## wpiea2021138-print-pdf - Section 6 concludes

## Source details

**Canonical URL:** [wpiea2021138-print-pdf - Section 6 concludes](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021138-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021138-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021138-print-pdf.pdf.json)

---

### Related literature and contribution
- Connects to literature linking structural transformation to gender inequality in labor markets (Akbulut, 2011; Ngai and Petrongolo, 2017; Rendall, 2018; Rendall, 2013; Ostry, Alvarez, Espinoza, and Papageorgiou, 2018).
- Fills a gap: no systematic studies linking structural transformation and gender inequality in China; this study documents facts and calibrates a model of structural transformation for China.
- Relates to literature estimating female labor supply elasticities and the role of cultural norms (Blau and Kahn, 2007; Heim, 2007; Chen and Ge, 2018). Documents a reverse trend in China—female labor supply elasticities shifted in the opposite direction relative to US evidence.
- Contributes to literature on economic growth, globalization, and income inequality in China by analyzing drivers of rising gender inequality from macro and micro perspectives and quantifying policy gains using a structural transformation framework.
- Notes: service sector labor productivity and TFP growth slowed since the Global Financial Crisis; rapid population aging likely exacerbates gender inequality through disproportionate elderly care burdens on women.

### Stylized facts and anomalies (structural transformation and gender gaps)
- Fact 1: Declining labor force participation (LFP) and widening gender LFP gap (ILO, population aged 15+ , 2000–2019)
  - Gender LFP gap: about 11 percentage points in 2000; widened to about 14 percentage points by 2013 and remained at that level.
  - Both men’s and women’s LFP rates decline after state-owned enterprise reforms; female LFP declines significantly faster, especially after 2010.
  - Women experienced a decline in average paid work hours relative to men.
- Fact 2: Widening conditional gender wage gap (CHIP microdata, working-age 15–64, non-zero wage incomes; compare 1995 and 2013)
  - Conditional gender wage gap increased from 12 percent in 1995 to 35 percent in 2013.
  - Selected regression coefficients (1995 sample / 2013 sample):
    - Female: -0.119 ∗∗∗ / -0.351 ∗∗∗ (standard errors (0.017) (0.021))
    - Age: 0.056 ∗∗∗ / 0.078 ∗∗∗ (standard errors (0.014) (0.006))
    - (Age)2: -0.001 ∗∗∗ / -0.001 ∗∗∗ (standard errors (0.000) (0.000))
    - Child in HH: -0.060 ∗ / -0.081 ∗∗∗ (standard errors (0.034) (0.020))
    - Education (College): 0.402 ∗∗∗ / 0.570 ∗∗∗ (standard errors (0.026) (0.043))
    - Service sector: 0.059 ∗∗ / -0.011 (standard errors (0.026) (0.021))
    - Observations: 13,015 / 40,620
    - R-squared: 0.095 / 0.152
  - Dependent variable = real annual log wages. Omitted education = “Less than High School.” Omitted sector = goods sector.
- Fact 3: Rising service sector share (ILO sectoral employment by gender, 2000–2019)
  - Reallocation: female (male) goods sector share decreased from 73 (75) percent in 1995 to 52 (57) percent in 2019.
- Puzzle: Structural transformation (declining goods share, rising services) typically reduces gender inequality in other economies, yet in China gender inequality widened.

### Growth–FLFP theory and province-level evidence
- Theoretical U-shaped relationship between economic growth and FLFP (Sinha, 1965; Goldin, 1994; Mammen and Paxson, 2000; Olivetti, 2013; Olivetti and Petrongolo, 2016).
- Cultural norms mediate the growth–FLFP link (Jayachandran, 2020); evidence in China shows male attitudes toward wives’ work matter (Chen and Ge, 2018).
- Empirical: province-level variation (1995–2013) shows a negative correlation between output per capita and female employment shares—richer provinces have lower female employment rates.

