## 18. Empirical specification, data, and variable definition

## Source details

**Canonical URL:** [18. Empirical specification, data, and variable definition](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019243-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019243-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019243-print-pdf.pdf.json)

---

### I. High-level objectives and findings
- Research objectives:
  - Comparative analysis of Luxembourg labor market developments using microdata from the Eurostat Labor Force Survey.
  - Estimate probabilities of being unemployed or employed controlling for other factors.
  - Analyze work-welfare trade-offs embedded in the tax-benefit system for young, low skilled, older workers and women.
  - Use an empirical model to assess the effect of the tax-benefit system on unemployment and activity rates.
- Novelty:
  - Micro-level data controls for overlap between vulnerable groups.
  - Comparison before and after the crisis (2006 and 2014).
  - Combines micro-level and macro-level analysis to study interactions between structural characteristics and macroeconomic policies.
- High-level findings preview:
  - Work disincentives in the tax-benefits system contribute to structural unemployment in Luxembourg.
  - Most vulnerable groups: youth, low-skilled, and non-EU born.
  - Low employment of older workers and women largely driven by low participation rates.
  - Low-skilled workers face both higher involuntary unemployment and lower participation.
  - Relative importance of benefit schemes varies across groups:
    - Unemployment traps explain relatively high unemployment among young and low-skilled.
    - Disincentives for second earners contribute to lower participation of women.
    - Generosity of the pensions system largely drives weak labor market attachment of seniors.

### II. Empirical model and microdata
- Model specification:
  - Probit regression on EU LFS microeconomic data for Luxembourg and neighboring countries (Germany excluded in some comparisons).
  - Latent variable 푌푖∗ with observed status Y = 1 if 푌푖∗ > 0, else 0.
  - 푌푖∗ = 훽0 + Σ푗=1푘 훽푗 푋푖푗 + 휇푖.
  - X: age, gender, education attainment, migration status, years of residency, household composition.
- Outcomes:
  - Unemployment probability: Y = 1 if individual i is unemployed, 0 if employed.
  - Employment probability: Y = 1 if individual i is employed, 0 if unemployed or inactive.
- Years compared: 2006 (pre-crisis) and 2014 (post-crisis).
- Subgroups: age (15–24, 25–54, 55–64), education (lower secondary, upper secondary, tertiary), migration status (native, EU born, non-EU born), gender.
- Sample sizes:
  - Unemployed models — 36,396 (2006), 6,106 (2014).
  - Employed models — 67,868 (2006), 11,085 (2014).
  - Micro-level dataset coverage: nearly 85,000 individual responses for 2006 and 14,000 for 2014 (EU LFS yearly dataset).

### III. Selected probit estimation results (preserve reported values)
- Unemployed — Probability (selected):
  - 2006: Age 25–54 (base) 0.036; 15–24 0.139; 55–64 -0.021.
  - 2014: Age 25–54 (base) 0.044; 15–24 0.126; 55–64 -0.004*.
- Employed — Probability (selected):
  - 2006: Age 25–54 (base) 0.794; 15–24 -0.535; 55–64 -0.495.
  - 2014: Age 25–54 (base) 0.817; 15–24 -0.561; 55–64 -0.418.
- Gender (base = Female):
  - Unemployed probability (2006) base 0.056; Male -0.024 (2006), 0.002* (2014).
  - Employed probability (2006) base 0.440; Male 0.152 (2006), 0.096 (2014).
- Country of birth (base = Native):
  - Unemployed probability (2006) base 0.032; EU born 0.020 (2006), 0.035 (2014); Non-EU born 0.088 (2006), 0.097 (2014).
  - Employed probability (2006) base 0.510; EU born 0.026 (2006), 0.005* (2014); Non-EU born -0.071 (2006), -0.072 (2014).
- Education (base = Lower secondary):
  - Unemployed probability (2006) base 0.057; Upper secondary -0.017 (2006), -0.013* (2014); Tertiary -0.029 (2006), -0.037 (2014).
  - Employed probability (2006) base 0.447; Upper secondary 0.085 (2006), 0.082 (2014); Tertiary 0.179 (2006), 0.182 (2014).
- Years of residency (base = ≤ 1 year):
  - Unemployed probability (2006) base 0.054; 2 or 3 years 0.004*; 4 years or more -0.011*.
  - Employed probability (2006) base 0.501; 2 or 3 years 0.003*; 4 years or more 0.014*.

### IV. Interpretation of micro-results (key magnitudes and contrasts)
- Youth and skill effects:
  - Youth unemployment penalty: 15–24 is 12.6 percentage points more likely to be unemployed in 2014 than 25–54.
  - Skill unemployment penalty: In 2006, not finishing upper-secondary raised unemployment risk by 2.9 percentage points relative to having a university degree; this rose to 3.7 percentage points in 2014.
  - Low-skilled workers are considerably more vulnerable than high-skilled workers; the skill penalty in Luxembourg is lower than in Belgium and France.
- Migration and residency:
  - Non-EU born migrants’ unemployment risk is more than three times that of natives (conditional on other factors).
  - EU born migrants: absolute unemployment risk in Luxembourg was 6.7 percent in 2014 (base probability plus marginal effect), the lowest among neighbors, but EU-born migrants incur the highest marginal unemployment penalty relative to natives.
  - Being resident for 4+ years increases employment probability by 10 percentage points relative to newcomers.
    - Contrast: In France, staying 4+ years reduces unemployment risk by 22 percentage points compared to recent immigrants.
- Age and gender dynamics:
  - Older workers (55–64) lost their pre-crisis premium: in 2006 they were 2.1 percentage points less likely to be unemployed than 25–54; this advantage disappeared by 2014.
  - Gender: pre-crisis Luxembourg had highest marginal unemployment risk for females relative to males; by 2014 this gender difference vanished. In 2014 males were 9.6 percentage points more likely to be employed, indicating lower activity for females.
- Joint subgroup contrasts:
  - Young, non-EU migrants and low-skilled workers underperform across joint subgroup comparisons.
  - Young workers with a university degree: unemployment probability 13.1 percent — more than double unemployment risk of low-skilled prime-age workers.
  - Employment probabilities: high-skilled young workers have a 37 percent probability to get a job, while low-skilled prime-age workers have a 72.2 percent employment probability.
  - Non-EU migrants with a tertiary degree have unemployment probabilities more than twice that of low-skilled natives. EU migrants with a tertiary degree perform similarly to low-skilled natives in employment probability.

