## _wp15151

## Source details

**Canonical URL:** [_wp15151](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15151.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15151.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15151.pdf.json)

---

### I. Introduction — objectives and context
- China initiated the comprehensive, third-plenum reform blueprint to move toward more inclusive and sustainable growth through better allocation of credit and resources and improved social welfare.
- Priority during the “new normal” transition: maintaining stable growth and ensuring ample employment while pursuing reforms (State Council, 2015).
- Labor market snapshot:
  - Newly created urban jobs exceeded official targets: 13.6 million in 2014 (official target 10 million).
  - New jobs reached 3.2 million in 2015:Q1.
  - Registered unemployment rate around 4 percent; official surveyed unemployment rate about 5 percent in 2015:Q1.
- Key concerns:
  - Structural resilience supported by demographic shifts, services-sector expansion, migrant flows, and labor hoarding in SOEs.
  - Prolonged labor hoarding by SOEs could reduce labor flexibility and limit productivity gains.
- Empirical and scenario analysis quantify employment effects of reform timing and sectoral adjustments, highlighting trade-offs between near-term unemployment rises and medium-term structural gains.

### II. Labor market developments — recent facts and indicators
- Employment and jobs:
  - Newly created urban jobs: 13.6 million in 2014.
  - New jobs: 3.2 million in 2015:Q1.
  - Over 1990–2014, total employment rose by about 250 million.
  - Nearly two-thirds of the gain in employment came from newly created jobs—at more than 10 million per year.
- Labor demand-supply:
  - Demand in urban labor markets has outpaced supply since the global financial crisis across regions.
  - Demand-supply ratio index indicates favorable demand when above 100 (Figure 1 referenced in source).
- Unemployment and measurement:
  - Official registered unemployment rate: about 4 percent.
  - Official surveyed unemployment rate: about 5 percent in 2015:Q1.
  - Tracking employment is difficult due to data shortcomings (Annex 1).
  - Manufacturing and services PMIs for employment fell below 50 in 2014, indicating contraction and some softening.
- Wages and productivity:
  - Average monthly income of migrant workers grew 9.5 percent in 2014, higher than nominal GDP growth of 8.2 percent.
  - Migrant wages have stayed at about 60 percent of urban workers’ wages over the past few years.
  - Labor force participation rate remains near 80 percent.

### III. Explaining labor market resilience — structural and unique buffers
- Structural trends:
  - China may be at a Lewis turning point, with less surplus rural labor.
  - Demographic headwinds: population aging, low fertility, rising dependency ratio, and a soon-to-contract working-age population.
  - Dependency ratio projected to reach nearly 50 percent by 2030.
  - Working-age population growth will slow and "begin to shrink in 2015."
  - Labor force participation remains relatively high at nearly 80 percent; easing one-child policy may mitigate long-term effects.
  - Incoming cohorts have more schooling, likely raising average labor productivity.
- Unique buffers and risks:
  - Migrant flows and SOE labor hoarding buffer employment against shocks.
  - Reliance on SOE labor hoarding risks inefficient allocation and lower productivity if prolonged.

### IV. Migrant flows — scale, characteristics, determinants, and implications
- Scale and urbanization:
  - About 270 million migrant workers in 2013, about a third of the total labor force and half of urban employment.
  - Urbanization rate: 54.8 percent (current), expected to rise to about 60 percent by 2020.
  - Urban employment: about 393 million in 2014, and for the first time in 2014 exceeded rural employment.
- Migrant worker characteristics (table values preserved):
  - Labor force participation rate: Migrants 95.9 ; Urban hukou residents 69.5
  - Employment rate: Migrants 94.3 ; Urban hukou residents 62.9
  - Self-employed share: Migrants 27.7 ; Urban hukou residents 8.4
  - Average weekly hours: Migrants 63.2 ; Urban hukou residents 43.8
  - Average hourly wage (2013): Migrants 55.6 ; Urban hukou residents 100.0
  - Years of schooling (average): Migrants 9.2 ; Urban hukou residents 12.3
  - Share of senior high and above: Migrants 33.0 ; Urban hukou residents 77.7
  - Employment industries (percent): Professional and office work: Migrants 10.5 ; Urban hukou residents 52.9
    - Sales / services workers: Migrants 55.9 ; Urban hukou residents 24.7
    - Manufacturing: Migrants 32.7 ; Urban hukou residents 15.5
  - Social welfare and benefits access:
    - Access to unemployment insurance (2008-2010): Migrants 12.0 - 13.5 ; Urban hukou residents 60.0 - 66.0
    - Access to urban health insurance (2010): Migrants 20.0 ; Urban hukou residents 87.0
  - Average duration stayed in cities (years): Migrants 7.0
- Dynamics and determinants:
  - Migrant flows closely related to GDP growth and better reflect short-term labor-market dynamics than unemployment rates (correlation with GDP growth 0.8 vs 0.4 for unemployment rate).
  - Empirical cross-province determinants include urban–rural income gap, GDP growth, infrastructure (road density), TFP, agricultural labor productivity, openness, SOE share, financial sector size, public education spending, returns to capital, and inflation.
  - Main empirical findings:
    - Urban–rural income gap is a key driver of migrant flows.
    - Higher GDP growth increases urban migration.
    - Better infrastructure reduces mobility costs and supports migrant flows.
    - Higher provincial SOE employment share associated with lower migrant flows.
    - Public expenditure on education found negative and significant, interpreted as urban-biased current spending reducing rural competitiveness.
- Policy data needs:
  - Broaden coverage and timeliness of migrant-flow data to facilitate policy design and assessment.
  - Authorities plan to expand surveyed unemployment coverage from 65 large cities to all prefecture-level cities at monthly frequency.