### Female labor supply elasticities — micro evidence and interpretation
- Focus: married vs unmarried women; sample for elasticity estimation = married women aged 15–64 residing in urban areas (exclude students, retired, disabled).
- Gender hours and earnings gaps (Table 2), 1995 / 2013:
  - Gender Hours Gap:
    - Married: -0.027 ∗∗∗ / -0.033 ∗ (standard errors (0.007) (0.017))
    - Unmarried: -0.006 / 0.066 ∗∗∗ (standard errors (0.012) (0.020))
    - Rural: -0.027 / -0.024 (standard errors (0.023) (0.019))
    - Urban: -0.019 ∗∗∗ / -0.053 ∗∗∗ (standard errors (0.006) (0.008))
  - Gender Earnings Gap:
    - Married: -0.153 ∗∗∗ / -0.395 ∗∗∗ (standard errors (0.022) (0.023))
    - Unmarried: -0.024 / -0.090 ∗∗∗ (standard errors (0.036) (0.019))
    - Rural: -0.073 / -0.357 ∗∗∗ (standard errors (0.064) (0.026))
    - Urban: -0.142 ∗∗∗ / -0.314 ∗∗∗ (standard errors (0.017) (0.020))
- Three-stage IV estimation (Heckman selection correction; instruments include experience quadratic and county public-sector share) — selected coefficients and elasticities for married women:
  - Log(Own Wage) coefficients:
    - 1995 OLS: 306.941 ∗∗ (139.178)
    - 1995 Three-Stage IV: 110.385 (167.927)
    - 2013 OLS: 939.414 ∗∗∗ (181.013)
    - 2013 Three-Stage IV: 699.724 ∗∗∗ (172.555)
  - Log(Spouse Wage) coefficients:
    - 1995 OLS: -112.564 ∗∗∗ (13.397)
    - 1995 Three-Stage IV: -161.875 (163.813)
    - 2013 OLS: -188.730 ∗∗∗ (19.120)
    - 2013 Three-Stage IV: -398.227 ∗∗∗ (116.162)
  - Log(Non-wage HH Income):
    - 1995 OLS: -3.205 (8.249)
    - 1995 Three-Stage IV: -0.431 (9.647)
    - 2013 OLS: -7.087 ∗∗∗ (2.702)
    - 2013 Three-Stage IV: -18.532 ∗∗∗ (4.215)
  - Child in HH:
    - 1995 OLS: 3.211 (13.232)
    - 1995 Three-Stage IV: 25.908 (30.647)
    - 2013 OLS: -39.564 ∗ (21.678)
    - 2013 Three-Stage IV: -86.595 ∗∗ (34.640)
  - Observations: 1995 OLS/IV: 3,768 / 3,768; 2013 OLS/IV: 5,604 / 5,604.
  - Elasticities (at mean annual hours):
    - Own Log Wage:
      - 1995 OLS: 0.140
      - 1995 Three-Stage IV: 0.050
      - 2013 OLS: 0.407
      - 2013 Three-Stage IV: 0.303
    - Spouse Log Wage:
      - 1995 OLS: -0.051
      - 1995 Three-Stage IV: -0.074
      - 2013 OLS: -0.082
      - 2013 Three-Stage IV: -0.173
- Key empirical interpretation:
  - Female own-wage elasticity increased from 0.05 in 1995 to 0.3 in 2013.
  - Spouse’s wage elasticity declined from -0.07 to -0.17 between 1995 and 2013.
  - Presence of a child: in 1995 the child indicator is not statistically significant; by 2013 presence of a child has a negative and statistically significant effect on women’s labor hours.
  - Indicates a shift toward a one-earner household model between 1995 and 2013, with falling prevalence of dual-earner households driven by rising incomes and stronger child-related care constraints in urban areas.
  - Evidence of rising assortative matching in education from 1995 to 2013, contributing to widening gender wage and hours gaps.

### Model of structural transformation — framework, calibration, and mechanisms
- Framework: adaptation of Ngai and Petrongolo (2017) with female and male workers allocating time to goods (j = g), market services (j = s), or non-market home sector.
  - Production: Y_j = A_j L_j, where L_j = [ξ_j L_fj^{(η−1)/η} + (1−ξ_j) L_mj^{(η−1)/η}]^{η/(η−1)}.
  - Female comparative advantage in services: ξ_s > ξ_g.
  - Productivity growth rates: γ_g > γ_s; market services productivity grows faster than home services (γ_h < γ_s).
- Preferences and home services:
  - Utility: U(c_g, c_s, c_h) = lnc, with c a CES composite; impose  < 1 and σ > 1 (services and goods are complements; home- and market-produced services are substitutes).
  - Home production: c_h = A_h [ξ_h L_fh^{(η−1)/η} + (1−ξ_h) L_mh^{(η−1)/η}]^{η/(η−1)}.
- Productivity wedges (sector-specific barriers): ξ_j = π_j χ_j, j = g,s. Wedges π_j < 1 reduce women’s productivity in sector j and distort gender wage ratios.
- Calibration targets and parameter values:
  - Targets: changes in sectoral employment shares, female-to-male wage ratio, gender-specific market and home production hours between 2000 and 2013.
  - Data sources: ILO, CHIP, CHNS, World Bank WDI, Bridgman et al. (2018).
  - Calibrated parameters:
    - γ_g − γ_s = 0.037
    - γ_s − γ_h = 0.041
    - σ = 2.0
    -  = 0.002
    - η = 2.27
    - χ_g = 0.29
    - χ_s = 0.43
    - L_m/L_f = 1.05
    - ξ_h = 0.48

