## wpiea2024002-print-pdf

## Source details

**Canonical URL:** [wpiea2024002-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024002-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024002-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024002-print-pdf.pdf.json)

---

### 5.1 Specification — Key findings and concepts
- By 2022, total hours worked exceeded their pre-COVID-19 level, driven by a strong bounce back in the employment rate (extensive margin), while average hours worked (intensive margin) remained below pre-pandemic levels.
- The recent fall in working hours is predominantly structural rather than cyclical, with average hours back at their long-term pre-pandemic trend.
- The decline in average hours worked since 2003 has been widespread across demographic groups, industries, and occupations.
- Within-group declines account for up to 80 percent of the aggregate decline in hours worked; compositional shifts (between-group) contribute less.
- The decline in average hours worked has been most pronounced among the young, men, and men with young children; these groups account for much of the overall decrease in average actual hours worked since 2003.
- For the young, an increased incidence of part-time work while enrolled in education helps explain the decline.
- For men (including those with young children), the decline cuts across full-time and part-time workers and reflects a broad downward shift in the distribution of hours worked.
- Measures of hours:
  - Actual hours worked — “number of hours actually worked during the reference week in the main job.”
  - Usual hours worked — “number of hours per week usually worked in the main job” (modal value excluding absence weeks).
  - Desired hours worked — “number of [weekly] hours that the person would like to work in total.”
- Gaps and interpretation:
  - Focusing on gaps between desired and actual or usual hours suggests the bulk of the trend decline primarily reflects worker preferences.
  - Youth and men: actual hours fell since 2003, but there has not been any increase in the gap between desired and actual hours; desired hours fell alongside actual hours.
  - Women with young children: report a growing gap between desired and actual hours, but not between desired and usual hours—slightly increased usual hours were offset by more (parental) leave depressing actual hours.
  - Fully eliminating the persistent positive gap between desired and usual hours would, in principle, increase total labor input by around 1.3 percent (upper bound thought experiment).
- Drivers and interpretation:
  - Increased income and wealth is likely the main force behind declines in desired and actual hours worked.
  - The trend growth slowdown in much of Europe may explain why the long-term decline in hours flattened out in recent decades.
  - Some evidence suggests post-COVID factors such as sick leave or long COVID could depress hours in the short run.

### 5.1 Specification — Policy implications and options
- Policy should align with worker preferences if aiming to boost working hours.
- Tax and benefit systems—including unemployment insurance, health and pension schemes—should avoid penalizing full-time relative to part-time work to prevent disincentives for full-time job take-up.
- Possible measures to narrow the hours gap:
  - Active labor market policies (e.g., retraining programs) to help part-time workers qualify for higher-skill full-time jobs.
  - Targeted policies for mothers with young children, such as expanded childcare and reducing the marginal taxation of second earners as needed.
  - Mainstreaming flexible work arrangements, including teleworking.
- Regulation historically played a role in reducing hours (examples: 8-hour day, 48-hour week, France’s 35-hour work week), but explicit changes to statutory working hours have not been a major driver in Europe over the last two decades.

### 5.2 Results — Which groups and magnitudes
- Regression and descriptive findings:
  - Men, particularly those with young children, and young workers have seen a sharper decline in hours worked.
  - Hours worked have been declining annually by about 0.07 hours per week more for men without young children than for the reference group.
  - Men with young children have experienced an additional 0.03-0.05 annual decline in their hours per week.
  - Young workers have seen an additional 0.07-0.11 hours decline; the estimated effect for the young implies a decline in average actual hours of up to 2 hours compared to the reference group over 2003-2019.
- Macro controls:
  - Coefficients on both the output gap and GDP per capita are highly significant, indicating hours are procyclical (negative output gap coefficient) and that the income effect dominates the substitution effect in aggregate (negative GDP per capita coefficient).
- Full-time vs part-time:
  - Results hold for both full-time and part-time workers, but tend to be weaker among full-time workers, particularly for youth.
  - Exception: female workers with young children—for whom only full-time workers have reduced their hours.
- Zero-hour weeks and parental leave:
  - Restricting to workers with non-zero hours or using usual hours as the dependent variable leaves main results unchanged for young and men with children.
  - For women with young children, the trend decline in actual hours appears to reflect a rise in zero-hour weeks (in 2019, 8.8 percent of all workers reported zero actual hours in the reference week, while 25.2 percent of women with children under 5 years did).
  - Appendix evidence shows zero-hour weeks and their increase over time are mostly explained by parental leave.
- Young workers and schooling:
  - Share answering “in school” as reason for zero hours increased from 7.3 percent in 2006 to 12.6 percent in 2019.
  - Young workers in school: 2003 actual hours = 26.68; 2019 actual hours = 23.18; ∆ = -3.49.
- Hours gap evidence and interpretation:
  - The gap between desired and actual hours has a positive mean of about 4 hours; the gap between desired and usual hours is much smaller.
  - The gap between desired and actual hours has grown over the sample period, but the gap between desired and usual hours has barely risen.
  - For main declining groups (men, youth, men with young children), desired hours fell alongside usual hours; trend coefficients for desired-usual gaps are at most very small and insignificant.
  - For mothers of young children, rising zero-hour weeks largely explain the widening desired-actual gap despite no increase in desired-usual gap.
- Decomposition result:
  - The within-component accounts for most—over four fifths—of the decline in average hours worked for 2003-2019.