### V. Empirical analysis — Okun’s law, spatial models, and elasticity estimates
- Okun’s law and migrant flows:
  - Model relates real GDP growth gyt to unemployment measures ut, a reform dummy Dt, structural reform interactions, and Migt (annual change in migrants as share of total employment).
  - Registered unemployment rate has little relationship with GDP growth; survey-based unemployment rates show negative and significant relationship.
  - Example: a 1 percentage point increase in unemployment after 1993 associated with a reduction in growth by about 0.8–1.0 percentage point.
  - Inclusion of migrant share improves regression fit: a 1 percentage point increase in migrant flows associated with GDP growth of nearly 2 percentage points.
- Selected regression statistics (Table 1, exact entries preserved):
  - Observations: 21 (certain columns).
  - R-squared across reported columns: 0.208; 0.467; 0.547; 0.660.
  - Coefficients (selected):
    - Δut: -5.503* (std. err. (3.103)), -3.090 (2.637), 6.242*** (1.436), 4.543*** (1.419)
    - (1993) tD: 2.293 (2.460), 0.793 (2.090), 3.998*** (1.367), 2.399 (1.407)
    - (1993)*ttDuΔ: 3.741 (4.364), 0.529 (4.185), -7.154*** (1.489), -5.489*** (1.534)
    - tMig: 2.750** (1.061), 1.950** (0.849)
    - Constant: 8.246*** (2.362), 6.851*** (1.755), 6.553*** (1.265), 6.041*** (1.235)
  - Note: Standard errors in parentheses. *, **, *** indicate statistical significance at 10 percent, 5 percent, and 1 percent levels.
  - Chow test indicates a structural break occurred in 1993 (F-statistic referenced in source: 3.67 with p-value of 0.047 when using the Urban Household Survey urban unemployment rate).
- Spatial econometric models (SAR and SEM) — selected results (Table 3, exact coefficients preserved):
  - Urban–rural income gap coefficients: 0.399***, 0.919***, 0.853***, 0.524***, 0.943***, 0.872*** (std. errs. shown).
  - GDP growth rate coefficients: 3.147***, 3.469***, 3.279***, 3.301**, 3.011***, 2.861***.
  - Infrastructure (log) coefficients: 0.500***, 0.549***, 0.545***, 0.518***, 0.547***, 0.534***.
  - TFP (log) coefficients: -0.580***, -0.620***, -1.045***, -0.643***, -0.631***, -1.056***.
  - Rural productivity (log) coefficients: -0.462***, -0.289***, -0.351***, -0.549***, -0.345***, -0.400***.
  - Moran’s I: 0.248***, 0.217***, 0.173***, 0.253***, 0.229***, 0.195***.
  - R2: 0.891, 0.885, 0.892, 0.890, 0.883, 0.889.
  - Spatial parameters: ρ: 0.230***, 0.114***, 0.096**; λ: 0.257***, 0.217***, 0.196***.
  - Observations: 589 (all models).
- Elasticity estimates (1993–2013, Table 4):
  - Aggregate employment elasticity: 0.0762***.
  - Primary (agriculture and mining): -0.459***.
  - Manufacturing: 0.212***.
  - Services (tertiary): 0.313***.
  - Aggregate employment elasticity fell to 0.04 after the global financial crisis (about half its historical level).
  - Services sector elasticity about 0.1 percentage point higher than manufacturing, indicating services are more labor intensive.

### VI. Scenario analysis — reform timing, sectoral dynamics, and simulation outcomes
- Scenario framework and inputs:
  - Historical employment–growth relationships applied across sectors.
  - Speed of services expansion estimated via cross-country regression on per-capita income.
  - Simulation uses the Flexible System of Global Models and incorporates financial, fiscal, SOE, and hukou reforms.
  - Hukou reform target: raise urbanization rate to about 60 percent by 2020 (about 1 percentage point per year).
  - Elasticities used: aggregate 0.076, manufacturing 0.21, services 0.31.
  - Agricultural employment assumed to decline to fewer than 200 million by 2020 (about 3 percent per year decline).
- Baseline scenario assumptions and results:
  - Growth slows from 6.8 percent in 2015 to about 6 percent by 2017, picks up to about 6.3 percent by 2020.
  - Services sector expands to nearly 52.4 percent of output and 46 percent of employment by 2020.
  - Unemployment rate rises by about ½ percentage point but remains stable in the medium term.
  - Net increase in urban employment just exceeds 10 million people each year.
- Slow reform scenario assumptions and results:
  - Inadequate reform progress; persistence of unsustainable growth patterns and rising vulnerabilities.
  - Migrant flows slow as services expansion stalls and hukou restrictions persist.
  - Net increase in urban employment would decline, at times about 10 million workers a year, while unemployment would spike from initially stable levels.
- Sensitivity and risks:
  - If aggregate productivity falls short of expectations, urban employment increases could miss targets or GDP and wage growth could be much lower even if employment targets are met.
  - Sensitivity analysis: absent reform-led productivity gains in services, GDP growth could slow by 0.2–0.4 percentage point even if urban employment increases are maintained.