### Estimated sector-specific barriers (productivity wedges)
- Sector-specific barriers π_j (matched to hours and wage ratios, 2000 and 2013):
  - π_g: 2000 = 1.27; 2013 = 1.03
  - π_s: 2000 = 0.86; 2013 = 0.70
- Interpretation:
  - Wedges widened in the service sector between 2000 and 2013; implied increase in barriers to female workers’ entry into goods and service sectors by about one third between 2000 and 2013.
  - Increasing barriers can reverse expected narrowing of gender wage and market hours gaps associated with structural transformation.

### Counterfactual exercises and quantitative results (Section 5.2)
- Counterfactual 1: Remove productivity wedges (πg = πs = 1; US 2004–2008 benchmark for χg and χs)
  - Model table entries for Counterfactual 1: 77.9 and 72.8 (table entries preserved verbatim).
  - Narrative results:
    - Matching comparative advantage parameters to the US, China’s unconditional wage gap would be reduced from about 40 percent to about 22 percent.
    - Reduction in labor market barriers would result in a reduction in the market hours gap by almost 32 percentage points.
- Counterfactual 2: Accelerate marketization of services ((γs − γh) increased from 4.1 percentage points to 6.2 percentage points)
  - Model table entries for Counterfactual 2: 78.1 and 77.7 (table entries preserved verbatim).
  - Narrative results:
    - Faster marketization has a relatively small additional effect on the gender wage ratio.
    - Faster marketization results in a 5 percentage points decline in the gender market hours gap as women exit home production and enter the market sector.
- Baseline and model fit to data moments:
  - Data moments:
    - Wage Gap: 2000 = 84.4; 2013 = 64.9
    - Market Hours Gap: 2000 = 46.6; 2013 = 41.3
  - Model baseline:
    - Wage Gap: 2000 = 84.4; 2013 = 69.1
    - Market Hours Gap: 2000 = 46.6; 2013 = 35.9
- Aggregate output effect:
  - Narrowing gender gaps in hours and wages to the levels described in Table 6 would result in 4.9 percent higher output.

### Policy dimensions, institutional drivers, and recommended interventions
- Identified institutional/policy distortions:
  - Decline in state childcare support increased time women spend on child care.
  - Nursery fees have risen to around 15–20 percent of a middle-class family’s annual income.
  - Job advertisements and hiring practices show rising gender-based hiring biases despite legal prohibitions.
  - Motherhood penalty persists; protections for pregnant women and mothers often not effectively enforced.
  - Female underrepresentation in professional and managerial jobs remains large (women’s share in management was still one of four in 2010).
  - Differential retirement ages by gender (women around 50 and 55 in blue-collar and white-collar work, respectively, versus male retirement age of 60) reduce women’s time to climb career ladders.
- Recommended interventions:
  - Childcare and family-friendly policies:
    - Increase affordable childcare provision, public subsidies, tax credits, and care credits.
    - Subsidize child and elderly care costs to increase marginal returns from paid work for mothers.
    - Consider parental leave policies that encourage sharing between parents, reserve non-transferable paid leave for fathers, mandate minimum nationwide paternity leave, and promote flexible/part-time arrangements.
    - Accelerate digitalization and training to close the gender digital divide and enable flexible work arrangements.
  - Enforce anti-discrimination and reduce workplace bias:
    - Strengthen enforcement of laws prohibiting gender discrimination in hiring, firing, promotion, and access to jobs, including protections for pregnant women and mothers.
    - Remove discriminatory hiring, firing, and promotion practices linked to fertility decisions.
    - Align retirement ages by gender.
  - Support women’s entrepreneurship and access to finance:
    - Strengthen enforcement of laws prohibiting gender discrimination in access to credit.
    - Improve availability of formal financing for women via technology, targeted funding, and alternative credit risk assessments.
  - Ease regulations and entry barriers to boost productivity in the service sector and accelerate marketization of services.
  - Support female entrepreneurship and strengthen social safety nets to facilitate reallocation from home to market production.

