## 3.1–5.3 Worker Characteristics through Expected Impact of AI Lifetime Earnings

## Source details

**Canonical URL:** [3.1–5.3 Worker Characteristics through Expected Impact of AI Lifetime Earnings](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024116-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024116-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024116-print-pdf.pdf.json)

---

### Worker demographic and education summary
- Median Worker Age: UK 42, Brazil 36
- Share of Women in Employment (%): UK 47, Brazil 42.5
- Share of Workers with a College Degree (%): UK 36.7, Brazil 17.3
- Share of Workers in High AI Exposure Occupations (%): UK 66.4, Brazil 39.7
- Median Hourly Wage (2019 PPP Dollars): UK 18.4, Brazil 3.8
- Informality Rate (%): Brazil 42 (note: informality rate defined only for Brazil)

### Key implications from characteristics
- Median worker age: Brazil 36 versus UK 42.
- Female participation: UK 47 percent versus Brazil 42 percent.
- Educational attainment: UK 36.7 percent college-educated versus Brazil 17.3 percent.
- Median wage disparity: UK median wage 18.4 (2019 PPP) versus Brazil 3.8 (2019 PPP).
- AI exposure concentration: UK 66.4 percent in high-exposure occupations versus Brazil 39.7 percent.
- Informality: Brazil near 42 percent of workers classified as informal.
- Cross-country observation: Educational disparity correlates with wage differences and with differing exposure to AI.

### Aggregate quarterly transition matrices (shares in subsequent quarter)
- UK Employment Flows (Table 2)
  - Employed → Employed: 97.8
  - Employed → Unemployed: 0.9
  - Employed → Not in labor force: 1.3
  - Unemployed → Employed: 24.5
  - Unemployed → Unemployed: 60.1
  - Unemployed → Not in labor force: 15.4
  - Not in labor force → Employed: 3.7
  - Not in labor force → Unemployed: 4.3
  - Not in labor force → Not in labor force: 92.3

- Brazil Employment Flows (Table 3)
  - Employed → Employed: 90.7
  - Employed → Unemployed: 3.2
  - Employed → Not in labor force: 6.1
  - Unemployed → Employed: 31.9
  - Unemployed → Unemployed: 42.9
  - Unemployed → Not in labor force: 25.2
  - Not in labor force → Employed: 13.3
  - Not in labor force → Unemployed: 7.0
  - Not in labor force → Not in labor force: 79.7

### Comparative findings on dynamics
- Higher status persistence in the UK across employed, unemployed, and NLF categories.
- UK employed retention: 97.8 versus Brazil 90.7.
- Unemployed→Employed: UK 24.5 versus Brazil 31.9 (longer unemployment duration implied in the UK).
- Brazil: higher short-term separation risk but more frequent transitions from non-employment to employment, indicating greater short-term dynamism.
- Gender patterns (Annex A.1): men more likely to stay employed than women; women more likely to leave the labor force.

### Education — transition matrices and insights
- College-educated workers (Table 4)
  - UK (college)
    - Employed → Employed: 98.0
    - Employed → Unemployed: 0.8
    - Employed → Not in labor force: 1.2
    - Unemployed → Employed: 34.8
    - Unemployed → Unemployed: 50.8
    - Unemployed → Not in labor force: 14.8
    - Not in labor force → Employed: 8.4
    - Not in labor force → Unemployed: 2.6
    - Not in labor force → Not in labor force: 87.4
  - Brazil (college)
    - Employed → Employed: 96.3
    - Employed → Unemployed: 1.5
    - Employed → Not in labor force: 2.2
    - Unemployed → Employed: 31.2
    - Unemployed → Unemployed: 49.5
    - Unemployed → Not in labor force: 19.3
    - Not in labor force → Employed: 15.0
    - Not in labor force → Unemployed: 8.7
    - Not in labor force → Not in labor force: 76.2

- Non-college workers (Table 5)
  - UK (non-college)
    - Employed → Employed: 97.7
    - Employed → Unemployed: 0.9
    - Employed → Not in labor force: 1.4
    - Unemployed → Employed: 21.7
    - Unemployed → Unemployed: 62.6
    - Unemployed → Not in labor force: 15.7
    - Not in labor force → Employed: 3.1
    - Not in labor force → Unemployed: 3.5
    - Not in labor force → Not in labor force: 93.4
  - Brazil (non-college)
    - Employed → Employed: 89.6
    - Employed → Unemployed: 3.5
    - Employed → Not in labor force: 6.9
    - Unemployed → Employed: 31.9
    - Unemployed → Unemployed: 42.3
    - Unemployed → Not in labor force: 25.8
    - Not in labor force → Employed: 13.2
    - Not in labor force → Unemployed: 6.8
    - Not in labor force → Not in labor force: 80.0

- Education-related insights
  - Employed retention among college-educated: UK 98.0 versus Brazil 96.3.
  - Non-college retention: UK 97.7 versus Brazil 89.6.
  - UK exhibits a large education gap in transitions from non-employment to employment; Brazil shows a smaller gap.
  - Suggests higher return to education for job security in Brazil and more dynamic opportunities for lower-education workers in Brazil.

### Occupation and job transitions — definitions and country comparisons
- Definitions:
  - "switch employer": current employer tenure < three months and employed previous quarter.
  - "switch occ.": occupation change (4-digit ISCO-08 for Brazil and UK).
- UK job & occupation switching (Table 6)
  - Same Employer & Same Occ.: 87.3
  - Same Employer & Switch Occ.: 10.7
  - Switch Employer & Same Occ.: 0.8
  - Switch Employer & Switch Occ.: 1.2
  - Marginal occupation switching rate: 11.9
  - Marginal job switching rate: 2.0
- Brazil job & occupation switching (Table 7)
  - Same Employer & Same Occ.: 65.2
  - Same Employer & Switch Occ.: 32.4
  - Switch Employer & Same Occ.: 0.8
  - Switch Employer & Switch Occ.: 1.6
  - Marginal occupation switching rate: 34.0
  - Marginal job switching rate: 2.4

- Comparative findings on mobility
  - UK: high job and occupational stability (87.3 percent same employer & same occupation).
  - Brazil: lower combined stability (65.2 percent) and higher occupation switching (34 percent).
  - Robustness: higher occupation switching in Brazil persists at 3-digit and 2-digit ISCO levels.
  - Annex A.2: college-educated slightly more likely to switch occupations than non-college in both countries.

### AI exposure classification and earnings patterns
- Occupation categories:
  - Low Exposure (LE)
  - High Exposure, High Complementarity (HEHC)
  - High Exposure, Low Complementarity (HELC)
- Earnings patterns:
  - HEHC occupations: tend to have the highest wages.
  - HELC occupations: fairly homogeneous across the income distribution.
  - LE occupations: concentrated in elementary and agricultural sectors, tend to be lower-paid.

### Distribution of employment by AI exposure and education (Table 8, percent)
- UK (No College): LE 29, HELC 19.1, HEHC 14.9, Total 63.7
- UK (College): LE 4.3, HELC 12.6, HEHC 20.1, Total 37.2
- UK Total: LE 33.3, HELC 31.7, HEHC 35.0, Total 100
- Brazil (No College): LE 35.7, HELC 15.8, HEHC 9.2, Total 82.7
- Brazil (College): LE 2.6, HELC 5.0, HEHC 9.7, Total 17.3
- Brazil Total (reported rows in table): Total 20.8, 18.9 (preserve reported row entries)
- Note from table: AI exposure is highly correlated with education: over 85% of college-educated workers are in highly exposed occupations in both the UK and Brazil.