### V. Tax-benefit indicators, cross-country panel regressions and elasticities
- Data and model:
  - Estimation sample: 35 advanced OECD countries over 2001–15.
  - Tax-benefits indicators from OECD; unemployment and activity rates from OECD and Eurostat.
  - Reduced-form panel regressions with country and time fixed effects; standard errors corrected using Hubert/White method.
- Definitions preserved (OECD-based): Total labor cost; Personal average tax rate; Net personal average tax rate; Net personal average tax rate of the second earner; Tax wedge (average and marginal); Marginal effective tax rate; Net replacement rate; Average effective age of retirement; Participation tax rate (PTR); Trap measures (inactivity/unemployment trap, low-wage trap).
- Cross-country elasticity evidence (selected coefficients — effect of a 1-percentage point increase):
  - Unemployment rate regressions:
    - Age 15–64: Average Participation Tax Rate from Unemployment 0.221***, Average Tax Wedge 0.327*, Average 5-Years Net Replacement Rate 0.264***.
    - Age 15–24: 0.446***, 0.726**, 0.556***.
    - Age 25–54: 0.191**, 0.298**, 0.230***.
    - Age 55–64: 0.151**, 0.304**, 0.203***.
    - Lower Secondary: 0.227**, 0.247, 0.338**.
    - Upper Secondary: 0.224**, 0.356*, 0.300***.
    - Tertiary: 0.069, 0.113, 0.120***.
    - Men: 0.246***, 0.404**, 0.312***.
    - Women: 0.180**, 0.221, 0.208***.
  - Participation rate regressions (effect of a 1-percentage point increase):
    - PTR Inactivity / Average Tax Wedge (selected):
      - Age 55–64: PTRIn -0.363***, ATW -0.729***.
      - Women: PTRIn 0.153***, ATW 0.022.
- Aggregate effect magnitudes:
  - A 10-percentage points reduction in the participation tax rate from unemployment benefits would reduce the overall unemployment rate by 2.2 percentage points.
  - A 10-percentage points reduction in the net replacement rate of unemployment benefits would lower the unemployment rate by 2.6 percentage points.
  - A 10-percentage point reduction in the average tax wedge would reduce the unemployment rate by 3.3 percent.

### VI. Marginal effective tax rates (METR), second earners, childcare, and seniors
- METR and low-wage traps:
  - High METR reduce incentives to work more hours, generating low-wage traps; in Luxembourg METR are high across family situations for part-time workers, especially at the bottom of the income distribution.
  - Example: a one-earner married couple with two children moving from 33 percent to 67 percent of the average wage faces a METR of more than 100 percent (no increase in net income after accounting for loss of benefits).
- Second earners and female labor supply:
  - Work disincentives are substantial for second earners (mostly women); higher tax wedges on second earners are associated with higher unemployment rates for women working part-time.
  - 2016 tax reform introduced optional individual taxation for married or co-habiting workers; planned move to fully individual taxation over the medium term noted.
  - Net cost of childcare relatively high; recently introduced free 20 hours/week multilingual childcare is a step forward.
  - Expansion of daycare and after-school programs recommended to encourage female participation.
- Older workers and pensions:
  - Generous pension system with several early retirement options discourages older workers’ participation.
  - Labor supply of older workers is significantly responsive to PTR from inactivity and the tax wedge.
  - A 10-percentage point reduction in the participation tax rate from inactivity would increase seniors’ labor market participation rate by 3.6 percentage points.
  - Responsiveness of seniors’ activity rate to the tax wedge is large (coefficient reported: -0.729*** for tax wedge on participation for 55–64).

### VII. Luxembourg-specific results, interactions, and robustness
- Interaction terms PTRUB#LUX, ATW#LUX, NRR5Y#LUX generally not statistically significant (large standard errors), implying cross-country coefficients broadly apply to Luxembourg.
- Selected baseline coefficients (PTRUB, ATW, NRR5Y) for Luxembourg similar to baseline panels (examples preserved in regression tables).
- Sensitivity checks:
  - Excluding Luxembourg leaves results broadly unchanged; key coefficients on PTRUB, ATW, NRR5Y remain similar.
  - Full estimation results and robustness checks presented in Annexes C, D, E, F.

### VIII. Policy-relevant evidence and recommendations (preserve phrasing and numbers)
- Labor market context:
  - Net employment creation accelerating to 3.8 percent in 2017.
  - Cross-border workers represent more than 40 percent of the employed.
  - Resident employment rate aged 20–64 years was 70.7 percent in 2016, below national target of 73 percent.
  - Overall participation rate was 70 percent, below neighboring countries' average of 74 percent.
  - Youth participation rate was 30.7 percent.
  - Women represent more than 80 percent of part-time workers.
  - Unemployment rate in Luxembourg stood at 5.8 percent in 2017.
  - Workers with less than a secondary education represent less than 20 percent of the labor force but more than 50 percent of registered jobseekers at ADEM at end-December 2017.
  - Share of unemployed out of a job for 1 year or more accounted for more than 45 percent of the unemployed in 2016.
- Work incentives and tax-benefit design:
  - Luxembourg’s average tax wedge for single workers without children at average earnings is 38 percent.
  - Marginal tax wedge highest for single parents with two children above 60 percent of average earnings.
  - Benefits can be equivalent to almost 50 percent of median income for not working.
  - For a one-earner married couple with 2 children taking a job at 67 percent of the average wage, taxes and benefits reduce the financial gain by more than 86 percent.
  - A one-earner married couple with 2 children earning previously 67 percent of the average wage can be better off living on unemployment benefits than taking a job: net replacement tax rate is more than 100 percent.
  - Replacement of guaranteed minimum income (RMG) by Social Inclusion Income (REVIS) in January 2019 can increase incentives to search and accept jobs by allowing beneficiaries to keep a higher part of welfare benefits after accepting a job.
  - Luxembourg’s maximum duration of unemployment insurance reported as 12 months (as in Germany) versus 24 months in France.
  - Recipients in 2014 were more than 1.5 times recipients in 2007; jobseekers more than doubled between December 2007 and December 2014.
- Policy recommendations:
  - Targeted labor market interventions:
    - Increase targeting of ALMPs to the young, low-skilled, non-EU immigrants, and refugees.
    - Expand job search assistance and enhance the apprenticeship system.
  - Education and training:
    - Focus on upgrading education outcomes in a multi-lingual society and improving vocational training to reduce skills mismatches.
  - Make work more rewarding, especially for low earners:
    - Refocus unemployment and welfare benefits to promote active job search and vacancy acceptance.
    - Greater use of in-work tax credits to ensure unemployed are better off taking up a job than remaining unemployed, reducing unemployment traps for the low-skilled.
    - Note: introduction of Revenu d’Inclusion Sociale (REVIS) described as a positive step.
  - Additional measures:
    - Further expand childcare and after-school availability to raise female labor market participation.
    - Reform pension system incentives and early-retirement options to increase seniors’ labor market participation.
  - Improve integration of migrants:
    - Targeted policies to accelerate migrants’ integration given rising participation with longer residency but persistent unemployment gaps.
- Activation and ALMPs:
  - ALMPs by ADEM are innovative but unemployment for low-skilled, young and older workers worsened compared to pre-crisis; non-EU migrants are less integrated.
  - Effective ALMPs can boost employment but are unlikely alone to substantially reduce unemployment given vulnerable group characteristics and low participation rates.

