## Appendix I. Additional Tables and Figures (_wp12248)

## Source details

**Canonical URL:** [Appendix I. Additional Tables and Figures (_wp12248)](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12248.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12248.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12248.pdf.json)

---

### I. Context and motivation
- Japan’s potential growth rate is steadily falling with the aging of its population; the paper asks whether raising female labor participation (FLP) can help slow this trend.
- Population projections and labor force trends:
  - Share of the population over age 65: from 9 percent in 1980 to 36 percent in 2040.
  - Working-age population (ages 15–64): peak of 87 million in 1995 falling to about 55 million in 2050.
- Japan’s FLP and immigration are well below OECD averages; attitudes on immigration change slowly, so increasing FLP is a near-term policy lever.

### II. Quantified scenarios and macroeconomic impact
- Back-of-the-envelope estimates of raising FLP:
  - Raise FLP ratio from 63 percent in 2010 to 70 percent in 2030 → GDP per capita permanently higher by approximately 4 percent versus baseline.
  - Raise FLP further—to the level of northern Europe—→ additional 4 percent increase in GDP per capita.
- Impact on potential GDP growth (in transition years):
  - About 0.2 percentage point under the first scenario.
  - About 0.4 percentage point under the second scenario.
- Note: Estimates reflect increased labor input only (no productivity gains) and are considered a lower bound.

### III. Empirical findings on FLP patterns and trends
- Cross-country/time patterns:
  - Mean FLP across OECD prime-age women: increased from 61.2 to 76.9 percent between 1985 and 2005.
  - Japan: FLP increased from 60.3 to 68.8 percent over the same period.
  - Gender gaps in prime-age labor participation:
    - Japan: 25 percentage points lower female participation relative to males (2009).
    - Most countries: differences around 10 percentage points; some northern European countries: as low as 5 percentage points.
  - Education: 2010 cohort of women in their late 20s averaged 14.3 years of schooling (surpassed among advanced economies only by New Zealand).
- Demographic drivers:
  - Family size and education explain many within-country changes over time.
  - Family-friendly policies (e.g., child care provision) are important in explaining differences across countries.
- Distributional observation:
  - Participation rates converged across countries between 1985 and 2005; distribution width narrowed.

### IV. Determinants and modeling approach
- Conceptual framework:
  - Becker’s time allocation model (1965): women allocate time between leisure, market labor, and home production (child rearing); decisions depend on market returns and home production costs/quantity.
- Econometric strategy:
  - Focus on prime-age women aged 25–54.
  - Model FLP as function of demographics (D), policies (Z), and controls (X), with country and time fixed effects.
  - To mitigate omitted-variable bias and nonstationarity, specification differences out male labor participation (MLP) and models FLP minus MLP as function of female-specific demographics and policies.
- Time-series properties and estimation choices:
  - Panel: small in N, large in T (22 OECD countries, 1960–2008).
  - FLP, fertility, and education likely I(1); marriage rates possibly I(0) or I(1).
  - Variables differenced to create I(0) series and eliminate country fixed effects; final estimator is a difference-in-difference (∆) specification.
  - Cointegration tests unable to reject null of no cointegration between demographic variables and FLP; differencing adopted with caution about omitted error-correction terms.

### V. Key policy-relevant empirical insights
- Demographics vs. policy:
  - Demographics (family size, education) explain within-country temporal changes.
  - Policy variables (childcare availability, tax treatment, maternity/parental leave) explain cross-country differences.
- Magnitudes and statistical findings (selected, preserved exactly as reported):
  - Measured in standard deviation terms: a 1 standard deviation decline in number of children per woman and a 1 standard deviation increase in education are each roughly associated with a 3 percentage point increase in FLP.
  - Coefficients (Table 1 examples):
    - Log(Number of children per woman): -12.386**, -16.345***, -25.009***, -37.345  [standard errors: 4.839, 6.017, 7.375, 29.206]
    - Log(Education): 11.627***, 10.765***, 5.586, 3.054  [standard errors: 3.390, 3.428, 3.487, 6.771]
    - Marriage rate: -0.760* , -3.872  [standard errors: 0.410, 7.125]
    - Observations: 231, 229, 174, 174
  - Time-span effects (Table 2 examples):
    - Log(Number of children per woman): -10.014**, -14.197**, -16.345***, -21.905***, -24.977**  [standard errors: 4.936, 6.135, 6.017, 7.354, 10.913]
    - Log(Education): 10.932***, 10.945***, 10.765***, 8.271**, 4.484  [standard errors: 3.609, 3.269, 3.428, 3.592, 3.256]
    - Observations: 729, 348, 229, 169, 118
- Policy variable effects (pull vs push; Table 3 and related):
  - Pull policies: raise return to work (e.g., tax wedge).
  - Push policies: reduce child-rearing costs (e.g., child care subsidies, parental leave).
  - No single policy silver bullet; policy effects varied, generally smaller and less robust than demographic effects.
  - Statistically significant positive effects from parental leave and family allowances; statistically significant negative effect from tax wedges.
  - A 1 standard deviation change in policy is associated with less than a 0.5 percentage point increase in FLP (sample mean).
  - Marginal effects (Table 3, preserved):
    - Wage gap: -0.032 ; Mean 7.391 ; S.D. 0.04 ; Effect by one S.D.: -0.27 (percentage point)
    - Log(Family allowance): 0.201 ; Mean 1.20 ; S.D. 0.700 ; Effect by one S.D.: 0.14
    - Log(Leave): 0.233 ; Mean 3.671 ; S.D. 1.610 ; Effect by one S.D.: 0.36
    - Log(Child benefits): -0.224 ; Mean 4.19 ; S.D. 0.58 ; Effect by one S.D.: -0.13
    - Part-time incidence: -0.031 ; Mean 3.275 ; S.D. 5.59 ; Effect by one S.D.: -0.19
    - Log(Childcare per child): -0.026 ; Mean 6.88 ; S.D. 1.69 ; Effect by one S.D.: -0.03
    - Log(Tax wedge): -2.024 ; Mean 4.67 ; S.D. 0.36 ; Effect by one S.D.: -0.73
  - Parental leave nonlinearity:
    - Estimated with a squared term (Log(Leave) negative, Squared Log(Leave) positive).
    - Leave policies effective only if sufficiently generous—typically more than one and a half years; very short leaves likely no effect or negative effect.
  - Part-time work:
    - FLP higher in countries with higher share of part-time workers; endogeneity likely (higher FLP → higher part-time incidence). Instrumentation can render part-time coefficient insignificant.

### VI. Policy recommendations emphasized for Japan
- Two overarching goals: end gender gaps in hiring/promotion; support working mothers better.
- Measures proposed (explicit items preserved):
  - Reallocate public resources from monetary benefits to in-kind benefits such as child care facilities.
  - Deregulate the child care industry to increase the number of facilities.
  - Extend the duration and broaden the coverage of parental leave policies.
  - Eliminate institutional exemptions on spousal income in social security and tax systems.
  - Reduce disparities between part-time and full-time workers.
  - Encourage firms to adopt more flexible work environments.
  - Ensure that promotion and employment policies are enforced equitably to increase female career employees and role models (Japan has by far the lowest rate of female managers among advanced economies).
  - Introduce a more flexible labor contract for career employees to reduce hiring risks for firms.
  - Consider establishing new rules for the number of female directors on corporate boards.
- Expected social benefits:
  - These policies would help increase FLP and could reduce the high incidence of poverty among single mothers.

### VII. Japan-specific empirical and institutional findings
- Decompositions and country results:
  - Demographics account for much of within-country variation:
    - Number of children per woman explains on average about one-fifth of within-variable variation across the 22 OECD countries and more than one-half in Finland, Japan, Sweden.
    - With education, demographics explain about one-quarter of within-variable variation.
  - Japan: FLP increased from 56.7 in 1980 to 70.3 in 2008; much of rise linked to decline in number of children per woman.
    - Percentage of unmarried women age 25–29: rose to 59 percent in 2010 from 24 percent in 1980.
    - Example calculation: Japan decline in number of children per woman from 4.2 to 3.7 (1980–2008) → predicted increase in FLP of 8 percentage points (more than half of the full change).
  - Cross-country decomposition:
    - Policies matter more for cross-country differences than for within-country changes.
    - Child care expenditure helps explain higher participation in Scandinavian countries; part-time work important for Netherlands.
    - For Japan, model explains little of the cross-country difference in participation rates; small negatives from lower child care support and higher wage gap, demographics positive; large unexplained residual remains.
- Labor market features and hurdles:
  - Hiring and promotion concentrated at postsecondary entry with implicit lifetime employment; limited mid-career mobility.
  - Career vs. noncareer positions: career (sogoshoku) vs noncareer (ippanshoku) split, with women concentrated in noncareer jobs.
  - Survey facts:
    - Women make up just 6 percent of career employees in 2010 (sample sizes: 215 firms in 2000, 129 in 2010; promotion question sample: 75 firms in 2000, 24 in 2010).
    - Share of women in career positions rose from 2.2 percent in 2000 to 6 percent in 2010; share of newly hired sogoshoku was 12 percent in 2010.
    - Gender wage gap in median wages: declined from 42 percent in 1980 to 28 percent in 2009.
    - 52 percent of women hold nonregular positions versus 17 percent of men.
  - Recommendations for employment/promotion:
    - Enforce equal employment laws more effectively.
    - Modify regular work contracts to include phased-in employment protection (reduce hiring risk).
    - Promote diversity and female role models; consider minimum female director rules.

