## 1 Introduction — wpiea2021142-print-pdf

## Source details

**Canonical URL:** [1 Introduction — wpiea2021142-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021142-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021142-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021142-print-pdf.pdf.json)

---

### Paper goals and main findings
- Goals:
  - Provide key statistics on evolution of distribution of earnings levels and earnings growth rates in Italy between 1985 and 2016, focusing on earnings inequality, volatility, and mobility.
  - Explore the role of key institutional elements of the Italian labor market in shaping these trends.
- Main empirical findings:
  - Significant rise in common measures of inequality and volatility between 1985 and 2016; earnings mobility does not change significantly over the sample period.
  - Rise in inequality is mostly driven by a decline in earnings in the bottom part of the distribution.
  - Earnings volatility, measured by dispersion in earnings growth rates, is concentrated in cohorts who entered the labor market after 2000.
  - Increase in part-time work explains much of the rise in earnings inequality; rise in fixed-term contracts explains much of the rise in volatility. Both trends operate primarily through hours worked.
  - Fixed-term contracts increased from 8% in 1998 to 18% of all employment contracts in 2016.
  - Cohort analysis: fixed-term contracts are substantially more prevalent among cohorts entering after 2000.
  - Individual-level analysis: the increase in earnings volatility associated with the end of a fixed-term contract is as large as the increase associated with a layoff.
  - Controlling for observable characteristics, workers who start careers with fixed-term contracts do not improve (and if anything, worsen) their chances of securing an open-ended contract by the end of their first decade.
  - Cohort born in the 1980s had accumulated 15% less work experience than previous cohorts by age 35 (equivalent of a full year of work).

*Source: wpiea2021142-print-pdf — 1. Introduction*

### Institutional context and links to macro trends
- Labor productivity and international comparison:
  - Labor productivity (GDP per hour worked) growth in Italy was half as large as in the G7 during the sample period; after the mid-1990s labor productivity halted in Italy while it kept increasing in the rest of the G7.
- Major structural labor market reforms (mid-1990s onward) and key features:
  - “Social Pact” of 1993: reformed collective bargaining and eliminated automatic indexation of wages to inflation (“scala mobile”).
  - “Treu reform” of 1997: liberalized entry wages for first-job seekers and reduced constraints on fixed-term and part-time use.
  - “Biagi reform” of 2003: introduced a wide variety of atypical contracts (e.g., lavoro intermittente, job sharing, lavoro a progetto).
  - “Fornero reform” of 2012: widened applicability of temporary employment and introduced renewal limits (fixed-term contracts could only be renewed once and have maximum 36 months duration).
  - Poletti decree and Jobs Act (both in 2014): Poletti allowed fixed-term contracts with no obligation to justify use and permitted up to five renewals with a constraint that ratio of fixed-term to open-ended contracts not exceed 20%; Jobs Act changed protection against “just cause” dismissals and introduced “contratto a tutele crescenti”.
- Pensions and social contributions:
  - Pension reforms shifted from defined benefit to defined contribution.
  - Social security contributions in temporary jobs are smaller than in open-ended contracts.
- Labor force participation and unemployment trends (25–54 age group, OECD statistics):
  - Male labor force participation: declined from 95% in 1985 to 88% in 2016.
  - Female labor force participation: expanded from 48% in 1985 to 67% in 2016.
  - Women’s unemployment rate: fell from 11% in 1993 to 7% before the Great Recession; by 2016 unemployment rates were 10% for men and 13% for women.
- Proposed channel linking reforms to weak productivity:
  - Changes in job composition reduce accumulation of work experience and firm-specific human capital among young workers, lowering productivity.

### Data, sample, and measurement highlights
- Data source: INPS administrative data covering 1985–2016.
- Sample design and coverage:
  - 6.6% sample of the Italian population based on 24 randomly selected birth dates.
  - Public sector jobs not reported and account for 16% of total employment.
  - Self-employment not reported and accounts for 20% of total employment.
- Earnings and top-coding:
  - Main earnings measure: sum of all regular and irregular income across all jobs within a year.
  - Data are top-coded for confidentiality at a job-level daily maximum; the topcode always exceeds the 99.5th percentile of the earnings distribution per authors’ calculation.
  - Example: in 2012 INPS reports Y* = min{Y, 645*Days}; with maximum working days 312 the 2012 topcode threshold is approximately 200,000 euros.
  - Annual maximums in other years adjusted using the “Indice delle retribuzioni contrattuali orarie lorde ISTAT”.
- Hours and hourly earnings:
  - Dataset reports number of weeks worked, cause of separation, and whether job is part- or full-time.
  - Imputed hours assumptions: part-time = 25 hours/week; full-time = 40 hours/week (based on Italian SHIW averages).
  - Annual hours worked: sum across jobs within a year, capped at 2,080 hours (38 hours×52 weeks).
  - Average hourly earnings: annual earnings divided by annual hours worked.
- Sample selection and size:
  - Restricted to workers aged 25 to 55 with positive earnings and at least 4 weeks worked in the year; no additional minimum earnings threshold.
  - First percentile of earnings in the sample never falls below 800 euros.
  - Sample includes 2.3 million unique workers (1.4 million men and 0.9 million women) and a total of 22.4 million worker-year observations — approximately 700,000 observations per year.

*Source: wpiea2021142-print-pdf — 1. Introduction (data, sample, measurement)*

### 2.3 INPS labor force trends and comparison with official statistics

### INPS labor market trends (aggregate patterns)
- Key trends (1985 to 2016, INPS series):
  - Share of female workers: increases from about 30% in 1985 to more than 40% in 2016.
  - Share of part-time workers: increases from virtually 0% in 1985 to almost 30% in 2016.
  - Share of workers working 52 weeks during the year: decreases from approximately 88% in 1985 to 77% in 2016.
  - Share of workers with an open-end contract (first observed in 1998): decreases from more than 90% in 1998 to 82% in 2016.
- Interpretation: transition away from mostly men employed full-time all year toward greater female participation, more part-time work, less full-year work, and more fixed-term contracts.

### Comparability of INPS to ISTAT and adjustments
- Two adjustments needed to compare INPS earnings to aggregate ISTAT series:
  - Account for rise in part-time employment.
  - Account for decline in fraction of workers working 52 weeks a year.
- Empirical note: multiplying earnings of part-time workers by factors of 1.5, 1.8 and 2 yields minimal changes in comparisons.
- Data sources compared: ISTAT “indice delle retribuzioni lorde per ULA” (Oros survey) vs INPS micro data for individuals aged 25 to 55.