*Source: IMF Staff calculations and analysis as presented in the chapter "18. Empirical specification, data, and variable definition."*

### REFERENCES  __________________________________________________________ 28

### REFERENCES

### I. INTRODUCTION
- Individual labor market performance is driven by individual characteristics and behavior as well as economic developments.
- People fit and willing to work may still be unemployed due to skills mismatches, deficiency of aggregate demand, or difficulties to integrate the labor market.
- Social safety nets create a work-welfare trade-off when net income gains from work do not justify leisure sacrifice.
- Luxembourg labor market characteristics:
  - Robust employment growth, but resident employment rate is relatively low.
  - Many newly created jobs go to cross-border workers.
  - Employment is shifting to high-skilled jobs, worsening prospects for workers lacking adaptation capabilities.
  - Unemployment rate remains higher than its pre-crisis level while the job vacancy rate is increasing, suggesting skills mismatches.
  - Active Labor Market Policies (ALMP) by ADEM are innovative but unemployment for low-skilled, young and older workers has worsened compared to pre-crisis; non-EU migrants are less integrated.
- Research objectives and methods:
  - Comparative analysis of Luxembourg labor market developments.
  - Use microdata from the Eurostat Labor Force Survey to estimate probabilities of being unemployed or employed, controlling for other factors.
  - Analyze work-welfare trade-offs embedded in the tax-benefit system for young, low skilled, older workers and women.
  - Use an empirical model to assess the effect of the tax-benefit system on unemployment and activity rates.
- Novelty:
  - Micro-level data controls for overlap between vulnerable groups.
  - Comparison before and after the crisis to assess differential impacts.
  - Combines micro-level and macro-level analysis to study interactions between structural characteristics and macroeconomic policies.
- High-level findings preview:
  - Work disincentives in the tax-benefits system contribute to structural unemployment in Luxembourg.
  - Most vulnerable groups: youth, low-skilled, and non-EU born.
  - Low employment of older workers and women largely driven by low participation rates.
  - Low-skilled workers face both higher involuntary unemployment and lower participation.
  - Relative importance of benefit schemes varies across groups:
    - Unemployment traps explain relatively high unemployment among young and low-skilled.
    - Disincentives for second earners contribute to lower participation of women.
    - Generosity of the pensions system largely drives weak labor market attachment of seniors.
- Paper structure:
  - Section II: literature overview.
  - Section III: key determinants of labor market performance.
  - Section IV: effects of tax and benefit system on labor supply.
  - Section V: policy implications.

### II. LITERATURE REVIEW
- Empirical approaches:
  - Many studies use risk-probit models to assess socio-economic determinants of individual labor market outcomes (examples cited).
  - Studies of aggregate unemployment/employment drivers include indicators of tax and benefit systems and characteristics of labor markets for different worker groups.
- Findings summarized from literature:
  - High replacement levels in unemployment benefits create financial disincentives to work.
  - Longer duration of unemployment benefits may generate benefit dependency and increase unemployment duration.
  - Stricter initial entitlement criteria reduce unemployment inflows; periodic re-examination of eligibility may increase outflows.
- Gap addressed by this paper:
  - Disentanglement of impacts of individual structural characteristics and work disincentives inherent to the tax-benefit system on different worker groups’ labor market outcomes.
  - Analysis includes a broad range of tax and benefits variables rather than a single aggregate indicator.

### III. WHAT MAKES AN INDIVIDUAL MORE VULNERABLE IN THE LABOR MARKET?
A. Who Are the Most Vulnerable in the Labor Market?
- Aggregate and group-level unemployment patterns:
  - Unemployment rate in Luxembourg stood at 5.8 percent in 2017.
  - Unemployment has more than doubled over the last seventeen years and is two percentage points or more above its pre-crisis level.
  - Workers with less than a secondary education represent less than 20 percent of the labor force but more than 50 percent of registered jobseekers at ADEM at the end of December 2017.
  - Share of unemployed out of a job for 1 year or more steadily increased since 2009 and accounted for more than 45 percent of the unemployed in 2016.
  - Unemployment persistence is highest for workers older than 45 years.
- Labor market overview (select points from figures/text):
  - Net employment creation accelerating to 3.8 percent in 2017.
  - Cross-border workers represent more than 40 percent of the employed.
  - Resident employment rate aged 20–64 years was 70.7 percent in 2016, below the national target of 73 percent.
  - Overall participation rate was 70 percent, below neighboring countries' average of 74 percent.
  - Youth participation rate was 30.7 percent.
  - Women represent more than 80 percent of part-time workers.
  - Low participation among specific groups (young, low-skilled, women, older workers) is a main factor driving their low employment.
- Education and skills issues:
  - Luxembourg spends almost the double per student than neighboring countries and more than almost all OECD countries, even after controlling for living standards.
  - 2015 PISA results: less than a quarter of the stock of teachers in primary education meet all the qualification requirements.
  - PISA 2015 indicates students in Luxembourg had lower performance in math, science and reading than in neighboring countries.
  - Average PISA scores of natives and second-generation immigrants are broadly in line with neighboring countries.
  - Socioeconomically disadvantaged students underperform more than in neighboring countries.
  - High spending on education does not translate into higher average students’ test scores.
  - Digital Skills Bridge program implemented to help adapt to digital transformation.
- Skills mismatches and structural change:
  - Employment is shifting to high value-added sectors requiring abstract and non-routine skills, reducing demand for routine manual jobs.
  - Job vacancy rate is increasing while integration of low-skilled and long-term unemployed remains challenging, indicating substantial mismatches.
  - Skills mismatches can lead to lower job satisfaction, higher unemployment risk, discouragement, and labor market withdrawal.
B. Key Determinants of Labor Market Performance Using Micro-Data
- Methodology preview:
  - Use probit regressions to estimate the relative likelihood of being out or in a job conditional on socioeconomic group membership.
  - Estimating both unemployment and employment probabilities allows assessment of labor force participation effects and controls for overlap between vulnerable sub-groups.