### VIII. Leave policy, child care supply, and time-use findings
- Usage and effects of leave:
  - Proportion of eligible female workers taking child care leave: from 49 percent in 1996 to 88 percent in 2011.
  - Fewer than 3 percent of fathers use child care leave (versus 70 percent in Sweden).
  - Micro-data: longer parental leave increases probability mothers return to jobs; effect particularly strong in Japan.
  - Probability of finding full-time work after career interruption in Japan (Ueda, 2007):
    - 18 percent for university educated women.
    - 12 to 13 percent for less educated women.
  - Recommendation: extend leave duration to levels similar to France, Germany, Scandinavian countries.
- Child care supply and fragmentation:
  - System split between day care centers (full-day, Ministry of Health, Labor, and Welfare) and kindergartens (part-day, Ministry of Education).
  - Demand outstrips supply: wait-listed children ~25,000 despite target capacity increase from 203,000 to 215,000 by 2009; informal reports suggest unmet demand up to one-third of capacity.
  - Kindergartens underutilized (~70 percent of capacity).
- Time-use and barriers:
  - Having an infant reduces participation by about 30 percent.
  - Mothers living with parents/in-laws more likely to participate.
  - Time men spend on child care is same regardless of woman’s work decision → market work is additional burden for women.
  - Survey: working hours and housework burden are top reasons for not participating; 87 percent indicated interest mainly in part-time work if participating.
- Policy levers:
  - Increase child care supply via deregulation, merging day care and kindergarten systems, small reallocation of spending toward child care.
  - Promote flexible work arrangements (e.g., Dutch model of part-time equality; laws granting right to request flexible hours).
  - Address tax and benefit design (the “¥1.03 million barrier” and child-rearing allowances): consider targeting allowances by income; in-kind benefits likely more effective for participation.

### IX. Data, coverage, strengths, and limitations
- Coverage and data sources:
  - Cross-country data set: 22 OECD countries; period 1960–2008 (data unbalanced).
  - Main sources: OECD.Stat, Comparative Family Policy Database ver.3 (Gauthier, 2010/2011), Bassanini and Duval (2006) for relative marginal tax rates.
- Dataset observation counts (Table 5 totals):
  - FLP: 770; Marriage rate: 797; Number of children per woman: 1091; Education: 1122; Wage gap: 1078; Family allowance: 1078; Leave: 1121; Child benefits: 814; Part-time incidence: 509; Childcare per child: 536; Tax wedge: 342.
- Country-level latest statistics (selected Table 6 entries, preserved exactly):
  - JPN: FLP 70.30, Marriage rate 5.80, Number of children per woman 0.42, Education (Year) 11.02, Wage gap 23.15, Family Allowance 4.64, Leave 66, Child benefits 26.20, Part-time incidence 15.20, Childcare per child 2151, Tax wedge 0.88
  - OECD 22 Mean: FLP 78.72, Marriage rate 5.03, Number of children per woman 0.51, Education 10.54, Wage gap 18.46, Family Allowance 2.38, Leave 84.82, Child benefits 47.34, Part-time incidence 14.41, Childcare per child 3807.58, Tax wedge 1.13
- Correlation structure (selected, Table 7):
  - Correlation(FLP, Marriage rate) = -0.28
  - Correlation(FLP, Log(Number of children per woman)) = -0.46
  - Correlation(FLP, Log(Education)) = 0.44
  - Correlation(FLP, Log(Childcare per child)) = 0.65
  - Three-year change correlations: Correlation(FLP-DLP, Log(Number of children per woman)) = -0.27; Correlation(FLP-DLP, Log(Education)) = 0.23; Correlation(FLP-DLP, Part-time incidence) = 0.42
- Regression evidence and robustness (selected Table 8/9/10 excerpts):
  - Log(Number of children per woman): consistently negative and highly significant across specifications (examples reported: -16.345***, -16.533***, -15.625***, -16.103***, -21.012***, -26.522***, -19.142*...).
  - Log(Education): positive and often significant (examples: 10.765***, 10.826***, 13.568***...).
  - Log(Leave): negative with positive squared term (examples: Log(Leave) -0.934***, Squared Log(Leave) 0.158**).
  - Part-time incidence: positive and significant in many cross-section specifications (examples: 0.649***, 0.643***, 0.576***...).
  - Log(Tax wedge): large negative coefficients in some specifications (example: -20.821***); interaction Log(Tax wedge) * Log(Education) positive in some specifications (examples: 9.056***, 13.276***).
  - Instrumentation: Number of children per woman instrumented in relevant columns; part-time incidence instrumented in specified columns; F-tests for instruments often large (examples: F-test Number of children per woman 217.76, 215.64, 176.45...; F-test Part-time incidence 24.64, 26.86...).
- Marginal effects for ten-year averages (Table 11, preserved exactly):
  - Marriage rate: Marginal Effect -4.55; Mean 6.08; S.D. 1.33; Effect by one S.D. Change: -6.06
  - Log(Number of children per woman): Marginal Effect -43.85; Mean 4.16; S.D. 0.23; Effect by one S.D. Change: -10.11
  - Log(Education): Marginal Effect 27.86; Mean 2.12; S.D. 0.32; Effect by one S.D. Change: 8.83
  - Wage gap: Marginal Effect -0.79; Mean 27.03; S.D. 9.59; Effect by one S.D. Change: -7.57
  - Log(Family allowance): Marginal Effect 7.28; Mean 1.20; S.D. 0.60; Effect by one S.D. Change: 4.40
  - Log(Leave): Marginal Effect 3.90; Mean 3.86; S.D. 1.54; Effect by one S.D. Change: 6.01
  - Log(Childcare per child): Marginal Effect 4.90; Mean 6.93; S.D. 1.56; Effect by one S.D. Change: 7.62
- Robustness and limitations:
  - Policy variables often have limited within-country variation (many changes stacked at zero), leading to weaker within-change estimates than cross-sectional estimates.
  - Cross-section (decade-average) regressions show larger policy effects (examples: 1 S.D. decline in wage gap or 1 S.D. increase in child care spending can increase participation by as much as 7.5 percentage points).
  - Some regressions face multicollinearity (e.g., parental leave with other family-friendly policies).
  - Omitted factors (e.g., household technologies, unmeasured market child care availability, labor-market institutions around initial entry) likely explain large residuals.

### X. Key interpretive takeaways (concise)
- Demographics (fertility declines and rising education) are robustly associated with large increases in FLP; quantified relationships preserved:
  - 1 standard deviation decline in number of children per woman ≈ 3 percentage point increase in FLP.
  - 1 standard deviation increase in education ≈ 3 percentage point increase in FLP.
  - 10 percent decline in number of children per woman → 1.6 percentage point increase in FLP.
  - 10 percent increase in education → 1.1 percentage point increase in FLP.
- Policy effects are smaller in within-country change analysis but can be sizable in cross-section; targeted and sufficiently generous parental leave and family allowances help, particularly for lower-education/low-income groups.
- For Japan:
  - Demographic changes explain much of the recent rise in FLP but the model leaves a large cross-country residual.
  - Institutional labor-market features (hiring/promotion around entry, career/noncareer bifurcation, low female managers) and fragmented child care supply are key complementary explanations and policy targets.

*Italic source: IMF staff estimates and analysis in “Japan: Can Women Save Japan?” (Appendix I. Additional Tables and Figures) — content unit _wp12248.*

### Appendix I. Additional Tables and Figures .............................................................................2

### Appendix I. Additional Tables and Figures

### I. Context and motivation
- Japan’s potential growth rate is steadily falling with the aging of its population; the paper asks: can raising female labor participation (FLP) help slow this trend?
- Population projections:
  - Share of the population over age 65 will rise from 9 percent in 1980 to 36 percent in 2040.
  - Size of Japan’s working-age population (ages 15–64) will fall from its peak of 87 million in 1995 to about 55 million in 2050.
- Japan’s FLP and immigration are well below OECD averages; attitudes on immigration change slowly, so increasing FLP is a near-term policy lever.

### II. Quantified scenarios and macroeconomic impact
- Back-of-the-envelope estimates of raising FLP:
  - If Japan raised its FLP ratio from 63 percent in 2010 to 70 percent in 2030 (to the level of the G7 excluding Italy and Japan), GDP per capita would be permanently higher by approximately 4 percent than under the baseline scenario.
  - Raising FLP further—to the level of northern Europe—could increase GDP per capita by an additional 4 percent.
  - Impact on potential GDP growth (in transition years):
    - About 0.2 percentage point under the first scenario.
    - About 0.4 percentage point under the second scenario.
- These estimates reflect the impact of increased labor input only and do not include potential productivity gains; thus they are considered a lower bound.