### Core statistics: categories and scope
- Analysis classified into three broad categories: inequality, volatility, and mobility.

### Inequality — cross-sectional distribution and time patterns
- Two phases in earnings percentiles (indexed to 1985):
  - 1985–1992: all percentiles rise, with percentiles above the median rising more.
  - 1993–2016: all percentiles gradually decline, more pronounced for the bottom of the distribution.
- Gender-specific patterns:
  - Women: median earnings declined as much as the 10th percentile in the second phase.
  - Men: median earnings much more stable.
- Direct measures:
  - P90-P10 gap: shows rising trend in earnings inequality.
  - 2.56 times the standard deviation: consistently higher than P90-P10, suggesting more pronounced tail features than the normal distribution.
- Decomposition:
  - Women: most increase in inequality comes from the top of the distribution, driven by a decline in median earnings (P90-P50).
  - Men: increase in top inequality between 1985–1992 and then a rise in bottom inequality between 1992–2016 (P50-P10).

### Top shares and top-coding adjustment (method and findings)
- INPS top-coded above the 99.5th percentile; simple sums bias top-income shares downward.
- Adjustment method: assume earnings above top-code cutoff are Pareto distributed with same Pareto-tail index as just below cutoff; estimate Pareto-tail index via maximum likelihood on data between the 90th and 99th percentiles each year.
- Estimated Pareto-tail index:
  - Falls sharply between 1985–1992 (from 3.3 to 2.7).
  - Remains rather stable in the 2.6–2.8 range thereafter.
- Impact:
  - Adjustment matters little for top 10% and top 5% shares.
  - Adjustment is crucial for documenting rise in top 1% and top 0.1% shares.
- Data limitation: INPS includes only compensation subject to social security contributions; certain benefits (e.g., stock options treated as ordinary income for tax purposes after 2008) are excluded and may understate top income inequality.

### Inequality trends across cohorts
- Initial inequality at age 25 for cohorts entering between 1985 and 2016:
  - Initial bottom inequality (P50-P10): higher and fairly similar over time, with a spike for cohorts entering around the Great Recession.
  - Initial top inequality (P90-P50): on the rise since the late 1990’s.
- Life-cycle cohort patterns (cohorts entering in 1985, 1995, 2005, 2010):
  - Cohorts entering before 2000: inequality relatively stable over life cycle.
    - Women in these cohorts: inequality remained almost the same throughout the sample period.
    - Men in these cohorts: inequality rises around the Great Recession and remains elevated.
  - Cohorts entering after 2000: faced rising initial inequality.
  - Cohort differences in level of inequality decline as time passes.

### Volatility of earnings growth
- Volatility measure: cross-sectional dispersion in one-year residual earnings growth g1it after controlling for age and year.
- Overall volatility (P90-P10):
  - Increases significantly over the sample period.
  - Women: increase concentrated in the first 15 years of the period.
  - Men: increase lasts until the Great Recession with slight decline thereafter.
  - Women face 50% more volatile earnings than men; difference remains stable over time.
- Decomposition:
  - Both upside (P90-P50) and downside (P50-P10) volatility increase over time.
  - After detrending, upside dispersion is procyclical and downside dispersion is countercyclical.
- Distributional shape measures:
  - Kelley skewness: nearly always lower in absolute terms for women than men; procyclical.
  - Crow-Siddiqui excess kurtosis: values above zero indicate stronger central tendency than normal.
- Cohort-age volatility:
  - Volatility declines as people age.
  - Increase in volatility over time partly reflects compositional changes: more recent cohorts face higher volatility.

### Heterogeneity in volatility by permanent income and age
- Permanent income: average residualized log-earnings over previous 3 years; percentiles calculated each sample year then pooled.
- Dispersion (P90-P10) by permanent income percentile and age groups (25–34, 35–44, 45–55):
  - Volatility declines with age for both genders.
  - Women: dispersion concentrated in early career (likely due to intermittent participation).
  - Men: higher dispersion among low-permanent income workers.
- Kelley skewness by permanent income percentile:
  - Men: distribution roughly symmetric in middle; skewness positive at bottom and top; minimal age effects.
  - Women: younger women show left skewness that steeply declines with permanent income percentile, suggesting reversion to the mean at the top (consistent with intermittent participation).
- Excess kurtosis by permanent income percentile:
  - Inverted-U shape for both genders.
  - Stronger age effects and much higher leptokurtosis for women.
- Time-evolution summary (Appendix A.8):
  - 1990–2000: volatility increases only for bottom third of permanent income distribution.
  - 2000–2010: all groups experience increase in volatility except those in the top third.
  - High-permanent income workers experienced little increase in volatility between 1990 and 2010 and start with the least volatility.
  - Interpretation: risk appears to have shifted over time toward people less equipped to self-insure.

*Source: wpiea2021142-print-pdf — 2.3 INPS labor force trends and comparison with official statistics*

### 3.6 Mobility

### Concept and measurement
- Mobility measures changes in the permanent component of income over extended periods (contrast with volatility).
- Method: allocate workers to permanent income percentiles in yeart and plot average permanent income percentile in yeart+10 conditional on percentile in yeart.
- Benchmarks:
  - “Perfectly mobile”: flat horizontal line at the 50th percentile.
  - “Perfectly stagnant”: curve coincides with the 45 degrees line.

### Empirical patterns and levels
- Mobility is higher early in the career for both women and men (25–34 vs 35–44).
- Low income workers: expected to rise up to 20 percentage points in the permanent earnings distribution.
- Top of distribution: mean reversion observed.
- Mobility patterns do not change significantly between 1995 and 2005 for both women and men.
- Rank-rank correlation analysis:
  - One-year apart rank-rank correlation coefficient: 0.89 (s.e. 0.0001).
  - Four-year apart rank-rank correlation coefficient: 0.79 (s.e. 0.0003).
  - Comparative note: PSID sample (US) for 1985–2012, employed workers aged 25–55, four-year apart rank-rank = 0.74 (s.e. 0.0037), suggesting mobility in Italian sample appears lower than in the US.
- Mobility trends over time:
  - Rank-rank coefficient is very stable over time, ranging between 0.87 and 0.92 throughout the sample period.