*Source: Excerpt from wpiea2019243-print-pdf (REFERENCES and preceding sections).*

### 18.      Empirical specification, data, and variable definition.

### 18.      Empirical specification, data, and variable definition.

### Empirical model and data
- Probit regression model estimated on microeconomic data from the European Union Labor Force Survey (EU LFS) for Luxembourg and neighboring countries.
- Latent variable formulation:
  - Observed (un)employment status Y linked to latent 푌푖∗ by: Y = 1 if 푌푖∗ > 0, else 0.
  - Probit model: 푌푖∗ = 훽0 + Σ푗=1푘 훽푗 푋푖푗 + 휇푖.
  - 푋: individual characteristics — age, gender, education attainment, migration status, years of residency in the country, and household composition.
- Outcomes considered:
  - Unemployment probability: Y = 1 if individual i is unemployed, 0 if employed.
  - Employment probability: Y = 1 if individual i is employed, 0 if unemployed or inactive.
- Years compared to assess crisis effects: 2006 (pre-crisis) and 2014 (post-crisis).
- Subgroup analyses by age (15–24, 25–54, 55–64), education (lower secondary, upper secondary, tertiary), migration status (native, EU born, non-EU born), and gender.

### Estimation strategy and robustness
- Two-step strategy:
  - Estimate probability of being (un)employed in 2006 and 2014 for Luxembourg and neighboring countries (Germany excluded).
  - Estimate likelihoods across sub-groups and joint sub-groups (age × education × migration × gender) to assess combined effects.
- Interpretation:
  - Absolute probability for base category shown in bold; marginal effects shown for other categories — marginal effects interpret as change relative to base category.
- Robustness notes:
  - Estimates robust to heteroskedasticity.
  - Full estimation results presented in Annex C.

### Key estimation results (selected coefficients and probabilities)
- Table 1. Probit Regression — Selected reported values (change in probability compared to base category unless otherwise noted). Observations noted per specification.
  - Unemployed — Probability (2006): 0.036 (25–54 years, base); 0.139 (15–24 years); -0.021 (55–64 years).
  - Unemployed — Probability (2014): 0.044 (25–54 years, base); 0.126 (15–24 years); -0.004* (55–64 years).
  - Employed — Probability (2006): 0.794 (25–54 years, base); -0.535 (15–24 years); -0.495 (55–64 years).
  - Employed — Probability (2014): 0.817 (25–54 years, base); -0.561 (15–24 years); -0.418 (55–64 years).
  - Gender (base = Female): Unemployed probability (2006) base 0.056; Male -0.024 (2006), 0.002* (2014).
  - Gender (base = Female): Employed probability (2006) base 0.440; Male 0.152 (2006), 0.096 (2014).
  - Country of birth (base = Native): Unemployed probability (2006) base 0.032; EU born 0.020 (2006), 0.035 (2014); Non-EU born 0.088 (2006), 0.097 (2014).
  - Country of birth (base = Native): Employed probability (2006) base 0.510; EU born 0.026 (2006), 0.005* (2014); Non-EU born -0.071 (2006), -0.072 (2014).
  - Education (base = Lower secondary): Unemployed probability (2006) base 0.057; Upper secondary -0.017 (2006), -0.013* (2014); Tertiary -0.029 (2006), -0.037 (2014).
  - Education (base = Lower secondary): Employed probability (2006) base 0.447; Upper secondary 0.085 (2006), 0.082 (2014); Tertiary 0.179 (2006), 0.182 (2014).
  - Years of residency (base = ≤ 1 year): Unemployed probability (2006) base 0.054; 2 or 3 years 0.004*; 4 years or more -0.011*.
  - Years of residency (base = ≤ 1 year): Employed probability (2006) base 0.501; 2 or 3 years 0.003*; 4 years or more 0.014*.
  - Observations: Unemployed models — 36,396 (2006), 6,106 (2014). Employed models — 67,868 (2006), 11,085 (2014).
  - * Indicates result not significant for p < 0.1.

### Main findings and interpretation
- Youth and skill effects:
  - Youth unemployment penalty: An individual aged 15–24 is 12.6 percentage points more likely to be unemployed in 2014 than an individual aged 25–54.
  - Skill unemployment penalty: In 2006, not finishing upper-secondary school raised unemployment risk by 2.9 percentage points relative to having a university degree; this rose to 3.7 percentage points in 2014.
  - Low-skilled workers are considerably more vulnerable than high-skilled workers; the skill penalty in Luxembourg is lower than in Belgium and France.
- Migration and residency:
  - Non-EU born migrants’ unemployment risk is more than three times that of natives (conditional on other factors).
  - EU born migrants: absolute unemployment risk in Luxembourg was 6.7 percent in 2014 (base probability plus marginal effect), the lowest among neighbors, but EU-born migrants incur the highest marginal unemployment penalty relative to natives.
  - Years of residency: No statistical difference in unemployment risk between newcomers and those who stayed more than 4 years in Luxembourg; however, being resident for 4+ years increases the probability of being employed by 10 percentage points relative to newcomers.
    - Contrast: In France, staying 4+ years reduces unemployment risk by 22 percentage points compared to recent immigrants.
- Age and gender dynamics:
  - Older workers (55–64) lost their pre-crisis premium: in 2006 they were 2.1 percentage points less likely to be unemployed than 25–54 year-olds; this advantage disappeared by 2014.
  - Gender: Luxembourg had the highest marginal unemployment risk for females relative to males before the crisis; by 2014 this gender difference vanished. In 2014 males were 9.6 percentage points more likely to be employed, indicating a lower activity rate for females.
- Joint subgroup results:
  - Young, non-EU migrants and low-skilled workers underperform across joint subgroup comparisons.
  - Young workers with a university degree: unemployment probability 13.1 percent — more than double the unemployment risk of low-skilled prime-age workers.
  - Employment probabilities: high-skilled young workers have a 37 percent probability to get a job, while low-skilled prime-age workers have a 72.2 percent employment probability.
  - Non-EU migrants with a tertiary degree have unemployment probabilities more than twice that of low-skilled natives. EU migrants with a tertiary degree perform similarly to low-skilled natives in employment probability.