### III. Empirical findings on FLP patterns and trends
- Cross-country and time-series patterns:
  - Mean FLP across OECD prime-age women:
    - Increased from 61.2 to 76.9 percent between 1985 and 2005.
  - Japan:
    - FLP increased from 60.3 to 68.8 percent over the same period, a slower pace than the median country.
  - Gender gaps in prime-age labor participation:
    - Japan: 25 percentage points lower female participation relative to males (2009).
    - Korea exhibits an even larger gap.
    - Most countries show differences around 10 percentage points; some northern European countries show differences as low as 5 percentage points.
  - Education:
    - In 2010, the cohort of women in their late 20s had on average 14.3 years of schooling, surpassed among advanced economies only by New Zealand.
- Demographic drivers:
  - Family size and education explain many of the changes within countries over time.
  - Family-friendly policies (e.g., child care provision) are important in explaining differences across countries.
- Distributional observations:
  - Participation rates have converged across countries between 1985 and 2005, with the width of the distribution narrowing.

### IV. Determinants and modeling approach
- Conceptual framework:
  - Becker’s time allocation model (1965): women allocate time between leisure, market labor, and home production (child rearing); decisions depend on market returns and home production costs/quantity.
- Econometric strategy:
  - Main focus: prime-age women aged 25–54 to abstract from education and retirement choices.
  - FLP modeled as a function of demographics (D), policies (Z), and other controls (X), with country and time fixed effects.
  - To mitigate omitted-variable bias and nonstationarity, specification differences out male labor participation (MLP) to control for factors affecting both genders; resulting specification models FLP minus MLP as a function of female-specific demographics and policies plus any residual differentials.

### V. Key policy-relevant empirical insights
- Demographics vs. policy:
  - Demographics (family size, education) explain within-country temporal changes.
  - Policy variables (childcare availability, tax treatment, maternity/parental leave) explain cross-country differences.
- Specific policy instruments identified in existing literature as positively associated with FLP:
  - More neutral tax treatment of second earners.
  - Child care subsidies and broader public expenditure on child care and early education services.
  - Paid maternity and parental leave.
  - Note: Japan’s spending on maternity and parental leave payments per child is less than one-half the OECD average; Japan is in the bottom quarter of the distribution for both maternity/parental leave payments and for public expenditure on child care and early education services.

### VI. Policy recommendations emphasized for Japan
- Two overarching goals: end gender gaps in hiring/promotion; support working mothers better.
- Measures that could be considered:
  - Reallocate public resources from monetary benefits to in-kind benefits such as child care facilities.
  - Deregulate the child care industry to increase the number of facilities.
  - Extend the duration and broaden the coverage of parental leave policies.
  - Eliminate institutional exemptions on spousal income in social security and tax systems.
  - Reduce disparities between part-time and full-time workers.
  - Encourage firms to adopt more flexible work environments.
  - Ensure that promotion and employment policies are enforced equitably to increase female career employees and role models (Japan has by far the lowest rate of female managers among advanced economies).
  - Introduce a more flexible labor contract for career employees to reduce hiring risks for firms.
  - Consider establishing new rules for the number of female directors on corporate boards.
- Expected social benefits:
  - These policies would help increase FLP and could reduce the high incidence of poverty among single mothers.

### VII. Data, coverage, and strengths of the analysis
- Coverage: OECD countries during 1960–2008.
- Strengths:
  - Large cross-country coverage and extended time period.
  - Use of updated comparative family policy data (Gauthier’s database) and an improved estimation technique relative to earlier macro studies.
- Focus of policy analysis: primarily married women with children, where actual participation rates are well below expressed preferences.

*Source: IMF staff estimates and analysis in “Japan: Can Women Save Japan?” (Appendix I. Additional Tables and Figures).*

### 16.      We next difference the equation over time. Because the panel is small in N, but large

### 16.      We next difference the equation over time. Because the panel is small in N, but large

### Methodology and time-series properties
- Panel: small in N, large in T; many variables exhibit unstable time-series properties and are integrated of order 1 (I(1)).
- Im-Pearsan-Shin panel unit root tests (estimated with trend) for FLP, fertility, and education cannot reject the null that all panels contain unit roots; the same test for marriage rates cannot be rejected when estimated without a trend but can be rejected with a trend. The null can be rejected for all the variables in differences. Conclusion: FLP, fertility, and education are likely I(1) with marriage rates of either I(0) or I(1).
- Differencing over time:
  - Creates variables integrated of order zero (I(0)).
  - Eliminates country fixed effects.
  - Final difference-in-difference estimator (equation (3)): ∆... (3) where Δ represents changes over time.
- Cointegration: tests unable to reject the null hypothesis that there is no cointegration between the demographic variables and FLP. (If cointegration existed, differencing could induce omitted-variable bias if the error-correction term is excluded.)

### Data
- Cross-country data set covering 22 OECD countries between 1960 and 2008.
- Data include labor force participation, demographics, and policies.
- Data set is unbalanced; demographic variables and the G7 countries generally cover longer time periods.
- Main sources: OECD.Stat, Comparative Family Policy Database ver.3 (Gauthier, 2010 and 2011), and relative marginal tax rates on second earners from Bassanini and Duval (2006).
- Appendix II contains a full description of variables and sources.

### Empirical results — Role of demographics
- Focus variables: marriage rates, number of children per woman, and education levels.
- Theoretical expectations (Becker’s FLP time allocation model):
  - Marriage rate: marriage tends to increase reservation wage and elasticity to market wage; declining marriage rates could be associated with higher FLP.
  - Number of children per woman: more children increase home production opportunity, raising reservation wage; declining fertility could be associated with higher FLP.
  - Education: higher education increases potential earnings, reducing scope for specialization, strengthening attachment to labor market and raising FLP.
- Endogeneity concerns:
  - Marriage and children variables may be endogenous (rising FLP may lead to fewer marriages and fewer children).
  - Instruments: lagged changes used as instruments (instrument for change from t to t+1 is lagged change from t–1 to t).
- Statistical significance and magnitudes:
  - Both log(number of children per woman) and log(education) are statistically significant across specifications and have expected signs.
  - Instrumentation increased the estimated coefficient on number of children per woman; reported F-statistics greater than 200.
  - Measured in standard deviation terms: both a 1 standard deviation decline in the number of children per woman and a 1 standard deviation increase in education are roughly associated with a 3 percentage point increase in FLP.
  - Interpreted individually:
    - Coefficient on number of children per woman: every 10 percent decline in the ratio corresponds to a 1.6 percentage point increase in the FLP rate (Table 1).
    - Coefficient on education: every 10 percent increase in education levels corresponds to a 1.1 percentage point increase in the FLP rate (Table 1).
  - Marriage results:
    - Before instrumentation, the marriage rate coefficient is significant and of the expected sign.
    - Magnitude: a 1 standard deviation decline in the marriage rate is associated with an approximately 1 percentage point increase in FLP (Table 1).
    - After instrumentation the marriage coefficient is not significantly different from zero; instruments are weak, so conclusions are not robust.

- Time-span effects:
  - Reestimated model over one- to five-year changes (Table 2): elasticity of FLP to number of children increases over longer time spans, interpreted as difference between short-term and long-term elasticities for life decisions (e.g., number of children). Coefficients on education remained largely unchanged.
  - Baseline regressions use three-year differences as a balance between long-term impact and sample size.

- Selected reported coefficient values (from Table 1 and Table 2 excerpts):
  - Table 1 (selected coefficients, OLS/IV specifications):
    - Log (Number of children per woman): -12.386**, -16.345***, -25.009***, -37.345  [standard errors in brackets: 4.839, 6.017, 7.375, 29.206]
    - Log (Education): 11.627***, 10.765***, 5.586, 3.054  [standard errors in brackets: 3.390, 3.428, 3.487, 6.771]
    - Marriage rate: -0.760* , -3.872  [standard errors in brackets: 0.410, 7.125]
    - Observations: 231, 229, 174, 174 (for respective columns)
  - Table 2 (effects over different change intervals):
    - Log (Number of children per woman): -10.014**, -14.197**, -16.345***, -21.905***, -24.977**  [standard errors: 4.936, 6.135, 6.017, 7.354, 10.913]
    - Log (Education): 10.932***, 10.945***, 10.765***, 8.271**, 4.484  [standard errors: 3.609, 3.269, 3.428, 3.592, 3.256]
    - Observations: 729, 348, 229, 169, 118

### Empirical results — Role of policies (pull vs push)
- Conceptual classification:
  - Pull policies: raise return to work (e.g., tax penalty/second-earner tax wedge; wage gap reductions).
  - Push policies: reduce child-rearing costs and raise relative return to work (e.g., child care subsidies, parental leave, part-time work availability, income support for children).
- Key empirical findings:
  - No single policy silver bullet; policy effects are varied, less robust and economically smaller than demographic effects.
  - Statistically significant and positive effects from parental leave and family allowances.
  - Statistically significant but negative effect from tax wedges.
  - Coefficients on wage gap, tax benefits, and child care are inconsistent and for the most part do not differ significantly from zero.
  - A 1 standard deviation change in policy is associated with less than a 0.5 percentage point increase in the FLP rate (Table 3), measured at the sample mean.
  - Part-time work:
    - FLP is significantly higher in countries with a higher share of part-time workers.
    - Endogeneity likely: higher FLP leads to higher shares of part-time workers.
    - Instrumented regression (regression 7 in Table 8) makes the coefficient on part-time work insignificant; F-test indicates instruments are sufficiently powerful.
  - Income dependence:
    - Family allowances and tax wedges interact strongly with education levels (Figure 9).
    - For women in low-paying jobs, tax incentives and cash payments can affect participation; for higher-educated women in higher-paying positions the effects are less important.
    - Tax benefits effect is close to zero or slightly negative.
  - Parental leave:
    - Estimated with a squared term: first term negative, squared term positive.
    - Leave policies are effective only if sufficiently generous—typically more than one and a half years; very short leaves (e.g., two weeks or one month) likely have no effect or a negative effect on participation.