### Historical occupation switching across AI exposure categories
- Conditional persistence: when workers switch occupations they are more likely to move within the same broad category (LE, HELC, HEHC) than across categories.
- Brazil: greater probability of moving to LE occupations conditional on switching (more “downward” moves) compared with the UK.
- UK: workers more likely to move to HEHC occupations conditional on switching, including movement from LE to HEHC.

### Annex A.3 and location heterogeneity
- Transition probabilities robust to exclusions (e.g., public sector).
- Rural vs urban: digital infrastructure limitations in rural areas affect job composition and likelihood of transitions to or from AI-exposed jobs.
- College-educated: higher chances of moving “upwards” to HEHC occupations in both countries; Brazil shows a slightly higher upward mobility premium for education.
- Non-college: higher chances of moving “downwards” in Brazil; example life-cycle stat: an estimated 80 percent of Brazilian workers without a college degree at age 40 are in jobs that have low exposure to AI.

### Regression specification for occupational switches
- Main estimating equation (occupational-switch dummy outcome):
  - y_kirt = α + Σ_{j≠HEHC} δ_j C_jir(t−1) + β X_irt + γ_t + η_r + ε_irt
  - y_kirt: dummy = 1 if worker switched occupations to category k (LE, HELC, or HEHC).
  - X_irt includes age, gender, education, and informality for Brazil.
  - γ_t and η_r are time and region fixed effects.
  - Base category: employed in HEHC, aged below 25, male, middle school education or below.

### Key regression results — UK (Tables 9)
- Education effects:
  - Holding a degree increases probability of moving into HEHC by 2.26 percentage points (relative to middle school or below).
  - Higher education decreases likelihood of remaining in or moving to LE occupations.
- Gender and age:
  - Being a woman reduces probability of switching to HEHC by approximately 0.63 p.p. and to LE by about 1.12 p.p.
  - Older workers (45-59 and 60+) show reduced likelihood of switching occupations relative to young workers.
- Model and sample:
  - Observations: 4552677, 4552677, 4552677 (columns (1)(2)(3))
  - R^2: 0.08, 0.058, 0.061
  - State FE: Yes; Year-Quarter FE: Yes

### Key regression results — Brazil (Tables 10)
- Education effects:
  - Holding a college degree increases likelihood of moving into HEHC by 11.4 percentage points (relative to middle school or below).
  - Higher education decreases probabilities of staying in or moving to LE occupations.
- Gender and age:
  - Being a woman reduces probability of switching to HEHC by about 1.02 p.p.
  - Being a woman increases likelihood of transitioning to HELC by 1.10 p.p. and decreases probability of moving to LE by 7.37 p.p.
  - Older workers (45-59 and 60+) more inclined to switch to HEHC relative to young workers but less likely to switch to HELC and LE.
- Model and sample:
  - Observations: 4552677, 4552677, 4552677 (columns (1)(2)(3))
  - R^2: 0.08, 0.058, 0.061
  - State FE: Yes; Year-Quarter FE: Yes

### Life-cycle dynamics — employment shares and interpretation
- Estimation: polynomial in age with female dummy and time FE.
- Findings:
  - Young-age transitions common: movement from HELC to HEHC before age 40 (steep HEHC increase, HELC decline).
  - LE share relatively constant over age.
  - Life-cycle profiles for college-educated very similar in UK and Brazil; non-college profiles differ markedly (more so in Brazil).
- Interpretation:
  - HELC occupations likely serve as entry-level stepping stones into HEHC jobs in late twenties and thirties.
  - AI disruption to HELC jobs could impede entry into HEHC roles and wage growth, particularly for college graduates.
- Caveat: results may reflect cohort or composition effects; data covers less than a full decade.

### Life-cycle dynamics — wages and wage premia for transitions
- Wage polynomial: log hourly earnings as polynomial in age plus controls.
- Descriptive wage dynamics:
  - Most wage growth up to 35-40 years old, coinciding with career changes.
  - Higher average wage growth for occupations more exposed to AI, particularly for college graduates.
- Panel regression for wage changes (annual panel) specification:
  - Δ log(y_irt) = δ1 J2J_irt + δ2 OS_irt × J2J_irt + δ3 EUE_irt + Σ_k θ_k C_kir(t−1) C_kirt + Σ_k Σ_j OS_irt φ_kj C_kir(t−1) C_jirt + β X_irt + γ_t + η_r + ε_irt
  - OS: occupation switch dummy; J2J: job switch dummy; EUE: transitions through unemployment.
  - θ_k: average log wage change for stayers in k; φ_kj: average log wage change for switchers k→j.
  - Wage premium for HELC→HEHC: φ_HELC,HEHC − θ_HELC.

- Empirical findings:
  - Premium for switching upwards (LE→HE and HELC→HEHC) in both countries.
  - Upward transitions especially benefit young college-educated workers.
  - Downward transitions in Brazil associated with wage penalty; non-college workers face greater risk of moving to LE and wage compression.
- Switching statistics:
  - Average yearly J2J switching rate: 9.2 percent (UK) and 11.7 percent (Brazil).
  - Average occupation switching rate: 28 percent (UK) and 39 percent (Brazil).

### Annex B wage variation and premia (selected estimates)
- Table B.1 selected coefficients:
  - Switch Occ: UK 0.0166 *** ; Brazil 0.00560 ***
  - Switch Emp: UK 0.0454 *** ; Brazil 0.00876 *
  - Occup. and Employer Change: UK -0.0224 ; Brazil -0.0150 ***
  - EUE Employer Change: UK -0.115 *** ; Brazil -0.0280 ***
  - Age 25-44: UK -0.0234 *** ; Brazil -0.0423 ***
  - College: UK 0.0692 ** ; Brazil 0.0553 ***
  - Is Informal: Brazil -0.0739 ***
  - Observations: UK 50700 ; Brazil 687501
  - R2: UK 0.007 ; Brazil 0.005
- Table B.2 selected transition premia:
  - Switch Emp: UK 0.0461 *** ; Brazil 0.00858 *
  - Switch Occ: UK 0.00582 ; Brazil 0.00391
  - Switch Emp × Switch Occ: UK -0.0255 ; Brazil -0.0172 ***
  - Observations: UK 50700 ; Brazil 687501
  - R2: UK 0.007 ; Brazil 0.007
- Interpretation:
  - Wage growth greater for younger college workers.
  - Informality carries a wage penalty in Brazil (-0.0739 log points).
  - Transitioning through unemployment raises probability of becoming informal and reduces wage growth.

### AI exposure employment flows (Annex A.3 selected)
- UK (Table A.5)
  - Employed (HE) → Employed (HE): 96.2
  - Employed (HE) → Employed (LE): 1.7
  - Employed (HE) → Unemployed: 0.8
  - Employed (HE) → NLF: 1.3
  - Employed (LE) → Employed (HE): 3.3
  - Employed (LE) → Employed (LE): 94.2
  - Unemployed → Employed (HE): 13.7
  - Unemployed → Employed (LE): 10.8
  - Unemployed → Unemployed: 60.1
  - NLF → Employed (HE): 2.6
  - NLF → Employed (LE): 1.4
  - NLF → Unemployed: 3.7
  - NLF → NLF: 92.3

- Brazil (Table A.6)
  - Employed (HE) → Employed (HE): 81.4
  - Employed (HE) → Employed (LE): 11.2
  - Employed (HE) → Unemployed: 7.4
  - Employed (HE) → NLF: 4.9
  - Employed (LE) → Employed (HE): 7.1
  - Employed (LE) → Employed (LE): 82.3
  - Employed (LE) → Unemployed: 3.7
  - Employed (LE) → NLF: 6.9
  - Unemployed → Employed (HE): 10.2
  - Unemployed → Employed (LE): 21.7
  - Unemployed → Unemployed: 42.9
  - Unemployed → NLF: 25.2
  - NLF → Employed (HE): 4.3
  - NLF → Employed (LE): 77
  - NLF → Unemployed: 79.7 (table formatting indicates high shares for LE from NLF)

- Interpretations:
  - LE workers face higher transitions to unemployment and NLF.
  - In Brazil, non-working individuals more likely to return via LE jobs; opposite in the UK.
  - Upward LE→HE transitions more common in the UK: UK HE→LE 1.7 percent vs LE→HE 3.3 percent; Brazil HE→LE 11.2 percent vs LE→HE 7.1 percent.