### VII. Policy implications and recommendations
- Overarching recommendation: eliminate impediments to labor market flexibility while providing on-budget, targeted social safety nets to facilitate transition to the "new normal."
- Key policy priorities:
  - Strengthen labor market flexibility instead of overrelying on buffers (migrant flows, SOE labor hoarding).
  - Implement retraining programs for jobs in the services sector to enhance reallocation and productivity.
  - Fiscal reforms: revenue reforms, pension portability, higher social spending to narrow urban–rural income gap.
  - Broadening the value-added tax to help services sector expansion by removing cascading effects on investment.
  - Social security reforms (including pension portability) to increase labor mobility.
  - Open up the services sector to encourage entry, competition, productivity gains, and job creation.
  - Hukou and rural land reforms to remove mobility obstacles and clarify property rights, speeding urbanization and improving migrant benefits.
  - On-budget, targeted social safety nets and retraining programs to facilitate labor-market adjustment.
- Data and institutional recommendations:
  - Improve data coverage and timeliness: wider surveyed unemployment coverage and public release of labor and household surveys.
  - Authorities intend to subscribe to the Special Data Dissemination Standard and expand survey coverage.

### References .............................................................................................................

### _wp15151 - References

### I. Introduction — objectives and context
- China initiated the comprehensive, third-plenum reform blueprint to move toward more inclusive and sustainable growth through better allocation of credit and resources and improved social welfare.
- Priority during the “new normal” transition: maintaining stable growth and ensuring ample employment while pursuing reforms (State Council, 2015).
- Labor market snapshot:
  - Newly created urban jobs have exceeded official targets by a significant margin.
  - The registered unemployment rate remains stable at about 4 percent.
  - The official surveyed unemployment rate was also stable, at about 5 percent, in the first quarter of 2015.
- Key concerns:
  - Structural resilience is supported by demographic shifts, services-sector expansion, migrant flows, and labor hoarding in SOEs.
  - Prolonged labor hoarding by SOEs could reduce labor flexibility and limit productivity gains.
- Empirical and scenario analysis aim to quantify employment effects of reform timing and sectoral adjustments, showing trade-offs between near-term unemployment rises and medium-term structural gains.

### II. Labor market developments — recent facts and indicators
- Employment and jobs:
  - Newly created urban jobs reached 13.6 million in 2014, exceeding the official target of 10 million.
  - New jobs reached 3.2 million in the first quarter of 2015; slightly lower than 2014:Q1 but estimated to exceed the target in that year.
  - Over 1990–2014, total employment rose by about 250 million.
  - Nearly two-thirds of the gain in employment was from newly created jobs—at more than 10 million per year.
- Labor demand-supply:
  - Demand in urban labor markets has outpaced supply since the global financial crisis across regions in China.
  - The demand-supply ratio index (Figure 1) indicates favorable demand conditions when above 100.
- Unemployment and measurement:
  - The official registered unemployment rate around 4 percent; the official surveyed unemployment rate around 5 percent.
  - Tracking employment is difficult due to data shortcomings (Annex 1).
  - High-frequency indicators (manufacturing and services PMIs for employment) fell below 50 in 2014, indicating contraction and some softening.
- Wages and productivity:
  - Average wage growth has slowed but outpaced nominal GDP growth and labor productivity in recent years.
  - The average monthly income of migrant workers grew 9.5 percent in 2014, higher than nominal GDP growth of 8.2 percent.
  - Migrant wages have stayed at about 60 percent of urban workers’ wages over the past few years, after significant convergence during the late 1990s and early 2000s.
  - Labor force participation rate remains near 80 percent.

### III. Explaining labor market resilience — structural and unique buffers
- Structural trends supporting resilience:
  - Demography: China may be at a Lewis turning point, with less surplus rural labor (Das and N’Diaye 2013; Zhang, Yang, and Wang, 2011).
  - A decline in surplus labor could dampen new pressures on employment as the economy slows.
  - Demographic headwinds: population aging, low fertility, rising dependency ratio, and a soon-to-contract working-age population.
  - Countervailing factors: high labor force participation (~80 percent); plans to raise retirement age could boost the labor force; incoming cohorts have more schooling, likely raising average labor productivity.
- Unique labor-market buffers:
  - Migrant flows and labor hoarding in SOEs and overcapacity sectors help buffer employment against adverse shocks.
  - However, reliance on SOE labor hoarding risks inefficient allocation and lower productivity if prolonged.

