## wpiea2022124-print-pdf

## Source details

**Canonical URL:** [wpiea2022124-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022124-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022124-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022124-print-pdf.pdf.json)

---

### Measurement approach and data sources
- Combines European Labor Force Survey (EU-LFS) and European Union Statistics on Income and Living Conditions (EU-SILC).
- EU-LFS: repeated cross-section; constructs job-to-job (EE_job) and sectoral (EE_sec) switches; coverage from 1983/1989 to 2019 for most countries.
- EU-SILC: panel structure; computes occupational switches (EE_occ, UE_occ) and earnings-related outcomes; available 2005–2018 and restricted to individuals who appear for 4 years consecutively (France exception).
- Industrial classification: ISIC Rev. 4. Occupation codes: one-digit ISCO-08. Definition of “routine occupations” follows Carrillo-Tudela and others (2016).
- Business cycle phases computed using the Hardin and Pagan (2002) algorithm for annual data. Recessions defined as years with negative GDP growth.

### Definitions of reallocation outcomes
- EE_job: dummy = 1 if employed this and last year and hired within the last 12 months.
- EE_sec: dummy = 1 if employed this and last year and changed sector this year (one-digit ISIC Rev. 4).
- EE_occ: dummy = 1 if employed this and last year and changed occupations (one-digit ISCO-08).
- UE_occ: dummy = 1 if unemployed last year and employed this year and changed occupation this year.
- Earnings changes: log changes in real earnings (nominal earnings deflated by CPI), computed as specified in the source.
  - For on-the-job occupational switches (EE): earnings changes over a two-year period among those continuously employed.
  - For via-unemployment switches (EUE): earnings changes over three years for those employed two years ago, unemployed last year, and employed this year.
- EU-LFS used for job-to-job and sectoral outcomes; EU-SILC used for occupation and panel-based outcomes.

### Key incidence statistics (long-run averages)
- Job-to-job (EE_job) probability: 0.060
  - 95 Percent confidence bands: [0.059, 0.061]
- Sector-to-sector (EE_sec) probability: 0.033
  - 95 Percent confidence bands: [0.032,0.035]
- Occupations — Employed workers (EE_occ) probability: 0.123
  - 95 Percent confidence bands: [0.116,0.130]
- Occupations — Unemployed workers (UE_occ) probability: 0.466
  - 95 Percent confidence bands: [0.450,0.481]

### Earnings consequences associated with occupational switches (long-run averages)
- Employed workers (EE_occ): Earning change due to switch = 0.019
  - 95 Percent confidence bands: [0.029, 0.009]
  - Interpretation: on-the-job occupational switches associated with earnings gains of about 2 percent.
- Unemployed workers reemployed with occupation change (UE_occ): Earning change due to switch = -0.158
  - 95 Percent confidence bands: [-0.078, -0.238]
  - Interpretation: occupational switches via an unemployment spell associated with an average earnings penalty of about 15.8 percent.
- Empirical estimates from linear probability models with country, year and sector/occupation t-1 fixed effects; standard errors clustered at country-year level.

### Cyclical fluctuations — transition probabilities by business cycle phase
- Estimation: linear probability model with recession and recovery dummies per Hardin and Pagan (2002); standard errors clustered at country-year level.
- Sectors — Job-to-job (EE_job)
  - Expansion 0.062 [0.061, 0.064]
  - Recession 0.055 [0.051, 0.059]
- Sectors — Sector-to-sector (EE_sec)
  - Expansion 0.035 [0.032, 0.038]
  - Recession 0.029 [0.020, 0.037]
- Occupations (all workers)
  - Expansion 0.127 [0.114, 0.139]
  - Recession 0.131 [0.110, 0.152]
- Occupations (employed workers, EE_occ)
  - Expansion 0.119 [0.107,0.132]
  - Recession 0.124 [0.103,0.145]
- Occupations (unemployed workers, UE_occ)
  - Expansion 0.430 [0.384,0.477]
  - Recession 0.444 [0.353,0.535]
- Interpretation:
  - Business cycle does not materially impact sectoral and occupational switch probabilities in aggregate; few statistically significant differences between expansions and recessions.
  - Mechanical effect: higher unemployment in recessions increases incidence of occupational switches via unemployment, generating more workers suffering earnings penalties upon reemployment.

### Cyclical fluctuations — earnings changes by business cycle phase
- All workers (EE_occ combined)
  - Expansion 0.018 [-0.004, 0.041]
  - Recession 0.021 [-0.011, 0.052]
- Employed workers (EE_job / on-the-job switches)
  - Expansion 0.029 [0.014, 0.043]
  - Recession 0.001 [-0.019, 0.020]
- Unemployed workers (EUE_occ / via-unemployment)
  - Expansion -0.172 [-0.305, -0.038]
  - Recession -0.150 [-0.304, 0.005]
- Interpretation:
  - Earnings gains vary over the business cycle for on-the-job occupational switches (EE_occ) but not for via-unemployment switches (EUE_occ).
  - On-the-job switchers gain less during recessions; reemployed switchers suffer losses in both phases.