*Source: wpiea2021138-print-pdf — Section 6 concludes.*

### Section 6 concludes.

### wpiea2021138-print-pdf - Section 6 concludes.

### Related Literature
- Connects to literature linking structural transformation to gender inequality in labor markets (Akbulut, 2011; Ngai and Petrongolo, 2017; Rendall, 2018; Rendall, 2013; Ostry, Alvarez, Espinoza, and Papageorgiou, 2018).
- Fills a gap: no systematic studies linking structural transformation and gender inequality in China; this study documents facts and calibrates a model of structural transformation for China.
- Relates to literature estimating female labor supply elasticities and the role of cultural norms (Blau and Kahn, 2007; Heim, 2007; Chen and Ge, 2018). Unlike findings for the US (Blau and Kahn, 2007; Heim, 2007), the study documents a reverse trend in China—female labor supply elasticities have shifted in the opposite direction.
- Contributes to literature on economic growth, globalization, and income inequality in China (Han, Liu, and Zhang, 2012; Li, Loungani, and Ostry, 2017; Dong and Joffre, 2019; Liu, Li, and Yang, 2014; Dasgupta, Matsumoto, and Xia, 2015) by analyzing drivers of rising gender inequality from macro and micro perspectives and quantifying policy gains using a structural transformation framework.
- Notes: labor productivity and TFP growth in the service sector have slowed since the Global Financial Crisis (Zhu, Zhang, and Peng, 2019). Rapid population aging—one legacy of China’s one-child policy—likely exacerbates gender inequality through disproportionate elderly care burdens on women.

### Structural Transformation and Gender Gaps in China — Stylized Facts and Anomalies
- Fact 1: Declining LFP and Widening Gender LFP Gap
  - Data: ILO labor force participation rates for population aged 15 or older, 2000–2019.
  - Both men’s and women’s LFP rates decline following state-owned enterprise reforms; female LFP declines significantly faster, especially after 2010.
  - Gender LFP gap: about 11 percentage points in 2000; widened to about 14 percentage points by 2013 and remained at that level.
  - Women have also experienced a decline in average paid work hours relative to men.

- Fact 2: Widening Gender Wage Gap (CHIP microdata, working-age 15–64, non-zero wage incomes; compare 1995 and 2013)
  - Regression: log(wi) = αg + βg1 Femalei + βg2 Xg i + εg i.
  - Controlling for demographics, education, and sectoral composition, the conditional gender wage gap increased from 12 percent in 1995 to 35 percent in 2013.
  - Table 1 selected coefficients (1995 sample / 2013 sample):
    - Female: -0.119 ∗∗∗ / -0.351 ∗∗∗ (standard errors (0.017) (0.021))
    - Age: 0.056 ∗∗∗ / 0.078 ∗∗∗ (standard errors (0.014) (0.006))
    - (Age)2: -0.001 ∗∗∗ / -0.001 ∗∗∗ (standard errors (0.000) (0.000))
    - Child in HH: -0.060 ∗ / -0.081 ∗∗∗ (standard errors (0.034) (0.020))
    - Education (College): 0.402 ∗∗∗ / 0.570 ∗∗∗ (standard errors (0.026) (0.043))
    - Service sector: 0.059 ∗∗ / -0.011 (standard errors (0.026) (0.021))
    - Observations: 13,015 / 40,620
    - R-squared: 0.095 / 0.152
  - Notes: dependent variable = real annual log wages. Robust standard errors in parentheses. Omitted education = “Less than High School.” Omitted sector = goods sector.

- Fact 3: Rising Service Sector Share (ILO sectoral employment by gender, 2000–2019)
  - Two broad sectors: goods (agriculture, mining, construction, utilities, manufacturing) and services (wholesale and retail, transportation, finance, health, education, etc.).
  - Female and male employment shares plotted 2010–2019 show women concentrated in services, men in goods.
  - Reallocation: female (male) goods sector share decreased from 73 (75) percent in 1995 to 52 (57) percent in 2019.

- Puzzle: Structural transformation in China resembles other economies (declining goods share, rising services) yet gender inequality widened—contradicting literature showing service-sector growth usually decreases gender inequality (Akbulut, 2011; Ngai and Petrongolo, 2017; Rendall, 2013, 2018).

### Economic Growth and Female Labor Force Participation (FLFP)
- Theory: U-shaped relationship between economic growth and FLFP (Sinha, 1965; Goldin, 1994; Mammen and Paxson, 2000; Olivetti, 2013; Olivetti and Petrongolo, 2016).
  - Early development: high female participation (home and agriculture).
  - Industrialization: men’s comparative advantage in industry leads women to exit labor force as single-earner households become affordable.
  - Advanced stage: rising education and services lead to upswing in FLFP.
- Cultural norms mediate growth–FLFP link (Jayachandran, 2020); in China, cultural norms influence male attitudes toward wives’ work (Chen and Ge, 2018).
- Empirical: province-level variation (1995–2013) shows negative correlation between output per capita and female employment shares—richer provinces have lower female employment rates (Figure 3).