### 5.2 Results — Key descriptive and regression statistics (preserved exactly)
- Aggregate average actual hours worked: 2003 = 35.37; 2019 = 32.88; ∆ = -2.49**.
- Men: 2003 = 38.78; 2019 = 36.03; ∆ = -2.75**; employment shares E2003 = .56; E2019 = .54; contribution Contr. = -2.35.
- Women: 2003 = 31.01; 2019 = 29.21; ∆ = -1.81; employment shares E2003 = .44; E2019 = .46; Contr. = -.14.
- Young: 15-29 yrs: 2003 = 34.41; 2019 = 31.67; ∆ = -2.74**; E2003 = .22; E2019 = .18; Contr. = -2.03.
- Prime: 30-54 yrs: 2003 = 36; 2019 = 33.8; ∆ = -2.2**; E2003 = .66; E2019 = .62; Contr. = -2.75.
- Older: 55-64 yrs: 2003 = 34.46; 2019 = 32.18; ∆ = -2.28; E2003 = .10; E2019 = .11; Contr. = .72.
- Elderly: 65+ yrs: 2003 = 29.03; 2019 = 24.23; ∆ = -4.8; E2003 = .02; E2019 = .03; Contr. = .18.
- Married: 2003 = 35.61; 2019 = 33.09; ∆ = -2.52**; E2003 = .61; E2019 = .54; Contr. = -3.99.
- Child u5: 2003 = 34.21; 2019 = 31.57; ∆ = -2.64**; E2003 = .17; E2019 = .16; Contr. = -.78.
- Men w. Child u5: 2003 = 39.77; 2019 = 37.02; ∆ = -2.75**; E2003 = .10; E2019 = .09; Contr. = -.68.
- Women w. Child u5: 2003 = 26.07; 2019 = 24.55; ∆ = -1.52; E2003 = .07; E2019 = .07; Contr. = -.09.
- Table A.1 (Usual hours, selected): Total: ̄h_usual_2003 = 38.19; ̄h_usual_2019 = 37.11; ∆̄h′_19−′_03 = -1.08.
- Table A.5 (Young in school): 2003 h_actual = 26.68; 2019 h_actual = 23.18; ∆ = -3.49.
- Table A.6 (Hours Gap Desired − Actual): Trend (Column (1)): 0.0808*** (0.00182); Women w. Child u5: 2.622*** (0.448); Trend x Women w. Child u5: 0.115** (0.0407).
- Table A.7 (Hours Gap Desired − Usual): Trend (Column (1)): 0.0181 (0.0259); Men: -0.791*** (0.137); Young 15-29 yrs: 0.936** (0.439).
- Table A.8 (Hours Gap Desired − Usual among Full-Time): Trend (Column (1)): 0.0153*** (0.0000569); Young: 0.405*** (0.132).
- Table A.9 (Reasons not working despite having a job): Parental leave (Women w YC: 2006 = 0.42; 2019 = 0.56).
- In 2019, 8.8 percent of all workers reported zero actual hours in the reference week; 25.2 percent of women with children under 5 years reported zero actual hours.

### 5.4 Further Insights — Cross-country, education, industry, occupation
- Education, industry, occupation (2011-2019 subsample):
  - Most key findings still hold when incorporating industry and occupation variables.
  - Men and men with young children have seen a sharper decline in average hours worked than other groups.
- Cross-country dimension:
  - Declines in hours worked share important common features across countries; the pattern that lower average hours have been driven by the young, men, and men with young children holds in nearly every country (magnitudes differ).
  - Cross-country evidence consistent with a dominant income effect:
    - Countries with higher GDP per capita tend to have fewer average working hours.
    - There is convergence in average hours worked across countries over time, with larger declines in countries that had longer average hours and, for many, lower GDP-per-capita levels at the beginning of the sample.

### 6 Concluding remarks — Outlook, aggregate impacts, and policy guidance
- Aggregate outcomes and outlook:
  - Total hours worked recovered after the COVID-19 pandemic, primarily driven by increased employment rates.
  - Average hours extended their pre-COVID trend decline, with little evidence of remaining slack along the intensive margin by mid-2023.
  - The decline in average hours is likely to continue in European countries; pace depends on trend productivity and wage growth and will vary with economic convergence paths.
- Aggregate counterfactuals and back-of-the-envelope calculations (preserved exactly):
  - Fully closing the gap between desired and actual hours would increase total hours worked by around 0.5 hours, or 1.3 percent.
  - Enabling all involuntary part-time workers to switch to full-time jobs would bring about an average increase of 0.42 hours or 1.1 percent in working hours.
  - Aggregate usual hours worked in 2019: 37.11 hours.
  - Aggregate desired working hours in 2019: 37.58 hours.
  - The aggregate gap in 2019: 0.47 hours (around 1.3 percent of the aggregate usual hours worked in 2019).
  - Usual hours worked gap for involuntary part-time workers in 2019: 9.6 hours; its share in total employment: around 4.4 percent in the sample. This translates into 1.1 percent of the total usual hours worked (37.11 hours) in 2019 (= 9.6×0.044/37.11).
  - Bringing the trend decline in hours gap back to 2003 level for women with young children would increase aggregate average hours worked by 0.13 hours or 0.38 percent (calculation shown: (0.112 hrs)×16 years×its employment share (0.07) / aggregate average hours in 2019 (32.88 hours)×100%).
- Policy measures aligned with preferences and reducing involuntary part-time:
  - Neutral tax and benefit systems with respect to workers’ hours choices.
  - Targeted measures to help part-time working mothers who would like to switch to full-time jobs (neutral tax treatment of second earners; higher child care subsidies or services; enhanced paid (pa)maternity leave; more flexible work arrangements including teleworking).
  - Active labor market policies to enhance matching between involuntary part-time workers and available full-time jobs (including retraining programs).
  - Design social protection to avoid incentives that unduly favor part-time work (e.g., contributory pension crediting and non-contributory benefit thresholds); consider slow phase-outs and reform of in-work tax credits while weighing fiscal costs.
  - Policy design must consider interactions between intensive and extensive margins and joint household labor supply decisions.