### Heterogeneity in sectoral transitions by demographic groups (selected coefficients)
- Regression: Equation (2) with individual controls, country, year, and sector in t-1 fixed effects; SE clustered at country-year level.
- Key coefficients (standard errors):
  - Female: EE_job -0.00360*** (0.000337); EE_sec -0.000312 (0.000500)
  - Youth: EE_job 0.0651*** (0.00135); EE_sec 0.0273*** (0.000794)
  - Old: EE_job -0.0326*** (0.000673); EE_sec -0.0124*** (0.000661)
  - Low skilled: EE_job -0.00480*** (0.000513); EE_sec -0.00175*** (0.000511)
  - Recession X Female on EE_job: 0.00283*** (0.00108)
  - Recession X Young on EE_job: -0.00736** (0.00358)
  - Recession X Old on EE_sec: 0.00426* (0.00243)
- Sample: N 29,684,842; R2 0.039 (EE_job), R2 0.056 (EE_sec).
- Interpretation:
  - Women, old, and low-skilled have lower chances of switching jobs; youth have higher chances.
  - Heterogeneous cyclical responses: women more likely to switch job during recessions; young worse prospects for job switches during recessions; old marginally more likely to change sector during recessions.

### Heterogeneity in occupational switches by demographic groups (selected coefficients)
- Dependent vars: EE_occ and UE_occ; country, year, occupation t-1 fixed effects; SE clustered at country-year.
- Key coefficients (standard errors):
  - Female: EE_occ -0.0174*** (0.00237); UE_occ -0.0189 (0.0186)
  - Youth: EE_occ 0.0399*** (0.00316); UE_occ 0.0562** (0.0223)
  - Old: EE_occ -0.0195*** (0.00185); UE_occ -0.0757*** (0.0225)
  - Low skilled: EE_occ -0.00668** (0.00268); UE_occ 0.0125 (0.0217)
  - Recession on UE_occ: -0.0598 (0.0440) and -0.126** (0.0603) across models
  - Recession x Low skilled on UE_occ: 0.128** (0.0556)
- Sample sizes and fit:
  - EE_occ: N 439,004; R2 0.084
  - UE_occ: N 8,189; R2 0.064–0.065
- Interpretation:
  - Women less likely than men to switch occupations.
  - Youth more likely, old less likely to switch occupations.
  - Low-skilled less likely to switch via employment; no significant difference via unemployment.
  - During recessions, low-skilled more likely to switch occupation via-unemployment (distributional concern).

### Earnings consequences and heterogeneity (selected regression results)
- Dependent var: Change in earnings (EE)
  - Switch: 0.0186** (0.00875)
  - Female: 0.0176*** (0.00293)
  - Young: 0.180*** (0.00616)
  - Old: -0.0331*** (0.00251)
  - Low skill: -0.0145*** (0.00337)
  - Low skill x Switch: -0.0194** (0.00868)
  - N 369,381; R2 0.027
- Dependent var: Change in earnings (EUE)
  - Switch: -0.366*** (0.117)
  - Female: -0.00889 (0.0611)
  - Young: 0.265*** (0.0756)
  - Old: -0.0879 (0.0911)
  - Low skill: -0.227** (0.0885)
  - Low skill x Switch: 0.200* (0.117)
  - N 4,616; R2 0.096
- Interpretation and examples:
  - Continuously employed workers (EE) experience average earning gains; females and young experience larger gains; old and low-skilled experience smaller gains.
  - Reemployed after one-year unemployment (EUE) generally suffer earning losses relative to pre-displacement earnings.
  - Example: for young workers in EUE, -0.233 + 0.265 = 0.035 (small gain); for low-skilled, -0.233 - 0.227 = -0.46 (severe loss).
  - Additional average earning loss of –0.37 if the worker also switches occupation upon reemployment (EUE_occ).
- Summary bullets from text:
  - EE: earning gains for workers continuously employed for two consecutive years.
  - EUE: earning losses upon reemployment after unemployment.
  - EE_occ associated with gains; EUE_occ associated with losses.
  - Low-skilled: smaller EE gains and larger EUE losses; do not gain much from EE_occ but suffer less when changing occupation via unemployment.
  - Young: larger EE gains, smaller EUE penalties, larger EE_occ increases, less severe EUE_occ losses.
  - Women: larger EE gains and less likely to switch occupation.
  - During recessions, young experience smaller EE gains from switching occupation and larger losses if they go through unemployment.