### Female Labor Supply Elasticities: Evidence from Micro Data
- Focus: drivers of women’s labor supply changes between 1995 and 2013; married vs unmarried women compared (sample: married women aged 15–64 residing in urban areas for elasticity estimation).
- Findings on hours and earnings gaps (Table 2):
  - Gender Hours Gap (1995 / 2013):
    - Married: -0.027 ∗∗∗ / -0.033 ∗ (standard errors (0.007) (0.017))
    - Unmarried: -0.006 / 0.066 ∗∗∗ (standard errors (0.012) (0.020))
    - Rural: -0.027 / -0.024 (standard errors (0.023) (0.019))
    - Urban: -0.019 ∗∗∗ / -0.053 ∗∗∗ (standard errors (0.006) (0.008))
  - Gender Earnings Gap (1995 / 2013):
    - Married: -0.153 ∗∗∗ / -0.395 ∗∗∗ (standard errors (0.022) (0.023))
    - Unmarried: -0.024 / -0.090 ∗∗∗ (standard errors (0.036) (0.019))
    - Rural: -0.073 / -0.357 ∗∗∗ (standard errors (0.064) (0.026))
    - Urban: -0.142 ∗∗∗ / -0.314 ∗∗∗ (standard errors (0.017) (0.020))
  - Interpretation:
    - Between 1995 and 2013, hours gaps reversed for unmarried women while earnings gaps slightly widened.
    - Married workers’ gender gaps widened significantly—changes in married women’s labor supply drove aggregate gender wage inequality.
    - Results consistent with a motherhood penalty and with childcare constraints being more binding in urban areas and for migrants (limited grandparents’ support).
- Estimating labor supply elasticities (three-stage procedure based on Mroz (1987); Heckman selection correction):
  - Sample: urban married women 15–64; exclude students, retired, disabled. Monetary variables deflated by province-level CPI.
  - Construction: annual hours computed differently in 1995 vs 2013 due to survey questions; hourly wage = annual wage ÷ annual hours (division bias acknowledged).
  - First stage: probit for labor force participation p_i = αp + βp1 log(I_i) + βp2 Xp i + εp i. Controls: age, age squared, education, presence of children, spouse’s age and education, county dummies, province population size. Compute inverse Mills ratio λ_i.
  - Second stage: selection-corrected wage regression log(w_i) = αw + βw1 λ_i + βw2 Xw i + εw i. Impute wages ŵ_i.
  - Third stage: hours equation h_i = αh + βh1 log(ŵ_i) + βh2 log(ws_i) + βh3 log(I_i) + βh4 λ_i + βh5 Xh i + εh i.
  - Instruments for own and spouse wages: individual-level experience quadratic; local labor market share of county employment in public sector (address division bias and unobserved ability).
- Table 3: OLS and Three-Stage IV results for married women (1995 / 2013)
  - Coefficients (selected):
    - Log(Own Wage):
      - 1995 OLS: 306.941 ∗∗ (139.178)
      - 1995 Three-Stage IV: 110.385 (167.927)
      - 2013 OLS: 939.414 ∗∗∗ (181.013)
      - 2013 Three-Stage IV: 699.724 ∗∗∗ (172.555)
    - Log(Spouse Wage):
      - 1995 OLS: -112.564 ∗∗∗ (13.397)
      - 1995 Three-Stage IV: -161.875 (163.813)
      - 2013 OLS: -188.730 ∗∗∗ (19.120)
      - 2013 Three-Stage IV: -398.227 ∗∗∗ (116.162)
    - Log(Non-wage HH Income):
      - 1995 OLS: -3.205 (8.249)
      - 1995 Three-Stage IV: -0.431 (9.647)
      - 2013 OLS: -7.087 ∗∗∗ (2.702)
      - 2013 Three-Stage IV: -18.532 ∗∗∗ (4.215)
    - Child in HH:
      - 1995 OLS: 3.211 (13.232)
      - 1995 Three-Stage IV: 25.908 (30.647)
      - 2013 OLS: -39.564 ∗ (21.678)
      - 2013 Three-Stage IV: -86.595 ∗∗ (34.640)
  - Observations: 1995 OLS/IV: 3,768 / 3,768; 2013 OLS/IV: 5,604 / 5,604.
  - Elasticities (at mean annual hours):
    - Own Log Wage:
      - 1995 OLS: 0.140
      - 1995 Three-Stage IV: 0.050
      - 2013 OLS: 0.407
      - 2013 Three-Stage IV: 0.303
    - Spouse Log Wage:
      - 1995 OLS: -0.051
      - 1995 Three-Stage IV: -0.074
      - 2013 OLS: -0.082
      - 2013 Three-Stage IV: -0.173
  - Notes: dependent variable = married women’s annual hours. Models include inverse Mills ratio, education, age, experience, spouse’s age and education, county dummies, province population. Instruments: experience squared and share of public sector in county for own wages; spouse’s experience, experience squared, and share of public sector in county for spouse wages. Sample = urban population only. Bootstrapped standard errors in parentheses.
- Key interpretation:
  - Substantial increase in both own-wage and spouse-wage elasticities between 1995 and 2013.
  - Contrary to evidence for the US (Blau and Kahn, 2007; Heim, 2007) which documents declining own-wage elasticities over time, China shows the opposite trend: women’s labor supply became more responsive to own wage offers and spouse’s wages, intensifying selection of women into the labor force.
  - The text continues: "Specifically, own wage elasticity increased from" — (content ends here).