### Policy-relevant evidence and implications
- Activation and ALMPs:
  - Effective active labor market policies (ALMPs) can boost employment but are unlikely to alone substantially reduce unemployment given the vulnerable group characteristics and low participation rates of some groups.
- Integration of migrants:
  - Results point to scope for targeted policies to accelerate migrants’ integration into the labor market given rising participation with longer residency but persistent unemployment gaps.
- Work incentives and tax-benefit design:
  - Luxembourg’s average tax wedge:
    - For single workers without children at average earnings, the tax wedge is 38 percent.
    - Marginal tax wedge highest for single parents with two children above 60 percent of average earnings; married couples face the lowest marginal wedge across earnings levels.
  - Welfare generosity and participation traps:
    - Benefits can be equivalent to almost 50 percent of median income for not working.
    - For a one-earner married couple with 2 children taking a job at 67 percent of the average wage, taxes and benefits reduce the financial gain by more than 86 percent.
    - A one-earner married couple with 2 children earning previously 67 percent of the average wage can be better off living on unemployment benefits than taking a job: net replacement tax rate is more than 100 percent.
    - High participation tax rates create substantial inactivity and unemployment traps, especially for low-skilled, low-wage workers.
  - Reforms:
    - Replacement of guaranteed minimum income (RMG) by Social Inclusion Income (REVIS) in January 2019 can increase incentives to search and accept jobs by allowing beneficiaries to keep a higher part of welfare benefits after accepting a job.
  - Unemployment benefit conditionality and duration:
    - Luxembourg’s unemployment benefit entitlement and eligibility criteria are among the strictest across OECD countries.
    - Maximum duration of unemployment insurance reported as 12 months in Luxembourg (as in Germany), versus 24 months in France.
    - Despite strict conditionalities, Luxembourg recorded the fastest increase in recipients among neighboring peers: recipients in 2014 were more than 1.5 times recipients in 2007; jobseekers more than doubled between December 2007 and December 2014.
    - Growth in old-age pension recipients is also significantly higher, reflecting early retirement schemes; disability recipients decreased following reclassification programs.

### Synthesis of policy conclusions
- Targeted supply-side measures are needed alongside ALMPs to address:
  - Youth unemployment (especially for high-skilled youth).
  - Integration of non-EU migrants and EU migrants’ higher marginal unemployment penalty.
  - Low participation rates among certain demographic groups, in particular female activity.
- Recalibrating tax-benefit incentives:
  - Reduce participation tax rates and inactivity traps for low-wage workers and single parents.
  - Use reforms like REVIS to preserve work incentives by allowing beneficiaries to retain a larger share of benefits once employed.
- Monitor and adapt unemployment benefit design and activation policies given the paradox of strict conditionality but rapid growth in recipients.

*Source: IMF Staff calculations and analysis as presented in the chapter "18. Empirical specification, data, and variable definition."*

### 35.      The labor supply of young and low-skilled workers is sensitive to the generosity of

### 35. The labor supply of young and low-skilled workers is sensitive to the generosity of the tax benefits system

### Sensitivity of unemployment and participation to tax-benefits generosity
- Cross-country evidence: the elasticity of the unemployment rate to the generosity of the tax-benefit system differs across worker groups.
  - The unemployment rate of young workers is twice more responsive to the generosity of the unemployment benefits system than that of prime-age workers.
  - Workers with lower secondary education are more sensitive to unemployment benefit generosity than workers with a university degree.
  - One reason: high-skilled workers earn high wages and are less likely to be content with unemployment benefits and actively search for a new position when they lose their job.
- Aggregate relationship:
  - Higher unemployment rates are likely in countries with more generous unemployment benefit systems, but the slopes are not steep.
- Data note:
  - For 2015, the average wage used in the OECD tax-benefit model is 55,858 euros for Luxembourg.
- Estimated elasticities (selected results from Tables):
  - Coefficients on unemployment rate regressions (effect of a 1-percentage point increase in each indicator):
    - Age 15–64: Average Participation Tax Rate from Unemployment 0.221***, Average Tax Wedge 0.327*, Average 5-Years Net Replacement Rate 0.264***
    - Age 15–24: 0.446***, 0.726**, 0.556***
    - Age 25–54: 0.191**, 0.298**, 0.230***
    - Age 55–64: 0.151**, 0.304**, 0.203***
    - Lower Secondary: 0.227**, 0.247, 0.338**
    - Upper Secondary: 0.224**, 0.356*, 0.300***
    - Tertiary: 0.069, 0.113, 0.120***
    - Men: 0.246***, 0.404**, 0.312***
    - Women: 0.180**, 0.221, 0.208***
  - Coefficients on participation rate regressions (effect of a 1-percentage point increase):
    - Average Participation Tax Rate from Inactivity and Average Tax Wedge (selected):
      - Age 15–64: 0.029, -0.108
      - Age 15–24: 0.059, -0.216
      - Age 25–54: 0.061, -0.095
      - Age 55–64: -0.363***, -0.729***
      - Lower Secondary: -0.030, -0.055
      - Upper Secondary: 0.067, 0.080
      - Tertiary: 0.103, -0.075
      - Men: 0.012, -0.181**
      - Women: 0.153***, 0.022
  - Aggregate effect magnitudes:
    - A 10-percentage points reduction in the participation tax rate from unemployment benefits would reduce the overall unemployment rate by 2.2 percentage points.
    - A 10-percentage points reduction in the net replacement rate of unemployment benefits would lower the unemployment rate by 2.6 percentage points.
    - A 10-percentage point reduction in the average tax wedge would reduce the unemployment rate by 3.3 percent.

### Marginal effective tax rates (METR) and low-wage traps
- High METR reduce incentives to work more hours, generating low-wage traps.
  - Cross-country comparison suggests higher METR are associated with higher shares of women working part-time.
  - In Luxembourg, METR are high across family situations for part-time workers, especially at the bottom of the income distribution.
  - Example: among OECD countries, Luxembourg has the highest METR for a one-earner married couple with two children moving from 33 percent to 67 percent of the average wage — the METR is more than 100 percent (no increase in net income after accounting for loss of benefits).
- Consequence:
  - With such METRs, part-time work is widespread, especially among women.