*Source: wpiea2024002-print-pdf — https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024002-print-pdf.pdf*

### 5.1  Specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .   17

### 5.1  Specification

### Key findings
- By 2022, total hours worked exceeded their pre-COVID-19 level, driven by a strong bounce back in the employment rate (extensive margin), while average hours worked (intensive margin) remained below pre-pandemic levels.
- The recent fall in working hours is predominantly structural rather than cyclical, with average hours back at their long-term pre-pandemic trend.
- The decline in average hours worked since 2003 has been widespread across demographic groups, industries, and occupations.
- Within-group declines account for up to 80 percent of the aggregate decline in hours worked; compositional shifts (between-group) contribute less.
- The decline in average hours worked has been most pronounced among the young, men, and men with young children. These groups account for much of the overall decrease in average actual hours worked since 2003.
- For the young, an increased incidence of part-time work while enrolled in education helps explain the decline.
- For men (including those with young children), the decline cuts across full-time and part-time workers and reflects a broad downward shift in the distribution of hours worked.
- Focusing on gaps between desired and actual or usual hours suggests the bulk of the trend decline in hours primarily reflects worker preferences:
  - Youth and men (including with young children): actual hours fell since 2003, but there has not been any increase in the gap between desired and actual hours; desired hours fell alongside actual hours.
  - Women with young children: report a growing gap between desired and actual hours, but not between desired and usual hours—slightly increased usual hours were offset by more (parental) leave depressing actual hours.
- Fully eliminating the persistent positive gap between desired and usual hours would, in principle, increase total labor input by around 1.3 percent (noting this is an upper bound thought experiment).

### Trends and cross-country evidence
- Average working hours across developed economies have been on a long-term declining trend since the 19th century (example: roughly halving in Germany between 1870 and 2000).
- More broadly, average working hours across OECD countries decreased by roughly 0.5 percent per year between the 1870s and the early 2000s, with the postwar United States an important exception.
- There is a strong negative cross-country correlation between GDP per capita and average hours:
  - Hours have tended to fall more in European countries where average hours were initially longest and in those that experienced the highest growth rates in GDP per capita.
  - Both actual and desired hours dropped in these countries, pointing to preferences driven by income growth.
- Cross-country convergence in average hours worked is evident; countries farther from the technological frontier and projected to grow faster have larger scope for further drops in average hours.

### Concepts of hours analyzed
- Three concepts are distinguished:
  1. Actual hours worked — determine labor input and are affected by annual leave, holidays, and exceptional changes in weekly hours.
  2. Usual hours worked — reflect a typical work week and are not affected by temporary absences like parental leave.
  3. Desired hours worked — capture potentially available labor supply and inform scope for raising hours.
- Changes in parental leave and similar practices can create a wedge between actual and usual hours.

### Drivers and interpretation
- Increased income and wealth is likely the main force behind declines in desired and actual hours worked.
- The trend growth slowdown in much of Europe may explain why the long-term decline in hours flattened out in recent decades.
- Some evidence suggests post-COVID factors such as sick leave or long COVID could depress hours in the short run (references to Arce et al. (2023) and Hernández de Cos (2023) noted in the text).

### Policy implications and options
- Policy should align with worker preferences if aiming to boost working hours.
- Tax and benefit systems—including unemployment insurance, health and pension schemes—should avoid penalizing full-time relative to part-time work to prevent disincentives for full-time job take-up.
- Possible measures to narrow the hours gap include:
  - Active labor market policies (e.g., retraining programs) to help part-time workers qualify for higher-skill full-time jobs.
  - Targeted policies for mothers with young children, such as expanded childcare and reducing the marginal taxation of second earners as needed.
  - Mainstreaming flexible work arrangements, including teleworking.
- Regulation historically played a role in reducing hours (examples: 8-hour day, 48-hour week, France’s 35-hour work week), but explicit changes to statutory working hours have not been a major driver in Europe over the last two decades.

### Outlook and scenarios
- Given Europe’s modest projected GDP per capita growth rates over the medium term, the more likely scenario is continued declines in average hours along a possibly flatter downward trend.
- Average hours returned to their pre-pandemic trend level by mid-2023, suggesting little to no remaining short-term slack along the intensive margin.
- Heterogeneity across countries is expected: countries farther from the technological frontier with faster projected growth may see larger further declines in average hours.

*Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024002-print-pdf.pdf*

### 2014. Bick, Fuchs-Schündeln and Lagakos (2018) find that average hours worked are sub-

### wpiea2024002-print-pdf - 2014. Bick, Fuchs-Schündeln and Lagakos (2018) find that average hours worked are sub-

### Literature and motivation
- Prior findings:
  - Bick, Fuchs-Schündeln and Lagakos (2018): average hours worked are substantially higher in lower-income countries; working hours fall with income except in the richest countries.
  - Bick et al. (2022): between 1999 and 2019 average hours worked per worker declined in all 19 countries (US and 18 European countries) in their sample, while employment rates increased in most countries.
  - Theoretical interpretation: decline (increase) in hours (employment) explained by a decrease in the fixed costs of heterogeneous preferences among workers.
- Related strands highlighted:
  - Role of parental leave and zero-hour weeks, especially for women (references to Kleven, Landais and Søgaard (2019); Angelov, Johansson and Lindahl (2016)).
  - Desired working hours and labor under-utilization: Böheim and Taylor (2004); Faberman et al. (2020) construct an “aggregate hours gap” from desired and actual hours.
- Paper focus:
  - Investigate long-term trends in average, usual, and desired hours worked in European countries, quantify between- vs within-group contributions to the decline in average hours, and analyze micro-level drivers using EU-LFS microdata.