*Source: wpiea2021142-print-pdf — 3.6 Mobility*

### 4.3 Are fixed-term contracts stepping stones to permanent jobs?

### Prevalence and transitions involving fixed-term contracts
- Share of fixed-term contracts: rose from 8% of employment contracts in 1998 to 18% in 2016.
- Of workers on a fixed-term contract in yeart:
  - 63% continue to work with the same type of contract at timet+1 (44% with the same employer, 19% with a different one).
  - 19% transition to an open-ended contract (10% with the same employer, 9% with a different one).
  - 17% leave employment altogether.
- Cohort (2016 snapshot):
  - More than one-third of workers who entered the labor market in that year held fixed-term contracts.
  - 20% among those who entered a decade earlier.
  - 15% among those who entered two decades earlier.
- Intermittent workers (employed in yeart but with no record in yeart−1 ort+1) are more common among cohorts entering after the labor market reforms.

### Evidence on the “stepping stone” hypothesis (econometric patterns)
- Empirical strategy uses two samples:
  - Sample 1: “Early starters” = employed at age 25 with a fixed-term contract; comparison = unemployed at age 25.
  - Sample 2: adds condition that both are employed at age 26.
- Coefficients for probability of working on an open-ended contract between age 27 and age 35 (selected specs from Table 2):
  - Early starters:
    - Sample 1, spec (1): 0.0764
    - Sample 1, spec (2): -0.0029
    - Sample 2, spec (3): 0.0103
    - Sample 2, spec (4): -0.0157
  - Average log income (included in spec (2) and (4)): 0.1393 and 0.2093.
  - Male: 0.0258 and 0.0077 (specs with gender).
  - Age: 0.2504 and 0.0547.
  - Age squared/100: -0.0037 and -0.0010.
  - Fixed-contract at 26: 0.0317 (included in Sample 2 specifications).
  - Open-ended contract at 26: 0.2352 and 0.1918.
  - Observations:
    - Sample 1: 3,029,242 (both specs).
    - Sample 2: 1,021,157 (both specs).
- Interpretation:
  - Unconditional estimates suggest early starters may have a slight advantage in securing permanent jobs.
  - Once conditioning on average log-income in the first decade, gender, age, and employment status at age 26, the early-starter advantage disappears or reverses slightly.
  - Conclusion: the “stepping stone” mechanism, if it exists, is rather weak; composition/selection issues remain and evidence is not causal.

### Volatility and separations (Table 1 highlights)
- Regressions of annual earnings volatility on causes of job separation and controls; two specifications reported: (1) without individual fixed effects, (2) with individual fixed effects.
- Key coefficients (Cause of separation):
  - Layoff:
    - (1): 0.267 (standard error 0.002)
    - (2): 0.135 (standard error 0.002)
  - Quit:
    - (1): 0.169 (0.002)
    - (2): 0.073 (0.002)
  - End Contract:
    - (1): 0.202 (0.002)
    - (2): 0.133 (0.002)
  - Other:
    - (1): 0.171 (0.004)
    - (2): 0.093 (0.004)
- Other covariates (selected):
  - Age: -0.015 (0.000) and -0.003 (0.001)
  - Age squared /100: 0.013 (0.000) and 0.009 (0.001)
  - Female: 0.052 (0.001) [reported in column (1)]
  - Job switcher: 0.119 (0.001) and 0.060 (0.001)
  - Sector switcher: 0.095 (0.002) and 0.057 (0.001)
  - Full time to part time: 0.234 (0.003) and 0.168 (0.003)
  - Part time to full time: 0.153 (0.003) and 0.054 (0.003)
- Model fit and samples:
  - Individual fixed effects: NO in column (1); YES in column (2).
  - Adj. R2: 0.084 (col 1); 0.449 (col 2).
  - Observations: 5,789,700 (col 1); 5,836,552 (col 2).
  - Mean volatility stayers: 0.097 (both columns).
  - P-value Layoff=End-contract: 0.000 (col 1); 0.602 (col 2).
- Note: end of fixed-term contract is an increasingly common cause of separation; separations including end-contract are associated with higher earnings volatility.

### Human capital, experience, and productivity growth
- Despite increases in average years of schooling, labor productivity in Italy has been stagnant for more than two decades.
- Human capital decomposition (Mincerian framework):
  - Human capital modeled as function of schooling S, cumulative weeks of experience X, weeks of tenure T, and cohort-quality Q.
  - Log earnings equation estimated to recover parameters β1, β2, γ1, γ2 and cohort fixed effects.
- Empirical findings:
  - The “quality/schooling” component (φ(Qi(b)t)+α(Si(b))) shows overall increase in schooling but a steep decline in “quality” term for cohorts entering between mid-1990s and late 2000s, so overall quality/schooling component declines for these cohorts.
  - Experience component (β(X)+γ(T)) is much larger in level than schooling component and shows recent cohorts have systematically lower predicted human capital profiles over the life cycle.
  - Extrapolating to age 55:
    - Cohort entering in 2005 would accumulate 13% less human capital than cohort entering in 1995.
    - Cohort entering in 2005 would accumulate 9% less human capital than cohort entering in 2000.
- Mechanism: reforms shifted younger workers toward atypical contracts with less attachment to employers, reducing general and firm-specific experience accumulation mechanically (shorter job duration) and behaviorally (lower incentives to invest).

### Policy-relevant implications and summary conclusions
- Expansion of fixed-term and part-time contracts since late 1990s is associated with:
  - Rising inequality and rising earnings volatility for men and women over 1985–2016.
  - A substantial change in dispersion of annual hours worked across jobs, contributing more to inequality than dispersion in average hourly earnings.
  - Increased earnings volatility from one year to the next, implying higher uninsurable risk for individual workers.
  - Reduced on-the-job human capital accumulation (general and firm-specific), plausibly slowing labor productivity growth.
- On the “stepping stone” question:
  - Evidence suggests early entry via fixed-term contracts provides, at best, a weak stepping-stone to permanent employment once observable early-career outcomes are controlled for; selection and unobserved heterogeneity complicate causal interpretation.
- Overall assessment:
  - Labor market reforms facilitating fixed-term contracts likely increased labor market flexibility and participation for some groups but also contributed to persistent two-tier dynamics, higher volatility, and lower human capital accumulation among younger cohorts—factors that plausibly contributed to Italy’s long-run productivity stagnation.