### Routinization — aggregate trend
- Average share of employment in routine occupations fell from around 61 percent in 1995 to 55 percent in 2019.
- Average calculated across countries using routine definition from Carrillo-Tudela and others (2016) and normalized time fixed effects per Karabarbounis and Neiman (2014).

### Transitions between routine and non-routine jobs (selected transition and regression results)
- Workers previously in non-routine occupations:
  - Much less likely to lose jobs, much more likely to find jobs, and more likely to stay in non-routine occupations than those previously in routine occupations (Table 7 summary).
- Transition matrix highlights (table layout preserved in source):
  - Origin Non-Routine (Total): 0.86 non-employment? 0.05 routine? 0.09
  - Origin Routine (Total): 0.04 non-employment? 0.77 routine? 0.19
- Heterogeneity (Table 8 selected coefficients):
  - Routine t-1 on EE_job: 0.0436*** (Column 1)
  - Routine t-1 on EE_sec: 0.0246*** (Column 4)
  - Interactions:
    - Female X Routine t-1 on EE_job: -0.00431***
    - Young X Routine t-1 on EE_job: 0.0598***
    - Old X Routine t-1 on EE_job: -0.0152***
    - Low skill X Routine t-1 on EE_job: 0.0176***
  - Recession coefficient example: EE_job recession -0.00785 (SE 0.00892) in one specification.
  - Observations: 29,684,842 (full), 4,138,623 (recession subsample). Country FE and Year FE included.

### Occupational switches and demographics (selected coefficients)
- Routine t-1 associated with lower occupational switching:
  - Routine t-1 on EE_occ: -0.0181*** (Column 1)
  - Routine t-1 on UE_occ: -0.0480** (Column 4)
- Selected interaction effects:
  - Routine t-1 on EE_occ (alternative spec): 0.0509*** (Column 2)
  - Routine t-1 on UE_occ (alternative spec): 0.0955**
  - Low skill X Routine t-1 on EE_occ: -0.119***
  - Low skill X Routine t-1 on UE_occ: -0.271***
- Sample sizes:
  - EE_occ regressions: N 438,971 (full)
  - UE_occ regressions: N 8,189 (full)
  - R2 examples: 0.073, 0.079, 0.084 (EE_occ); 0.040, 0.052, 0.119 (UE_occ)

### Earnings consequences related to routine jobs (selected results)
- On-the-job EE_occ:
  - Switch coefficient (Column 1): 0.0107**
  - Switch coefficient (Column 2 including SwitchNon2Routine): 0.0200***
  - SwitchNon2Routine (EE_occ, Column 2): -0.0502***
  - Sum of Switch + SwitchNon2Routine is negative and significant: movement from non-routine to routine can more than offset on-the-job gains.
  - During recessions, downgrading loss stronger (Column 3 shows larger negative estimates).
  - N for EE_occ: 369,381 (All), 71,289 (Recession). R-sq examples: 0.026, 0.041.
- Via-unemployment EUE_occ:
  - Switch coefficient (Column 4): -0.187***
  - SwitchNon2Routine effect (Column 5): -0.242*** (additional ~24 percent loss noted)
  - Business cycle does not appear to affect EUE_occ downgrading penalty materially.
  - N for EUE_occ: 4,616 (All), 910 (Recession). R-sq examples: 0.081, 0.144.

### Earnings consequences across demographic groups (selected highlights)
- EE_occ — SwitchNon2Routine penalties (Table 11):
  - Women SwitchNon2Routine (All): -0.0544***; during recession -0.124*** ("-12.4 percentage points during the recession relative to -5.4 percentage points for the whole sample").
  - Old SwitchNon2Routine during recession: -0.0620* (text: -6.4 percentage points in recession vs -3.8 for whole sample).
  - Low skill SwitchNon2Routine during recession: -0.0673*** (text: -6.7 percentage points in recession vs -3.5 for whole sample).
  - Baseline Switch coefficients examples:
    - Women Switch: 0.0305***; Young Switch: 0.0235*; Old Switch: 0.0508***; Low skill Switch: 0.0169***.
- EUE_occ — SwitchNon2Routine penalties:
  - Women Switch (All): -0.134**.
  - Low skill SwitchNon2Routine (All): -0.454*** (substantial negative coefficient in one specification).
  - Older and low-skilled workers suffer large earning penalties from switching non-routine to routine via unemployment.
  - Sample sizes for EUE_occ demographic regressions smaller (e.g., Women N = 2,096; Young N = 426; Old N = 1,024; Low skill N = 3,449).
- Summary:
  - EE transitions: non-routine→routine earning losses larger for young and women; during recessions worse for women, older, low-skilled.
  - EUE transitions: older and low-skilled suffer the largest penalties from non-routine→routine switching.

