobs:
  - Although polluting occupations have lower overall AI exposure than green jobs, within polluting jobs women exhibit higher exposure and greater complementarity with AI than men.
  - Overall pattern: women occupy more cognitive roles (higher exposure and complementarity), men occupy more manual and non-cognitive roles (lower exposure).
- Education and AI exposure in green jobs:
  - College-educated workers in green jobs face higher AI exposure than those with lower educational levels.
  - Around 90 percent of college-educated workers, across most countries, are engaged in occupations with high AI exposure (primarily professional roles).
  - Workers without a high school diploma are largely in elementary occupations and have significantly lower AI exposure; in Brazil, Colombia, and the US, fewer than 20 percent of these workers are in high-exposure occupations.
  - UK and South Africa: over 40 and 30 percent, respectively, of workers with only a middle school education or less are in high-exposure occupations.
  - Complementarity potential by education:
    - Across all countries, over 70 percent of college-educated workers in green jobs could benefit from high AI complementarity.
    - Workers with middle school education are typically in occupations less exposed to AI, but among those exposed, the majority stand to benefit from AI complementarity.

*Source: wpiea2024156-print-pdf - 5.3    Regression analysis*

### 6.4    Green Jobs, AI, and Labor Market Returns

### 6.4    Green Jobs, AI, and Labor Market Returns

### Regression specification and variables
- Expanded regression equation (3) (Mincerian log-wage specification) estimates ln(w_{i,t,c}) with the following indicators and controls:
  - I[Female_{i,t,c} = 1]: woman indicator
  - I[Green_{i,t,c} = 1]: green job indicator
  - I[HELC_{i,t,c} = 1]: high-exposure and low-complementarity to AI indicator
  - I[HEHC_{i,t,c} = 1]: high-exposure and high-complementarity to AI indicator
  - Interaction terms: Female×Green, Female×HELC, Female×HEHC, Green×HELC, Green×HEHC, Female×Green×HELC, Female×Green×HEHC
  - X_{i,t,c}: control vector including age, age-squared, age-cohorts, education, male and female marriage status, sectors, and informality (for emerging countries)
- Outcome: percentage change in wage-per-hour remuneration associated with each indicator and their interactions.
- Note: As in Section 5 the analysis does not consider South Africa.

### Table 6 — Coefficient estimates (Gender, Greenness, and AI exposure)
- Coefficients and standard errors reported per country (numbers in parentheses are standard errors). Statistical significance indicated by
  - ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
- Brazil
  - Female (β_f_c): -0.15 ∗∗∗ (0.004)
  - Green (β_g_c): 0.04 ∗∗∗ (0.004)
  - HELC (β_l_c): 0.07 ∗∗∗ (0.004)
  - HEHC (β_h_c): 0.10 ∗∗∗ (0.004)
  - Female×Green (β_fg_c): -0.09 ∗∗∗ (0.018)
  - Female×HELC (β_fl_c): -0.12 ∗∗∗ (0.005)
  - Female×HEHC (β_fh_c): -0.21 ∗∗∗ (0.005)
  - Green×HELC (β_gl_c): -0.03 ∗∗ (0.012)
  - Green×HEHC (β_gh_c): 0.02 ∗∗ (0.008)
  - Female×Green×HELC (β_fgl_c): 0.08 ∗∗∗ (0.027)
  - Female×Green×HEHC (β_fgh_c): 0.22 ∗∗∗ (0.021)
  - Observations: 565,452
  - R-squared: 0.395
- Colombia
  - Female (β_f_c): -0.12 ∗∗∗ (0.007)
  - Green (β_g_c): 0.07 ∗∗∗ (0.007)
  - HELC (β_l_c): 0.17 ∗∗∗ (0.007)
  - HEHC (β_h_c): 0.16 ∗∗∗ (0.008)
  - Female×Green (β_fg_c): -0.07 ∗ (0.037)
  - Female×HELC (β_fl_c): -0.07 ∗∗∗ (0.009)
  - Female×HEHC (β_fh_c): -0.06 ∗∗∗ (0.009)
  - Green×HELC (β_gl_c): -0.29 ∗∗∗ (0.021)
  - Green×HEHC (β_gh_c): 0.11 ∗∗∗ (0.013)
  - Female×Green×HELC (β_fgl_c): 0.43 ∗∗∗ (0.050)
  - Female×Green×HEHC (β_fgh_c): 0.07 ∗ (0.041)
  - Observations: 119,801
  - R-squared: 0.543
- United Kingdom
  - Female (β_f_c): -0.07 ∗∗∗ (0.011)
  - Green (β_g_c): 0.06 ∗∗∗ (0.019)
  - HELC (β_l_c): 0.11 ∗∗∗ (0.012)
  - HEHC (β_h_c): 0.00 0.18 ∗∗∗ (0.013)  [table formatting shows "0.000.18 ∗∗∗"; interpreted here as HEHC coefficient reported as 0.18 ∗∗∗ with (0.013)]
  - Female×Green (β_fg_c): 0.10 0.06 ∗∗∗ (0.077)  [table formatting shows "0.100.06 ∗∗∗"; gender×green row contains formatting inconsistencies; preserve numbers as presented]
  - Female×HELC (β_fl_c): -0.04 ∗∗∗ (0.014)
  - Female×HEHC (β_fh_c): -0.03 ∗∗ (0.014)
  - Green×HELC (β_gl_c): -0.02 (0.032)
  - Green×HEHC (β_gh_c): 0.01 (0.023)
  - Female×Green×HELC (β_fgl_c): -0.02 (0.089)
  - Female×Green×HEHC (β_fgh_c): -0.04 (0.080)
  - Observations: 30,028
  - R-squared: 0.399
- United States
  - Female (β_f_c): -0.16 ∗∗∗ (0.001)
  - Green (β_g_c): 0.11 ∗∗∗ (0.002)
  - HELC (β_l_c): 0.11 ∗∗∗ (0.001)
  - HEHC (β_h_c): 0.18 ∗∗∗ (0.001)
  - Female×Green (β_fg_c): 0.06 ∗∗∗ (0.006)
  - Female×HELC (β_fl_c): 0.00 ∗∗ (0.002)
  - Female×HEHC (β_fh_c): 0.02 ∗∗∗ (0.002)
  - Green×HELC (β_gl_c): -0.11 ∗∗∗ (0.003)
  - Green×HEHC (β_gh_c): -0.03 ∗∗∗ (0.002)
  - Female×Green×HELC (β_fgl_c): -0.02 ∗∗ (0.008)
  - Female×Green×HEHC (β_fgh_c): 0.01 (0.007)
  - Observations: 5,563,305
  - R-squared: 0.296
- Controls included in all regressions: Age, Age-Squared, Age-Cohort, Education, Married, Married-Female, Sector. Informality included for Brazil and Colombia; not included for UK and US.