*Source: wpiea2021142-print-pdf — section 4.3 “Are fixed-term contracts stepping stones to permanent jobs?”*

### 1. Introduction_________________________________________________________2

### 1    Introduction

### Paper goals and main findings
- The paper has two goals:
  - Provide key statistics on the evolution of the distribution of earnings levels and earnings growth rates in Italy between 1985 and 2016, focusing on earnings inequality, volatility, and mobility.
  - Explore the role of key institutional elements of the Italian labor market in shaping these trends.
- Main empirical findings:
  - Document a significant rise in common measures of inequality and volatility between 1985 and 2016, while earnings mobility does not change significantly over the sample period.
  - The rise in inequality is mostly driven by a decline in earnings in the bottom part of the distribution.
  - Earnings volatility, measured by dispersion in earnings growth rates, is concentrated in the cohorts of workers who entered the labor market after 2000.
  - The increase in part-time work explains much of the rise in earnings inequality, while the rise in fixed-term contracts explains much of the rise in volatility. Both trends operate primarily through hours worked.
  - The increase in fixed-term contracts (from 8% in 1998 to 18% of all employment contracts in 2016) raises the question of whether they serve as “stepping stones” to stable employment.
  - Cohort analysis shows fixed-term contracts are substantially more prevalent among cohorts entering the labor market after 2000.
  - Individual-level analysis shows the increase in earnings volatility associated with the end of a fixed-term contract is as large as the increase associated with a layoff.
  - Controlling for observable characteristics, workers who start their careers with fixed-term contracts do not improve (and if anything, worsen) their chances of securing an open-ended contract by the end of their first decade in the labor market.
  - The cohort born in the 1980s had accumulated 15% less work experience than previous cohorts by age 35 (the equivalent of a full year of work), reflecting reductions in tenure and early-career accumulation of firm-specific human capital.

### Institutional context and links to macro trends
- Italy experienced a divergence in labor productivity (GDP per hour worked) relative to the G7 during the sample period; labor productivity growth in Italy was half as large as in the G7 during the sample period, and after the mid-1990s labor productivity halted in Italy while it kept increasing in the rest of the G7.
- A string of structural labor market reforms starting in the mid-1990s reshaped the Italian labor market and contributed to a de facto dual labor market:
  - Major reforms listed: the “Social Pact” of 1993, the “Treu reform” of 1997, the “Biagi reform” of 2003, the “Fornero reform” of 2012, the “Poletti decree” and the Jobs Act (both passed in 2014).
  - The Social Pact reformed collective bargaining and eliminated automatic indexation of wages to inflation (“scala mobile”).
  - The Treu reform liberalized entry wages for first-job seekers and reduced constraints on fixed-term and part-time use.
  - The Biagi reform introduced a wide variety of atypical employment contracts (e.g., lavoro intermittente, job sharing, lavoro a progetto).
  - The Fornero reform widened applicability of temporary employment while introducing renewal limits (fixed-term contracts could only be renewed once and have maximum 36 months duration).
  - The Poletti decree allowed use of fixed-term contracts with no obligation to justify their use and permitted up to five renewals, with a constraint that the ratio of fixed-term contracts to open-ended contracts not exceed 20%.
  - The Jobs Act changed protection against “just cause” dismissals and introduced the “contratto a tutele crescenti” to ease transitions from fixed-term to open-ended contracts.
- Pension reforms shifted from defined benefit to defined contribution and social security contributions in temporary jobs are smaller than in open-ended contracts.
- These reforms increased the fraction of workers employed through atypical (part-time or fixed-term) contracts, affecting primarily cohorts entering the labor market after the reforms.
- Unemployment and participation trends (25-54 age group, OECD statistics):
  - Male labor force participation declined from 95% in 1985 to 88% in 2016.
  - Female labor force participation expanded from 48% in 1985 to 67% in 2016.
  - Women’s unemployment rate fell from 11% in 1993 to 7% before the Great Recession, but by 2016 unemployment rates were 10% for men and 13% for women.
- Possible mechanisms discussed for Italy’s weak productivity relative to peers include the expansion of less productive jobs due to job flexibility, failure of firms to adopt ICT, high share of small firms, exposure to the China trade shock, and decline in government efficiency; the paper highlights an additional channel: changes in job composition reduce accumulation of work experience and firm-specific human capital among young workers, lowering productivity.

### Data, sample, and measurement highlights
- Data source: INPS (Istituto Nazionale di Previdenza Sociale) administrative data covering 1985-2016.
- Sample design and coverage:
  - 6.6% sample of the Italian population based on 24 randomly selected birth dates.
  - Public sector jobs not reported and account for 16% of total employment.
  - Self-employment not reported and accounts for 20% of total employment.
- Earnings and top-coding:
  - Main earnings measure is the sum of all regular and irregular income across all jobs within a year.
  - Data are top-coded for confidentiality at a job-level daily maximum; the topcode always exceeds the 99.5th percentile of the earnings distribution according to the authors’ calculation.
  - Example of top-coding mechanism: in 2012 INPS reports Y* = min{Y, 645*Days}; with maximum working days 312 the 2012 topcode threshold is approximately 200,000 euros. In other years the maximum is adjusted using the “Indice delle retribuzioni contrattuali orarie lorde ISTAT”.
- Hours and hourly earnings:
  - The dataset reports number of weeks worked, cause of separation, and whether job is part- or full-time.
  - Imputed hours assumptions: part-time = 25 hours/week; full-time = 40 hours/week (based on Italian SHIW averages).
  - Annual hours worked constructed as sum across jobs within a year, capped at 2,080 hours (38 hours×52 weeks).
  - Average hourly earnings computed as annual earnings divided by annual hours worked.
- Sample selection and size:
  - For comparability with other countries, sample restricted to workers aged 25 to 55 with positive earnings and at least 4 weeks worked in the year; no additional minimum earnings threshold.
  - The first percentile of earnings in the sample never falls below 800 euros.
  - Sample includes 2.3 million unique workers (1.4 million men and 0.9 million women) and a total of 22.4 million worker-year observations — approximately 700,000 observations per year.