### Industry distribution of AI exposure (Annex C)
- Services sectors concentrate most AI-exposed employment; some services sections have HE shares up to around 90 percent.
- Brazil: jobs more concentrated in trade with higher share of low-complementarity occupations.
- UK: more jobs in communications and finance with higher complementarity.
- Education sector: almost 85 percent of workers in high-exposure, high-complementarity jobs.

### Methodology for lifetime expected earnings (Section 5.3)
- Lifetime expected earnings for entrant at age a0:
  - Ea0[wa] = Σk pk a wk a
  - Wa0 = Σa≥a0 βa−a0 Ea[ w a ]
  - Counterfactual ˆWa0 = Σa≥a0 βa−a0 Σk ˆpk a ˆwk a using fitted values from Equations 2 and 3
- Education split: college vs non-college.
- Discount rates: 2.5% for the UK and 5% for Brazil.
- Baseline: historical occupational shares and wages.
- Four counterfactual exercises isolating channels:
  - Exercise 1: All HELC jobs destroyed; displaced workers become unemployed.
  - Exercise 2: All HELC jobs destroyed; displaced workers reallocate to LE jobs (total employment constant).
  - Exercise 3: All HELC jobs destroyed; displaced workers move to HEHC jobs.
  - Exercise 4: HEHC wages rise by 10%.

### Baseline lifetime expected earnings (2019 thousand PPP dollars) and counterfactual percent changes (Table 11)
- Baseline Wa0:
  - College, Brazil: 423.3
  - College, UK: 1567.5
  - No College, Brazil: 251.5
  - No College, UK: 1128.8

- Percent changes in expected lifetime earnings from counterfactuals (values as printed)
  - HELC to Unemployment (% Change): -29.3 -24.1 -33.6 -28.5
  - HELC to LE (% Change): -6.5 -6.6 -2.2 -2.7
  - HELC to HEHC (% Change): 54 11.4 10.8
  - HEHC Wage Increase (% Change): 5.8 6.9 3.9 4.9

- Additional exercise-specific outcomes (text summary)
  - Exercise 1 (HELC → Unemployment):
    - College: -29.3% (Brazil) vs. -24.1% (UK)
    - No college: -33.6% (Brazil) vs. -28.5% (UK)
  - Exercise 2 (HELC → LE):
    - College: -6.5% (Brazil), -6.6% (UK)
    - No college: -2.2% (Brazil), -2.7% (UK)
  - Exercise 3 (HELC → HEHC): positive outcomes; larger improvements for those without a college degree.
  - Exercise 4 (HEHC wages +10%):
    - College: 5.8% (Brazil), 6.9% (UK)
    - No college: 3.9% (Brazil), 4.9% (UK)

### Combined scenarios (example calibrations)
- Example: 20% of HELC jobs displaced; half (10%) become permanently unemployed; the other half relocate between HEHC and LE per transition probabilities.
  - UK: expected lifetime earnings change by -2.4% (non-college) and -2.2% (college).
  - Brazil: expected lifetime earnings change by -3.1% (non-college) and -2.6% (college).
  - Brazil worse on average because unemployment losses represent a higher share of lifetime earnings and relocation probabilities to HEHC are limited for non-college workers.
- Variant with HEHC wages +10% (combined with above displacement):
  - UK: lifetime earnings increase by 5% (college) and 2.75% (non-college).
  - Brazil: lifetime earnings increase by 3.4% (college) and 1% (non-college).
  - Net effect can be positive on average but increase inequality.

### Interpretation of mechanisms and heterogeneity
- Job destruction raises unemployment and disproportionately lowers lifetime earnings, especially in Brazil and among non-college workers.
- Reallocation to LE occupations reduces lifetime earnings due to lower pay relative to HELC.
- Reallocation to HEHC or HEHC wage growth increases lifetime earnings, benefiting college-educated workers more because of higher representation in HEHC.
- Non-college workers are concentrated in HELC and thus suffer larger percent losses when HELC is destroyed.
- Young college-educated face both opportunities (AI-complementary jobs) and risks (disruption in entry HELC jobs).

### Limitations of analysis
- Cohort effects not estimated; starting prospects for new entrants may differ from historical life-cycle patterns.
- NLF excluded as an earnings category; modeling wages in NLF would require endogenous selection modeling.
- Partial-equilibrium analysis: does not account for productivity changes, general equilibrium wage adjustments, or emergence of new occupations.
- Predictive power for AI-driven structural changes is uncertain; focus is on relative magnitudes of transition probabilities.

### Policy-relevant takeaways and recommendations
- Structural change from AI can have large, heterogeneous effects on lifetime earnings across countries and education groups.
- Education is key: marginal effect of a college degree larger in Brazil (11.4 p.p.) than in the UK (2.26 p.p.) for moving into HEHC.
- Policies should target demographic intersections (e.g., young college graduates and non-college workers) because vulnerability and adjustment capacity differ by age and education.
- Interventions should:
  - Limit income losses after displacement (income support, retraining).
  - Prevent displacement by boosting ability to move from shrinking occupations to growing ones (reskilling, improving transition probabilities to HEHC, digital infrastructure).
- Wage growth in AI-complementary occupations can offset displacement on average but may increase inequality if gains concentrate among those already likely to be in HEHC jobs.
- Gender and age-specific measures may be warranted to support female and older workers’ mobility into higher-complementarity occupations.

*Source: wpiea2024116-print-pdf — sections 3.1–5.3 (Worker Characteristics through Expected Impact of AI Lifetime Earnings).*

### 3.1  Worker Characteristics

### 3.1  Worker Characteristics

### Worker demographic and education summary
- Median Worker Age: UK 42, Brazil 36
- Share of Women in Employment (%): UK 47, Brazil 42.5
- Share of Workers with a College Degree (%): UK 36.7, Brazil 17.3
- Share of Workers in High AI Exposure Occupations (%): UK 66.4, Brazil 39.7
- Median Hourly Wage (2019 PPP Dollars): UK 18.4, Brazil 3.8
- Informality Rate (%): Brazil 42 (note: informality rate defined only for Brazil)

### Key implications from characteristics
- Median worker age is markedly lower in Brazil (36 years) than in the UK (42 years).
- Female participation is higher in the UK (47 percent) than in Brazil (42 percent).
- Educational attainment is substantially higher in the UK (36.7 percent college-educated) than in Brazil (17.3 percent).
- Median wage disparity: UK median wage (2019 PPP) 18.4 versus Brazil 3.8 — nearly fivefold difference as reported.
- AI exposure concentration: higher share of employment in high-exposure occupations in the UK (66.4 percent) than Brazil (39.7 percent).
- High informality in Brazil: almost 42 percent of workers classified as informal.

### Notable cross-country observation
- Educational disparity correlates with wage differences and with differing exposure to AI across the two labor markets.