### Key implications and conclusions
- Aggregate decline in routine employment from 61 percent (1995) to 55 percent (2019) interacts with reallocation to produce distributional consequences.
- Workers moving from non-routine to routine occupations:
  - Do not enjoy typical on-the-job earning gains.
  - Suffer higher earning penalties when reemployed after unemployment.
- Heterogeneity:
  - Young workers: more job-to-job switches with EE gains; less severe EUE losses.
  - Low-skilled workers: smaller EE gains; larger EUE losses; during recessions large penalties when switching non-routine to routine.
  - Older workers: less occupational mobility and large EUE penalties when downgrading.
  - Women: larger EE gains but more likely to incur SwitchNon2Routine losses during recessions.
- Overall: declining routine-job share generates significant distributional impacts via lower job-finding probabilities and worse earning outcomes for affected groups.

*Source: Working Paper No. WP/2022/124 — "The Distributional Impacts of Worker Reallocation: Evidence from Europe" (selected sections 2.1, 3.2, 5.1).*

### 2.1 Measuring the incidence and earnings consequences of worker reallocation ......................................... 7

### 2.1 Measuring the incidence and earnings consequences of worker reallocation

### Measurement approach and data sources
- Combines two main data sources: the European Labor Force Survey (EU-LFS) and the European Union Statistics on Income and Living Conditions (EU-SILC).
- EU-LFS: repeated cross-section; used to construct job-to-job (EE_job) and sectoral (EE_sec) switches; coverage from 1983/1989 to 2019 for most countries.
- EU-SILC: panel structure; used to compute occupational switches (EE_occ, UE_occ) and earnings-related outcomes; available 2005–2018 in the analysis and restricted to individuals who appear for 4 years consecutively (France is an exception).
- Industrial classification: ISIC Rev. 4. Occupation codes: one-digit ISCO-08. Definition of “routine occupations” follows Carrillo-Tudela and others (2016).
- Business cycle phases computed using the Hardin and Pagan (2002) algorithm for annual data. Recessions defined as years with negative GDP growth.

### Definitions of reallocation outcomes (as used throughout the empirical exercises)
- EE_job: dummy equal to one if an individual, conditional on being employed this and last year, was hired within the last 12 months.
- EE_sec: dummy equal to one if an individual, conditional on being employed this and last year, changed sector this year (one-digit ISIC Rev. 4).
- EE_occ: dummy equal to one if an individual, conditional on being employed this and last year, changed occupations (one-digit ISCO-08).
- UE_occ: dummy equal to one if an individual, conditional on being unemployed last year and employed this year, changed occupation this year.
- Earnings changes: log changes in real earnings (nominal earnings deflated by the consumer price index), computed as:
  - Δ푙푙푙푙푙푙푙푙푙푙푙푙푙푙푙푙푔푔
    푡푡
    =100∗(푙푙푙푙
    (
    푙푙푙푙푙푙푙푙푙푙푙푙푔푔
    푡푡
    )
    −푙푙푙푙
    (
    푙푙푙푙푙푙푙푙푙푙푙푙푔푔
    푡푡−푙푙
    )
    )
  - For “on-the-job” occupational switches (EE): earnings changes defined over a two-year period among those continuously employed.
  - For “via unemployment” switches (EUE): earnings changes defined over three years (change between this year and two years ago) for those employed two years ago, unemployed last year, and employed this year.
- Outcomes requiring job-to-job and sectoral information use EU-LFS; outcomes requiring occupation and panel structure use EU-SILC.

### Key incidence statistics (long-run averages)
- Job-to-job (EE_job) probability: 0.060
  - 95 Percent confidence bands: [0.059, 0.061]
- Sector-to-sector (EE_sec) probability: 0.033
  - 95 Percent confidence bands: [0.032,0.035]
- Occupations — Employed workers (EE_occ) probability: 0.123
  - 95 Percent confidence bands: [0.116,0.130]
- Occupations — Unemployed workers (UE_occ) probability: 0.466
  - 95 Percent confidence bands: [0.450,0.481]

### Earnings consequences associated with occupational switches (long-run averages)
- Employed workers (EE_occ): Earning change due to switch = 0.019
  - 95 Percent confidence bands: [0.029, 0.009]
  - Interpretation: on-the-job occupational switches are on average associated with earnings gains of about 2 percent.
- Unemployed workers reemployed with occupation change (UE_occ): Earning change due to switch = -0.158
  - 95 Percent confidence bands: [-0.078, -0.238]
  - Interpretation: occupational switches via an unemployment spell are associated with an average earnings penalty of about 15.8 percent.