### IV. Migrant flows — scale, determinants, and implications
- Scale:
  - Number of migrant workers about 270 million in 2013, or a third of the total labor force (Meng, 2012) and half of urban employment.
- Dynamics and determinants:
  - Migrant flows are closely related to GDP growth and better reflect short-term labor-market dynamics than unemployment rates.
  - Empirical estimates suggest the urban-rural income gap and economic growth are key determinants of migrant flows.
  - Hukou restrictions and lack of social services for migrants could weaken long-term labor-market flexibility.
- Policy data needs:
  - Broadening the coverage and timeliness of data, especially related to migrant flows, will facilitate policy design and assessment.

### V. Scenario analysis — reform timing and employment outcomes
- Main findings (scenario analysis summarized):
  - Delays in reform implementation could cause further distortions and weaken medium-term employment prospects.
  - New employment levels risk falling below the current official job target if reforms are delayed.
  - Faster reforms in overcapacity sectors and SOEs may, in the near term, release excess labor and push up the interim unemployment rate by ½‒¾ percentage point, but facilitate structural transition (urbanization and services-sector expansion) and more sustainable medium-term job creation.

### VI. Policy implications and recommendations
- Strengthen labor market flexibility to facilitate China’s transition to the new normal.
- Prioritize policies that foster reallocation of surplus labor through effective, on-budget social policies rather than relying solely on inherent buffers (e.g., SOE labor hoarding).
- Steadfast implementation of reforms to:
  - Facilitate migrant flows and structural trends that generate jobs and urban employment in the medium term.
  - Open up the services sector and reform hukou regulations to enhance labor-market flexibility (Whalley and Zhang, 2007).
- Fiscal and social policy adjustments:
  - Fiscal reforms on taxation, pension portability, and higher social spending to help narrow the urban–rural income gap (Lam and Wingender, 2015).
- Improve data:
  - Broaden coverage and timeliness of labor-market and migrant-flow data to support policy design and assessment.

*Source: _wp15151 - References.*

### 1.2 percent per year), but will begin to shrink in 2015. Easing of the one-child policy may eventually mitigate the

### _wp15151 - 1.2 percent per year), but will begin to shrink in 2015. Easing of the one-child policy may eventually mitigate the

### Demography and labor participation
- Population and working-age trends
  - Working-age population growth shown to slow and "begin to shrink in 2015."
  - Dependency ratio is set to rise further, "reaching nearly 50 percent by 2030."
  - Labor participation rate has fallen but "remains relatively high at nearly 80 percent."
- Labor force levels and participation (chart indicators preserved from source)
  - Labor force and labor participation rate series spanning 1990–2013 displayed (millions and percent).

### Services sector expansion and implications for employment and productivity
- Key findings on services sector
  - Services sector employment accounted for "about 40 percent of the labor force in 2014."
  - Value-added from the services sector reached "48.2 percent in 2014, surpassing that of the manufacturing sector."
  - Jobs created from a "1 percentage point increase of the services sector share in GDP could offset the employment loss from a 0.4 percentage point decline in GDP growth (Ma and others, 2014)."
  - Employment in and output of the services sector expanded rapidly, "particularly after 2008."
- Sectoral productivity and employment mix
  - Labor productivity in the services sector is, in general, "lower than that in manufacturing."
  - Services sector contributions to total employment "are large, often exceeding half in most provinces."
- Sectoral employment dynamics (charts and series preserved)
  - Annualized growth in employment by sector shown for "Between 2002 and the latest year available" and "Between 2008 and 2013."
  - Growth of employment by sector and share of output across sectors series (1990s–2014) presented in source.

### Short-term buffers in labor markets: migrant flows and SOEs
- Rural–urban migrant flows as shock absorbers
  - Migrants account for "about 35.5 percent of total employment and 50.9 percent of nonagricultural employment."
  - Migrant flows slow before unemployment rises; "about 20–45 million migrant workers returned to their rural homes" when the global financial crisis hit in mid-2008 (Meng, 2012).
  - Migrant worker jobs are "largely in the private sector and in low-skill industries" and "usually more vulnerable to a growth slowdown than are urban workers’ jobs."
- State-owned enterprises (SOEs) and labor hoarding
  - SOEs provide buffers by "hoarding excess labor instead of laying off workers during downturns."
  - SOE adjustments tend to be gradual via "relocation, buyouts, and severance pay."
  - The employment share of the state sector has declined, "falling below 50 percent in recent years," with almost half of urban hukou workers shifting to the private sector.
- Risks of persistent buffers
  - Persistent limited migrant flows could imply "inefficient allocation of labor that limits productivity gains."
  - SOEs holding excess labor could "delay the unwinding of overcapacity sectors" and delay necessary reforms.