### Second earners, childcare, and female labor supply
- Work disincentives are substantial for second earners (mostly women).
  - High elasticities of hours worked of second earners to tax-benefits disincentives.
  - Higher tax wedges on second earners are associated with higher unemployment rates for women working part-time.
- Luxembourg tax reform context:
  - The 2016 tax reform in Luxembourg will likely have a limited impact on the tax wedge on second earners.
  - Under optional individual taxation, any additional incentive for the second earner can be at the cost of the principal earner because the tax scale applied to the first earner under joint taxation is more favorable.
  - The new government plans to move to fully individual taxation over the medium term (noted as planned).
- Childcare and participation:
  - Net cost of childcare borne by workers in Luxembourg is relatively high, reducing incentives for women with young children to participate in the labor market.
  - The recently introduced free 20 hours/week multilingual childcare is a step in the right direction.
  - Additional expansion of daycare and after-school programs could further encourage women’s labor market participation.
- Empirical associations:
  - Higher METR and higher tax wedge on second earners correlate with higher shares of women in part-time work and lower employment outcomes for sole-parent mothers when childcare costs are high.

### Older workers, pensions, and work incentives
- Incentives for older workers to keep working could be improved.
  - Financial incentives to retire play a significant role for both men and women.
  - Taxation affects the financial return to continued work and the level of net retirement income.
  - In Luxembourg, the generous pension system with several options of early retirement discourages older workers’ labor market participation and encourages early exit.
- Empirical magnitudes:
  - Only the labor supply of older workers is significantly responsive with the expected sign to either the participation tax rate from inactivity or the tax wedge.
  - A 10-percentage point reduction in the participation tax rate from inactivity would increase seniors’ labor market participation rate by 3.6 percentage points.
  - The responsiveness of seniors’ activity rate to the tax wedge is even higher (coefficient reported: -0.729*** for tax wedge on participation for 55–64 in participation regressions).

### Modeling approach, data, and robustness
- Empirical specification:
  - Reduced-form panel regression relating participation or unemployment rates of worker group g in country k at time t to tax-benefits indicators X, controls Z (output gap/GDP growth rate, education shares), country fixed effects, and time fixed effects.
  - Tax-benefits indicators: participation tax rates from unemployment or inactivity, the tax wedge, and the net replacement rate of unemployment benefits over 5 years.
- Data, sample, and variables:
  - Estimation sample covers 35 advanced OECD countries over the period 2001–15.
  - Tax-benefits indicators come from the OECD tax-benefits data and are simple averages across income levels and family situations.
  - Worker groups: 15–64 (aggregate), young (15–24), prime-age (25–54), senior (55–64); education: low secondary, upper secondary, tertiary; gender.
- Estimation strategy and robustness:
  - Cross-country panel regressions with country and time fixed effects.
  - Standard errors corrected using the Hubert/White method.
  - Excluding Luxembourg from the sample leaves results broadly unchanged.
  - Interaction terms between tax-benefits variables and a Luxembourg dummy are generally not statistically significant, implying cross-country coefficients apply to Luxembourg.
  - Full estimation results presented in Annexes D, E, and F (referenced in source).

### Key policy findings and recommendations
- Labor market outcomes summary:
  - Job creation is strong in Luxembourg, but unemployment of young and low-skilled declines only gradually; activity rates of women and seniors remain low; non-EU migrants and refugees are less integrated.
  - Resident employment remains below the national target and lags European peers; rising shares of unemployed face longer spells without a job.
  - Skills mismatches explain much structural unemployment; more than half of new jobs go to cross-border commuters due to skills mismatches.
- Policy recommendations:
  - Targeted labor market interventions:
    - Increase targeting of ALMPs to the most vulnerable groups: the young, low-skilled, non-EU immigrants, and refugees.
    - Expand job search assistance and enhance the apprenticeship system.
  - Education and training:
    - Education reforms should focus on upgrading education outcomes in a multi-lingual society with diverse pupil backgrounds and improving the quality of vocational training to reduce skills mismatches.
  - Make work more rewarding, especially for low earners:
    - Refocus unemployment and welfare benefits to promote active job search and vacancy acceptance.
    - Greater use of in-work tax credits to ensure unemployed are better off taking up a job than remaining unemployed, reducing unemployment traps for the low-skilled.
    - The recent introduction of the Revenu d’Inclusion Sociale (REVIS) is noted as a positive step.
  - Additional measures:
    - Further expand childcare and after-school availability to raise female labor market participation.
    - Reform pension system incentives and early-retirement options to increase seniors’ labor market participation.

*Source: IMF staff analysis (pages 35–27 of the provided content unit).*

### 51.      Improving participation of women and seniors. The 2016 tax reform has introduced

### 51. Improving participation of women and seniors

### Policy changes to taxation and gender neutrality
- The 2016 tax reform has introduced optional individual taxation for married or co-habiting workers.
- Consideration should be given to increasing the second-earner income tax-deduction.
- Moving to fully individual income taxation would make the tax system more gender neutral by reducing the marginal tax rate applied to the earnings of second earners, often women.

### Labor market participation supports for women
- Further expanding the availability of daycare and after-school programs could also improve women labor market participation.

### Increasing participation of seniors
- Raising the participation of seniors would require significantly limiting access to benefits for early retirement.

### Related microdata and coverage (analysis basis)
- The micro-level analysis is based on data from the European Union Labor Force Survey form Eurostat.
- The analysis focuses on the yearly dataset for 2006 and 2014.
- The database contains nearly 85,000 individual responses for 2006 and 14,000 for 2014.
- Data collection covers the years from 1983 and onwards, and LFS data cover residents—natives and former migrants living in the country, but does not cover cross border workers.
- For the purposes of this study, “natives” are identified as all LFS respondents born in the country, and “migrants” as all the respondents who moved to the country at some point in the past.