### Data
- Main data sources:
  - Aggregate quarterly Eurostat data from 2003Q1 to 2023Q1 covering the EU27 countries.
  - OECD data for longer-term trends beyond the early 2000s for selected European countries.
  - EU Labour Force Survey (EU-LFS) microdata for 25 European countries over 2003-2019 in annual cross-section.
- EU-LFS details and sample:
  - Covers all EU countries plus Iceland, Norway, Switzerland, United Kingdom; this study covers 25 European countries over 2003-2019.
  - Sample selection based on availability of gender, age, marital status and children.
  - Sample covers employed workers aged 15 and up focusing on the intensive margin, yielding 24 million individual observations over the baseline sample period.
- Hours measures used:
  - Actual hours: “number of hours actually worked during the reference week in the main job.”
  - Usual hours: “number of hours per week usually worked in the main job” (modal value excluding absence weeks).
  - Desired hours: “number of [weekly] hours that the person would like to work in total” (reflects preferences and external factors).
- Limitations:
  - Eurostat: up-to-date but limited granularity.
  - EU-LFS: rich microdata but available with significant lag; used until 2019 (pre-pandemic).

### Descriptive statistics and trends
- Long-term trends:
  - Average hours worked have trended down consistently over past decades with flattening in several countries and substantial cross-country heterogeneity between 1970 and 2022.
  - OECD (1998) noted a flattening in pace of reduction as of the 1980s; statutory regular weekly hours stabilized with a dominant 40-hour norm since late 1990s/early 2000s (exceptions: Belgium 38 hours, France 35 hours). Spanish government announced plans to reduce working week from 40 to 37.5 hours in the next two years (statement present in source).
- Contribution to aggregate labor input (2003-2019 and 2019-2023):
  - Drop in average hours corresponded to a reduction of about 0.2 percent per year in total hours worked.
  - Aggregate labor input still grew by around 0.4 percent per year due to falling unemployment and rising labor participation.
  - Growth rate would have been 50 percent higher had average hours remained stable.
- Group-level patterns:
  - Aggregate average actual hours worked fell from 35.37 hours per week in 2003 to 32.88 hours per week by 2019, a change of -2.49 hours.
  - Declines observed for all demographic groups; decline most pronounced for men—particularly with young children—and for young workers (15-29 years).
  - Men work more than women on average, but the gender gap has shrunk over time; the gender gap in employment rate also shrunk.
  - Young workers saw biggest decline in hours alongside a decline in their employment rate (and rise in schooling).
  - Older workers (55-64) and elderly workers (65+) increased employment shares but also experienced drops in average hours.
  - Men with young children under 5 years saw a sharp decline in average hours; the corresponding fall for women with young children was much milder.
- Usual vs actual hours:
  - Much of the decline in actual hours reflects an increase in non-worked periods; decline in usual hours is much smaller.
  - Drop in usual hours is about one-hour-per-week smaller on average than fall in actual hours, and not statistically significant for the overall population.
  - Significant negative changes in usual hours observed for men and men with young children; women with young children saw an increase in usual hours despite a fall in actual hours.

- Table 1 key numbers (selected, preserved exactly as in source):
  - Aggregate average actual hours worked: 2003 = 35.37; 2019 = 32.88; ∆ = -2.49**.
  - Men: 2003 = 38.78; 2019 = 36.03; ∆ = -2.75**; employment shares E2003 = .56; E2019 = .54; contribution Contr. = -2.35.
  - Women: 2003 = 31.01; 2019 = 29.21; ∆ = -1.81; employment shares E2003 = .44; E2019 = .46; Contr. = -.14.
  - Young: 15-29 yrs: 2003 = 34.41; 2019 = 31.67; ∆ = -2.74**; E2003 = .22; E2019 = .18; Contr. = -2.03.
  - Prime: 30-54 yrs: 2003 = 36; 2019 = 33.8; ∆ = -2.2**; E2003 = .66; E2019 = .62; Contr. = -2.75.
  - Older: 55-64 yrs: 2003 = 34.46; 2019 = 32.18; ∆ = -2.28; E2003 = .10; E2019 = .11; Contr. = .72.
  - Elderly: 65+ yrs: 2003 = 29.03; 2019 = 24.23; ∆ = -4.8; E2003 = .02; E2019 = .03; Contr. = .18.
  - Married: 2003 = 35.61; 2019 = 33.09; ∆ = -2.52**; E2003 = .61; E2019 = .54; Contr. = -3.99.
  - Child u5: 2003 = 34.21; 2019 = 31.57; ∆ = -2.64**; E2003 = .17; E2019 = .16; Contr. = -.78.
  - Men w. Child u5: 2003 = 39.77; 2019 = 37.02; ∆ = -2.75**; E2003 = .10; E2019 = .09; Contr. = -.68.
  - Women w. Child u5: 2003 = 26.07; 2019 = 24.55; ∆ = -1.52; E2003 = .07; E2019 = .07; Contr. = -.09.
  - Note on Table 1: standard errors in parentheses; significance denoted *p <0.10, **p <0.05, ***p <0.01.

### Between- vs. within-group decomposition
- Decomposition identity used:
  - h_t − h_{t−1} = Σ_i (ω_it − ω_i,t−1) h_i,t (between) + Σ_i ω_i,t−1 (h_i,t − h_i,t−1) (within).
  - h_i,t denotes group i aggregate average actual hours at time t; ω_it denotes employment share of group i at time t.
- Baseline grouping: cross-product of 32 groups = 2 gender × 4 age groups × 2 marital status × young child dummy.
- Main result:
  - The within-component accounts for most—over four fifths—of the decline in average hours worked for 2003-2019.
  - Industry-level decomposition (21 industry groups, 2008-2019) and occupation-level decomposition (9 occupation groups, 2011-2019) also show within-group component dominating, though slightly smaller than demographic decomposition.
- Implication:
  - Dominant role of within-group declines motivates deeper micro-level investigation of within-group dynamics rather than compositional shifts.