### Characteristics and conditions of migrant workers
- Scale and urbanization
  - "There were about 270 million migrant workers in China in 2013, about a third of the total labor force (Meng, 2012) and half of urban employment."
  - Urbanization rate "now at 54.8 percent, is expected to rise to about 60 percent by 2020."
  - Urban employment "has more than doubled during the past two decades to about 393 million, and for the first time, in 2014, exceeded rural employment."
- Migrant worker labor market indicators (table values preserved)
  - Labor force participation rate: Migrants "95.9" ; Urban hukou residents "69.5"
  - Employment rate: Migrants "94.3" ; Urban hukou residents "62.9"
  - Self-employed share: Migrants "27.7" ; Urban hukou residents "8.4"
  - Average weekly hours: Migrants "63.2" ; Urban hukou residents "43.8"
  - Average hourly wage (2013): Migrants "55.6" ; Urban hukou residents "100.0"
  - Years of schooling (average): Migrants "9.2" ; Urban hukou residents "12.3"
  - Share of senior high and above: Migrants "33.0" ; Urban hukou residents "77.7"
  - Employment industries (percent): Professional and office work: Migrants "10.5" ; Urban hukou residents "52.9"
    - Sales / services workers: Migrants "55.9" ; Urban hukou residents "24.7"
    - Manufacturing: Migrants "32.7" ; Urban hukou residents "15.5"
  - Social welfare and benefits access:
    - Access to unemployment insurance (2008-2010): Migrants "12.0 - 13.5" ; Urban hukou residents "60.0 - 66.0"
    - Access to urban health insurance (2010): Migrants "20.0" ; Urban hukou residents "87.0"
  - Average duration stayed in cities (in years): Migrants "7.0" (measured in calendar year; subject to selection bias)
- Coverage, wages, and trends
  - Migrant workers "often have limited access to social welfare and services" due to hukou restrictions and lack of labor contracts.
  - Migrants’ wages "have also increased in line with urban workers in recent years," partly driven by services sector expansion and rise of minimum wages.
  - Migrant workers still account for most employment in the informal sector.
- Additional source charts and figures preserved in the source
  - Charts on migrant workers’ wages by industry (wages in RMB and growth rates), aging trends, access to social welfare and benefits in 2014, and labor contract types for migrants as of 2014.

### Empirical analysis: migrant flows and Okun’s law
- Migrant flows and GDP growth
  - Migrant flows "are closely related to GDP growth and better reflect short-term dynamics in labor markets than unemployment rates."
  - Correlation between GDP growth and migrant flows is "0.8, relative to 0.4 for the unemployment rate."
- Model specification (Okun’s law adaptation)
  - The estimation model used (equation (1) in source) relates real GDP growth gyt to unemployment measures ut (official registered urban unemployment rate or estimated rate from Urban Household Survey 1989–2009), a dummy Dt for the year of urban employment reform, year of structural reform k, and Migt denoting annual change in migrants as a share of total employment.
  - The empirical results "suggest a correlation between the fluctuations of output and the cyclical conditions of China’s labor market."
  - The Chow test implies "the structural break occurred in 1993 (F-statistic is" [chart/data continuation in source]).

*Italic source: Excerpt from IMF working paper content (_wp15151) provided in the source PDF content unit.*

### 3.67 with p-value of 0.047 when using the Urban Household Survey urban unemployment rate; the F-

### _wp15151 - 3.67 with p-value of 0.047 when using the Urban Household Survey urban unemployment rate; the F-

### Okun’s Law Estimates for China
- The Okun coefficient is β1 before the structural reform and β1 + β3 afterward.
- Estimates suggest the registered unemployment rate has little relationship with GDP growth, while estimates using unemployment rates from surveys show a negative and significant relationship (Table 1).
- For instance, a 1 percentage point increase in unemployment after 1993 is associated with a reduction in the growth rate by about 0.8–1.0 percentage point.
- Inclusion of the migrant share in employment improves the overall fit of the regression.
- Growth in migrant flows is strongly correlated with GDP growth: a 1 percentage point increase in migrant flows is associated with GDP growth of nearly 2 percentage points.
- Migrant workers have a closer link to economic fluctuations, possibly because they are more vulnerable to job losses; migrant flows may better reflect labor market conditions.
- Footnote evidence:
  - Okun (1962) estimates that a 1 percentage point rise in the unemployment rate is associated with about 3 percentage points fall in output.
  - Other China studies using the official unemployment rate find significant deviations from Okun’s results.

Key regression statistics from Table 1 (selected entries, exact formatting preserved from source):
- Observations: 21 (columns O.2 and S.2 also 21)
- R-squared: 0.208; 0.467; 0.547; 0.660 (across reported columns)
- Coefficients (selected):
  - Δut: -5.503* (std. err. (3.103)), -3.090 (2.637), 6.242*** (1.436), 4.543*** (1.419)
  - (1993) tD: 2.293 (2.460), 0.793 (2.090), 3.998*** (1.367), 2.399 (1.407)
  - (1993)*ttDuΔ: 3.741 (4.364), 0.529 (4.185), -7.154*** (1.489), -5.489*** (1.534)
  - tMig: 2.750** (1.061), 1.950** (0.849)
  - Constant: 8.246*** (2.362), 6.851*** (1.755), 6.553*** (1.265), 6.041*** (1.235)
- Note: Standard errors in parentheses. *, **, *** indicate statistical significance at 10 percent, 5 percent, and 1 percent levels, respectively.
- Dependent variables in columns (O.2) and (S.2) are authors’ calculations based on Urban Household Survey data; others are from NBS.