- Marginal effects from Table 3 (effects by one standard deviation change; columns correspond to: Marginal Effect, Mean, S.D., Effect by one S.D. Change (In percentage point)):
  - Wage gap: -0.032 ; Mean 7.391 ; S.D. 0.04 ; Effect by one S.D. Change (In percentage point) -0.27
  - Log (Family allowance): 0.201 ; Mean 1.20 ; S.D. 0.700 ; Effect by one S.D. Change (In percentage point) 0.14
  - Log (Leave): 0.233 ; Mean 3.671 ; S.D. 1.610 ; Effect by one S.D. Change (In percentage point) 0.36
  - Log (Child benefits): -0.224 ; Mean 4.19 ; S.D. 0.58 ; Effect by one S.D. Change (In percentage point) -0.13
  - Part-time incidence: -0.031 ; Mean 3.275 ; S.D. 5.59 ; Effect by one S.D. Change (In percentage point) -0.19
  - Log (Childcare per child): -0.026 ; Mean 6.88 ; S.D. 1.69 ; Effect by one S.D. Change (In percentage point) -0.03
  - Log (Tax wedge): -2.024 ; Mean 4.67 ; S.D. 0.36 ; Effect by one S.D. Change (In percentage point) -0.73

### Robustness, limitations, and cross-section versus within-country results
- Data coverage is uneven across policy variables and countries; many policy variables have limited within-country variation over time (stacked at zero in changes).
- Within-variation (difference) regressions show weaker policy effects than cross-sectional analysis.
- Cross-section (decade-average) regressions (Table 10 in Appendix I) show wage gap and child care significant and economically large:
  - A 1 standard deviation decline in wage gap or a 1 standard deviation increase in child care spending can increase labor participation by as much as 7.5 percentage points (Table 11 in Appendix I).
  - Parental leave appears important in cross section but is insignificant in multivariate regressions due to multicollinearity with other family-friendly policies.
- Nonlinearity and additional omitted factors:
  - Squared terms tested for nonlinearity in demographics were insignificant.
  - Other likely contributors to FLP rise not included in regressions: new household technologies and improved working conditions.

### Key statistical and interpretive takeaways
- Demographics (fertility declines and rising education) are robustly associated with large increases in FLP; policy effects are smaller and less robust in within-country change analysis.
- Quantitative magnitudes preserved from the analysis:
  - Data span: 22 OECD countries; 1960–2008.
  - Measured impacts:
    - 1 standard deviation decline in number of children per woman ≈ 3 percentage point increase in FLP.
    - 1 standard deviation increase in education ≈ 3 percentage point increase in FLP.
    - 10 percent decline in number of children per woman → 1.6 percentage point increase in FLP.
    - 10 percent increase in education → 1.1 percentage point increase in FLP.
    - 1 standard deviation decline in marriage rate → approximately 1 percentage point increase in FLP (instrumentation weak; not robust).
    - Policy effect magnitudes (Table 3): largest in absolute terms is Log (Tax wedge) effect by one S.D. change = -0.73 percentage points; Log (Leave) effect by one S.D. change = 0.36 percentage points; Log (Family allowance) effect by one S.D. change = 0.14 percentage points.
- Policy implications implied by results:
  - Demographic trends are powerful drivers of FLP changes and explain a large share of the observed increases.
  - Policies can help raise FLP but typically produce smaller within-country changes than demographic shifts; targeted and sufficiently generous parental leave and family allowances can have positive effects, especially for lower-education/low-income groups.
  - Reducing the second-earner tax wedge and addressing gender wage gaps can matter, but cross-country and within-country evidence differ; tax and cash benefits interact with education and income levels.
  - Part-time work appears correlated with higher FLP but causality is ambiguous and likely runs from FLP to part-time incidence.

*Italic source: IMF staff analysis (excerpt from content unit _wp12248 - 16).*

### 29.      To explore why Japan is different, this section uses the estimated econometric results

### _wp12248 - 29.      To explore why Japan is different, this section uses the estimated econometric results

### Explaining Differences over Time within Countries
- Method: Decomposition of percentage point change in each country’s FLP using estimates in column 13 of Table 8 in Appendix I; interactive and squared terms evaluated at each country’s average value over time.
- Key findings:
  - Demographics account for much of within-country variation:
    - The number of children per woman explains on average about one-fifth of the within-variable variation in the 22 OECD countries in the sample and more than one-half of the within-variable variation in three countries (Finland, Japan, Sweden).
    - With education, demographics explain about one-quarter of the within-variable variation.
  - Japan: FLP increased from 56.7 in 1980 to 70.3 in 2008; much of this rise is linked to the decline in the number of children per woman.
    - Percentage of unmarried women age 25–29 rose to 59 percent in 2010 from 24 percent in 1980.
    - Resulted in a steady rise in single-person households (Matsui, 2010).
  - Part-time work has increased with rising female participation (notably Ireland, Italy, Netherlands). In the Netherlands, part-time work remains a popular, often high-quality, long-term choice.
  - For some countries with large FLP increases (Ireland, Netherlands, Spain), the estimated equation does not fully explain changes—demographics and policy help, but the equation misses deeper economic transformation.
- Quantitative example:
  - Japan: decline in number of children per woman from 4.2 to 3.7 from 1980 to 2008 is associated with a predicted increase in FLP of 8 percentage points (more than half of the full change).
  - United States: decline in number of children per woman from 4.2 to 4.1 from 1980 to 2008 explains 15 percent of the change in FLP; all demographic variables together explain close to 20 percent of the change.
- Interpretation: Demographic coefficients are economically significant for changes within countries over time.

### Explaining Differences across Countries
- Method: Decomposition of percentage deviation of each country’s FLP from the OECD average using estimates in column 13 of Table 8 in Appendix I; interactive and squared terms evaluated at each country’s current value. Also a cross-sectional decomposition using the less robust regressions in Table 9 (column 13).
- Key findings:
  - Policies matter more for cross-country differences than for within-country time-series changes.
    - Key policy variables: availability of part-time work, the wage gap in manufacturing, and public expenditure on child care.
    - Child care expenditure helps explain some of the higher participation rates in Scandinavian countries and partly why the United States is now below the OECD average.
    - Part-time work is important for the Netherlands in the cross section.
  - Demographic factors are relatively less important in the cross section than in the time series.
    - Example: relationship between number of children per woman and FLP changed from a somewhat negative correlation in 1980 cross sections to a seemingly positive correlation in 2008—possibly reflecting diminishing importance of demographics as countries’ demographics converge.
- Japan-specific cross-country result:
  - Very little of the difference in participation rates between Japan and the OECD is explained by the model.
    - Some small negatives: somewhat lower than average support for child care and a higher wage gap.
    - Demographics tend to be positives.
    - Net effect: resulting unexplained residual differs little from the starting percentage deviation from the OECD average.

### Explaining the Residuals (Country Fixed Effects)
- Observation: Large unexplained residuals in both within- and across-country decompositions indicate limits of the regression model; complementary qualitative analysis is needed.
- Examples of factors behind residuals:
  - United States and Canada: relatively high FLP despite low policy support—reflects high education levels and availability of market-based child care and other child support services not captured in policy variables.
  - Scandinavian countries: high FLP and higher fertility are linked to a family-friendly set of government policies that became cultural thresholds; this cultural effect is not easily captured by regression analysis.
  - Japan: a large residual in cross-country comparisons relates in part to Japan’s unique job market (discussed further below).
- Summary statement:
  - Demographic changes are strongly associated with changes in aggregate participation rates within countries over time.
  - Policies increasingly explain differences across countries.
  - For Japan, the model helps explain the recent rise in participation rates but does not convincingly capture characteristics that set Japan apart in cross-country comparisons—likely missing labor market elements around initial labor market entry and leave decisions.

*Italicized source attribution: Extracted from the provided content unit.*

### IV. WOMEN TO THE RESCUE: POLICIES TO RAISE FLP IN JAPAN
- Context and problem statement:
  - Japan has a paucity of female managers: ratio of female managers was 9 percent in 2009 compared with 43 percent in the United States (Figure 13).
  - Low FLP is linked to hiring practices, promotion policies, and lack of policies promoting work-family balance. Korea shares similar hiring practices and disparity.
  - Aim: review system and discuss changes to raise FLP and share of female managers.