*Source: wpiea2021142-print-pdf - 1. Introduction (https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021142-print-pdf.pdf)*

### 2.3    INPS labor force trends and comparison with official statistics

### 2.3    INPS labor force trends and comparison with official statistics

### INPS labor market trends (Figure 2)
- Share of female workers: increases from about 30% in 1985 to more than 40% in 2016.
- Share of part-time workers: increases from virtually 0% in 1985 to almost 30% in 2016.
- Share of workers working 52 weeks during the year: decreases from approximately 88% in 1985 to 77% in 2016 (right hand side axis).
- Share of workers with an open-end contract (first observed in 1998): decreases from more than 90% in 1998 to 82% in 2016 (right hand side axis).
- Interpretation: the labor market is transitioning away from being mostly men employed full-time all year toward greater female participation, more part-time work, less full-year work, and more fixed-term contracts.

### Comparability of INPS earnings to aggregate ISTAT series
- Two adjustments needed for comparability:
  - Account for the rise in part-time employment.
  - Account for the decline in the fraction of workers working 52 weeks a year.
- Empirical note: multiplying earnings of part-time workers by factors of 1.5, 1.8 and 2 yields minimal changes in comparisons.
- Data sources: ISTAT “indice delle retribuzioni lorde per ULA” from the Oros survey (integrating Social Security data and Large firms Survey data) vs INPS micro data for individuals aged 25 to 55.

### Core statistics: scope
- Analysis classified into three broad categories: inequality, volatility, and mobility.

### Inequality: cross-sectional distribution and time patterns (Figures 3–4)
- Two phases in the distribution of earnings (percentiles relative to 1985):
  - 1985-1992: all percentiles rise, with percentiles above the median rising more.
  - 1993-2016: all percentiles gradually decline, more pronounced for the bottom of the distribution.
- Gender-specific patterns:
  - Women: median earnings declined as much as the 10th percentile in the second phase.
  - Men: median earnings have been much more stable.
- Two direct measures of inequality:
  - P90-P10 gap: shows rising trend in earnings inequality.
  - 2.56 times the standard deviation: consistently higher than P90-P10, suggesting more pronounced tail features than the normal distribution.
- Decomposition (Panels (c) and (d) of Figure 4):
  - For women, most increase in inequality comes from the top of the distribution, driven by a decline in median earnings (P90-P50).
  - For men, there is an increase in top inequality between 1985-1992 and then a rise in bottom inequality between 1992-2016 (P50-P10).

### Top shares and top-coding adjustment (Section 3.2, Figure 5)
- INPS data are top-coded above the 99.5th percentile; simple sums bias top-income shares downward.
- Adjustment method:
  - Assume earnings above the top-code cutoff are Pareto distributed with the same Pareto-tail index as just below the cutoff.
  - Estimate Pareto-tail index using maximum likelihood on data between the 90th and 99th percentiles each year.
- Estimated Pareto-tail index:
  - Falls sharply between 1985-1992 (from 3.3 to 2.7).
  - Remains rather stable in the 2.6-2.8 range thereafter.
- Impact of adjustment on top shares:
  - Adjustment matters little for top 10% and top 5% shares.
  - Adjustment is crucial for documenting rise in top 1% and top 0.1% shares, which would be greatly underestimated without it.
- Data limitations noted:
  - INPS includes only compensation subject to social security contributions; certain benefits (e.g., stock options treated as ordinary income for tax purposes after 2008) are excluded and may understate top income inequality.

### Inequality trends across cohorts (Section 3.3, Figure 6)
- Initial inequality measured at age 25 for cohorts entering labor market between 1985 and 2016:
  - Initial bottom inequality (P50-P10): higher and fairly similar over time, with a spike for cohorts entering around the Great Recession.
  - Initial top inequality (P90-P50): on the rise since the late 1990’s.
- Life-cycle cohort patterns (cohorts entering in 1985, 1995, 2005, 2010):
  - Cohorts entering before 2000: inequality relatively stable over life cycle.
    - Women in these cohorts: inequality remained almost the same throughout the sample period.
    - Men in these cohorts: inequality rises around the Great Recession and remains elevated.
  - Cohorts entering after 2000: faced rising initial inequality.
  - Cohort differences in level of inequality decline as time passes; entering cohorts after 2000 face gradually changing labor market with relatively mild effects on older cohorts.

### Volatility of earnings growth (Section 3.4, Figures 7–9)
- Volatility measure: cross-sectional dispersion in one-year residual earnings growth g1it after controlling for age and year.
- Overall volatility (P90-P10):
  - Increases significantly over the sample period.
  - For women: increase concentrated in the first 15 years of the period.
  - For men: increase lasts until the Great Recession, with slight decline thereafter.
  - Women face 50% more volatile earnings than men; the difference remains stable over time.
- Decomposition into upside (P90-P50) and downside (P50-P10) volatility:
  - Both upside and downside volatility increase over time.
  - After detrending, upside dispersion is procyclical and downside dispersion is countercyclical.
- Distributional shape measures:
  - Kelley skewness (K = ((P90−P50)−(P50−P10))/(P90−P10)): nearly always lower in absolute terms for women than for men; procyclical.
  - Crow-Siddiqui excess kurtosis (CS = (P97.5−P2.5)/(P75−P25) − 2.91): values above zero indicate stronger central tendency than normal.
- Cohort-age volatility (Figure 9):
  - Volatility declines as people age.
  - Increase in volatility over time partly reflects compositional changes: more recent cohorts face higher volatility than older cohorts.

### Heterogeneity in volatility by permanent income and age (Section 3.5, Figure 10)
- Permanent income: defined as average residualized log-earnings over previous 3 years; percentiles calculated each sample year then pooled.
- Dispersion (P90-P10) by permanent income percentile and age groups (25-34, 35-44, 45-55):
  - Volatility declines with age for both genders.
  - Women: dispersion concentrated in early career, likely due to intermittent labor force participation.
  - Men: higher dispersion among low-permanent income workers.
- Kelley skewness by permanent income percentile:
  - Men: distribution roughly symmetric in middle; skewness positive at bottom and top; minimal age effects.
  - Women: older women similar to men; younger women show left skewness that steeply declines with permanent income percentile, suggesting reversion to the mean at the top—consistent with intermittent participation.
- Excess kurtosis by permanent income percentile:
  - Inverted-U shape for both genders.
  - Stronger age effects and much higher leptokurtosis for women.
- Time-evolution note (Appendix A.8 summary):
  - 1990–2000: volatility increases only for bottom third of permanent income distribution.
  - 2000–2010: all groups experience increase in volatility except those in the top third.
  - High-permanent income workers experienced little increase in volatility between 1990 and 2010 and start with the least volatility.
  - Interpretation: risk appears to have shifted over time toward people less equipped to self-insure (assuming permanent income measures self-insurance ability).