### Determinants of Migrant Flows (Cross-province analysis)
- Sample period begins in 1992; dependent variable is annual change in the rural labor force net of agricultural employment (assumes rural agricultural sector fully employed).
- Core explanatory variables:
  1. Urban–rural income gap (measured as gap between urban household income and rural household net income per capita)
  2. GDP growth rate
  3. Infrastructure level (proxied by road density)
  4. Total factor productivity (TFP, estimated using provincial panel data on industrial output, net values of fixed assets and labors with system GMM)
  5. Agricultural labor productivity (measured as ratio of total agricultural capital use to agricultural employment)
- Control variables include: degree of openness (FDI/GDP, Trade/GDP), share of SOE output in total industrial output, financial sector size (loans-to-GDP), per capita public expenditure on education, urban unemployment rate (registered and surveyed), rate of return on capital, and inflation rate.
- Main empirical findings:
  - Urban–rural income gap is a key driver of migrant flows: larger gaps encourage movement to cities for nonagricultural jobs.
  - Higher GDP growth is associated with shifting labor out of agriculture and increasing urban migration.
  - Infrastructure is statistically significant: better infrastructure reduces migrant mobility costs.
  - Higher share of SOE employment in a province is associated with lower migrant flows.
  - Negative coefficients on TFP may reflect replacement effects between capital and workers when technology is capital oriented.
  - Public expenditure on education is negative and significant, interpreted as current public education spending biased toward urban households, reducing rural labor competitiveness.
  - Size of provincial financial sector and agricultural labor productivity generally correlate with migrant flows.
  - Returns to capital have a strong positive effect on migrant flows, suggesting capital–labor complementarities during opening up.
  - Inflation coefficient is not significant across provinces.
  - Unemployment rates do not have a strong effect, perhaps due to data shortcomings.

Selected descriptive statistics from Table 2 (exact values preserved):
- Migrant Flows (log): Observations 530; Mean 2.694; Standard Error 1.454; Minimum -1.609; Maximum 5.205
- Urban-rural income gap (log): Observations 589; Mean 8.285; Standard Error 0.518; Minimum 6.975; Maximum 9.570
- GDP growth rate: Observations 589; Mean 0.108; Standard Error 0.045; Minimum -0.043; Maximum 0.345
- Infrastructure (log): Observations 584; Mean 7.825; Standard Error 0.932; Minimum 5.092; Maximum 9.839
- Loans/GDP: Observations 589; Mean 0.996; Standard Error 0.286; Minimum 0.533; Maximum 2.260
- Loans/savings: Observations 589; Mean 0.870; Standard Error 0.251; Minimum 0.233; Maximum 1.890
- TFP (log): Observations 587; Mean -1.001; Standard Error 0.344; Minimum -1.805; Maximum -0.070
- Rural productivity (log): Observations 583; Mean 2.882; Standard Error 0.698; Minimum 0.846; Maximum 4.364
- FDI/GDP: Observations 576; Mean 0.035; Standard Error 0.036; Minimum 0.000; Maximum 0.243
- Trade/GDP: Observations 589; Mean 0.299; Standard Error 0.397; Minimum 0.032; Maximum 2.173
- SOE share: Observations 584; Mean 0.511; Standard Error 0.202; Minimum 0.094; Maximum 0.899
- Public expenditure on education: Observations 483; Mean 3.280; Standard Error 3.099; Minimum 0.374; Maximum 20.15
- Urban registered unemployment rate (%): Observations 565; Mean 3.370; Standard Error 0.966; Minimum 0.400; Maximum 7.400
- Urban surveyed unemployment rate (%): Observations 162; Mean 6.367; Standard Error 3.184; Minimum 1.338; Maximum 14.49
- Capital returns: Observations 589; Mean 0.096; Standard Error 0.083; Minimum -0.055; Maximum 0.461
- CPI (%): Observations 589; Mean 5.178; Standard Error 7.021; Minimum -3.900; Maximum 29.70
- Regression sample spans from 1992 to 2010; due to data missing, numbers of observations differ across variables.

### Spatial Econometric Models and Results
- To account for spatial correlation of migrant flows across provinces, two spatial econometric models are used:
  - Spatial autoregressive model (SAR)
  - Spatial error model (SEM)
- Model specifications (as presented):
  - SAR: Y = ρ W Y + X β + ε
  - SEM: Y = X β + u, u = λ W u + ε
  - W is spatial weighting matrix with weight 1 for neighboring provinces and 0 otherwise; matrix standardized as in Luo (2010) and Zhang, Hong, and Chen (2013).
- Regression results summary (Table 3, selected entries — exact coefficients and significance preserved):
  - Urban–rural income gap coefficients across models: 0.399***, 0.919***, 0.853***, 0.524***, 0.943***, 0.872*** (with respective standard errors (0.133), (0.168), (0.166), (0.149), (0.177), (0.175))
  - GDP growth rate coefficients: 3.147***, 3.469***, 3.279***, 3.301**, 3.011***, 2.861*** (std. errs. shown)
  - Infrastructure (log) coefficients: 0.500***, 0.549***, 0.545***, 0.518***, 0.547***, 0.534*** (std. errs. shown)
  - TFP (log) coefficients: -0.580***, -0.620***, -1.045***, -0.643***, -0.631***, -1.056*** (std. errs. shown)
  - Rural productivity (log) coefficients: -0.462***, -0.289***, -0.351***, -0.549***, -0.345***, -0.400*** (std. errs. shown)
  - Moran’s I: 0.248***, 0.217***, 0.173***, 0.253***, 0.229***, 0.195***
  - R2: 0.891, 0.885, 0.892, 0.890, 0.883, 0.889
  - Adjusted R2: 0.884, 0.877, 0.883, 0.883, 0.874, 0.880
  - Log-likelihood: -721.7, -616.9, -603.6, -722.3, -611.7, -599.8
  - Observations: 589 (all models)
  - Spatial parameters:
    - ρ: 0.230***, 0.114***, 0.096** (standard errors (0.045), (0.041), (0.041))
    - λ: 0.257***, 0.217***, 0.196*** (standard errors (0.051), (0.049), (0.052))