#### A. Hurdle 1: Employment and Promotion Policies
- Institutional feature: major hiring and career decisions occur postsecondary graduation, filling jobs with implicit lifetime employment guarantees; employees make few job shifts during prime working years.
- Consequences:
  - Most women face a choice between noncareer positions and career positions (sogoshoku vs. ippanshoku).
    - Career positions pay more and include significant human capital investment over a lifetime.
    - Noncareer positions are predominantly filled by women, pay less, and include less investment in human capital.
  - Corporations select for long-term commitment to minimize risk of early retirement of women (Yamaguchi, 2008).
- Empirical facts:
  - Very few women in career path positions within large corporations (Figure 14).
    - 2010 survey: women make up just 6 percent of career employees (consistent with low level of female managers).
    - Share of women in career positions rose from 2.2 percent in 2000 to 6 percent in 2010; share of newly hired sogoshoku was 12 percent in 2010.
    - Sample sizes for MHLW survey: 215 firms in 2000, 129 firms in 2010.
    - Sample size for promotion question: 75 firms in 2000 and 24 firms in 2010.
  - Promotion disadvantage: at more than half the firms in the survey, top-performing male employees were one or more steps ahead of top-performing female employees in the promotion cycle.
  - Wage gap:
    - Gender gap in median wages declined from 42 percent in 1980 to 28 percent in 2009 (difference in median wages between men and women).
    - Japan’s gap nearly twice that of Sweden but smaller than Korea.
    - Micro panel studies find the wage gap cannot be explained by productivity differences (Abe, 2005; Kawaguchi, 2007).
    - Contribution of nonregular positions to the gap: 52 percent of women hold nonregular positions relative to 17 percent of men.
- Policy recommendations:
  - Enforce equal employment laws more effectively: government needs to better enforce laws on wages, employment, and promotion discrimination (Matsui, 2010).
  - Introduce more flexible employment contracts to reduce hiring risks:
    - One option: modify regular work contracts to include phased-in employment protection that gradually increases dismissal costs over tenure—reducing hiring risk while preserving protection for tenured employees.
  - Promote diversity and female role models:
    - Increase number of women in high-profile career positions to counter early self-selection out of career tracks (e.g., University of Tokyo with less than 20 percent female student body).
    - Examples of progress: Bank of Japan first female branch manager; Daiwa Securities placed four women on its board in 2009; Shiseido set a goal of 30 percent female managers by 2013.
    - Consider rules for minimum number of female directors on company boards (following Norway, Spain, France).

#### B. Hurdle 2: Balancing Family Responsibilities with Work
- Pattern: Japan exhibits an M-curve in female participation—participation similar to comparators in early twenties but drops sharply in late twenties and thirties (Figure 16).
  - Roughly 60 percent of Japanese women quit working after giving birth to their first child.
  - Causes: weaker labor-market attachment due to lower wages and fewer career opportunities, and weak support systems for working mothers.
- Policy focus areas: (1) leave policy to allow women to retain current positions, (2) child care policies to reduce time burden of family responsibilities, (3) flexible work arrangements to balance work and family.
- Parental leave provisions:
  - Japan’s leave provisions are near OECD averages but generally less than major European countries.
  - System includes maternity leave (established 1947) and child care leave for children under one year old (established 1991), which raised child-related leave from 14 to 58 weeks—bringing Japan broadly in line with OECD averages.
  - Working parents are entitled to 50 percent of their previous income up to an income ceiling of 52 weeks.
  - 2005 revision extended Act on Parental Leave to some nonregular workers, but their share in the total remains low, at 4.3 percent in 2007 (Oishi, 2011).

*Italicized source attribution: Extracted from the provided content unit.*

### 43.      Usage of leave policy has increased

### 43.      Usage of leave policy has increased

### Usage of leave policy
- The proportion of eligible female workers taking child care leave increased from 49 percent in 1996 to 88 percent in 2011.
- Fewer than 3 percent of fathers make use of child care leave (relative to 70 percent in Sweden) despite a system that allows both parents parental leave.

### Impact of leave length on return to work
- Micro-data evidence (Waldfogel, Higuchi, and Abe (1999)) finds that longer parental leave increases the probability that mothers will return to their jobs after childbirth in the United Kingdom, the United States, and Japan, with the effect particularly strong in Japan.
- Shigeno and Ohkusa (1998) and Suruga and Cho (2003) confirm that women working at companies that support parental leave are more likely to have a baby and return to their jobs; Suruga and Cho (2003) report 22 percent.
- Cross-country results in this chapter indicate that for leave policy to be effective it needs to be longer; this is particularly true for Japan where the probability of finding full-time work after a career interruption is:
  - 18 percent for university educated women (Ueda, 2007)
  - 12 to 13 percent for less educated women (Ueda, 2007)
- Recommendation arising from analysis: consider extending the duration of leave policy to levels similar to those in France, Germany, and the Scandinavian countries (reference to Figure 18 in source).

### Child care supply, fragmentation, and demand
- Usage of child care and early educational services in Japan is still low by international standards.
- The system is fragmented between:
  - Day care centers: full-day child care for working mothers with children between the ages of zero and 6; regulated and funded by the Ministry of Health, Labor, and Welfare.
  - Kindergartens: usually part-day for children ages 3 to 6; regulated and subsidized by the Ministry of Education; intended largely for traditional single-earner households.
- Demand has increased with rising two-earner households, with demand largely outstripping supply.
- The number of wait-listed children remained largely unchanged at about 25,000 children despite a government target to increase capacity from 203,000 to 215,000 children by 2009.
- Informal reports suggest potential unmet demand could be as high as one-third of current child care capacity (Nikkei, 2011).
- Kindergartens remain underutilized (approximately 70 percent of capacity) because the population has aged and more families require full-day child care.

### Time use and barriers to participation
- Having an infant child reduces participation rates by about 30 percent (Waldfogel, Higuchi, and Masahiro (1999)).
- Mothers living with parents or in-laws are more likely to participate in market work (Sasaki (2002)).
- Time men spend on child care is the same regardless of a woman’s work decision (Murakami (2007) and Sakamoto (2008)), implying market work represents an additional burden for women.
- Survey evidence: working hours was the second-highest reason for not participating in the workforce, behind only the additional burden of housework (Table 4).
- In one survey, 87 percent of respondents indicated that if they were to participate in the labor force they would be interested mainly in part-time work.

### Policy recommendations and options
- Extend leave policy duration to levels comparable with France, Germany, and Scandinavian countries to improve return-to-work probabilities.
- Encourage greater male participation in parental leave (current take-up in Japan: fewer than 3 percent of fathers; Sweden: 70 percent).
- Increase supply of child care facilities to reduce women’s child care burdens and support higher participation; policy levers include:
  - Deregulation to encourage private sector entry: current regulations include floor space requirements and stringent licensing processes limiting supply.
  - Merging/unifying day care centers and kindergartens to reduce fragmentation; progress has started but is likely to be slow due to different ministerial oversight responsibilities.
  - Consider a small reallocation of spending toward child care: Japan’s child care spending (as a percent of GDP) is still somewhat lower than in comparator countries (reference to Figure 22 in source).
- Promote more flexible work arrangements:
  - Adopt elements of the Dutch model emphasizing part-time but equal work (equal hourly wages and other full-time benefits such as parental leave and employment protection).
  - Consider laws granting the right to request more flexible working hours, as in the Netherlands (employees with more than one year can change working hours) and Sweden (parents eligible to work shorter hours until their child’s eighth birthday).
- Address the benefit gap between nonregular (part-time) and regular work by either:
  - Closing the benefit gap, or
  - Making regular work more flexible to accommodate caregivers’ needs.

### Tax system and child-rearing allowances (special issues for low-income households)
- Prior to 2004, a head of household could claim both a dependent exemption and a special dependent exemption of ¥380,000 each, as long as the spouse’s annual income was less than ¥1.03 million; ¥1.03 million is often referred to as the “barrier to full-time female employment.”
- Histogram evidence indicates just under one-third of female workers earn less than the ¥1.03 million threshold.
- 2004 reform: one of the special dependent exemptions was eliminated following the Basic Law for a Gender-Equal Society (1999); eliminating both the pension exemption and the other dependent exemption is currently under review.
- Micro-data analyses indicate:
  - Ishizuka (2003): eliminating tax distortions would lead to a small increase in regular full-time employment but a decrease in overall labor force participation.
  - Murakami (2008): the 2004 reforms had no discernible short-term impact on participation choices.
- Child allowances history and changes:
  - Until 2010: monthly ¥5,000 or ¥10,000 child allowances for children in elementary school or below, conditional on income.
  - In 2010: renamed “child-rearing allowance,” increased to ¥13,000 per month, eligibility raised to include junior high school students, and made no longer conditional on income levels; benefits were later reduced for a majority of households.
- Effectiveness of allowances:
  - Allowances are effective primarily for low-income households; if liquidity constrained, an increase in income could lead to higher FLP.
  - In-kind benefits, such as child care, are likely to be more effective.
  - Jaumotte (2003) finds a negative effect from tax benefits, likely due to income effects.
- Equity rationale: child-rearing allowances can lower child poverty; the relative poverty rate for single-parent households with children in Japan was the highest among OECD countries and its proportion is 10 percent higher than in the United States.
- Policy suggestion: consider better targeting child-rearing allowances by conditioning them on income.

### Conclusions related to leave policy and broader FLP goals
- Japan must consider policies to increase the number of career-track female employees and provide better support for working mothers.
- A more flexible work environment combined with better child care facilities and longer leave policies would help reduce the number of women who exit the workforce after childbirth.