*Source: wpiea2019243-print-pdf - 51. Improving participation of women and seniors. The 2016 tax reform has introduced optional individual taxation for married or co-habiting workers.*

### 2.      Tax-benefit regression. OECD is the primary data source for the tax-benefits indicators.

### 2. Tax-benefit regression. OECD is the primary data source for the tax-benefits indicators.

### Data, sample, and period
- Primary data sources: OECD (tax-benefits indicators), OECD and Eurostat (unemployment and activity rates, population by education attainment, GDP growth).
- Analysis period: 2001-2015.
- Sample: 35 advanced countries: Austria, Belgium, Bulgaria, Switzerland, Czech Republic, Germany, Denmark, Estonia, Greece, Spain, Finland, France, Hungary, Croatia, Cyprus, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherland, Norway, Poland, Portugal, Slovakia, Romania, Slovenia, Sweden, Turkey, U.K., U.S., Canada, and Japan.
- Number of countries used in each estimation depends on data availability for unemployment and activity rates and explanatory variables.

### Definition of tax and benefit indicators (OECD-based definitions preserved)
- Total labor cost: sum of gross wage earnings of employees, employer social security contributions and—in some countries—payroll taxes.
- Personal average tax rate: income tax plus employee social security contributions as a percentage of gross wage earnings.
- Net personal average tax rate: personal income tax and employee social security contributions net of cash benefits as a percentage of gross wage earnings.
- Net personal average tax rate of the second earner: increase in income tax and employee social contributions (net of in-work benefits) paid by the family because of the second earner entering workforce divided by the increase in family gross income because of the second earner entering in the workforce.
- Tax wedge: difference between the total labor cost of employing a worker and its net earnings; calculated as (personal income tax + employee + employer social security contributions + any payroll tax - benefits) as a percentage of labor costs.
  - Average tax wedge: part of total labor costs taken in tax and social security contributions net of cash benefits.
  - Marginal tax wedge: percentage of the marginal increase in labor costs that is deducted through increasing taxes and social security contributions and decreasing cash benefits.
- Marginal effective tax rate: part of an increase in earnings "taxed away" by personal income taxes and employee social security contributions, considering withdrawal of social and other earnings-related benefits.
- Net replacement rate: net income of an unemployed person receiving unemployment (and possibly other) benefits as a share of pre-unemployment income; calculated at different points because benefits decline over unemployment spell.
  - Net pension replacement rate: individual net pension entitlement divided by net pre-retirement earnings, considering personal income taxes and social security contributions paid by workers and pensioners.
- Average effective age of retirement: average age of exit from the labor force during a 5-year period calculated as a weighted average of net withdrawals from the labor market, with net labor force exits estimated by the difference in participation rates for each 5-year age group (40 and over) at the beginning and the corresponding age group aged 5-years older at the end.
- Participation tax rate (PTR): proportion of gross earnings taken in tax or reduced benefits; measured by one minus the financial gains to working (net income in work – net income out of work) as proportion of gross earnings, calculated for moving from inactivity (or unemployment benefits) to work.
- Trap measures:
  - Inactivity (unemployment) trap: share of additional gross income from transition that is taxed away by higher taxes and lower benefits; incentive for an inactive or unemployed person to move to paid employment (or to refrain from doing so).
  - Low-wage trap: financial incentive (or disincentive) to increase a low level of earnings by working additional hours.

### Determinants of individual labor market performance (summary of coefficient table structure)
- Table C.2 reports coefficients representing changes in probability (unemployment or employment) compared to base categories, estimates robust to heteroskedasticity; "* Indicates that the result is not significant for p < 0.1".
- Base categories used repeatedly: Age 25-54 years; Female; Native; Single, no child; Lower secondary education; Years of residency less than or equal to 1 year.
- Selected coefficient examples (preserve values exactly as in the tables):
  - Probability of being unemployed (aggregated examples):
    - Age 25-54 years (base) 0.036 0.044 0.075 0.080 0.082 0.105
    - 15-24 years 0.139 0.126 0.127 0.138 0.129 0.134
    - 55-64 years -0.021 -0.004* -0.028 -0.024 -0.030 -0.035
    - Female (base) 0.056 0.051 0.096 0.087 0.103 0.114
    - Male -0.024 0.002* -0.021 0.001* -0.019 -0.003
  - Probability of being employed (selected country panels):
    - Luxembourg (examples): Age 25-54 years (base) 0.794 0.817; 15-24 years -0.535 -0.561; 55-64 years -0.495 -0.418.
    - Belgium (examples): Age 25-54 years (base) 0.749 0.763; Female (base) 0.440 0.476; Male 0.152 0.096.
    - France (examples): Age 25-54 years (base) 0.786 0.751; 15-24 years -0.499 -0.479; Female (base) 0.453 0.473.

### Effects of tax-benefit system — Baseline specification (key regression coefficients preserved)
- Table D.1 — Dependent variable: Unemployment Rate. Notes: Standard errors in parentheses; * 10%, ** 5%, *** 1%. PTRUB = participation tax rate from unemployment benefits averaged across all income/family situations. NRR5Y = net replacement rate of unemployment benefits averaged across all income/family situations.
- PTRUB coefficients (selected):
  - 15-64: 0.221*** (0.076)
  - 15-24: 0.446*** (0.146)
  - 25-54: 0.191** (0.070)
  - 55-64: 0.151** (0.059)
  - Lower Secondary: 0.227** (0.106)
  - Upper Secondary: 0.224** (0.084)
  - Tertiary: 0.069 (0.046)
  - Men: 0.246*** (0.080)
  - Women: 0.180** (0.076)
- Average Tax Wedge coefficients (selected):
  - 15-64: 0.327* (0.163)
  - 15-24: 0.726** (0.340)
  - 25-54: 0.298** (0.144)
  - 55-64: 0.304** (0.143)
  - Lower Secondary: 0.247 (0.264)
  - Upper Secondary: 0.356* (0.175)
  - Tertiary: 0.113 (0.072)
  - Men: 0.404** (0.169)
  - Women: 0.221 (0.158)
- NRR5Y (Net Replacement Rate 5-year) coefficients (selected):
  - 15-64: 0.264*** (0.082)
  - 15-24: 0.556*** (0.175)
  - 25-54: 0.230*** (0.068)
  - 55-64: 0.203*** (0.070)
  - Lower Secondary: 0.338** (0.125)
  - Upper Secondary: 0.300*** (0.094)
  - Tertiary: 0.120*** (0.043)
  - Men: 0.312*** (0.096)
  - Women: 0.208*** (0.069)
- GDP Growth coefficients (negative association with unemployment rate, selected):
  - 15-64: -0.279** (0.100)
  - 15-24: -0.528** (0.207)
  - 25-54: -0.273** (0.100)
  - 55-64: -0.143** (0.066)
  - Lower Secondary: -0.353** (0.150)
  - Upper Secondary: -0.296** (0.110)
  - Tertiary: -0.181** (0.072)
  - Men: -0.319*** (0.108)
  - Women: -0.233** (0.092)
- Model statistics (examples):
  - Observations: 300 (most columns), R-squared examples: 0.324 (15-64 PTRUB), 0.391 (15-24 PTRUB), Number of countries: 21 (PTRUB panel).