### Regression analysis: specification and approach
- Preferred empirical specification (preserved exactly as in source):
  - h_ict = β_0 + X_ict β + γ Trend + X_ict × λ Trend + θ Z_ct + ε_it
  - h_ict: hours worked variable (actual hours, usual hours, or hours gap between desired and actual/usual) for individual i in country c in year t.
  - X_ict: vector of covariates including gender, age group (15-29, 30-54, 55-64, and 65+), marital status (married and non-married), men having young children (under 5 years) in household, and women having young children. Reference group: prime-aged working women who are unmarried and do not have young children.
  - Trend: yearly trend; λ captures differential trends across demographic groups relative to reference group (main coefficient of interest).
  - Z_ct: country-level controls in some specifications.
  - Standard errors clustered by country and year.
  - Additional robustness: include macro controls (output gap, GDP per capita, net exports), country fixed effects, and country-specific trends.
- Outcomes of interest:
  - Actual hours, usual hours, hours gap between desired and actual/usual hours.
- Purpose:
  - Isolate main groups driving decline in average hours while controlling for demographics, country characteristics, and country-specific trends and fixed effects.

### Key implications and directions (from text)
- Since within-group declines dominate the aggregate fall in average hours, policy and analysis should focus on within-group dynamics (e.g., changing preferences, non-work spells, parental leave impacts, zero-hour weeks).
- The hours gap (desired minus actual hours) is useful to identify which groups have scope to raise labor supply toward preferred levels and where policy could help align hours with preferences.
- The paper proceeds to present decomposition of changes in hours (between vs within) and microdata regression analysis to quantify drivers and inform policy (sections 4 and 5 lead to concluding remarks in section 6).

*Italic: Source — wpiea2024002-print-pdf - 2014. Bick, Fuchs-Schündeln and Lagakos (excerpt provided).*

### 5.2  Results

### 5.2 Results

### Which groups have seen a sharper decline?
- Regression results (Table 2) confirm descriptive findings: men, particularly those with young children, and young workers have seen a sharper decline in hours worked.
- Key estimated trends and group effects:
  - Hours worked have been declining annually by about 0.07 hours per week more for men without young children than for the reference group.
  - Men with young children have experienced an additional 0.03-0.05 annual decline in their hours per week.
  - Women with young children have seen a decline in average hours worked similar to that of the reference group.
  - Young workers have seen an additional 0.07-0.11 hours decline.
  - The estimated effect for the young implies a decline in average actual hours of up to 2 hours compared to the reference group over 2003-2019.
- Macroeconomic controls:
  - Coefficients on both the output gap and GDP per capita are highly significant, indicating hours are procyclical (negative output gap coefficient) and that the income effect dominates the substitution effect in aggregate (negative GDP per capita coefficient).
- Full-time vs part-time:
  - Results hold for both full-time and part-time workers, but tend to be weaker among full-time workers, particularly for youth.
  - Exception: female workers with young children—for whom only full-time workers have reduced their hours.
- Full-time-only regression results (Table A.2) highlights:
  - (i) Negative trend is less steep for full-time workers but still highly significant.
  - (ii) For men and men with young children, prior results hold but with somewhat smaller coefficients.
  - (iii) Young full-time workers have not seen a trend decline in average hours worked.
  - (iv) Full-time mothers with young children have seen a trend decline in average hours vis-à-vis the reference group (while the average woman has not).

### Are zero-hour weeks driving the results?
- Two exercises to test role of zero-hour weeks:
  - Restrict sample to workers who worked non-zero hours during the reference week.
  - Use usual hours worked as the dependent variable instead of actual hours.
- Findings:
  - Both exercises leave main results unchanged (Appendix Tables A.3 and A.4), indicating the broad reduction in hours among the young and especially men with children does not merely reflect increases in leave-periods or zero-hour weeks.
  - For women with young children, the trend decline in actual hours appears to reflect a rise in zero-hour weeks rather than a fall in usual weekly hours:
    - When excluding workers with zero actual hours, or when using usual hours, the trend coefficient for women with a child under 5 years becomes positive and significant (it was zero for the full sample using actual hours, and negative for full-time workers).
    - In 2019, 8.8 percent of all workers reported zero actual hours in the reference week, while 25.2 percent of women with children under 5 years old did.
  - Appendix Table A.9 shows zero-hour weeks and their increase over time are mostly explained by parental leave.

### Why sharper declines for young (part-time) workers?
- Rising school enrollment among young people contributes to the decline:
  - Share answering “in school” as reason for working zero hours increased from 7.3 percent in 2006 to 12.6 percent in 2019.
- Actual hours and evolution for young workers in school (Appendix Table A.5):
  - Young workers in school work much less than average young workers: 26.68 hours versus 34.41 hours per week in 2003.
  - Their average hours worked declined by 3.49 hours to 23.18 hours per week between 2003 and 2019.
  - This pattern of lower average hours and sharper decline is observed for both men and women.