*Source: wpiea2021142-print-pdf - 2.3    INPS labor force trends and comparison with official statistics*

### 3.6    Mobility

### 3.6    Mobility

### Concept and measurement
- Mobility measures changes in the permanent component of income over extended periods, as opposed to volatility, which measures changes in total income over short periods.
- To evaluate mobility, workers are allocated to permanent income percentiles in yeart and the average permanent income percentile in yeart+10 is plotted conditional on the percentile in yeart.
- Benchmarks:
  - “Perfectly mobile” distribution: mobility curve is a flat horizontal line at the 50th percentile.
  - “Perfectly stagnant” distribution: mobility curve coincides with the 45 degrees line.

### Empirical patterns (Figure 11)
- Mobility is higher early in the career for both women and men (comparing 25-34 year olds to 35-44 year olds).
- Low income workers are expected to rise up to 20 percentage points in the permanent earnings distribution.
- At the top of the distribution there is mean reversion.
- Mobility patterns do not change significantly between the years 1995 and 2005 for both women and men.

### Rank-rank correlation analysis and levels of mobility
- Procedure: allocate workers to earnings percentiles and regress the percentile in a year on the percentile in the previous year (rank-rank correlation).
- One-year apart rank-rank correlation coefficient: 0.89 (s.e. 0.0001), indicating a rather low level of earnings mobility.
- Four-year apart rank-rank correlation coefficient: 0.79 (s.e. 0.0003).
- Comparison note: In a PSID sample for 1985-2012 selecting employed workers aged 25-55, the four-year apart rank-rank correlation coefficient is 0.74 (s.e. 0.0037), suggesting mobility in the Italian sample appears lower than in the US.

### Mobility trends over time (Figure 12)
- A regression with interactions by calendar year shows the rank-rank coefficient is very stable over time.
- The coefficient ranges between 0.87 and 0.92 throughout the sample period.

*Source: wpiea2021142-print-pdf - 3.6    Mobility*

### 4.3    Are fixed-term contracts stepping stones to permanent jobs?

### 4.3    Are fixed-term contracts stepping stones to permanent jobs?

### Key empirical facts and trends
- The share of fixed-term contracts rose from 8% of employment contracts in 1998 to 18% in 2016.  
- Of workers on a fixed-term contract in yeart:
  - 63% continue to work with the same type of contract at timet+1 (44% with the same employer, 19% with a different one).
  - 19% transition to an open-ended contract (10% with the same employer, 9% with a different one).
  - 17% leave employment altogether.
- By cohort (2016 snapshot):
  - More than one-third of workers who entered the labor market in that year held fixed-term contracts.
  - 20% among those who entered a decade earlier.
  - 15% among those who entered two decades earlier.
- Intermittent workers (employed in yeart but with no record in yeart−1 ort+1) are more common among cohorts entering after the labor market reforms.

### Evidence on the “stepping stone” hypothesis
- The theoretical motivation: fixed-term contracts may allow earlier labor market entry for workers and potentially raise chances of later securing open-ended contracts (Booth, Francesconi, and Frank (2002) framing).
- Counterargument: fixed-term contracts can create a two-tiered labor market with “insiders” (open-ended contracts) and “outsiders” (fixed-term contracts) with limited mobility to permanent status.
- Empirical tests (Table 2) use two samples:
  - Sample 1: “Early starters” = employed at age 25 with a fixed-term contract; comparison group = unemployed at age 25.
  - Sample 2: Adds the condition that both are employed at age 26.

- Main patterns from Table 2 (coefficients for the probability of working on an open-ended contract between age 27 and age 35):
  - Early starters:
    - Sample 1, spec (1): 0.0764
    - Sample 1, spec (2): -0.0029
    - Sample 2, spec (3): 0.0103
    - Sample 2, spec (4): -0.0157
  - Average log income (included in spec (2) and (4)): 0.1393 and 0.2093.
  - Male: 0.0258 and 0.0077 (specs with gender).
  - Age: 0.2504 and 0.0547.
  - Age squared/100: -0.0037 and -0.0010.
  - Fixed-contract at 26: 0.0317 (included in Sample 2 specifications).
  - Open-ended contract at 26: 0.2352 and 0.1918.
  - Observations:
    - Sample 1: 3,029,242 (both specs).
    - Sample 2: 1,021,157 (both specs).

- Interpretation:
  - Unconditional estimates suggest early starters may have a slight advantage in securing permanent jobs.
  - Once conditioning on average log-income in the first decade, gender, age, and employment status at age 26, the early-starter advantage disappears or reverses slightly.
  - The authors conclude the “stepping stone” mechanism, if it exists, is rather weak and composition/selection issues remain (evidence is not causal).

### Volatility and separations (Table 1 highlights)
- Table reports coefficients from regressions of annual earnings volatility on causes of job separation and controls; two specifications: (1) without individual fixed effects, (2) with individual fixed effects.
- Key coefficients (Cause of separation):
  - Cause Separation: Layoff
    - (1): 0.267 (standard error 0.002)
    - (2): 0.135 (standard error 0.002)
  - Cause Separation: Quit
    - (1): 0.169 (0.002)
    - (2): 0.073 (0.002)
  - Cause Separation: End Contract
    - (1): 0.202 (0.002)
    - (2): 0.133 (0.002)
  - Cause Separation: Other
    - (1): 0.171 (0.004)
    - (2): 0.093 (0.004)
- Other selected covariates (both columns reported where available):
  - Age: -0.015 (0.000) and -0.003 (0.001)
  - Age squared /100: 0.013 (0.000) and 0.009 (0.001)
  - Female: 0.052 (0.001) [reported in column (1)]
  - Job switcher: 0.119 (0.001) and 0.060 (0.001)
  - Sector switcher: 0.095 (0.002) and 0.057 (0.001)
  - Full time to part time: 0.234 (0.003) and 0.168 (0.003)
  - Part time to full time: 0.153 (0.003) and 0.054 (0.003)
- Model fit and samples:
  - Individual fixed effects: NO in column (1); YES in column (2).
  - Adj. R2: 0.084 (column (1)); 0.449 (column (2)).
  - Observations: 5,789,700 (col 1); 5,836,552 (col 2).
  - Mean volatility stayers: 0.097 (both columns).
  - P-value Layoff=End-contract: 0.000 (col 1); 0.602 (col 2).
- Note on interpretation: end of fixed-term contract is an increasingly common cause of separation; separations (including end-contract) are associated with higher earnings volatility.