*Source: wpiea2021138-print-pdf - Section 6 concludes.*

### 0.05 in 1995 to 0.3 in 2013 while spouse’s wage elasticity declined from -0.07 to -0.17.

### 0.05 in 1995 to 0.3 in 2013 while spouse’s wage elasticity declined from -0.07 to -0.17

### Empirical findings on female labor supply and household structure
- Female own-wage elasticity increased from 0.05 in 1995 to 0.3 in 2013.
- Spouse’s wage elasticity declined from -0.07 to -0.17 between 1995 and 2013.
- Presence of a child in the household:
  - In 1995, the coefficient on the child indicator is not statistically significant (Table 3).
  - By 2013, presence of a child has a negative and statistically significant effect on women’s labor hours, indicating women are more likely to reduce work hours after having a child.
- Interpretation:
  - Results suggest a shift toward a one-earner household model between 1995 and 2013.
  - Falling prevalence of dual-earner households was driven by rising incomes across China, making female participation more a choice than a necessity, particularly in urban areas.
- Assortative matching:
  - Using the methodology in Greenwood, Guner, Kocharkov, and Santos (2014), regressions of women’s years of education on spouse’s years of education interacted with a 2013 dummy document a rise in assortative matching in China from 1995 to 2013 (results available upon request), suggesting rising selection contributes to widening gender wage and hours gaps.

### Model of structural transformation — framework and mechanisms
- Model framework:
  - Adaptation of Ngai and Petrongolo (2017) to China with female and male workers allocating time to goods (j = g), services (j = s), or non-market home sector.
  - Goods and services produced with technology:
    - Y_j = A_j L_j, where L_j = [ξ_j L_fj^{(η−1)/η} + (1−ξ_j) L_mj^{(η−1)/η}]^{η/(η−1)}.
  - Female labor has comparative advantage in services, implying ξ_s > ξ_g.
  - Productivity growth rates: γ_g > γ_s; market services productivity grows faster than home services (γ_h < γ_s).
- Preferences and production of home services:
  - Utility: U(c_g, c_s, c_h) = lnc, with c a CES composite of goods and services and c_z a composite of market- and home-produced services.
  - Imposed parameter restrictions:  < 1 and σ > 1, implying services and goods are complements while home- and market-produced services are substitutes.
  - Home services produced with same functional form as market services: c_h = A_h [ξ_h L_fh^{(η−1)/η} + (1−ξ_h) L_mh^{(η−1)/η}]^{η/(η−1)}.
- Equilibrium:
  - Competitive equilibrium clears markets and labor: c_j = Y_j for j = g,s; and L_ig + L_is = L_i − L_ih for i = f,m.
- Productivity wedges (barriers to FLFP):
  - Comparative advantage parameter ξ_j = π_j χ_j, j = g,s.
  - Wedges π_j < 1 reduce women’s productivity in sector j and distort the gender wage ratio:
    - w_f/w_m = [π_j χ_j / (1 − π_j χ_j)] (L_mj / L_fj)^{1/η}.
  - Wedges vary over time and capture changes in social norms and policies affecting women’s comparative advantage.

### Model predictions vs. stylized facts
- Prediction 1: Expansion of service sector raises relative wages for women (women’s comparative advantage in services).
  - Stylized fact contradiction: Despite rising service sector share between 1995 and 2013, relative female wages declined.
  - Role of productivity wedge: a reduction in female comparative advantage via π_j can reverse the predicted relationship.
- Prediction 2: Faster growth of market services vs. home sector leads to marketization of home-produced services and higher market services labor share.
  - Stylized fact contradiction: Declines observed in both female and male LFP despite marketization forces.
- Prediction 3: Removal of sector-specific barriers to FLFP (π_j changes) explains within-sector changes in female labor intensity not accounted for by productivity-driven sectoral reallocation.