### Stylized implications summarized in the section
- Workers have a higher tendency to remain in their current occupation when continuously employed; occupational switching is much more frequent following an unemployment spell (0.123 versus 0.466 probabilities).
- On-the-job occupational switches are generally associated with small average earnings gains (0.019), consistent with sequential bargaining theories.
- Occupational switches associated with reemployment after unemployment incur large average earnings penalties (-0.158).
- The empirical estimates are obtained from linear probability models including country, year and sector/occupation t-1 fixed effects; standard errors clustered at the country-year level.

*IMF WORKING PAPERS — The Distributional Impacts of Worker Reallocation: Evidence from Europe (Section 2.1)*

### 3.2 Cyclical fluctuations

### 3.2 Cyclical fluctuations

### Purpose and empirical approach
- Study whether stylized facts on worker reallocation vary over the business cycle.
- Competing views: recessions accelerate reallocation (Mortesen and Pissarides 1994; Groshen and Potter; Jaimovich and Siu 2014) versus expansions having more employment-to-employment transitions (Barlevy 2002; Carrillo-Tudela and others 2016).
- Estimation: linear probability model (Equation (1)) where the outcome is a labor market transition for individual i, country c, time t. Recession and recovery dummies follow Hardin and Pagan (2002). Standard errors clustered at the country-year level.
- Regressions weighted using individual-level weights rescaled to sum to one for each country-year; EU-SILC longitudinal sampling weights used for EU-SILC data.

### Transition probabilities by business cycle phase (Table 2)
- Reported transitions: job-to-job (EE_job), sector-to-sector (EE_sec), on-the-job occupational switch (EE_occ), via-unemployment occupational switch (UE_occ).
- Means and 95 percent confidence bands (obtained from estimated Equation (1)):

  - Sectors
    - Job-to-job (EE_job)
      - Expansion 0.062 [0.061, 0.064]
      - Recession 0.055 [0.051, 0.059]
    - Sector-to-sector (EE_sec)
      - Expansion 0.035 [0.032, 0.038]
      - Recession 0.029 [0.020, 0.037]
  - Occupations (all workers)
    - Expansion 0.127 [0.114, 0.139]
    - Recession 0.131 [0.110, 0.152]
  - Occupations (employed workers, EE_occ)
    - Expansion 0.119 [0.107,0.132]
    - Recession 0.124 [0.103,0.145]
  - Occupations (unemployed workers, UE_occ)
    - Expansion 0.430 [0.384,0.477]
    - Recession 0.444 [0.353,0.535]

- Interpretation:
  - The state of the business cycle does not appear to materially impact sectoral and occupational switch probabilities; generally no statistically significant differences between expansions and recessions.
  - Mechanical effect: because unemployment rises in recessions and occupational switches are more frequent after unemployment spells, recessions likely generate more occupational switches and more workers suffering earnings penalties upon reemployment.

### Earnings changes by business cycle phase (Table 3)
- Earnings associated with occupational switches considered for on-the-job (EE_occ) and via-unemployment (EUE_occ).
- Means and 95 percent confidence bands (from estimating Equation (1)):

  - All workers
    - Expansion 0.018 [-0.004, 0.041]
    - Recession 0.021 [-0.011, 0.052]
  - Employed workers (EE_job)
    - Expansion 0.029 [0.014, 0.043]
    - Recession 0.001 [-0.019, 0.020]
  - Unemployed workers (EUE_occ)
    - Expansion -0.172 [-0.305, -0.038]
    - Recession -0.150 [-0.304, 0.005]

- Interpretation:
  - Earnings gains vary significantly over the business cycle for “on-the-job” occupational switches (EE_occ) but not for switches “via-unemployment” (EUE_occ).
  - Workers who switch occupation “on-the-job” do not gain as much during recessions as during expansions.
  - Those reemployed after unemployment suffer earnings losses (EUE_occ), with losses evident in both expansions and recessions.

### Heterogeneity in sectoral transitions by demographic groups (Table 4)
- Regression framework: Equation (2) with individual characteristics (age, gender, skill), country, year, and sector in t-1 fixed effects, standard errors clustered at country-year level.
- Key coefficients and significance (standard errors in parentheses):

  - Female
    - EE_job: -0.00360*** (0.000337)
    - EE_sec: -0.000312 (0.000500)
  - Youth
    - EE_job: 0.0651*** (0.00135)
    - EE_sec: 0.0273*** (0.000794)
  - Old
    - EE_job: -0.0326*** (0.000673)
    - EE_sec: -0.0124*** (0.000661)
  - Low skilled
    - EE_job: -0.00480*** (0.000513)
    - EE_sec: -0.00175*** (0.000511)
  - Recession interactions (examples)
    - Recession X Female on EE_job: 0.00283*** (0.00108)
    - Recession X Young on EE_job: -0.00736** (0.00358)
    - Recession X Old on EE_sec: 0.00426* (0.00243)

- Sample size and fit:
  - N 29,684,842; R2 0.039 for EE_job models, R2 0.056 for EE_sec models.
- Interpretation:
  - Women, old, and low-skilled tend to have lower chances of switching jobs; only old and low-skilled have significantly lower probability of switching sectors.
  - Youth have higher chances of switching jobs and sectors.
  - No statistically significant evidence of overall cyclicality of job-to-job and sectoral switches during recessions, but heterogeneous responses by demographic groups: women more likely to switch job during recessions; young workers have worse prospects for job switches during recessions; young disadvantaged for sectoral switches during recessions; old workers marginally more likely to change sector during recessions.