*Source: Excerpt from IMF working paper content unit titled “43.      Usage of leave policy has increased” (provided content).*

### APPENDIX I. ADDITIONAL TABLES AND FIGURES

### _wp12248 - APPENDIX I. ADDITIONAL TABLES AND FIGURES

### Dataset composition and number of observations (Table 5)
- Countries covered include: AUS, AUT, BEL, CAN, CHE, DEU, DNK, ESP, FIN, FRA, GBR, GRC, IRL, ITA, JPN, LUX, NLD, NOR, NZL, PRT, SWE, USA.
- Total number of observations across variables: 
  - FLP: 770
  - Marriage rate: 797
  - Number of children per woman: 1091
  - Education: 1122
  - Wage gap: 1078
  - Family allowance: 1078
  - Leave: 1121
  - Child benefits: 814
  - Part-time incidence: 509
  - Childcare per child: 536
  - Tax wedge: 342
- Sources: OECD and Gauthier (2010 and 2011).

### Latest available country-level summary statistics (Table 6)
- Selected country-level indicators (latest data available); each entry shows the listed variable value as in Table 6 (examples):
  - AUS: FLP 75.41, Marriage rate 5.40, Number of children per woman 0.57, Education (Year) 12.30, Wage gap 5.10, Family Allowance 3.70, Leave (Week) 52, Child benefits 47.07, Part-time incidence 18.28, Childcare per child (PPP-USD) 2047, Tax wedge 1.43
  - AUT: FLP 81.53, Marriage rate 4.33, Number of children per woman 0.45, Education (Year) 8.72, Wage gap 23.97, Family Allowance 3.43, Leave 112, Child benefits 49.82, Part-time incidence 16.80, Childcare per child 2844, Tax wedge 0.82
  - BEL: FLP 78.99, Marriage rate 4.28, Number of children per woman 0.52, Education (Year) 10.45, Wage gap 14.86, Family Allowance 2.71, Leave 84, Child benefits 81.17, Part-time incidence 17.43, Childcare per child 4494, Tax wedge 1.16
  - CAN: FLP 81.97, Marriage rate 4.60, Number of children per woman 0.49, Education (Year) 11.40, Wage gap 34.28, Family Allowance 1.59, Leave 50, Child benefits 40.92, Part-time incidence 11.69, Childcare per child 964, Tax wedge 1.29
  - CHE: FLP 83.59, Marriage rate 5.34, Number of children per woman 0.45, Education (Year) 9.50, Wage gap 23.44, Family Allowance 2.47, Leave 14, Child benefits 63.98, Part-time incidence 24.41, Childcare per child 2267, Tax wedge 1.08
  - DEU: FLP 80.51, Marriage rate 4.48, Number of children per woman 0.41, Education (Year) 11.84, Wage gap 23.70, Family Allowance 3.40, Leave 162, Child benefits 90.30, Part-time incidence 20.97, Childcare per child 2596, Tax wedge 1.46
  - DNK: FLP 86.98, Marriage rate 6.70, Number of children per woman 0.56, Education (Year) 9.97, Wage gap 14.14, Family Allowance 1.77, Leave 36, Child benefits 45.48, Part-time incidence 9.49, Childcare per child 8572, Tax wedge 1.61
  - ESP: FLP 74.73, Marriage rate 4.61, Number of children per woman 0.43, Education (Year) 9.90, Wage gap 22.98, Family Allowance 0.59, Leave 172, Child benefits 21.50, Part-time incidence 9.87, Childcare per child 2575, Tax wedge 0.86
  - FIN: FLP 85.89, Marriage rate 5.58, Number of children per woman 0.51, Education (Year) 10.24, Wage gap 16.03, Family Allowance 1.98, Leave 196, Child benefits 23.79, Part-time incidence 6.35, Childcare per child 5274, Tax wedge 0.44
  - FRA: FLP 83.25, Marriage rate 4.21, Number of children per woman 0.56, Education (Year) 9.91, Wage gap 14.79, Family Allowance 1.59, Leave 162, Child benefits 34.36, Part-time incidence 11.67, Childcare per child 5511, Tax wedge 0.52
  - GBR: FLP 78.29, Marriage rate 5.23, Number of children per woman 0.53, Education (Year) 9.31, Wage gap 19.94, Family Allowance 1.90, Leave 65, Child benefits 30.47, Part-time incidence 18.60, Childcare per child 4910, Tax wedge 1.26
  - GRC: FLP 69.45, Marriage rate 5.16, Number of children per woman 0.43, Education (Year) 9.90, Wage gap 18.03, Family Allowance 0.30, Leave 34, Child benefits 45.70, Part-time incidence 7.45, Childcare per child 613
  - IRL: FLP 71.83, Marriage rate 5.13, Number of children per woman 0.61, Education (Year) 11.60, Wage gap 15.97, Family Allowance 2.99, Leave 30, Child benefits 60.05, Part-time incidence 18.70, Childcare per child 1304, Tax wedge 3.07
  - ITA: FLP 65.24, Marriage rate 4.21, Number of children per woman 0.43, Education (Year) 8.78, Wage gap 18.02, Family Allowance 3.54, Leave 48, Child benefits 39.73, Part-time incidence 15.12, Childcare per child 3474, Tax wedge 0.78
  - JPN: FLP 70.30, Marriage rate 5.80, Number of children per woman 0.42, Education (Year) 11.02, Wage gap 23.15, Family Allowance 4.64, Leave 66, Child benefits 26.20, Part-time incidence 15.20, Childcare per child 2151, Tax wedge 0.88
  - LUX: FLP 72.93, Marriage rate 4.10, Number of children per woman 0.54, Education (Year) 9.67, Wage gap 27.10, Family Allowance 4.80, Leave 164, Child benefits 115.52, Part-time incidence 13.62, Childcare per child 4228
  - NLD: FLP 81.61, Marriage rate 4.49, Number of children per woman 0.53, Education (Year) 10.81, Wage gap 17.83, Family Allowance 1.86, Leave 58, Child benefits 37.01, Part-time incidence 28.92, Childcare per child 7662, Tax wedge 1.00
  - NOR: FLP 85.59, Marriage rate 4.98, Number of children per woman 0.59, Education (Year) 12.69, Wage gap -6.93, Family Allowance 1.98, Leave 158, Child benefits 34.90, Part-time incidence 13.28, Childcare per child 6287, Tax wedge 0.98
  - NZL: FLP 77.42, Marriage rate 5.13, Number of children per woman 0.61, Education (Year) 12.30, Wage gap 18.64, Family Allowance 4.12, Leave 14, Child benefits 47.53, Part-time incidence 17.06, Childcare per child 2689, Tax wedge 1.01
  - PRT: FLP 82.95, Marriage rate 4.37, Number of children per woman 0.45, Education (Year) 7.27, Wage gap 30.91, Family Allowance 0.92, Leave 30, Child benefits 29.44, Part-time incidence 5.53, Childcare per child 1301, Tax wedge 0.46
  - SWE: FLP 87.55, Marriage rate 5.24, Number of children per woman 0.52, Education (Year) 11.84, Wage gap 10.13, Family Allowance 2.13, Leave 147, Child benefits 26.84, Part-time incidence 9.25, Childcare per child 10334, Tax wedge 0.60
  - USA: FLP 75.81, Marriage rate 7.30, Number of children per woman 0.60, Education (Year) 12.46, Wage gap 20.03, Family Allowance 0.00, Leave 12, Child benefits 49.71, Part-time incidence 7.31, Childcare per child 1668, Tax wedge 1.94
- OECD 22 Mean (as reported): FLP 78.72, Marriage rate 5.03, Number of children per woman 0.51, Education 10.54, Wage gap 18.46, Family Allowance 2.38, Leave 84.82, Child benefits 47.34, Part-time incidence 14.41, Childcare per child 3807.58, Tax wedge 1.13
- JPN / OECD 22 Mean standardized differences reported in Table 6 (Japan / OECD 22 Mean): -0.11, 0.95, -1.34, 0.34, 0.54, 1.69, -0.31, -0.91, 0.13, -0.64, -0.42
- G5 Mean (Canada, France, Germany, United Kingdom, United States) reported: FLP 79.96, Marriage rate 5.16, Number of children per woman 0.52, Education 10.99, Wage gap 22.55, Family Allowance 1.70, Leave 90.20, Child benefits 49.15, Part-time incidence 14.05, Childcare per child 3129.95, Tax wedge 1.29
- JPN / G5 Mean standardized differences reported: -3.26, 0.51, -1.28, 0.03, 0.08, 2.44, -0.35, -0.95, 0.21, -0.49, -0.80
- Notes: 1/ (Japan-OECD 20 Mean) / SD. 2/ G5: Canada, France, Germany, United Kingdom, United States. 3/ (Japan-G5 Mean) / SD.