### Human capital, experience, and productivity growth
- Despite increases in average years of schooling, labor productivity in Italy has been stagnant for more than two decades.
- Human capital decomposition (Mincerian framework):
  - Human capital modeled as function of schooling S, cumulative weeks of experience X, weeks of tenure T, and cohort-quality Q.
  - Log earnings equation estimated to recover parameters β1, β2, γ1, γ2 and cohort fixed effects.
- Empirical findings:
  - The “quality/schooling” component (φ(Qi(b)t)+α(Si(b))) shows an overall increase in schooling but a steep decline in the “quality” term for cohorts entering between the mid-1990s and late 2000s, such that the overall quality/schooling component actually declines for these cohorts.
  - The experience component (β(X)+γ(T)) is much larger in level than the schooling component and shows that recent cohorts have systematically lower predicted human capital profiles over the life cycle.
  - Extrapolating to age 55:
    - The cohort entering the labor market in 2005 would accumulate 13% less human capital than the cohort entering in 1995.
    - The cohort entering in 2005 would accumulate 9% less human capital than the cohort entering in 2000.
- Mechanisms: labor market reforms shifted younger workers toward atypical contracts with less attachment to employers, reducing general and firm-specific experience accumulation mechanically (shorter job duration) and behaviorally (lower incentives for firms and workers to invest in one another).

### Policy-relevant implications and summary conclusions
- The expansion of fixed-term and part-time contracts since the late 1990s is associated with:
  - Rising inequality and rising earnings volatility for men and women over 1985–2016.
  - A substantial change in the dispersion of annual hours worked across jobs, contributing more to inequality than dispersion in average hourly earnings.
  - Increased earnings volatility from one year to the next, implying higher uninsurable risk for individual workers.
  - Reduced on-the-job human capital accumulation (general and firm-specific), plausibly slowing labor productivity growth.
- On the “stepping stone” question:
  - Evidence suggests that early entry via fixed-term contracts provides, at best, a weak stepping-stone to permanent employment once observable early-career outcomes are controlled for; selection and unobserved heterogeneity complicate causal interpretation.
- Overall assessment:
  - Labor market reforms facilitating fixed-term contracts likely increased labor market flexibility and participation for some groups but also contributed to persistent two-tier dynamics, higher volatility, and lower human capital accumulation among younger cohorts—factors that plausibly contributed to Italy’s long-run productivity stagnation.

*Source: IMF Working Paper section 4.3, “Are fixed-term contracts stepping stones to permanent jobs?”*

### References

### References

### Key citations
- Atkinson, A. B., T. Piketty, and E. Saez(2011): “Top Incomes in the Long Run of History,” Journal of Economic Literature, 49, 3–71.
- Beran, J. and D. Schell(2012): “On robust tail index estimation,” Computational Statistics & Data Analysis, 56, 3430–3443.
- Boeri, T. and P. Garibaldi(2007): “Two Tier Reforms of Employment Protection: a Honeymoon Effect?” The Economic Journal, 117, F357–F385.
- Booth, A. L., M. Francesconi, and J. Frank(2002): “Temporary Jobs: Stepping Stones or Dead Ends?” The Economic Journal, 112, F189–F213.
- Bugamelli, M., S. Fabiani, S. Federico, A. Felettigh, C. Giordano, and A. Linarello (2017): “Back on track? A macro-micro narrative of Italian exports,” Questioni di Economia e Finanza (Occasional Papers) 399, Bank of Italy, Economic Research and International Relations Area.
- DiNardo, J., N. M. Fortin, and T. Lemieux(1996): “Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach,” Econometrica, 64, 1001–1044.
- Finkelstein, M., H. G. Tucker, and J. Alan Veeh(2006): “Pareto tail index estimation revisited,” North American actuarial journal, 10, 1–10.
- Giordano, R., S. Lanau, P. Tommasino, and P. Topalova(2015): “Does Public Sector Inefficiency Constrain Firm Productivity; Evidence from Italian Provinces,” IMF Working Papers 15/168, International Monetary Fund.
- Hoffmann, E. B. and D. Malacrino(2019): “Employment time and the cyclicality of earnings growth,” Journal of Public Economics, 169, 160–171.
- Miniaci, R. and G. Weber(1999): “The Italian Recession of 1993: Aggregate Implications of Microeconomic Evidence,” Review of Economics and Statistics, 81, 237–249.
- Pellegrino, B. and L. Zingales(2017): “Diagnosing the Italian Disease,” Working Paper 23964, National Bureau of Economic Research.
- Rosolia, A. and R. Torrini(2007): “The generation gap: relative earnings of young and old workers in Italy,” Temi di discussione (Economic working papers) 639, Bank of Italy.
- Tealdi, C.(2011): “Typical and atypical employment contracts: the case of Italy,” Questioni di Economia e Finanza (Occasional Papers) 39456, Munich Personal RePEc Archive.

### Appendix A.1 — Other forms of employment: public sector and self employment
- Data source: Survey of Household Income and Wealth (SHIW), Bank of Italy; representative cross-section with a rotating panel since 1989; analysis uses workers aged 25–55 pooled over years since 2000.
- SHIW variables used:
  - identifies self employment and self employment income, with subcategories (sole proprietor, free lance, “member of the professions”);
  - industry variable indicating private vs public employee (broad industry categories);
  - self-reported worker status; contract type; full time vs part time.
- Method: estimate 2-year transition probabilities among four labor market states (permanent contract, temporary/fixed-term contract, self employment, non-employment) for men and women; infer continuous-time Markov chain via unique generator matrix matching observed transition probabilities; use generator to compute conditional state distributions up to 20 years.
- Empirical findings (preserve exact reported percentages):
  - For both women and men, probability that a fixed-term contract holder has an open-ended contract within 5 years is 50%, and to have a fixed-term contract 15%.
  - Women with fixed-term contract have a 30% probability of being non employed within 5 years, compared to 22% for men.
  - Self employment persistence after 5 years: 50% for women, 75% for men.
  - Non-employment persistence after 5 years: 75% for women, 50% for men.
- Interpretations:
  - Workers in open-ended jobs (men and women) are unlikely to transition to fixed-term contracts.
  - Workers in fixed-term employment often end up in an open-ended contract.
  - Temporary contracts serve their “stepping stone” purpose better for men than for women.
  - Transitions into self employment are rare; self employed men are likely to transition directly into a permanent contract, while self employed women are equally likely to be employed in a temporary contract, suggesting different roles of self employment by gender.
- Sectoral transitions (private vs public vs self employment vs non-employment):
  - Probability of a private sector worker being in the public sector after 20 years: 21% for women and 16% for men.
  - Transitions from public to private: after 5 years, 22% of women and 24% of men in public sector jobs transition to private sector jobs; after 20 years this grows to 30% for women and 44% for men.
  - Possible driver: gradual reduction in size of the Italian public sector over the sample period.