### Heterogeneity in occupational switches by demographic groups (Table 5)
- Dependent variables: EE_occ and UE_occ. Regression includes country, year, and occupation in t-1 fixed effects. Standard errors clustered at country-year level.
- Key coefficients and significance (standard errors in parentheses):

  - Female
    - EE_occ: -0.0174*** (0.00237)
    - UE_occ: -0.0189 (0.0186)
  - Youth
    - EE_occ: 0.0399*** (0.00316)
    - UE_occ: 0.0562** (0.0223)
  - Old
    - EE_occ: -0.0195*** (0.00185)
    - UE_occ: -0.0757*** (0.0225)
  - Low skilled
    - EE_occ: -0.00668** (0.00268)
    - UE_occ: 0.0125 (0.0217)
  - Recession effects
    - Recession on UE_occ: -0.0598 (0.0440) and -0.126** (0.0603) across models
    - Recession x Low skilled on UE_occ: 0.128** (0.0556)

- Sample sizes and fit:
  - EE_occ models: N 439,004; R2 0.084
  - UE_occ models: N 8,189; R2 0.064–0.065
- Interpretation:
  - Women are less likely than men to switch occupations regardless of labor force transition history.
  - Young workers more likely, older workers less likely to switch occupations than prime-aged workers.
  - Low-skilled workers less likely to switch via employment; no significant difference via unemployment.
  - Over the business cycle: young workers are less likely to switch “on-the-job” during recessions; low-skilled workers have a higher chance of switching occupation via-unemployment during recessions, pointing to potential distributional impacts.

### Earnings consequences and heterogeneity (Table 6 and summary)
- Regression of change in earnings; models include country, year, and occupation in year t-1 fixed effects. Standard errors clustered at country-year level.
- Key coefficients and standard errors:

  - Dep. Var: Change in earnings (EE)
    - Switch: 0.0186** (0.00875)
    - Female: 0.0176*** (0.00293)
    - Young: 0.180*** (0.00616)
    - Old: -0.0331*** (0.00251)
    - Low skill: -0.0145*** (0.00337)
    - Low skill x Switch: -0.0194** (0.00868)
    - N 369,381; R2 0.027
  - Dep. Var: Change in earnings (EUE)
    - Switch: -0.366*** (0.117)
    - Female: -0.00889 (0.0611)
    - Young: 0.265*** (0.0756)
    - Old: -0.0879 (0.0911)
    - Low skill: -0.227** (0.0885)
    - Low skill x Switch: 0.200* (0.117)
    - N 4,616; R2 0.096

- Interpretation (from text and coefficient sums):
  - Continuously employed workers (EE) experience average earning gains; magnitude varies by group:
    - Females and young experience larger earning gains while old and low-skilled experience smaller earning gains.
    - Young workers who switch occupations “on-the-job” gain more; low-skilled workers gain less from EE_occ.
  - Workers reemployed after one-year unemployment (EUE) generally suffer earning losses relative to pre-displacement earnings:
    - Example calculations: for young workers in EUE, -0.233 + 0.265 = 0.035 (a small gain); for low-skilled workers, -0.233 - 0.227 = -0.46 (a severe loss).
    - Additional average earning loss of –0.37 if the worker also switches occupation upon reemployment (EUE_occ).
  - Summary bullets from the text:
    - Workers continuously employed for two consecutive years (EE) experience earning gains; those who lose their job tend to earn less upon reemployment (EUE) relative to pre-unemployment earnings.
    - “On-the-job” occupational switches (EE_occ) are associated with earning gains; “via-unemployment” occupational switches (EUE_occ) are associated with earning losses.
    - Low-skilled workers experience smaller earning gains while continuously employed (EE) and larger earning losses upon reemployment after an unemployment spell (EUE).
    - Low-skilled workers also do not experience substantial earning gains from “on-the-job” occupational switches (EE_occ) but do not suffer as much when changing occupation via unemployment (EUE_occ).
    - Young workers experience larger earning gains for employment-to-employment transitions (EE), smaller earning penalties when reemployed after unemployment (EUE), larger increases from on-the-job occupation changes (EE_occ), and less severe losses from switches via-unemployment (EUE_occ).
    - Women experience larger gains while continuously employed (EE) and are less likely to switch occupation (EUE_occ).
    - During recessions, young workers experience smaller earning gains than other demographic groups when they switch occupation and remain employed, and their earning losses are more substantial when they go through an unemployment spell.