### Correlation structure among variables (Table 7)
- Level correlations (selected):
  - Correlation(FLP, Marriage rate) = -0.28
  - Correlation(FLP, Log(Number of children per woman)) = -0.46
  - Correlation(FLP, Log(Education)) = 0.44
  - Correlation(FLP, Wage gap) = -0.34
  - Correlation(FLP, Log(Family allowance)) = 0.26
  - Correlation(FLP, Log(Leave)) = 0.35
  - Correlation(FLP, Log(Child benefits)) = 0.05
  - Correlation(FLP, Part-time incidence) = 0.19
  - Correlation(FLP, Log(Childcare per child)) = 0.65
  - Correlation(FLP, Log(Tax wedge)) = 0.08
- Three-year change correlations (selected):
  - Correlation(FLP-DLP, Marriage rate) = -0.12
  - Correlation(FLP-DLP, Log(Number of children per woman)) = -0.27
  - Correlation(FLP-DLP, Log(Education)) = 0.23
  - Correlation(FLP-DLP, Part-time incidence) = 0.42
  - Correlation(FLP-DLP, Log(Tax wedge)) = 0.11
- Ten-year average correlations (selected):
  - Correlation(FLP-DLP, Marriage rate) = -0.39
  - Correlation(FLP-DLP, Log(Number of children per woman)) = -0.58
  - Correlation(FLP-DLP, Log(Education)) = 0.46
  - Correlation(FLP-DLP, Wage gap) = -0.44
  - Correlation(FLP-DLP, Log(Leave)) = 0.37
  - Correlation(FLP-DLP, Log(Childcare per child)) = 0.58
  - Correlation(FLP-DLP, Log(Tax wedge)) = 0.12
- Source: Fund staff calculations.

### Regression results: Gap between FLP and MLP and demographic/policy variables (Table 8 and Table 9)
- Instrumentation and model notes:
  - Number of children per woman is instrumented in all relevant columns.
  - Part-time incidence is instrumented in columns 7, 13, 14, and 15 (Table 8) and in columns 6, 11, 12, and 13 (Table 9) as noted.
  - Structural break terms introduced for Germany (1991), Ireland (1985), Japan (1968), Netherlands (1987), and Portugal (1978).
  - Clustered standard errors in brackets; significance levels: *** p<0.01, ** p<0.05, * p<0.1.
- Key coefficient magnitudes and significance (selected from Tables 8 and 9; coefficients are reported exactly as in the tables):
  - Log(Number of children per woman): consistently negative and highly significant across specifications (examples):
    - Table 8 (column examples): -16.345***, -16.533***, -15.625***, -16.103***, -21.012***, -26.522***, -19.142*...
    - Table 9 (examples): -24.070***, -24.004***, -23.321***, -23.574***, -24.084***, -24.939**, -21.625***...
  - Log(Education): positive and often significant (examples):
    - Table 8 (examples): 10.765***, 10.826***, 13.568***, 10.076***, 9.015***...
    - Table 9 (examples): 6.549**, 6.484**, 9.340***, 6.165**, 6.573**...
  - Log(Family allowance): positive and sometimes significant; examples: 3.995, 4.240, 5.061**, 10.134***, 9.715***, 10.889***...
  - Log(Leave): negative and often significant with a positive Squared Log(Leave) term:
    - Log(Leave) examples: -0.934***, -0.948***, -0.726**, -1.016**, -1.096**, -1.726***...
    - Squared Log(Leave) examples: 0.158**, 0.162**, 0.137*, 0.228***, 0.245***, 0.320***...
  - Part-time Incidence: positive and significant in some specifications (examples): 0.649***, 0.643***; in Table 9 part-time incidence coefficients include 0.576***, 0.594***, 0.632***, 0.537***.
  - Log(Tax wedge): large negative coefficients in some specifications (example from Table 8): -20.821***; interaction Log(Tax wedge) * Log(Education) positive and significant in some specifications (examples): 9.056***, 13.276***.
  - Observations vary by specification (examples): 229, 214, 146, 124, 90; Table 9 observations examples: 184, 103, 135, 85.
  - F-test (Number of children per woman) reported across columns (examples): 217.76, 215.64, 176.45, 223.38, 230.31, 40.49...
  - F-test (Part-time incidence) examples: 24.64, 26.86, 13.24, 6.16 (Table 8); 14.94, 14.75, 12.14, 7.08 (Table 9).
- Table 10 (Regression of FLP on demographic and policy variables) reports multiple specifications with Log(Family allowance) * Log(Education) included; selected coefficients (examples):
  - Log(Number of children per woman): -16.740***, -16.734***, -16.346***, -16.940***, -20.372***...
  - Log(Education): 7.728*, 7.725*, 9.664**, 7.079*...
  - Log(Leave): -0.774***, -0.802***, -0.819***...
  - Part-time incidence: 0.576***, 0.594***, 0.632***, 0.537*** (when included).
  - Log(Tax wedge) and its interaction with Log(Education) reported in some specifications: Log(Tax wedge) -16.028**, interaction 6.778** (Table 10 examples).
- Notes: All tabulated coefficient values and standard errors should be read from the tables for exact specification-by-specification estimates.

### Marginal effects (Table 11)
- Effects on FLP by one standard-deviation change in ten-year average variables (reported exactly):
  - Marriage rate: Marginal Effect -4.55; Mean 6.08; S.D. 1.33; Effect by one S.D. Change (In percentage point) -6.06
  - Log(Number of children per woman): Marginal Effect -43.85; Mean 4.16; S.D. 0.23; Effect by one S.D. Change -10.11
  - Log(Education): Marginal Effect 27.86; Mean 2.12; S.D. 0.32; Effect by one S.D. Change 8.83
  - Wage gap: Marginal Effect -0.79; Mean 27.03; S.D. 9.59; Effect by one S.D. Change -7.57
  - Log(Family allowance): Marginal Effect 7.28; Mean 1.20; S.D. 0.60; Effect by one S.D. Change 4.40
  - Log(Leave): Marginal Effect 3.90; Mean 3.86; S.D. 1.54; Effect by one S.D. Change 6.01
  - Log(Child benefits): Marginal Effect -1.04; Mean 4.20; S.D. 0.52; Effect by one S.D. Change -0.54
  - Part-time incidence: Marginal Effect 0.52; Mean 13.00; S.D. 5.54; Effect by one S.D. Change 2.87
  - Log(Childcare per child): Marginal Effect 4.90; Mean 6.93; S.D. 1.56; Effect by one S.D. Change 7.62
  - Log(Tax wedge): Marginal Effect 4.03; Mean 4.68; S.D. 0.35; Effect by one S.D. Change 1.40

### Figures and graphical diagnostics (Figures 28–33)
- Figures include scatter plots of each variable in levels (Figure 28), scatter plots of changes (Figure 29), scatter plots of ten-year mean variables at level (Figure 30), within-variable explanation plots (Figure 31), cross-section explanation plots (Figure 32), and cross-section explanation using ten-year average (Figure 33).
- Graphical displays label country observations and show the distribution of FLP-MLP against each explanatory variable (examples of x-axis scales used in figures):
  - Marriage Rate axes: e.g., 4 6 8 10 12 in Figure 28
  - Log(Number of Children per Woman) axes: e.g., 3.6 3.8 4.2 4.4 4.6 in Figure 28
  - Log(Education) axes: e.g., 11.5 22.5 in Figure 28
  - Wage Gap axes: e.g., -200 20 40 60 in Figure 28
  - Log(Family Allowance), Log(Leave), Log(Child Benefits), Part-time Incidence, Log(Childcare per Child), Log(Tax Wedge) axes shown with ranges as in the figures.
- Figures 31–33 present country-by-country contribution decompositions for FLP and explanatory variables (within-variable and cross-section explanations), with axes showing percentage-point contributions (e.g., -100 to 50, -15 to 15) and country labels AUS, AUT, BEL, CAN, CHE, DEU, DNK, ESP, FIN, FRA, GBR, GRC, IRL, ITA, JPN, LUX, NLD, NOR, NZL, PRT, SWE, USA.

*Source: Fund staff calculations (tables and figures as presented in APPENDIX I).*

### APPENDIX II. DEFINITION AND SOURCES OF DATA

### APPENDIX II. DEFINITION AND SOURCES OF DATA

### 1. Dependent Variable
- Total female prime-age labor force participation rates:
  - Definition: Sum of unemployed and employed female workers as a share of the female labor force ages 25–54, in percent (0–100).
  - Mathematical expression (as presented in source):  
    ܲܮܨ  
    ൌ  
    ݏݎ݁݇ݎ݋ܹ ݈݁ܽ݉݁ܨ ݀݁ݕ݋݈݌݉ܧ54൅ ݋ݐ 25 ݏ݁݃ܣ ݏݎ݁݇ݎ݋ܹ ݈݁ܽ݉݁ܨ ݀݁ݕ݋݈݌ܷ݉݁݊54 ݋ݐ 25 ݏ݁݃ܣ   
    54 ݋ݐ 25 ݏ݁݃ܣ ݁ܿݎ݋ܨ ݎ݋ܾܽܮ ݈݁ܽ݉݁ܨ  
    ൌ100
  - Available Period: 1960–2008.
  - Source: Organization for Economic Cooperation and Development, Database on Labour Force Statistics.

### 2. Demographic Variables
- Crude marriage rate:
  - Definition: Annual number of new marriages divided by population, in thousands.
  - Mathematical expression (as presented in source):  
    ൌ݁ݐܴܽ ݁݃ܽ݅ݎݎܽܯ  
    ݊݋݅ݐ݈ܽݑ݌݋ܲ ݀݁݅ݎݎܽܯ ݕ݈ݓ݁ܰ  
    ݊݋݅ݐ݈ܽݑ݌݋ܲ ݈ܽݐ݋ܶ  
    ൌ1,000
  - Available Period: 1970–2007.
  - Source: Organization for Economic Cooperation and Development, Society at a Glance, 2009.