---

### 3.2  Labor Market Transitions

### Aggregate quarterly transition matrices (shares in subsequent quarter)
- UK Employment Flows (Table 2)
  - Employed → Employed: 97.8
  - Employed → Unemployed: 0.9
  - Employed → Not in labor force: 1.3
  - Unemployed → Employed: 24.5
  - Unemployed → Unemployed: 60.1
  - Unemployed → Not in labor force: 15.4
  - Not in labor force → Employed: 3.7
  - Not in labor force → Unemployed: 4.3
  - Not in labor force → Not in labor force: 92.3

- Brazil Employment Flows (Table 3)
  - Employed → Employed: 90.7
  - Employed → Unemployed: 3.2
  - Employed → Not in labor force: 6.1
  - Unemployed → Employed: 31.9
  - Unemployed → Unemployed: 42.9
  - Unemployed → Not in labor force: 25.2
  - Not in labor force → Employed: 13.3
  - Not in labor force → Unemployed: 7.0
  - Not in labor force → Not in labor force: 79.7

### Comparative findings on dynamics
- Labor market status persistence is higher in the UK across employed, unemployed, and not in labor force categories.
- UK employed workers are more likely to remain employed across two quarters (97.8 versus 90.7).
- Unemployment duration appears longer on average in the UK: lower transition from unemployment to employment in the UK (24.5) than in Brazil (31.9).
- Brazil shows higher short-term employment risk (higher separation) but transitions from non-employment to employment are more frequent, suggesting greater short-term dynamism.
- Gender-specific patterns (Annex A.1): similar male/female patterns in both countries; men more likely to stay employed than women, women more likely to leave the labor force.

---

### 3.2.1  Education

### College-educated workers (Table 4)
- UK (college)
  - Employed → Employed: 98.0
  - Employed → Unemployed: 0.8
  - Employed → Not in labor force: 1.2
  - Unemployed → Employed: 34.8
  - Unemployed → Unemployed: 50.8
  - Unemployed → Not in labor force: 14.8
  - Not in labor force → Employed: 8.4
  - Not in labor force → Unemployed: 2.6
  - Not in labor force → Not in labor force: 87.4

- Brazil (college)
  - Employed → Employed: 96.3
  - Employed → Unemployed: 1.5
  - Employed → Not in labor force: 2.2
  - Unemployed → Employed: 31.2
  - Unemployed → Unemployed: 49.5
  - Unemployed → Not in labor force: 19.3
  - Not in labor force → Employed: 15.0
  - Not in labor force → Unemployed: 8.7
  - Not in labor force → Not in labor force: 76.2

### Non-college workers (Table 5)
- UK (non-college)
  - Employed → Employed: 97.7
  - Employed → Unemployed: 0.9
  - Employed → Not in labor force: 1.4
  - Unemployed → Employed: 21.7
  - Unemployed → Unemployed: 62.6
  - Unemployed → Not in labor force: 15.7
  - Not in labor force → Employed: 3.1
  - Not in labor force → Unemployed: 3.5
  - Not in labor force → Not in labor force: 93.4

- Brazil (non-college)
  - Employed → Employed: 89.6
  - Employed → Unemployed: 3.5
  - Employed → Not in labor force: 6.9
  - Unemployed → Employed: 31.9
  - Unemployed → Unemployed: 42.3
  - Unemployed → Not in labor force: 25.8
  - Not in labor force → Employed: 13.2
  - Not in labor force → Unemployed: 6.8
  - Not in labor force → Not in labor force: 80.0

### Education-related insights
- College-educated workers show remarkably similar retention across countries: employed retention UK 98.0 versus Brazil 96.3.
- Among non-college workers, the UK shows higher employment retention (97.7) than Brazil (89.6).
- Transition rates from non-employment to employment differ by education group: large gap by education in the UK but small or non-existent gap in Brazil.
- Suggests higher return to education for job security in the emerging market context (Brazil) and more dynamic opportunities for lower-education workers in Brazil.

---

### 3.3  Occupation and Job Transitions

### Definitions and measurement
- Distinguish "switch employer" (current employer tenure < three months and employed previous quarter) vs "same employer".
- Distinguish "switch occ." (occupation change) vs "same occ.".
- Occupation switching rates estimated at 4-digit ISCO-08 for Brazil and UK.

### UK job and occupation switching probabilities (Table 6)
- Same Employer & Same Occ.: 87.3
- Same Employer & Switch Occ.: 10.7
- Switch Employer & Same Occ.: 0.8
- Switch Employer & Switch Occ.: 1.2
- Marginal occupation switching rate: 11.9
- Marginal job switching rate: 2.0 (implied by text: "Overall, 98 percent of workers either stay with the same employer or within the same occupation, with only 2 percent engaging in both employer and occupation switches.")

### Brazil job and occupation switching probabilities (Table 7)
- Same Employer & Same Occ.: 65.2
- Same Employer & Switch Occ.: 32.4
- Switch Employer & Same Occ.: 0.8
- Switch Employer & Switch Occ.: 1.6
- Marginal occupation switching rate: 34.0
- Marginal job switching rate: 2.4 (from table marginal probabilities)

### Comparative findings on mobility
- UK exhibits high job and occupational stability: 87.3 percent retain same employer and same occupation.
- Brazil shows lower combined job-and-occupation stability (65.2 percent) and higher occupational mobility (34 percent switch occupations).
- Brazil has greater overall dynamism in job and occupation switching, which may reflect higher flexibility but may also include involuntary separations or lower-wage transitions.
- Robustness check referenced: occupation switching remains higher in Brazil even at more aggregate occupational levels (3-digit and 2-digit), reducing concern that measurement error at 4-digit ISCO-08 fully explains the difference.
- Annex A.2: college-educated workers slightly more likely to switch occupations than non-college in both countries; no clear pattern for job-switching probabilities by education.

---

### 4  Transitions and AI Exposure

### AI exposure classification and earnings context
- Occupations categorized into three groups (following Pizzinelli et al. (2023) and Felten et al. (2021)):
  - Low Exposure (LE)
  - High Exposure, High Complementarity (HEHC)
  - High Exposure, Low Complementarity (HELC)
- General earnings patterns by category (as discussed):
  - HEHC occupations tend to have the highest wages.
  - HELC occupations are fairly homogeneous across the income distribution.
  - LE occupations largely concentrated in elementary and agricultural sectors, tending to be lower-paid.

### Distribution of employment by AI exposure and education (Table 8, percent)
- UK (No College): LE 29, HELC 19.1, HEHC 14.9, Total 63.7
- UK (College): LE 4.3, HELC 12.6, HEHC 20.1, Total 37.2
- UK Total: LE 33.3, HELC 31.7, HEHC 35.0, Total 100 (reported as 33.3, 31.7, 35.0)
- Brazil (No College): LE 35.7, HELC 15.8, HEHC 9.2, Total 82.7
- Brazil (College): LE 2.6, HELC 5.0, HEHC 9.7, Total 17.3
- Brazil Total: LE 38.3?, HELC 20.8?, HEHC 18.9? (table reports Total 20.8 and 18.9 in row; preserve reported row: Total 20.8, 18.9) 

(Note: Table 8 shows AI exposure is highly correlated with education: over 85% of college-educated workers are in highly exposed occupations in both the UK and Brazil.)

### Historical occupation switching across AI exposure categories (figure summary)
- Conditional on switching occupations, each category (LE, HELC, HEHC) is persistent: workers more likely to switch within the same broad category than to another.
- Brazil displays greater probability of moving to LE occupations conditional on switching (more “downward” occupational movement) compared with the UK.
- In the UK, high-exposure workers are about half as likely as in Brazil to move to LE occupations upon switching; UK workers, including those in LE jobs, have a greater probability of moving to HEHC occupations conditional on switching.