### Routinization and cyclical hiring patterns (summary of Figures 1–2 discussion)
- Literature context: decline in routine-task jobs documented (Goos and others, 2009; Autor, 2003; Acemoglu and Autor, 2011; Autor and Dorn, 2013) and argued to occur disproportionately during recessions (Jaimovich and Siu, 2020; Gaggle and Kauman, 2020; Furukawa and Toyoda, 2013).
- Measurement note: routine occupations classified per Carrillo-Tudela and others (2016).
- Net hiring dynamics (Figure 2 interpretation):
  - Across all periods, net-hiring into routine jobs is on average positive.
  - Net hiring is relatively higher for non-routine sectors, predominantly through hiring from unemployment and non-participation, and also through direct hiring from routine-dominated sectors.
  - During recessions, net-hiring into routine sectors turns negative, led by net movements into unemployment.
  - Net hiring into non-routine sectors remains positive during recessions.

*Italic: Content unit: wpiea2022124-print-pdf - 3.2 Cyclical fluctuations*

### 5.1 Aggregate Level of Trend and Cyclicality of Routine Jobs

### 5.1 Aggregate Level of Trend and Cyclicality of Routine Jobs

### Aggregate trend in routinization
- The average share of employment in occupations characterized as routine has fallen from around 61 percent in 1995 to 55 percent in 2019.
- The average is calculated across countries in the sample using the definition of routine occupations from Carrillo-Tudela and others (2016).
- Note on measurement: To account for sample coverage changes, the average share of employment in working-age population across selected economies over time is calculated from the normalized, time fixed effects from a regression of the indicated variable on country and time fixed-effects (Karabarbounis and Neiman 2014).

### Transitions between routine and non-routine jobs
- Workers previously employed in non-routine occupations:
  - Much less likely to lose jobs, much more likely to find jobs, and more likely to stay in non-routine occupations than those previously employed in routine occupations (Table 7).
- Transition matrix highlights (Table 7):
  - Origin Non-Routine (Total): 0.86 non-employment? 0.05 routine? 0.09 (table layout preserved from source).
  - Origin Routine (Total): 0.04 non-employment? 0.77 routine? 0.19 (table layout preserved from source).
  - (Full transition matrix values appear in Table 7 of the source; these entries summarize the reallocation patterns.)
- Heterogeneity by demographics (Table 8):
  - Routine t-1 increases probability of switching jobs and sectors:
    - Routine t-1 coefficient in job transitions (EE_job): 0.0436*** (Column 1).
    - Routine t-1 coefficient in sector transitions (EE_sec): 0.0246*** (Column 4).
  - Interaction effects (selected coefficients, full table in source):
    - Female X Routine t-1 on EE_job: -0.00431***.
    - Young X Routine t-1 on EE_job: 0.0598***.
    - Old X Routine t-1 on EE_job: -0.0152***.
    - Low skill X Routine t-1 on EE_job: 0.0176***.
  - Recession effects:
    - Recession coefficient in EE_job (Column 2 or 3 depending) reported as -0.00785 (standard error 0.00892 in one column; alternative recession rows present).
  - Sample sizes and fixed effects:
    - Observations for Table 8: 29,684,842 (Full sample columns), 4,138,623 (Recession columns).
    - Country FE: Yes. Year FE: Yes. Sector FE: No.

### Occupational switches and demographic patterns
- Summary from Table 9:
  - Workers previously in routine jobs are on average less likely to switch occupations both “on-the-job” (EE_occ) and via-unemployment (UE_occ).
  - Young workers previously in routine jobs have a higher chance of switching occupations than comparable young workers previously in non-routine jobs, for both EE_occ and UE_occ.
  - Low-skilled workers previously in routine occupations have lower chances of switching occupations than low-skilled workers who did not previously work in routine occupations, regardless of transition type.
- Selected coefficients (Table 9):
  - Routine t-1 on EE_occ (Column 1): -0.0181***.
  - Routine t-1 on EE_occ (Column 2 full sample): 0.0509***.
  - Routine t-1 on UE_occ (Column 4): -0.0480**.
  - Routine t-1 on UE_occ (Column 5): 0.0955**.
  - Low skill X Routine t-1 on EE_occ: -0.119***.
  - Low skill X Routine t-1 on UE_occ: -0.271***.
  - N for occupational switch regressions: 438,971 (EE_occ full), 8,189 (UE_occ full); recession subsamples smaller.
  - R2 reported: e.g., 0.073, 0.079, 0.084 for EE_occ specifications; 0.040, 0.052, 0.119 for UE_occ specifications.