### Calibration details and parameters
- Targets: changes in sectoral employment shares, female-to-male wage ratio, gender-specific market and home production hours between 2000 and 2013.
- Data sources: ILO for sectoral employment; CHIP for market hours and wages; CHNS for home production hours; World Bank WDI for value added per hour; Bridgman et al. (2018) for home productivity benchmarks.
- Construction:
  - Total work hours across market and home sectors calculated by scaling average weekly hours by number of workers or adult population aged 15 to 64.
  - CHIP wage data used to compute gender wage gap.
  - Home production hours include childcare, cleaning, laundry, shopping, and food preparation.
- Table of calibrated parameters (as presented):
  - γ_g − γ_s = 0.037 (World Bank, ILO, CHIP)
  - γ_s − γ_h = 0.041 (World Bank, ILO, CHIP, Bridgman et al. (2018))
  - σ = 2.0 (Aguiar, Hurst, and Karabarbounis (2012))
  -  = 0.002 (Herrendorf, Rogerson, and Valentinyi (2013))
  - η = 2.27 (Ngai and Petrongolo (2017))
  - χ_g = 0.29 (Ngai and Petrongolo (2017))
  - χ_s = 0.43 (Ngai and Petrongolo (2017))
  - L_m/L_f = 1.05 (Match service share in 2000, given male and female service and market hours)
  - ξ_h = 0.48 (Match wage and hours ratios in 2013)
- Notes on calibration:
  - Sectoral productivity growth γ_g and γ_s calculated using growth of sectoral value added per hour of work.
  - Home sector productivity growth γ_h estimated using average growth rates among emerging and regional countries from Bridgman, Duernecker, and Herrendorf (2018).
  - Female comparative advantage parameters χ_g and χ_s taken from Ngai and Petrongolo (2017); ξ_h backed out by matching gender home hours ratio to gender wage ratio in 2013.

### Estimated sector-specific barriers to female employment (productivity wedges)
- Table 5: Sector-specific barriers π_j (derived by matching changes in goods and market services hours and wage ratios in 2000 and 2013)
  - π_g: 2000 = 1.27; 2013 = 1.03
  - π_s: 2000 = 0.86; 2013 = 0.70
- Interpretation:
  - Relative to US-based comparative advantage parameters, the wedge between the wage ratio and marginal rate of technical substitution in the goods sector is smaller in China in 2000, implying relatively more women in goods in China compared to the US.
  - Over time, productivity wedges widened in the service sector and women’s participation in the goods sector converged toward US levels.
  - Findings imply barriers to female workers’ entry into both goods and service sectors increased by a third between 2000 and 2013.
  - Conclusion: Increasing barriers to female employment can reverse the expected narrowing of gender wage and market hours gaps associated with structural transformation; counterfactual exercises quantify potential gains from removing these barriers.

### Policy dimensions, institutional drivers, and recommended interventions
- Cross-country comparisons:
  - China falls behind advanced and emerging economy averages on the Women, Business and the Law index (World Bank).
  - OECD Social Institutions and Gender Index (SIGI) ranks China with high gender biases among selected countries.
- Identified institutional/policy distortions:
  - Decline in state childcare support has increased time women spend on child care, limiting labor force participation and pushing women into lower-paid, informal, or home-based work.
  - Childcare costs have been rising faster than incomes; nursery fees have risen to around 15–20 percent of a middle-class family’s annual income.
  - Job advertisements and hiring practices show rising gender-based hiring biases despite legal prohibitions.
  - Motherhood penalty persists; protections for pregnant women and mothers often not effectively enforced.
  - Female underrepresentation in professional and managerial jobs remains large (women’s share in management was still one of four in 2010), much lower than many comparator economies.
  - Pension/retirement system distortions: different retirement ages by gender reduce women’s time to climb career ladders (noted retirement ages: women around 50 and 55 in blue-collar and white-collar work, respectively, versus male retirement age of 60).
- Policy recommendations and interventions:
  - Childcare and family-friendly policies:
    - Increase affordable childcare provision, public subsidies, tax credits, and care credits to expand supply and affordability of child and dependent care.
    - Subsidizing child and elderly care costs can increase marginal returns from paid work for mothers and support accumulation of work experience.
    - Consider parental leave policies that encourage sharing between parents, including reserving non-transferable paid leave for fathers to boost paternal uptake.
    - Mandate minimum nationwide paternity leave and encourage uptake by male workers.
    - Promote flexibility and part-time leave arrangements; accelerate digitalization and training to close the gender digital divide and enable flexible work arrangements.
  - Enforce anti-discrimination and reduce workplace bias:
    - Strengthen enforcement of laws prohibiting gender discrimination in hiring, firing, promotion, and access to jobs, including protections for pregnant women and mothers.
    - Remove discriminatory hiring, firing, and promotion practices linked to fertility decisions.
    - Align retirement ages by gender (remove differential retirement ages) to allow women more time for career progression and reduce leadership gaps.
  - Support women’s entrepreneurship and access to finance:
    - Strengthen enforcement of laws prohibiting gender discrimination in access to credit.
    - Improve availability of formal financing for women via technology, targeted funding, and alternative credit risk assessments that reduce reliance on land ownership or traditional collateral.
- Counterfactuals and gains:
  - The model performs counterfactual exercises to quantify effects of removing gender-biased policies (productivity wedges) on female wages, LFP, and aggregate growth; these exercises link identified institutional barriers to quantified potential gains in female employment and the aggregate economy.