### Scenario Analysis on the Labor Market under the "New Normal"
- Reform measures (third-plenum reform blueprint, State Council, 2013) and other reforms (hukou reforms, expanding social security coverage, raising the minimum wage) will affect economic growth and have direct effects on labor markets.
- Reform implementation may reinforce structural trends affecting labor market conditions.
- Scenario approach:
  - Historical estimates on relationship between employment and growth across sectors are obtained.
  - Speed of services sector expansion estimated using cross-country panel regression on per-capita income.
  - Scenario design follows IMF staff report on China (2015) and Lam and Maliszewski (2015).
  - Simulation based on the Flexible System of Global Models (Andrle and others, 2015).
  - Scenario incorporates key elements of the reform blueprint: financial, fiscal, SOE, and hukou reforms.
  - Hukou reforms will improve labor mobility and support urbanization (Annex 3).
  - The reform plan commits to raising the urbanization rate to about 60 percent by 2020 (about 1 percentage point per year).

### Elasticity between Employment and Growth across Sectors
- Elasticity measures how much employment in a sector will increase if growth in that sector rises by 1 percentage point.
- Average elasticity over sample period 1993–2013 is estimated for agriculture, manufacturing, and services (Table 4).
- Based on estimated aggregate elasticity, a 1 percentage point increase in employment is associated with GDP growth of 0.08 percentage point, on average.
- The elasticity declined to about (text ends at this point in the supplied content).

*Source: NBS, Urban Household Survey, IMF staff calculations.*

### 0.04 after the global financial crisis, about half its historical level. The elasticity for the primary sector

### _wp15151 - 0.04 after the global financial crisis, about half its historical level. The elasticity for the primary sector

### Elasticities and sectoral labor dynamics
- The elasticity for aggregate employment fell to 0.04 after the global financial crisis, about half its historical level.
- The elasticity for the primary sector is negative because rural workers moving to nonagricultural employment would likely boost growth.
- The elasticity of the services sector tends to be about 0.1 percentage point higher than elasticity of manufacturing, suggesting the services sector is more labor intensive and has lower labor productivity.
- Estimated elasticities (based on data from 1993–2013):
  - Aggregate employment elasticity: 0.0762***
  - Primary (agriculture and mining): -0.459***
  - Manufacturing: 0.212***
  - Services (tertiary): 0.313***
  - (Standard errors in parentheses in source; significance levels: *, **, *** indicate 10 percent, 5 percent and 1 percent level respectively.)

### Estimation of services sector share and income linkages
- International comparisons indicate a close, positive linkage between per capita income and services sector employment: estimates suggest that a 1 percent increase in per capita GDP would drive up the services sector share of employment and output by 0.09 and 0.06 percentage points, respectively.
- Table 5 estimates (selected coefficients) relating Ln(GDP per capita) to service sector measures:
  - Share of employment in services: coefficients include 0.0906*** and 0.0922*** in different specifications.
  - Share of GDP in services: coefficients include 0.0671*** and 0.0809*** in different specifications.
  - R-squared values reported: up to 0.892 for employment share regressions and up to 0.740 for GDP share regressions.
- Observed empirical relationship (text chart): y = 10.509ln(x) - 40.429, R² = 0.8817 (Per-capita Income and Share of Employment in Services Sector, in percent and in constant 2005 USD).

### Scenario analysis framework
- Baseline scenario assumptions:
  - Gradual yet steady progress in implementing reform.
  - Near-term slowdown as unsustainable demand is reduced: slower credit growth, multiyear residential real estate adjustment.
  - Growth falls to 6¼ percent in 2016 and 6 percent in 2017, cushioned by productivity gains from structural reforms.
  - Starting in 2018, growth picks up modestly as productivity gains dominate.
- Slow reform scenario assumptions:
  - Inadequate progress in reforms and containment of vulnerabilities.
  - Persistence of unsustainable growth patterns and rising vulnerabilities.
  - Higher likelihood of protracted weak growth and risk of a sharp and disorderly correction as buffers diminish.
- Use of elasticities in scenario simulations:
  - Estimated elasticity used to determine impact on employment in manufacturing and services for each scenario.
  - Estimated elasticities cited: aggregate 0.076, manufacturing 0.21, services 0.31 (see section V part B).
  - Agricultural employment taken as residual; agricultural employment expected to decline further to fewer than 200 million workers by 2020, a decline of about 3 percent per year.
- Other modeling inputs:
  - Per capita income growth used to pin down services sector share of employment and migrant flows.
  - Okun’s law estimates used to derive underlying unemployment rate.
  - Annual increase in urban employment or nonagricultural employment used as proxies for official job targets.