- Table D.2 — Dependent variable: Participation Rate. PTR Inactivity = participation tax rate from inactivity averaged across income/family situations.
- PTR Inactivity coefficients (selected):
  - 15-64: 0.029 (0.041)
  - 15-24: 0.059 (0.132)
  - 25-54: 0.061 (0.050)
  - 55-64: -0.363*** (0.086)
  - Lower Secondary: -0.030 (0.070)
  - Upper Secondary: 0.067 (0.052)
  - Tertiary: 0.103 (0.087)
  - Men: 0.012 (0.088)
  - Women: 0.153*** (0.043)
- Average Tax Wedge coefficients on participation (selected):
  - 15-64: -0.108 (0.100)
  - 15-24: -0.216 (0.246)
  - 25-54: -0.095 (0.101)
  - 55-64: -0.729*** (0.230)
  - Lower Secondary: -0.055 (0.169)
  - Upper Secondary: 0.080 (0.125)
  - Tertiary: -0.075 (0.088)
  - Men: -0.181** (0.087)
  - Women: 0.022 (0.153)
- GDP Growth coefficients on participation (selected; mostly small/insignificant):
  - 15-64: -0.043 (0.038)
  - 15-24: -0.072 (0.144)
  - 25-54: -0.024 (0.026)
  - 55-64: 0.036 (0.084)

### Luxembourg-specific effects (Tables E.1 and E.2: interaction terms with LUX preserved)
- Table E.1 — Dependent variable: Unemployment Rate; interaction terms PTRUB#LUX, ATW#LUX, NRR5Y#LUX included.
- PTR Unemployment Benefits (PTRUB) baseline coefficients (same as baseline):
  - 15-64: 0.222*** (0.076)
  - 15-24: 0.439*** (0.144)
  - 25-54: 0.193** (0.070)
  - 55-64: 0.151** (0.059)
  - Men: 0.251*** (0.081)
  - Women: 0.176** (0.075)
- PTRUB#LUX interaction coefficients (selected; large standard errors):
  - 15-64: -0.026 (0.301)
  - 15-24: 0.300 (0.557)
  - 25-54: -0.069 (0.272)
  - 55-64: 0.555 (0.442)
  - Men: -0.185 (0.334)
  - Women: 0.179 (0.271)
- Average Tax Wedge (ATW) baseline coefficients (selected):
  - 15-64: 0.345** (0.162)
  - 15-24: 0.757** (0.353)
  - 25-54: 0.318** (0.144)
  - 55-64: 0.304** (0.145)
  - Men: 0.426** (0.170)
  - Women: 0.236 (0.155)
- ATW#LUX interaction coefficients (selected):
  - 15-64: -0.303 (0.420)
  - 15-24: -0.529 (0.869)
  - 25-54: -0.341 (0.381)
  - 55-64: -0.024 (0.201)
  - Men: -0.367 (0.461)
  - Women: -0.238 (0.380)
- NRR5Y baseline coefficients (selected):
  - 15-64: 0.265*** (0.081)
  - 15-24: 0.558*** (0.173)
  - 25-54: 0.232*** (0.068)
  - 55-64: 0.203*** (0.070)
  - Men: 0.314*** (0.095)
  - Women: 0.209*** (0.068)
- NRR5Y#LUX interaction coefficients (selected; very large standard errors):
  - 15-64: -0.529 (1.305)
  - 15-24: -0.772 (2.404)
  - 25-54: -0.542 (1.168)
  - 55-64: -0.210 (1.203)
  - Men: -0.923 (1.414)
  - Women: -0.092 (1.195)
- Table E.2 — Dependent variable: Participation Rate. PTRIn#LUX and ATW#LUX interaction terms reported.
- PTR Inactivity (PTRIn) baseline examples:
  - 15-64: 0.024 (0.042)
  - 55-64: -0.376*** (0.086)
  - Women: 0.143*** (0.040)
- PTRIn#LUX interaction coefficients (selected):
  - 15-64: 0.285 (0.172)
  - 15-24: 0.687* (0.397)
  - 25-54: 0.333** (0.154)
  - 55-64: 0.811 (0.542)
  - Men: 0.616*** (0.196)
  - Women: 0.598*** (0.189)
- ATW#LUX interaction coefficients on participation (selected):
  - 15-64: 0.334 (0.218)
  - 15-24: 0.841 (0.575)
  - 25-54: 0.495*** (0.161)
  - 55-64: -0.830* (0.474)
  - Men: 0.424* (0.208)
  - Women: 0.632** (0.279)

### Excluding Luxembourg (sensitivity; Tables F.1 and F.2)
- Table F.1 — Dependent variable: Unemployment Rate (Luxembourg excluded). Key coefficients largely similar to baseline:
  - PTRUB (selected): 15-64: 0.221*** (0.076); 15-24: 0.435*** (0.144); 25-54: 0.192** (0.070).
  - Average Tax Wedge (selected): 15-64: 0.344** (0.162); 15-24: 0.748** (0.351); 25-54: 0.317** (0.144).
  - NRR5Y (selected): 15-64: 0.262*** (0.081); 15-24: 0.552*** (0.174); 25-54: 0.229*** (0.068).
  - GDP Growth (selected): 15-64: -0.277** (0.109); 15-24: -0.525** (0.223); 25-54: -0.271** (0.107).
- Table F.2 — Dependent variable: Participation Rate (Luxembourg excluded). Selected coefficients:
  - PTR Inactivity: 15-64: 0.024 (0.040); 55-64: -0.374*** (0.086); Women: 0.141*** (0.039).
  - Average Tax Wedge on participation: 15-64: -0.130 (0.106); 55-64: -0.655** (0.242); Men: -0.215** (0.092).
  - GDP Growth on participation: 15-64: -0.051 (0.036).

*Source: wpiea2019243-print-pdf - 2.      Tax-benefit regression. OECD is the primary data source for the tax-benefits indicators.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019243-print-pdf.pdf_