### Caution on interpretation
- Historical transition patterns provide suggestive evidence on adaptability to AI-induced labor demand shifts, but predictive power for AI-driven structural changes is uncertain.
- The analysis focuses on relative magnitudes of transition probabilities across countries and groups rather than absolute levels, given uncertainty about AI’s future impacts.

*Source: IMF working paper chapter section 3.1–4 (content unit: wpiea2024116-print-pdf - 3.1  Worker Characteristics).*

### Annex A.3 breaks down employment flows to and from unemployment and NLF for workers in high and

### Annex A.3 breaks down employment flows to and from unemployment and NLF for workers in high and low exposure occupations

### Transition probabilities and heterogeneity by location
- Transition probabilities are robust to several sample restrictions (e.g., excluding public sector workers).
- Rural versus urban location conditions results: digital infrastructure in rural areas is substantially more limited, affecting job composition and likelihood of transitions to or from jobs exposed to AI.
- Labor markets in Brazil are more dynamic and workers are more likely to change occupations between quarters (Figure B.1 illustrates transition probabilities conditional on the “from” category).

### Education and occupational mobility
- College-educated workers:
  - Have higher chances of moving “upwards” in terms of exposure (to HEHC occupations) in both countries.
  - The profile of transitions for college-educated workers is similar across the UK and Brazil; Brazil shows an even slightly higher upward mobility premium for education in descriptive charts.
- Non-college workers:
  - Have higher chances of moving “downwards” in Brazil compared to the UK; this is not solely a composition effect from lower education attainment in Brazil.
  - Example life-cycle statistic: an estimated 80 percent of Brazilian workers without a college degree at age 40 are in jobs that have low exposure to AI.

### Regression specification for occupational switches
- Main estimating equation (occupational-switch dummy outcome):
  y_kirt = α + Σ_{j≠HEHC} δ_j C_jir(t−1) + β X_irt + γ_t + η_r + ε_irt
  - y_kirt is a dummy equal to one if the worker switched occupations to category k (LE, HELC, or HEHC).
  - X_irt includes demographic covariates (age, gender, education, and informality for Brazil).
  - γ_t and η_r are time (year-quarter) and region fixed effects.
- Base category: workers employed in HEHC, aged below 25, male, with middle school education or below.

### Key regression results — UK (Tables 9)
- Education effects:
  - Holding a degree increases the probability of moving into HEHC occupations by 2.26 percentage points relative to workers with middle school education.
  - Higher educational attainment is linked with decreased likelihood of remaining in or transitioning to Low Exposed (LE) occupations.
- Gender and age:
  - Being a woman reduces the probability of switching to HEHC occupations by approximately 0.63 p.p. and to LE occupations by about 1.12 p.p.
  - Older workers (ages 45-59 and 60+) show reduced likelihood of switching occupations relative to young workers.
- Table 9 model fit and sample:
  - Observations: 4552677, 4552677, 4552677 (columns (1)(2)(3))
  - R^2: 0.08, 0.058, 0.061
  - State FE: Yes; Year-Quarter FE: Yes

### Key regression results — Brazil (Tables 10)
- Education effects:
  - Holding a college degree increases the likelihood of moving into HEHC occupations by 11.4 p.p. relative to workers with middle school education.
  - Higher educational levels are associated with decreased probabilities of staying in or moving to LE occupations.
- Gender and age:
  - Being a woman reduces the probability of switching to HEHC occupations by about 1.02 p.p. relative to men.
  - Being a woman significantly increases the likelihood of transitioning to HELC occupations by 1.10 p.p. and decreases the probability of moving to LE occupations by 7.37 p.p.
  - Older workers (ages 45-59 and 60+) are more inclined to switch to HEHC occupations relative to young workers but less likely to switch to HELC and LE occupations.
- Table 10 model fit and sample:
  - Observations: 4552677, 4552677, 4552677 (columns (1)(2)(3))
  - R^2: 0.08, 0.058, 0.061
  - State FE: Yes; Year-Quarter FE: Yes

### Life-cycle dynamics — employment shares
- Estimation approach:
  - Polynomial specification: C_kit = β0 + β1 age_it + β2 age_it^2 + β3 age_it^3 + δ female_it + γ_t + ε_it, where C_kit is a dummy for employment in category k∈{HEHC, HELC, LE}.
- Findings:
  - Young-age transitions are common and typically from HELC to HEHC jobs; these are captured by a steep upward slope in HEHC shares and steep decline in HELC shares before age 40.
  - The share of LE jobs is relatively constant over age.
  - Life-cycle profiles for college-educated workers are very similar in the UK and Brazil; non-college profiles differ markedly (more so in Brazil).
- Interpretation:
  - HELC occupations likely correspond to entry-level jobs serving as stepping stones into HEHC jobs in late twenties and thirties.
  - AI-led disruption to HELC jobs could hamper entry into the labor market and transition to better-earning HEHC jobs, particularly for college graduates.
- Caveat:
  - Analysis relies on average shares by age in the sample period and may be affected by cohort or composition effects; the time period covers less than a full decade and cannot fully capture cohort effects.

### Life-cycle dynamics — wages and wage premia for transitions
- Wage polynomial:
  - y_kit = β0 + β1 age_it + β2 age_it^2 + β3 age_it^3 + δ female_it + γ_t + ε_it, where y_kit is log hourly earnings of workers in exposure category k.
- Descriptive wage dynamics:
  - Individuals typically experience most wage growth up until 35 to 40 years old, coinciding with career changes.
  - Average wage growth is higher for those in occupations more exposed to AI, particularly for college graduates.
- Panel regression for wage changes (annual panel):
  - Δ log(y_irt) = δ1 J2J_irt + δ2 OS_irt × J2J_irt + δ3 EUE_irt + Σ_k θ_k C_kir(t−1) C_kirt + Σ_k Σ_j OS_irt φ_kj C_kir(t−1) C_jirt + β X_irt + γ_t + η_r + ε_irt
    - OS: occupation switch dummy; J2J: job switch dummy; EUE: transitions through unemployment.
    - θ_k: average log wage change for stayers in category k; φ_kj: average log wage change for workers switching from k to j.
    - Wage premium for switching from HELC to HEHC computed as φ_HELC,HEHC − θ_HELC.
- Empirical findings:
  - There is a premium associated with switching “upwards” (LE → HE and HELC → HEHC) in both countries.
  - Upward transitions particularly benefit young college-educated workers, who account for a large share of these transitions.
  - Downward transitions in Brazil are associated with a wage penalty; non-college workers face greater risk of moving to LE occupations and potential wage compression.
- Switching statistics:
  - Average yearly J2J switching rate: 9.2 percent (UK) and 11.7 percent (Brazil).
  - Average occupation switching rate: 28 percent (UK) and 39 percent (Brazil).

### Implications
- Education is a key determinant of occupational mobility and access to HEHC occupations, with a notably larger marginal effect of a college degree in Brazil than in the UK (11.4 p.p. in Brazil versus 2.26 p.p. in the UK).
- Gender and age shape mobility differently across countries; policy measures to support female and older workers’ mobility into higher-complementarity occupations may be warranted.
- Disruption to entry-level (HELC) occupations from AI could impede career progression for younger workers into HEHC roles and reduce potential wage growth.
- Wage dynamics imply that policies promoting transitions into HEHC occupations (training, education, digital infrastructure) could increase earnings, while measures to protect or re-skill workers at risk of downward transitions are important to avoid wage compression, especially in Brazil.