*Italic: Source — wpiea2021138-print-pdf.*

### 5.2  Model Counterfactuals and Results

### 5.2 Model Counterfactuals and Results

### Counterfactual exercises and setup
- Two counterfactual exercises using the model from Section 4:
  - Counterfactual 1: Policies aimed at eliminating barriers for women to participate in the labor market (support female entrepreneurship, strengthen child and elderly care provision, eliminate gender bias in the workforce). Operationalized by setting sector-specific productivity wedges πg and πs to unity (πg = πs = 1), using the US in 2004-2008 as the benchmark for comparative advantage parameters χg and χs.
  - Counterfactual 2: Faster marketization of services—raising relative productivity growth of market services so service production reallocates from home to the market (can be achieved via investing in childcare and elderly care, strengthening social safety nets, and deregulation of the service sector). Implemented by increasing the gap (γs − γh) from 4.1 percentage points in the baseline to 6.2 percentage points (a 50 percent increase in the relative productivity growth of market services).

### Baseline and data moments reproduced by the model
- Data moments (conditional earnings/hours as reported):
  - Wage Gap: 2000 = 84.4; 2013 = 64.9
  - Market Hours Gap: 2000 = 46.6; 2013 = 41.3
- Model baseline:
  - Wage Gap: 2000 = 84.4; 2013 = 69.1
  - Market Hours Gap: 2000 = 46.6; 2013 = 35.9

### Counterfactual 1 — Removing productivity wedges (πg = πs = 1)
- Implementation: set πg and πs to unity (assumes labor market bias in China similar to US in 2004-2008), all other parameters unchanged.
- Key model implications and numerical findings (as reported):
  - Table 6 entries for Counterfactual 1: 77.9 and 72.8 (table entries preserved verbatim).
  - Narrative summary in text:
    - Matching comparative advantage parameters to those of the US, China’s unconditional wage gap would be reduced from about 40 percent to about 22 percent.
    - Reduction in labor market barriers would result in a reduction in the market hours gap by almost 32 percentage points.
- Interpretation:
  - Removing within-sector productivity wedges that lower women’s perceived marginal product raises within-sector demand for female labor and substantially narrows both wage and hours gender gaps.

### Counterfactual 2 — Accelerating marketization of services (γs − γh = 6.2 percentage points)
- Implementation: increase (γs − γh) to 6.2 percentage points while keeping the relationship between market services and goods productivity rates unchanged.
- Key model implications and numerical findings (as reported):
  - Table 6 entries for Counterfactual 2: 78.1 and 77.7 (table entries preserved verbatim).
  - Narrative summary in text:
    - Faster marketization has a relatively small additional effect on the gender wage ratio.
    - Faster marketization results in a 5 percentage points decline in the gender market hours gap, as women exit home production and enter the market sector.
- Interpretation:
  - Reforms boosting overall productivity of the market sector and rebalancing toward a more robust service sector can have a sizeable positive effect on gender equality in China, primarily by increasing women’s market hours.

### Aggregate output effect
- Narrowing gender gaps in hours and wages to the levels described in Table 6 would result in 4.9 percent higher output.

### Policy implications and recommendations (drawn from model results and discussion)
- Remove discriminatory practices against women in the workforce and in pay to level the playing field.
- Increase availability of low cost, high quality child and elderly care to lift the burden on women seeking employment and reduce firms’ need to pay hiring premia for women.
- Ease regulations and entry barriers to boost productivity in the service sector and accelerate marketization of services.
- Support female entrepreneurship and strengthen social safety nets to facilitate reallocation from home to market production and to mitigate care-related constraints.

*Source: CHIP, CHNS, ILO, World Bank, and staff calculations*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021138-print-pdf.pdf_