### Appendix A.2 — Top income share estimation in right-censored data
- Objective: describe procedure to estimate top income shares in top-coded (right-censored) data and implement it on INPS data.
- Model assumption:
  - Let F(y) be the true CDF of earnings. If there exists y0 such that for all y > y0 earnings are Pareto:
    - 1 − F(y) = (1 − F(y0)) (y / y0)^−α, with α > 0 the Pareto tail index.
  - y0 assumed smaller than the top-coding threshold, so mean income below it E[y | y < y0] and F(y0) can be estimated from uncensored observations.
  - Quantile function for probabilities P > F(y0): Q(P) = y0 ((1 − P) / (1 − F(y0)))^(−1/α).
  - If α > 1, mean income above the P-quantile: E[y | y > Q(P)] = (α / (α − 1)) Q(P).
- Closed-form expression for top income share S(1 − P) (equation A.1):
  - S(1 − P) = (1 − P) E[y | y > Q(P)] / (F(y0) E[y | y ≤ y0] + (1 − F(y0)) E[y | y > y0])
  - Which simplifies to:
    - S(1 − P) = α (1 − P) y0 / ((α − 1) F(y0) E[y | y ≤ y0] + α (1 − F(y0)) y0 ((1 − P) / (1 − F(y0)))^(−1/α))
- Empirical check of Pareto-tail assumption:
  - Zipf (log-log) plots of log survival function vs log(y) between the 90th and 99th percentiles for years 1985, 1995, 2005, 2015 appear approximately linear, supporting Pareto-tail assumption.
- Methods adapted to estimate Pareto-tail index from right-censored data:
  1. Maximum Likelihood (ML)
  2. Quantile Slope (QS)
  3. Kernel Density Slope (KDS)
  4. Probability Integral Transform Statistic (PITS)
- Data censoring setup:
  - Assume true earnings ˜y Pareto for ˜y ≥ y_min and observed data y = ˜y if ˜y < y_max, and y = y_max if ˜y ≥ y_max. Consider N observations ≥ y_min, sorted, with last N_censored observations equal to y_max.
- Brief descriptions of estimators:
  - ML estimator: derived from likelihood of censored observations; closed-form ML estimator:
    - ˆα_ML = (N − N_censored) / (sum_{i=1}^N (log y_i − log y_max))
    - Alternate algebraic expression shown in text.
  - Quantile Slope (QS): regression-based estimator using sample quantiles Q_j and slope of log(1 − P_j) on log Q_j; closed form expression given for ˆα_QS.
  - Kernel Density Slope (KDS): works on x = log y, uses kernel density estimate ˆf(x) at J points and regress log ˆf(x_j) on x_j; KDS estimator ˆα_KDS defined as negative slope.
  - Probability Integral Transform Statistic (PITS): method-of-moments estimator using transformation G(y, a) with tuning parameter t > 0; choose ˆα_PITS such that (1/N) sum_i G(y_i, ˆα_PITS) = 1. Transformation defined and property E[(y/y0)^(−αt)] used.
- Comparison of estimators (qualitative):
  - QS and KDS: intuitive, transparent, effectively ignore censored observations; sampling points (P_j, x_j) choice can affect estimates; KDS may be biased with sparse data and requires kernel/bandwidth choices.
  - ML: most efficient and simple, but not robust when tail is only approximately Pareto (per Finkelstein et al. (2006) and Beran and Schell (2012)).
  - PITS: more robust than ML when Pareto mixed with other “noise” distributions, at mild cost in efficiency.
- Implementation choices on INPS 2015 earnings data:
  - Lower threshold set at the 0.9 sample quantile (focus on top 10%).
  - Upper threshold for estimators set at the 0.99 sample quantile (INPS top-codes daily earnings and reports annual earnings; annual records smaller than maximal top-coded threshold 365 times the daily threshold may be partially top coded).
  - QS method: choose 100 equally distanced cumulative probability points between 0 and 0.85.
  - KDS method: Epanechnikov kernel with bandwidth 0.05 and 100 equally spaced sample points in log earnings between the 90th percentile and the 98.5th percentile.
  - PITS tuning parameter: t = 0.5.
- Bootstrap experiment (100 bootstrapped samples):
  - Bootstrapped standard deviations reported:
    - ML estimator sd: 0.0089
    - KDS estimator sd: 0.0189
  - Point estimates differ across methods but economic implications small:
    - Top 1% income share based on PITS estimator: 6.63% (PITS yields lowest index)
    - Top 1% income share based on QS estimator: 6.43% (QS yields highest index)
- Pareto-tail index estimates across years:
  - Pareto-tail index estimates for all years in the sample are reported in Figure A.5 with shaded 95% confidence intervals based on 100 bootstrap replications.

### Appendix A.3 — Additional Figures (high-level notes)
- Figure A.6: Closing the gap in the evolution of earnings from 2000 to 2016 (index, 2015=100) — compares ISTAT index with various INPS adjustments (INPS raw; INPS weeks adjusted and PT × 1.5; INPS weeks adjusted and PT × 1.8; INPS weeks adjusted and PT × 2).
- Figure A.7: Inequality for public employees (SHIW) — presents 90-10th percentile and 90-50th/50-10th percentile differentials across years.
- Figure A.8: Volatility by permanent income for selected years — P90-P10 differential of g_it plotted across quantiles of permanent income P_it−1 for 1990, 2000, 2010.
- Figure A.9: Hours per week by employment status — average hours per week for full-time and part-time jobs based on SHIW for sample restricted to 25–55 year old workers, excluding self-employed.
- Figure A.10: Cause of separations — percent breakdown of separations by reason (Firing, Quits, End contract, Other reasons) for 2005, 2010, 2015.

*Content derived from wpiea2021142-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021142-print-pdf.pdf_