*Source: wpiea2024116-print-pdf - Annex A.3 breaks down employment flows to and from unemployment and NLF for workers in high and low exposure occupations*

### 5.3  Expected Impact of AI Lifetime Earnings

### 5.3  Expected Impact of AI Lifetime Earnings

### Methodology
- Lifetime expected earnings for a worker entering the labor market at age a0 are computed as:
  - Ea0[wa] = Σk pk a wk a, where pk a is the share of employment and wk a the average wage in occupation k at age a.
  - Lifetime expected earnings: Wa0 = Σa≥a0 βa−a0 Ea[ w a ].
  - Estimated counterfactual lifetime earnings: ˆWa0 = Σa≥a0 βa−a0 Σk ˆpk a ˆwk a, using fitted values from Equations 2 and 3.
- Education split: college-educated and non-college-educated.
- Discount rates: 2.5% for the UK and 5% for Brazil.
- Baseline refers to historical occupational shares and wages.
- Four illustrative counterfactual exercises isolate the channels of job destruction, reallocation, and wage growth:
  - Exercise 1: All HELC jobs destroyed; displaced workers become unemployed.
  - Exercise 2: All HELC jobs destroyed; displaced workers reallocate to LE jobs (total employment constant).
  - Exercise 3: All HELC jobs destroyed; displaced workers move to HEHC jobs.
  - Exercise 4: HEHC wages rise by 10%.

### Counterfactual Exercises and Key Results (Table 11)
- Baseline lifetime expected earnings (2019 thousand PPP dollars):
  - College, Brazil: 423.3
  - College, UK: 1567.5
  - No College, Brazil: 251.5
  - No College, UK: 1128.8
- Percent changes in expected lifetime earnings from counterfactuals (values presented in the table as printed):
  - HELC to Unemployment (% Change): -29.3 -24.1 -33.6 -28.5
  - HELC to LE (% Change): -6.5 -6.6 -2.2 -2.7
  - HELC to HEHC (% Change): 54 11.4 10.8
  - HEHC Wage Increase (% Change): 5.8 6.9 3.9 4.9
- Additional exercise-specific outcomes described in the text:
  - Exercise 1 (HELC → Unemployment): larger lifetime earnings declines for non-college workers than for college-educated; country comparison:
    - College: -29.3% (Brazil) vs. -24.1% (UK)
    - No college: -33.6% (Brazil) vs. -28.5% (UK)
  - Exercise 2 (HELC → LE): negative effects for all; larger losses for college-educated because pay gap between HELC and LE is larger:
    - College: -6.5% (Brazil), -6.6% (UK)
    - No college: -2.2% (Brazil), -2.7% (UK)
  - Exercise 3 (HELC → HEHC): positive outcomes, similar across countries but larger improvements for those without a college degree due to lower initial wages and lower probabilities of being in HEHC jobs.
  - Exercise 4 (HEHC wages +10%): positive impacts, larger for college-educated due to higher representation in HEHC:
    - College: 5.8% (Brazil), 6.9% (UK)
    - No college: 3.9% (Brazil), 4.9% (UK)

### Combined Scenarios (linear combinations of exercises)
- Example combination: 20% of HELC jobs displaced; half (10%) become permanently unemployed; the other half relocate between HEHC and LE according to transition probabilities.
  - With that calibration:
    - UK: expected lifetime earnings change by -2.4% for non-college and -2.2% for college graduates.
    - Brazil: expected lifetime earnings change by -3.1% for non-college and -2.6% for college graduates.
  - Explanation: Brazil exhibits worse average outcomes because lost income due to unemployment represents a higher share of lifetime earnings and limited relocation probabilities to HEHC jobs for non-college workers.
- Variant where HEHC jobs receive a 10% wage boost (combined with the displacement scenario above):
  - UK: lifetime earnings increase by 5% for college-educated and 2.75% for non-college.
  - Brazil: lifetime earnings increase by 3.4% for college-educated and 1% for non-college.
  - Note: net effect can be positive on average, but inequality increases (more people with lower wages and higher earnings at the top).

### Interpretation and Mechanisms
- Channels highlighted:
  - Job destruction raises unemployment and disproportionately lowers lifetime earnings, especially in Brazil and among non-college workers.
  - Reallocation to lower-exposure (LE) occupations reduces lifetime earnings due to lower average pay relative to HELC.
  - Reallocation to higher-complementarity (HEHC) occupations or HEHC wage growth can increase lifetime earnings, benefiting college-educated workers more because of their higher representation in HEHC.
- Heterogeneity:
  - Non-college workers are more concentrated in HELC jobs and thus suffer larger percentage losses when HELC jobs are destroyed.
  - Young college-educated workers face both opportunities from AI-complementary jobs and risks from disruption in low-complementarity jobs early in their careers.

### Limitations
- Cohort effects not estimated: starting occupational prospects for new entrants may differ from historical life-cycle patterns due to composition effects.
- “Not in the labor force” (NLF) excluded as a category; accounting for wages in NLF would require modeling endogenous selection across ages.
- Analysis is partial-equilibrium and does not account for general equilibrium effects:
  - Does not model productivity changes, wage adjustments in AI-impacted occupations, or how these would alter job search behavior and labor supply across occupations.
- Does not capture emergence of entirely new occupations related to AI or the potential role of policies and labor market institutions in shaping transitions.

### Policy-relevant Takeaways
- Structural change induced by AI can have large and heterogeneous effects on lifetime earnings across countries and education groups.
- Policies should target the intersection of demographic groups (e.g., young college graduates and non-college workers) because vulnerability and adjustment capacity differ by age and education.
- Interventions should aim both to:
  - Limit income losses after displacement (income support, retraining), and
  - Prevent displacement by boosting workers’ ability to move from shrinking occupations to growing ones (reskilling, improving transition probabilities to HEHC).
- Wage growth in AI-complementary occupations can offset negative displacement effects on average but can increase inequality if gains concentrate among workers already more likely to be in HEHC jobs.