### Earning consequences related to routine jobs
- General patterns:
  - “On-the-job” occupational switches (EE_occ) are on average associated with earning gains.
  - Occupational switches via-unemployment (EUE_occ) are on average correlated with an earning penalty.
- Impact of switching from non-routine to routine occupations (Table 10):
  - Switch coefficient for EE_occ (Column 1): 0.0107**.
  - Switch coefficient for EE_occ (Column 2 including SwitchNon2Routine): 0.0200***.
  - SwitchNon2Routine coefficient (EE_occ) in Column 2: -0.0502***.
  - The sum of Switch and SwitchNon2Routine is negative and statistically significant, implying that movement from non-routine to routine occupations can more than offset the earnings gains from on-the-job switches.
  - During recessions, the downgrading loss is particularly strong (Column 3 shows larger negative estimates in recession subsample).
- EUE_occ (via-unemployment) consequences:
  - Switch coefficient for EUE_occ (Column 4): -0.187***.
  - SwitchNon2Routine effect for EUE_occ (Column 5): -0.242*** (additional earning loss around 24 percent reported in text).
  - Business cycle does not seem to affect the earning penalty due to downgrading in EUE_occ (text statement).
- Sample sizes and model fit (Table 10):
  - N for EE_occ: 369,381 (All), 71,289 (Recession).
  - N for EUE_occ: 4,616 (All), 910 (Recession).
  - R-sq values reported (e.g., 0.026, 0.041 for EE_occ; 0.081, 0.144 for EUE_occ).

### Earning consequences across demographic groups
- On-the-job occupational switches (EE_occ) — SwitchNon2Routine penalties (Table 11):
  - Women: SwitchNon2Routine coefficient (All sample) -0.0544***; during recession -0.124*** (text: "-12.4 percentage points during the recession relative to -5.4 percentage points for the whole sample").
  - Young: SwitchNon2Routine shows losses; text notes young workers’ earning losses do not vary as much over business cycles.
  - Old: SwitchNon2Routine during recession -0.0620* (text highlights older workers have -6.4 percentage points during recession relative to -3.8 percentage points for whole sample).
  - Low skill: SwitchNon2Routine during recession -0.0673*** (text highlights -6.7 percentage during recession relative to -3.5 percentage point loss for whole sample).
  - Selected baseline Switch coefficients for EE_occ:
    - Women Switch: 0.0305***.
    - Young Switch: 0.0235*.
    - Old Switch: 0.0508***.
    - Low skill Switch: 0.0169***.
  - N for EE_occ demographic regressions vary by subgroup (e.g., Women N = 180,415; Young N = 34,344; Old N = 39,896; Low skill N = 80,17).
- Via-unemployment occupational switches (EUE_occ) — SwitchNon2Routine penalties:
  - Older and low-skilled workers suffer large earning penalties due to switching from non-routine to routine occupations (Table 11).
  - Examples of coefficients (EUE_occ):
    - Women Switch (All): -0.134**.
    - Old SwitchNon2Routine (All): -0.0348 (with large standard error in some columns); text emphasizes large penalty for older workers.
    - Low skill SwitchNon2Routine (All): -0.454*** (substantial negative coefficient in one reported specification).
  - Sample sizes for EUE_occ demographic regressions are small relative to EE_occ (e.g., Women N = 2,096; Young N = 426; Old N = 1,024; Low skill N = 3,449).
- Summary from demographic analysis:
  - EE transitions: earning losses from switching non-routine to routine are larger for young and women in general; during recessions they get worse for women, older, and low-skilled.
  - EUE transitions: older and low-skilled workers suffer large earning penalties from switching non-routine to routine occupations; women and young do not show the same pattern for EUE in the reported results.
  - No statistically significant difference in additional earning loss due to switching non-routine to routine between full sample and recession periods for some EUE specifications (text statement).

### Key implications and conclusions
- Aggregate decline in routine jobs (61 percent in 1995 to 55 percent in 2019) interacts with individual-level reallocation and produces substantial distributional consequences.
- Workers moving from non-routine to routine occupations:
  - Do not enjoy the typical earning gains from on-the-job occupational switches.
  - Suffer higher earning penalties when re-employed after unemployment.
- Heterogeneity:
  - Young workers more likely to experience job-to-job switches with earning gains via EE.
  - Low-skilled workers tend to suffer earning losses upon reemployment after unemployment spells.
  - Low-skilled and old workers during recessions experience severe earning penalties when switching from non-routine to routine occupations.
- Overall: economy-wide declining trend of routine jobs generates significant distributional impacts through lower job finding probabilities and worse earning consequences for affected workers.

*Source: Working Paper No. WP/2022/124 — "The Distributional Impacts of Worker Reallocation: Evidence from Europe" (selected sections 5.1–5.4 and Conclusion).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022124-print-pdf.pdf_