### Does the decline in average hours worked reflect preferences or constraints?
- Measures of hours gap constructed from desired hours:
  - Gaps defined as differences between (i) desired and actual hours, (ii) desired and usual hours, and (iii) desired and usual hours for full-time workers. Positive gap indicates working less than desired.
- Distributional patterns (Figure 7 and Figure 8):
  - Actual, usual, and desired hours distributions have a mode at 40 hours.
  - Actual hours distribution has much thicker tails, especially a thicker left tail.
  - The gap between desired and actual hours has a positive mean of about 4 hours.
  - The gap between desired and usual hours is consistently much smaller than between desired and actual hours.
  - While the gap between desired and actual hours has grown over the sample period, the gap between desired and usual hours has barely risen.
- Econometric evidence using hours gaps as dependent variables (Appendix Tables A.6–A.8):
  - (i) All groups have a positive gap between desired and actual hours.
  - (ii) The gap between desired and usual hours is much narrower.
  - (iii) For full-time workers, there is on average no gap between desired and usual hours; most groups work slightly more than desired but the young work slightly less.
  - The positive level of the usual-hours gap is driven by part-time work for all groups except the young (comparison of Appendix Tables A.7 and A.8).
- Interpretation:
  - The actual-hours gap is partly misleading because annual leave, parental leave and other exceptional leave enlarge the left tail of actual hours; annual and parental leave are often desired.
  - The gap between desired and usual hours is a more relevant gauge of unmet desire to work more.
  - For the main declining groups (men, youth, men with young children), trend declines appear likely driven by preferences: group-specific trend coefficients for the desired-usual gaps are at most very small and insignificant, suggesting desired hours fell alongside usual hours.
  - Married full-time workers have been closing their negative hours gap.
  - For mothers of young children, the rising incidence of zero-hour weeks (largely parental leave) explains the widening desired-actual gap despite no increase in the desired-usual gap.
- Reasons for part-time work have not shifted in ways suggestive of rising involuntary part-time (Appendix Table A.10).

*Source: EULFS and authors’ calculations (5.2 Results, wpiea2024002-print-pdf).*

### 5.4  Further Insights

### 5.4  Further Insights

### 5.4.1 The Roles of Education, Industry and Occupation
- Limited consistent data availability for industry and occupation variables since 2003 constrained earlier analysis.
- When the sample is restricted to 2011 to incorporate industry and occupation variables:
  - Most key findings still hold.
  - Men and men with young children have seen a sharper decline in average hours worked than other groups.
  - Young workers continue to show a negative trend, though it is no longer statistically significant due to the change in sample (see Appendix Table A.11).

### 5.4.2 The Cross-Country Dimension
- Two aims of cross-country investigation:
  - Test how broadly the demographic-group trends hold and whether they are driven by a subset of countries.
  - Compare aggregate dynamics of hours worked across European countries to assess consistency with basic labor supply theory.
- Findings:
  - Declines in hours worked share important common features across countries, indicating broad-based shifts.
  - The pattern that lower average hours have been driven by the young, men, and men with young children holds in nearly every country (country-by-country regressions), though magnitudes differ across countries.
  - Cross-country evidence consistent with a dominant income effect:
    - Countries with higher GDP per capita tend to have fewer average working hours (Figure 9a).
    - There is convergence in average hours worked across countries over time, with larger declines in countries that had longer average hours and, for many, lower GDP-per-capita levels at the beginning of the sample (Figure 9b).
- Data sources referenced:
  - EULFS, Eurostat, IMF World Economic Outlook, and authors’ calculations.

### 6 Concluding Remarks — Patterns, Interpretation, and Policy Implications
- Aggregate trends and micro evidence:
  - Total hours worked recovered remarkably after the COVID-19 pandemic, primarily driven by an increase in the employment rate.
  - Average hours worked extended their pre-COVID trend decline, with little evidence of remaining slack along the intensive margin by mid-2023.
  - The drop in working hours over the last two decades was driven by within-demographic-group declines rather than compositional effects.
  - Men—particularly with young children—and youth drove much of the decline.
  - Actual hours worked fell in line with desired hours worked for most groups.
  - For groups where actual hours did not fall in line with desired hours (most importantly, women with young children), a rising gap between desired and actual hours was largely driven by increased leave periods (annual, parental, or other types).
  - The gap between actual and usual weekly hours was broadly stable.
  - Overall inference: falling hours worked across Europe likely reflected preferences; results are consistent with a dominant role of the income effect over the substitution effect in determining workers’ labor supply at the intensive margin.
- Outlook:
  - The decline in average hours worked is likely to continue in European countries in the future.
  - The pace will depend on trend productivity and wage growth across the continent and will vary across countries depending on economic convergence paths.
  - Over the medium term, most economic forecasts foresee modest productivity gains for economies close to the technological frontier (advanced Europe), which, all else equal, would be expected to lead to modest reductions in working hours.
  - Over the longer term, sources and constraints on growth (e.g., artificial intelligence, climate change) will play a critical role.
  - For emerging European countries farther from the technological frontier, productivity growth prospects and scope for falling hours appear larger as these economies catch up to living standards in advanced Europe.
- Role for policy:
  - Preferences predominate, but policy can dampen (though unlikely to reverse) the trend fall in hours.
  - Desired hours typically exceed usual hours, implying scope for welfare-enhancing increases in hours and reduction of involuntary part-time work.
  - Back-of-the-envelope calculations:
    - Fully closing the gap between desired and actual hours would increase total hours worked by around 0.5 hours, or 1.3 percent.
    - Enabling all involuntary part-time workers to switch to full-time jobs would bring about an average increase of 0.42 hours or 1.1 percent in working hours.
  - Aggregate and subgroup statistics:
    - Aggregate usual hours worked in 2019: 37.11 hours.
    - Aggregate desired working hours in 2019: 37.58 hours.
    - The aggregate gap in 2019: 0.47 hours (around 1.3 percent of the aggregate usual hours worked in 2019).
    - Usual hours worked gap for involuntary part-time workers in 2019: 9.6 hours; its share in total employment: around 4.4 percent in the sample. This translates into 1.1 percent of the total usual hours worked (37.11 hours) in 2019 (= 9.6×0.044/37.11).
    - Bringing the trend decline in hours gap back to 2003 level for women with young children would increase aggregate average hours worked by 0.13 hours or 0.38 percent. The contribution is calculated as: (0.112 hrs)×16 years×its employment share (0.07) / aggregate average hours in 2019 (32.88 hours)×100%.
  - Policy measures to increase hours that align with worker preferences and reduce involuntary part-time include:
    - Neutral tax and benefit systems with respect to workers’ hours choices.
    - Targeted measures to help part-time working mothers who would like to switch to full-time jobs.
    - Active labor market policies to enhance matching between involuntary part-time workers and available full-time jobs (including retraining programs).
  - Social protection design considerations:
    - Avoid contributory pension and other social protection schemes that unduly incentivize part-time work through benefit formulas that excessively credit part-time work periods.
    - Design non-contributory benefits (e.g., housing allowances) to minimize threshold effects that disincentivize taking full-time jobs; slowly phasing out benefits can help, though there is a trade-off with fiscal costs and reform of in-work tax credits.
  - Measures for part-time working mothers:
    - Neutral tax treatment of second earners, higher child care subsidies or services, enhanced paid (pa)maternity leave, and more flexible work arrangements including teleworking may raise working hours towards desired levels.
    - Aggregate impacts are likely limited because the targeted group is relatively small and because policies may reshuffle hours between workers; effects are upper bounds given potential re-shuffling between mothers and fathers.
  - Active labor market policies should be available to part-time workers seeking full-time employment.
  - Policy design must consider interactions between intensive and extensive margins of labor supply and joint household labor supply decisions; e.g., more generous parental leave can affect both mothers’ and fathers’ hours and may incentivize labor force participation of young parents, potentially raising total hours even if average actual hours decline.