*Source: 5.3 Expected Impact of AI Lifetime Earnings — wpiea2024116-print-pdf*

### References

### wpiea2024116-print-pdf - References and Annexes

### References
- List of cited works includes (preserving original authorship and titles as presented):
  - Acemoglu, D., D. Autor, J. Hazell, and P. Restrepo (2022). Artificial Intelligence and Jobs: Evidence from Online Vacancies. Journal of Labor Economics 40(S1), 293–340.
  - Acemoglu, D. and P. Restrepo (2018). Artificial intelligence, automation, and work. In A. Agrawal, J. Gans, and A. Goldfarb (Eds.), The economics of artificial intelligence: An agenda, pp. 197–236. University of Chicago Press.
  - Ad ̃ao, R., M. Beraja, and N. Pandalai-Nayar (2024). Fast and slow technological transitions. Journal of Political Economy Macroeconomics 0(ja), null.
  - Albanesi, S., A. D. da Silva, J. F. Jimeno, A. Lamo, and A. Wabitsch (2023). New technologies and jobs in europe. Technical Report DP18220, CEPR Discussion Papers.
  - Alekseeva, L., J. Azar, M. Gine, S. Samila, and B. Taska (2021). The demand for ai skills in the labor market. Labour economics 71, 102002.
  - Bluedorn, J., F. Caselli, M. I. Shibata, and M. M. Tavares (2022). The distributional impacts of worker reallocation: Evidence from europe. IMF Working Paper 2022/124.
  - Bonfiglioli, A., G. Gancia, I. Papadakis, and R. Crin`o (2023). Artificial intelligence and jobs: Evidence from us commuting zones. CESifo Working Paper.
  - Brazil (2019). Anu ́ario estat ́ıstico da previdˆencia social.
  - Briggs, J. and D. Kodnani (2023). The potentially large effects of artificial intelligence on economic growth. Global Economics Analyst.
  - Carrillo-Tudela, C., B. Hobijn, P. She, and L. Visschers (2016). The extent and cyclicality of career changes: Evidence for the uk. European Economic Review 84, 18–41.
  - Cazzaniga, M., M. F. Jaumotte, L. Li, M. G. Melina, A. J. Panton, C. Pizzinelli, E. J. Rockall, and M. M. M. Tavares (2024). Gen-ai: Artificial intelligence and the future of work. IMF Staff Discussion Note 2024/01.
  - Constanzi, R. N. and C. F. dos Santos (2022). An ́alise dos impactos da reforma de 2019 sobre as idades de aposentadoria no regime geral de previdˆencia social. Informe de Previdˆencia Social 34(8), 5–13.
  - Copestake, A., M. Marczinek, A. Pople, and K. Stapleton (2023). Ai and services-led growth: Evidence from indian job adverts.
  - Cortes, G. M., C. J. Nekarda, N. Jaimovich, and H. E. Siu (2020, 8). The dynamics of disappearing routine jobs: A flows approach. Labour Economics 65.
  - Dabla-Norris, E., C. Pizzinelli, and J. Rappaport (2023). Job polarization and the declining wages of young female workers in the united kingdom. Oxford Bulletin of Economics and Statistics 85(6), 1185–1210.
  - DataZoom (2023). Data zoom: Simplifying access to brazilian microdata.
  - Eloundou, T., S. Manning, P. Mishkin, and D. Rock (2023). Gpts are gpts: An early look at the labor market impact potential of large language models.
  - Felten, E., M. Raj, and R. Seamans (2021). Occupational, industry, and geographic exposure to artificial intelligence: A novel dataset and its potential uses. Strategic Management Journal 42(12), 2195–2217.
  - Felten, E., M. Raj, and R. Seamans (2023). How will language modelers like chatgpt affect occupations and industries?
  - Gmyrek, P., J. Berg, and D. Bescond (2023). Generative ai and jobs: A global analysis of potential effects on job quantity and quality. ILO Working Paper 96.
  - Gmyrek, P., H. Winkler, and S. Garganta (2024). Buffer or bottleneck? generative ai, employment exposure and the digital divide in latin america. ILO Working Paper.
  - Huitfeldt, I., A. R. Kostøl, J. Nimczik, and A. Weber (2023). Internal labor markets: A worker flow approach. Journal of Econometrics 233(2), 661–688.
  - Lane, M., M. Williams, and S. Broecke (2023). The impact of ai on the workplace: Main findings from the oecd ai surveys of employers and workers.
  - Leyva, G. and C. Urrutia (2020). Informality, labor regulation, and the business cycle. Journal of International Economics 126, 103340.
  - Manca, F. (2023). Six questions about the demand for artificial intelligence skills in labour markets.
  - Milanez, A. (2023). The impact of ai on the workplace: Evidence from oecd case studies of ai implementation.
  - Moszoro, M., M. Nose, S. Nunhuck, S. Pattanayak, L. R. del Paso, F. Solomon, R. Sparkman, H. Tourpe, and G. U ̃na (2023). Transforming public finance through govtech. International Monetary Fund, STaff Discussion Note 2023/03.
  - OECD (2021). Pensions at a Glance 2021. OECD Publishing, Paris.
  - OECD (2023a). OECD Employment Outlook 2023.
  - OECD (2023b). Purchasing power parities (ppp) (indicator).
  - Pizzinelli, C., A. Panton, M. Mendes Tavares, M. Cazzaniga, and L. Li (2023). Labor market exposure to ai: Cross-country differences and distributional implications. IMF Working Paper 2023/216.
  - Ribas, R. P. and S. S. D. Soares (2008). Sobre o painel da pesquisa mensal de emprego (pme) do ibge. IPEA, Textos para Discuss ̃ao 1348.
  - Ulyssea, G. (2018). Firms, informality, and development: Theory and evidence from brazil. American Economic Review 108(8), 2015–2047.
  - Webb, M. (2020). The impact of artificial intelligence on the labor market.

### Annex A — Additional Employment Flow Analysis
- Structure:
  - Annex A contains analysis of worker flows with gender implications (A.1), occupation and job transition by education (A.2), and employment flows by AI exposure (A.3).
- A.1 Employment Flow Analysis by Gender and Education — Key statistics and findings:
  - Table A.1: Employment Flows by Gender: Females (UK; Brazil)
    - UK row/column shares (Females):
      - Employed → Employed: 97.4
      - Employed → Unemployed: 0.8
      - Employed → NLF: 1.8
      - Unemployed → Employed: 24.5
      - Unemployed → Unemployed: 55.8
      - Unemployed → NLF: 19.7
      - NLF → Employed: 43.3
      - NLF → Unemployed: 92.7
      - NLF → NLF: 11.4
    - Brazil row/column shares (Females):
      - Employed → Employed: 88.4
      - Employed → Unemployed: 3
      - Employed → NLF: 8.6
      - Unemployed → Employed: 26
      - Unemployed → Unemployed: 43.4
      - Unemployed → NLF: 30.6
      - NLF → Employed: 6.1
      - NLF → Unemployed: 82.5
      - NLF → NLF: (implicit from table)
  - Table A.2: Employment Flows by Gender: Males (UK; Brazil)
    - UK row/column shares (Males):
      - Employed → Employed: 98.1
      - Employed → Unemployed: 1.0
      - Employed → NLF: 0.9
      - Unemployed → Employed: 24.4
      - Unemployed → Unemployed: 63.6
      - Unemployed → NLF: 12
      - NLF → Employed: 44.3
      - NLF → Unemployed: 91.6
      - NLF → NLF: 17.9
    - Brazil row/column shares (Males):
      - Employed → Employed: 92.5
      - Employed → Unemployed: 3.3
      - Employed → NLF: 4.2
      - Unemployed → Employed: 38.3
      - Unemployed → Unemployed: 42.4
      - Unemployed → NLF: 19.3
      - NLF → Employed: 8.9
      - NLF → Unemployed: 73.2
      - NLF → NLF: (implicit from table)
  - Interpretations:
    - UK shows greater job retention for both genders: Females 97.4 percent and Males 98.1 percent remain employed quarter-to-quarter.
    - Brazil shows lower retention: Females 88.4 percent and Males 92.5 percent remain employed.
    - Unemployed-to-stay-unemployed: Males 63.6 percent (UK) vs Females 55.8 percent (UK); Brazil: Males 42.4 percent and Females 43.4 percent.
    - Females are more likely than males to transition from unemployment to NLF in both countries (UK: 19.7 percent vs 12 percent; Brazil: 30.6 percent vs 19.3 percent).
- A.2 Occupation and Job Transitions by Education — Key statistics and findings:
  - Table A.3: UK Job and Occupation Switching Probabilities by Education
    - No College:
      - Sm Occ. → Sm J: 87.7
      - Sm Occ. → Sw J: 0.7
      - Sm Occ. total: 88.4
      - Sw Occ. → Sm J: 10.4
      - Sw Occ. → Sw J: 1.2
      - Sw Occ. total: 11.6
      - Marginals: 98.1 (Sm J), 1.9 (Sw J), 97.8 (Sm Occ.), 2.2 (Sw Occ.)
    - College:
      - Sm Occ. → Sm J: 86.4
      - Sm Occ. → Sw J: 1.0
      - Sm Occ. total: 87.4
      - Sw Occ. → Sm J: 11.4
      - Sw Occ. → Sw J: 1.2
      - Sw Occ. total: 12.6
      - Marginals: 97.8 (Sm J), 2.2 (Sw J)
  - Table A.4: Brazil Job and Occupation Switching Probabilities by Education
    - No College:
      - Sm Occ. → Sm J: 65.9
      - Sm Occ. → Sw J: 0.9
      - Sm Occ. total: 66.8
      - Sw Occ. → Sm J: 31.4
      - Sw Occ. → Sw J: 1.8
      - Sw Occ. total: 33.2
      - Marginals: 97.3 (Sm J), 2.7 (Sw J)
    - College:
      - Sm Occ. → Sm J: 61.9
      - Sm Occ. → Sw J: 0.3
      - Sm Occ. total: 62.2
      - Sw Occ. → Sm J: 37.1
      - Sw Occ. → Sw J: 0.7
      - Sw Occ. total: 37.8
      - Marginals: 99.0 (Sm J), 1.0 (Sw J)
  - Interpretations:
    - College-educated workers switch occupations more frequently than those without a college degree: the probability is 1 percentage point higher in the UK and 4.6 percentage points higher in Brazil.
    - Job switching shows no clear pattern across countries: UK rates similar across education; Brazil shows Sw J: College 1% vs No College 3.2% (noting the table marginal values).
    - Suggested drivers: greater ease of career changes or advancement for higher-educated workers; informality and lower job stability for lower-educated workers in Brazil.
- A.3 Employment Flow Analysis by AI Exposure — Key statistics and findings:
  - Table A.5: AI Exposure Employment Flows for the UK (states: Employed (HE), Employed (LE), Unemployed, NLF)
    - Employed (HE) → Employed (HE): 96.2
    - Employed (HE) → Employed (LE): 1.7
    - Employed (HE) → Unemployed: 0.8
    - Employed (HE) → NLF: 1.3
    - Employed (LE) → Employed (HE): 3.3
    - Employed (LE) → Employed (LE): 94.2
    - Employed (LE) → Unemployed: 1.1
    - Employed (LE) → NLF: 1.4
    - Unemployed → Employed (HE): 13.7
    - Unemployed → Employed (LE): 10.8
    - Unemployed → Unemployed: 60.1
    - Unemployed → NLF: 15.4
    - NLF → Employed (HE): 2.6
    - NLF → Employed (LE): 1.4
    - NLF → Unemployed: 3.7
    - NLF → NLF: 92.3
  - Table A.6: AI Exposure Employment Flows for Brazil (states: Employed (HE), Employed (LE), Unemployed, NLF)
    - Employed (HE) → Employed (HE): 81.4
    - Employed (HE) → Employed (LE): 11.2
    - Employed (HE) → Unemployed: 7.4
    - Employed (HE) → NLF: 4.9
    - Employed (LE) → Employed (HE): 7.1
    - Employed (LE) → Employed (LE): 82.3
    - Employed (LE) → Unemployed: 3.7
    - Employed (LE) → NLF: 6.9
    - Unemployed → Employed (HE): 10.2
    - Unemployed → Employed (LE): 21.7
    - Unemployed → Unemployed: 42.9
    - Unemployed → NLF: 25.2
    - NLF → Employed (HE): 4.3
    - NLF → Employed (LE): 77
    - NLF → Unemployed: 79.7 (note: table formatting indicates high shares for LE from NLF)
  - Interpretations:
    - LE workers show higher transition rates to unemployment and outside the labor force compared to HE workers.
    - In Brazil, non-working individuals are more likely to return to employment via LE jobs; the opposite is found for the UK.
    - Upward switching in exposure (LE→HE) is more common in the UK relative to Brazil when compared to HE→LE flows:
      - UK: HE→LE 1.7 percent vs LE→HE 3.3 percent.
      - Brazil: HE→LE 11 percent vs LE→HE 7.1 percent.
    - Possible contributing factors: higher share of college-educated individuals and higher supply of HE jobs in the UK; informality in Brazil.