### Key empirical findings and interpretation
- Green wage premium for men:
  - Statistically significant positive green jobs premium across countries for men, ranging between 4 and 11 percent for workers in low-exposure to AI jobs.
  - For men in HEHC occupations (high-exposure and high-complementarity to AI), the green premium is generally still positive across most countries.
  - For men in HELC occupations (high-exposure and low-complementarity to AI), the green premium does not exist or becomes negative in Colombia.
  - These findings add granularity to the earlier result on the existence of a green wage premium (Section 5).
- Gender pay gap and AI complementarity in green jobs (summary of Figure 17 analysis):
  - Overall: the gender pay gap is lower in green jobs that have more potential to complement AI.
  - Low-exposure to AI occupations:
    - Brazil and Colombia: gender pay gap in green jobs is larger than in non-green counterparts by 9 and 7 percentage points, respectively.
    - United Kingdom and United States: gender pay gap in low-exposed green jobs is smaller than in non-green counterparts, a difference ranging between 7 and 10 percentage points.
    - Note: the gender pay gap difference between green and non-green jobs for Colombia and the UK are not statistically significant.
  - High-exposure and low-complementarity (HELC) occupations:
    - Colombia and the US show statistically significant differences: green jobs have a lower gender pay gap than non-green counterparts by 36 and 5 percentage points, respectively.
  - High-exposure and high-complementarity (HEHC) occupations:
    - Overall gender pay gap in green jobs is lower than in non-green counterparts.
    - Brazil and the United States exhibit statistically significant differences at 12 and 8 percentage points, respectively.

### Implications, unresolved questions, and research needs
- Main conclusions from the chapter:
  - The study documents similar proportions of workers in green jobs across five economies but large disparities in occupational distribution: in AEs green jobs concentrate among high-skilled and cognitive occupations; in EMs many green jobs are manual, especially in construction.
  - Green jobs are disproportionately held by men in both AEs and EMs, driven by underrepresentation of women in STEM fields and managerial roles.
  - There is a green wage premium and narrower gender pay gaps in green jobs.
  - Many green jobs are well-positioned to harness AI benefits; green jobs with greater capacity to leverage AI exhibit reduced gender pay gaps, potentially attracting more women to green occupations.
- Identified gaps and priorities for further work:
  - Need for further studies examining the resilience and expansion potential of green jobs over time.
  - Need to identify factors that could enhance or eliminate the current green jobs’ wage premium, particularly for women.
  - Need for in-depth analysis of the evolving impact of AI on the green economy, wage structures, and the gender gap to inform policies promoting an inclusive, equitable green job market.

*Italic: Source — wpiea2024156-print-pdf - 6.4    Green Jobs, AI, and Labor Market Returns*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024156-print-pdf.pdf_