*Source: wpiea2024002-print-pdf - 5.4  Further Insights*

### References

### References

### Appendix A — Additional figures and tables: Key empirical findings
- Table A.1: Summary Statistics: Usual Hours Worked
  - Total: ̄h_usual_2003 = 38.19; ̄h_usual_2019 = 37.11; ∆̄h′_19−′_03 = -1.08
  - Men: ̄h_usual_2003 = 41.33; ̄h_usual_2019 = 39.94; ∆̄h′_19−′_03 = -1.4**; E_2003 = .56; E_2019 = .54; Contr. = -1.63
  - Women: ̄h_usual_2003 = 34.17; ̄h_usual_2019 = 33.81; ∆̄h′_19−′_03 = -.36; E_2003 = .44; E_2019 = .46; Contr. = .56
  - Young: ̄h_usual_2003 = 37.23; ̄h_usual_2019 = 35.38; ∆̄h′_19−′_03 = -1.85; E_2003 = .11; E_2019 = .09; Contr. = -1.11
  - Prime: ̄h_usual_2003 = 38.77; ̄h_usual_2019 = 38.08; ∆̄h′_19−′_03 = -.69; E_2003 = .77; E_2019 = .71; Contr. = -2.59
  - Older: ̄h_usual_2003 = 37.61; ̄h_usual_2019 = 36.84; ∆̄h′_19−′_03 = -.77; E_2003 = .11; E_2019 = .17; Contr. = 2.40
  - Elderly: ̄h_usual_2003 = 30.74; ̄h_usual_2019 = 27.48; ∆̄h′_19−′_03 = -3.26; E_2003 = .02; E_2019 = .03; Contr. = 0.28
  - Married: ̄h_usual_2003 = 38.51; ̄h_usual_2019 = 37.45; ∆̄h′_19−′_03 = -1.05; E_2003 = .6; E_2019 = .53; Contr. = -3.30
  - Child u5: ̄h_usual_2003 = 38.21; ̄h_usual_2019 = 37.29; ∆̄h′_19−′_03 = -.91; E_2003 = .17; E_2019 = .16; Contr. = -0.43
  - Men w. Child u5: ̄h_usual_2003 = 42.35; ̄h_usual_2019 = 41.05; ∆̄h′_19−′_03 = -1.3**; E_2003 = .1; E_2019 = .09; Contr. = -0.51
  - Women w. Child u5: ̄h_usual_2003 = 32.14; ̄h_usual_2019 = 32.46; ∆̄h′_19−′_03 = .32; E_2003 = .07; E_2019 = .07; Contr. = .08
  - Note: Standard errors in parentheses; significance levels denoted by *p <0.10, **p <0.05, ***p <0.01. Source: EULFS and authors’ calculations.

- Table A.2: Actual Hours Worked — Full-Time (selected coefficients, period 2003-2019, 25 European countries)
  - Trend coefficient (Column (1)): -0.121 ∗∗∗ (0.0219)
  - Trend coefficient (Column (5), with macro controls, country FEs, country-specific trends): -0.180 ∗∗∗ (0.0142)
  - Men (Column (1)): 2.928 ∗∗∗ (0.278)
  - Women x Child u5 (Column (1)): -5.287 ∗∗∗ (0.704)
  - Trend x Men (Columns reported): -0.0290 ∗∗ (0.0126); -0.0371 ∗∗∗ (0.0102); -0.0364 ∗∗∗ (0.00817)
  - Trend x Men x Child u5 (Columns): -0.0298 ∗∗∗ (0.00853); -0.0373 ∗∗∗ (0.00711); -0.0485 ∗∗∗ (0.00660)
  - R-squared range reported: 0.00168 to 0.0373
  - N.Obs: up to 20,071,743 (varies by column); regressions run over 2003-2019.

- Table A.3: Actual Hours Worked — Nonzero hours (selected coefficients)
  - Trend (Column (1)): -0.0930 ∗∗∗ (0.00212)
  - Men: 6.822 ∗∗∗ (0.768)
  - Young: -4.502 ∗∗∗ (0.788)
  - Women w. Child u5: -2.757 ∗∗∗ (0.619)
  - Trend x Men: -0.0735 ∗∗∗ (0.00679)
  - Trend x Men w Child u5: -0.0330 ∗∗∗ (0.00452)
  - N.Obs up to 22,234,088.