### Annex B — Additional Job Transition Analysis
- Contents and methods:
  - Figure B.1: Unconditional transition probabilities between exposure categories (scaled by probability of changing occupations).
  - Figure B.2: Occupation transition probabilities for workers with a high school degree (UK vs Brazil).
  - Figure B.3: Wage lifecycle analysis restricted to “stayers” (workers who did not change occupation over one year).
  - Regression specification (simplified) for log wage changes (equation 5):
    - ∆ log(y_irt) = δ1 J2J_irt + δ2 OS_irt × J2J_irt + δ3 EUE_irt + β X_irt + γ_t + η_r + ε_irt
- Table B.1: Wage Variation — estimated coefficients (UK and Brazil)
  - Selected coefficient estimates and significance:
    - Switch Occ: UK 0.0166 *** ; Brazil 0.00560 ***
    - Switch Emp: UK 0.0454 *** ; Brazil 0.00876 *
    - Occup. and Employer Change: UK -0.0224 ; Brazil -0.0150 ***
    - EUE Employer Change: UK -0.115 *** ; Brazil -0.0280 ***
    - Age 25-44: UK -0.0234 *** ; Brazil -0.0423 ***
    - Age 45-59: UK -0.0347 *** ; Brazil -0.0507 ***
    - Age 60+: UK -0.0537 * ; Brazil -0.0404 ***
    - College: UK 0.0692 ** ; Brazil 0.0553 ***
    - Is Informal: Brazil -0.0739 ***
    - Was Informal: Brazil 0.0621 ***
    - Observations: UK 50700 ; Brazil 687501
    - R2: UK 0.007 ; Brazil 0.005
    - State FE: Yes ; Year-Quarter FE: Yes
  - Interpretation:
    - Wage growth is greater for younger workers with college degrees.
    - There is a penalty associated with becoming informal (-0.074 log points in Brazil) and informality affects wage growth.
- Table B.2: Wage Premia — estimates of transition terms (UK and Brazil)
  - Selected coefficient estimates (significance levels preserved):
    - Switch Emp: UK 0.0461 *** ; Brazil 0.00858 *
    - Switch Occ: UK 0.00582 ; Brazil 0.00391
    - Switch Emp × Switch Occ: UK -0.0255 ; Brazil -0.0172 ***
    - EUE: UK -0.114 *** ; Brazil -0.0277 ***
    - Multiple interaction terms shown involving Same Occ, Switch Occ, HE/LE, HC/LC (coefficients and significance preserved as in table).
    - Observations: UK 50700 ; Brazil 687501
    - R2: UK 0.007 ; Brazil 0.007
    - State FE: Yes ; Year-Quarter FE: Yes ; Individual Characteristics: Yes
  - Interpretations and findings:
    - Workers with a college degree experience higher gains when transitioning to high-exposure occupations, but larger losses when moving to low-exposure occupations.
    - Non-college individuals also see gains when moving to occupations more exposed to AI, but gains are more modest.
    - Figure B.5: Transition probabilities through unemployment (conditional on occupation switch, formality, and education) show higher probabilities of becoming informal when transitions involve unemployment:
      - Average probability of going informal after switching occupations through unemployment around 30% to 40%, up to 50% for HE→LE transitions.
    - Implication: Transitioning jobs through unemployment can carry higher risk of becoming informal and associated wage penalties (linked to Table B.1 estimates).

### Annex C — Industry Distribution of AI Exposure
- Summary findings:
  - Services sectors concentrate most AI-exposed employment; some services sections have shares of HE jobs up to around 90 percent.
  - Country differences:
    - Brazil: jobs more concentrated in trade, with a higher share of low complementarity occupations.
    - UK: more jobs in other services activities such as communications and finance, with higher complementarity.
  - Notable sector:
    - Education sector: almost 85 percent of workers are in high-exposure, high-complementarity jobs.
- Figures:
  - Figure C.1: High-Exposure Employment Shares by Industry Section (UK and Brazil) — displays shares of total employment in high exposure occupations, broken into low complementarity and high complementarity components.

*Source: wpiea2024116-print-pdf - References and Annexes*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024116-print-pdf.pdf_