- Number of Children per Woman:
  - Definition: Total population ages 0–14 divided by female population ages 15–64.
  - Mathematical expression (as presented in source):  
    ݃݋ܮ  
    ሺ  
    ܥ ݂݋ ݎܾ݁݉ݑ݄ܰ݊ܽ݉݋ܹ ݎ݁݌ ݊݁ݎ݈݀݅  
    ሻ  
    ሺ ݃݋ܮൌ  
    ܥ  
    ݄14 ݋ݐ 0 ݏ݁݃ܣ ݊݋݅ݐ݈ܽݑ݌݋ܲ ݈݀݅  
    64 ݋ݐ 15 ݏ݁݃ܣ ݊݋݅ݐ݈ܽݑ݌݋ܲ ݁܈ܽ݉݁ܨ  
    ሻ
  - Available Period: 1960–2008.
  - Source: Organization for Economic Cooperation and Development, Annual Labour Force Statistics.
  - Data adjustments: Some countries have a value of less than 1, so when transformed to a logarithmic scale, each value is multiplied by 100.

- Female education:
  - Definition: Average years of education of female population over age 25.
  - Available Period: 1960–2008.
  - Source: Barro and Lee (2010), Educational Attainment Dataset.
  - Data adjustments: Missing observations are obtained by linear interpolation when possible.

### 3. Policy Variables
- Gender wage gap:
  - Definition: Hourly wage gap in manufacturing between male and female workers in percentage of male wage.
  - Mathematical expression (as presented in source):  
    ൌ݌ܽܩ ܹ݁݃ܽ ݎ݁݀݊݁ܩ  
    ܹ݁݃ܽ ݕ݈ݎݑ݋ܪ ݈݁ܽ݉݁ܨ ܹ݁݃ܽ ݕ݈ݎݑ݋ܪ ݈݁ܽܯ  
    ܹ݁݃ܽ ݕ݈ݎݑ݋ܪ ݈݁ܽܯ  
    ൌ100
  - Available Period: 1960–2008.
  - Source: Gauthier (2010), Comparative Family Benefits Database, 1960–2008.

- Family allowances:
  - Definition: Monthly family allowances for children (assuming a two-child family) in purchasing-power-parity-adjusted U.S. dollars.
  - Available Period: 1960–2008.
  - Source: Gauthier (2010), Comparative Family Benefits Database, 1960–2008.
  - Data adjustments: Some countries have a value of zero, so when changed into a logarithmic scale, each value is transformed as follows (as presented in source):  
    ݃݋ܮ  
    ሺ  
    ݁ܿ݊ܽݓ݋݈݈ܣ ݕ݈݅݉ܽܨ  
    ሻ  
    ൅݁ܿ݊ܽݓ݋݈݈ܣ ݕ݈݅݉ܽܨሺ ݃݋ܮൌ  
    ሺ  
    ݁ܿ݊ܽݓ݋݈݈ܣ ݕ݈݅݉ܽܨ  
    ଶ  
    ൅1  
    ሻ  
    ଴.ହ  
    ሻ

- Number of parental leave weeks:
  - Definition: Maximum number of weeks a mother may take after the birth of a first child as maternity leave, parental leave, and child care leave.
  - Mathematical expression (as presented in source):  
    ܥ൅݁ݒܽ݁ܮ ݈ܽݐ݊݁ݎܽܲ൅݁ݒܽ݁ܮ ݕݐ݅݊ݎ݁ݐܽܯൌ݁ݒܽ݁ܮ݄݁ݒܽ݁ܮ ݁ݎܽܥ ݈݀݅
  - Available Period: 1960–2008.
  - Source: Gauthier (2011), Comparative Family Benefits Database, 1960–2008.
  - Data adjustments: Some countries have a value of zero, so when changed into a logarithmic scale, each value is transformed as follows (as presented in source):  
    ݃݋ܮ  
    ሺ  
    ݁ݒܽ݁ܮ  
    ሻ  
    ൅݁ݒܽ݁ܮሺ ݃݋ܮൌ  
    ሺ  
    ݁ݒܽ݁ܮ  
    ଶ  
    ൅1  
    ሻ  
    ଴.ହ  
    ሻ

- Child benefits:
  - Definition: Child benefits were calculated by subtracting the disposable income (after taxes and transfers) of a one-earner, two-parent, two-child family from that of a comparable childless single earner, converted to purchasing-power-parity-adjusted U.S. dollars.
  - Available Period: 1972–2008.
  - Source: Gauthier (2010), Comparative Family Benefits Database, 1960–2008.
  - Data adjustments: Some countries have a value of zero, so when changed into a logarithmic scale, each value is transformed as follows (as presented in source):  
    ݃݋ܮ  
    ሺ  
    ܥ  
    ݄ݏݐ݂݅݁݊݁ܤ ݈݀݅  
    ሻ  
    ܥሺ ݃݋ܮൌ݄൅ݏݐ݂݅݁݊݁ܤ ݈݀݅  
    ሺ  
    ܥ ݈݄݀݅ݐ݂݅݁݊݁ܤ  
    ଶ  
    ൅1  
    ሻ  
    ଴.ହ  
    ሻ

- Part-time incidence:
  - Definition: Part-time employment as a share of prime-age employment (25–54), in percent (0–100). Part-time employment is based on the Organization for Economic Cooperation and Development typical 30-hour minimum for full-time work.
  - Mathematical expression (as presented in source):  
    ൌ݁ܿ݊݁݀݅ܿ݊ܫ ݁݉݅ܶ‐ݐݎܽܲ  
    ݐ݊݁݉ݕ݋݈݌݉ܧ ݁݉݅ܶ‐ݐݎܽܲ  
    ݐ݊݁݉ݕ݋݈݌݉ܧ ݈ܽܐ݋ܶ  
    ൌ100
  - Available Period: 1976–2008.
  - Source: Organization for Economic Cooperation and Development, Database on Labour Force Statistics.

- Public expenditures on child care:
  - Definition: Public spending on formal child care in purchasing-power-parity-adjusted U.S. dollars is divided by the child population ages 0 to 4.
  - Mathematical expression (as presented in source):  
    ݃݋ܮ  
    ሺ  
    ܥ݄ܥ ݎ݁݌ ݁ݎܽܥ ݈݈݄݀݅݀݅  
    ሻ  
    ሺ ݃݋ܮൌ  
    ܥ  ݈ܽ݉ݎ݋ܨ ݊݋ ݃݊݅݀݊݁݌ܵ ݈ܾܿ݅ݑܲ ݈ܽܐ݋ܶ  
    ݄݁ݎܽܥ ݈݀݅  
    4 ݋ݐ 0 ݏ݁݃ܣ ݊݋݅ݐ݈ܽݑ݌݋ܲ  
    ሻ
  - Available Period: 1980–2007.
  - Source: Main data sources are the Organization for Economic Cooperation and Development, Social Expenditures Database. The population of children ages 0–4 is from the OECD Statistical Profiles 2010.

- Relative marginal tax rates on second earners:
  - Definition: Ratio of the marginal tax rate on the second earner to the tax wedge for a single-earner couple with two children earning 100 percent of average production worker (APW) earnings. The marginal tax rate on the second earner is in turn defined as the share of the wife’s earnings that goes into paying additional household taxes.
  - Mathematical expression (as presented in source):  
    ݃݋ܮ  
    ሺ  
    ݎ݁݊ݎܽ݁ ݀݊݋ܿ݁ݏ ݔܽܶ  
    ሻ  
    ሺ1െ ݃݋ܮൌ  
    ሺ  
    ݁݉݋ܿ݊ܫ ݐ݁ܰ ݈݀݋݄݁ݏݑ݋ܪ  
    ሻ  
    ஻  
    ൌ  
    ሺ  
    ݁݉݋ܿ݊ܫ ݐ݁ܰ ݈݀݋݄݁ݏݑ݋ܪ  
    ሻ  
    ஺  
    ሺ  
    ݁݉݋ܿ݊ܫ ݏݏ݋ݎܩ ݈݀݋݄݁ݏݑ݋ܪ  
    ሻ  
    ஻  
    ൌ  
    ሺ  
    ݁݉݋ܿ݊ܫ ݏݏ݋ݎܩ ݈݀݋݄݁ݏݑ݋ܪ  
    ሻ  
    ஺  
    ሻ
  - Notes from source text: A denotes a situation in which the wife does not earn any income and B denotes a situation in which the wife’s gross earnings are X% of APW earnings. Two different tax rates are calculated, depending on whether the wife is assumed to work full-time (X = 67 percent) or part-time (X = 33 percent). In all cases it is assumed that the husband earns 100 percent of APW earnings and that the couple has two children. The difference between gross and net income includes income taxes, an employee’s social security contribution, and universal cash benefits.
  - Available Period: 1982–2003.
  - Source: Bassanini and Duval (2006).
  - Data adjustments: Some countries have a value of less than 1, so when transformed to a logarithmic scale, each value is multiplied by 100.

*Source: _wp12248 - APPENDIX II. DEFINITION AND SOURCES OF DATA*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12248.pdf_