- Table A.4: Usual Hours Worked (selected coefficients)
  - trend (Column (1)): -0.0778 ∗∗∗ (0.00794)
  - Men: 6.909 ∗∗∗ (0.750)
  - Women w. Child u5: -2.242 ∗∗∗ (0.553)
  - Trend x Men: -0.0725 ∗∗∗ (0.00983)
  - Trend x Men w. Child u5: -0.0316 ∗∗∗ (0.00471)
  - Trend x Women w. Child u5: 0.0645 ∗∗∗ (0.0174)
  - R-squared range: 0.00109 to 0.165
  - N.Obs: 23,425,275 down to 21,182,287 by column.

- Table A.5: Actual Hours Worked of Young Workers (averages and changes)
  - Total h_actual_2003 = 35.37; h_actual_2019 = 32.88; ∆h_actual′_19−′_03 = -2.49** (.81) (.71) (1.06)
  - Young: 34.41 → 31.67; ∆ = -2.74** (1.02) (.8) (1.28)
  - Young in school: 26.68 → 23.18; ∆ = -3.49 (3.29) (1.47) (3.53)
  - Young Women: 31.37 → 28.88; ∆ = -2.49 (1.12) (.75) (1.33)
  - Young Men: 36.92 → 34.03; ∆ = -2.89** (.89) (.84) (1.21)

- Table A.6: Hours Gap (Desired − Actual) (selected coefficients)
  - Trend (Column (1)): 0.0808*** (0.00182)
  - Men: -0.774*** (0.167)
  - Women w. Child u5: 2.622*** (0.448)
  - Trend x Women w. Child u5: 0.115** (0.0407)
  - R-squared range: 0.000948 to 0.0627
  - N.Obs up to 20,328,560.

- Table A.7: Hours Gap (Desired − Usual) (selected coefficients)
  - Trend (Column (1)): 0.0181 (0.0259)
  - Men: -0.791 ∗∗∗ (0.137)
  - Young 15-29 yrs: 0.936 ∗∗ (0.439)
  - Women w. Child u5: -0.178 (0.113)
  - R-squared range: 0.000168 to 0.0966
  - N.Obs up to 19,827,882.

- Table A.8: Hours Gap (Desired − Usual) among Full-Time Workers (selected coefficients)
  - Trend (Column (1)): 0.0153 ∗∗∗ (0.0000569)
  - Young: 0.405 ∗∗∗ (0.132)
  - Older: -0.394 ∗∗∗ (0.0967)
  - Married: -0.411 ∗∗∗ (0.138)
  - Trend x Married: 0.0140 ∗∗∗ (0.00408)
  - R-squared range: 0.000188 to 0.0584
  - N.Obs up to 16,647,454.

- Table A.9: Reasons for Not Working Despite Having a Job (proportions reported)
  - Columns present values for Groups: Others, Men with YC, Women w YC for Years 2006 and 2019.
  - Selected entries (proportions): Own illness/injury (Others: 2006 = 0.21; 2019 = 0.24), Parental leave (Women w YC: 2006 = 0.42; 2019 = 0.56), Holidays (Others: 2006 = 0.63; 2019 = 0.61).

- Table A.10: Reasons for Working Part-Time (proportions reported)
  - Selected entries (Others / Men with YC / Women w YC across 2006 and 2019)
  - Person in edu./training (Others: 0.13 → 0.12); Looking after child/disabled (Women w YC: 0.65 → 0.65); Could not find a FT job (Others: 0.23 → 0.23; Men with YC: 0.41 → 0.38).

- Table A.11: Hours Worked — Education, Industry, and Occupation (selected coefficients; regressions include macro controls, country fixed effects, and country-specific trends)
  - Trend coefficients across columns: -0.230 ∗∗∗; -0.128 ∗∗∗; -0.146 ∗∗∗; -0.120 ∗∗∗; -0.125 ∗∗∗; -0.127 ∗∗∗ (standard errors reported per column)
  - Men: coefficients around 7.36 ∗∗∗ to 7.49 ∗∗∗ (different columns)
  - Women x Child u5: -5.147 ∗∗∗; -4.320 ∗∗∗; -4.570 ∗∗∗; -4.527 ∗∗∗; -4.326 ∗∗∗; -4.591 ∗∗∗
  - Highly Educ.: 1.685 ∗∗ (0.587) in one specification; Trend x High. Educ. = -0.0349 ∗ (0.0180)
  - Non-Routine: 2.347 ∗∗∗ (0.502); Trend x NonRout. = -0.0413 (0.0237)
  - Industry and sector trend interactions: Trend x Ind. = 0.0734 ∗ (0.0322); Service sector trend effects reported.
  - R-squared values between 0.0926 and 0.1070
  - N.Obs: 21,809,787; sub-sample N.Obs = 9,165,969 for columns analyzing 2011-2019 subset and controls.

### Notes on methodology and data (from appendix notes)
- Dependent variables across tables include: h_actual,ft (actual hours for full-time), h_actual,nonzero (actual hours > 0), h_usual (usual hours), hgap_actual (desired − actual), hgap_usual (desired − usual), and hgap_usual,ft (desired − usual for full-time).
- Reference group in regressions: prime-aged women who are unmarried workers with no young children in the household.
- “Child u5” is a dummy equal to 1 if a child under 5 years is in the household.
- Standard errors are clustered at the country-year level.
- Significance notation: * p <0.10, ** p <0.05, *** p <0.01.
- Regressions typically cover the period 2003-2019 for 25 European countries and use EULFS and authors’ calculations.

*Source: wpiea2024002-print-pdf - References (Appendix A tables and notes).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024002-print-pdf.pdf_