### Simulation results across scenarios (key quantitative findings)
- Baseline scenario results:
  - Baseline growth forecast slows from 6.8 percent in 2015 to about 6 percent by 2017 before picking up to about 6.3 percent by 2020.
  - Services sector expands to nearly 52.4 percent of output and 46 percent of employment by 2020.
  - Unemployment rate rises by about ½ percentage point but remains stable in the medium term.
  - Net increase in urban employment (proxy for new urban jobs) just exceeds 10 million people each year.
- Slow reform scenario results:
  - Investment-led measures may support near-term growth but medium-term risks rise.
  - Migrant flows would slow as services sector expansion stalls and hukou restrictions pose obstacles.
  - Net increase in urban employment would decline, at times about 10 million workers a year, while the unemployment rate would spike from initially stable levels.
- Sensitivity and risks:
  - If aggregate productivity falls short of expectations, the rise in urban employment could fall short of targets, or GDP and wage growth could be much lower even if employment targets are met.
  - Sensitivity analysis: if the increase in urban employment stays the same as in the baseline but without reform-led productivity gains in the services sector, then GDP growth could slow by 0.2–0.4 percentage point.

### Policy implications and recommendations
- Overarching implication: elimination of impediments to labor market flexibility with on-budget and targeted social safety nets will facilitate the economic transition to the new normal in China.
- Strengthen labor market flexibility rather than relying too much on buffers:
  - Buffers (migrant flows, SOEs’ capacity to hoard labor) can temporarily lessen unemployment pressures but hinder reform efforts and productivity gains.
  - Policies such as retraining for work in the services sector could strengthen labor market flexibility while enhancing productivity.
- Structural reforms priorities:
  - Continue reforms to contain vulnerabilities and move China toward a more sustainable growth path.
  - Fiscal reforms, including revenue reforms and pension portability, to support labor mobility across provinces.
  - Broadening the value-added tax can help services sector expansion by removing cascading effects on investment.
  - Social security reforms, including pension portability, would significantly increase labor mobility while strengthening social safety nets.
  - On-budget targeted social safety nets and retraining programs to facilitate labor market flexibility.
  - Opening up the services sector to encourage entry and competition, generating productivity gains and jobs.
  - Hukou and rural land reforms to remove labor mobility obstacles and clarify property rights, speeding urbanization and improving migrant workers’ social benefits.
- Data improvements for policy design:
  - Address data shortcomings: wider coverage of surveyed unemployment and public release of labor and household surveys.
  - Better data collection and coverage of migrant flows to improve understanding of China’s labor markets.
  - Authorities’ steps: intention to subscribe to the Special Data Dissemination Standard and plan to expand coverage of the surveyed unemployment rate from 65 large cities to all prefecture-level cities at a monthly frequency.

### Conclusions and data caveats
- Labor market stability is important while implementing structural reforms; labor market conditions have held up despite slowdown, but labor hoarding in overcapacity sectors is increasing.
- Migrant flows between rural and urban employment are more correlated with growth than measured unemployment.
- Delays in reforms could weaken labor market conditions over the medium term, causing sustained increases in the unemployment rate and job creation falling short of policy targets.
- For successful transition, prioritize labor reallocation to new growth sectors, labor market mobility, and increased productivity supported by on-budget targeted social safety nets, retraining programs, and acceleration of hukou reforms.
- Data limitations:
  - China’s labor market data have limited coverage and disclosure; official registered unemployment rate stayed at 4 percent for two decades while surveyed unemployment was about 5 percent in late 2014.
  - Many labor and household surveys are not publicly available; employment statistics across industries were discontinued in 2010.
  - Estimates of underlying unemployment vary widely (margin as wide as 6 percentage points in some studies).

*Source: IMF staff analysis and authors' estimates as presented in the provided content.*

### Annex Table 1.1. Labor Market Statistics in China

### Annex Table 1.1. Labor Market Statistics in China

### Data inventory: administrative and aggregate sources
- Principal sources: National Bureau of Statistics (NBS), Ministry of Human Resources and Social Security (MoHRSS), the State Administration for Industry and Commerce (SAIC), and CEIC.
- Aggregate data collection systems:
  - Reporting Form System on Labour Statistics.
  - Sample Survey System on Labour Force.
  - System of Rural Social and Economic Surveys.
  - Reporting Form System on Training and Employment Statistics (for employment services, change of labour force, number of registered unemployed persons in urban areas).
- Key administrative/aggregate coverage and vintage:
  - Employment: Monthly, Quarterly, and Annually; Covers nationwide with provincial and industrial data; Data starting from 1952.
  - Wage: Quarterly, and Annually; Covers nationwide with provincial and city-level data; Data starting from 1952.
  - Migrant worker: Quarterly; Covers nationwide with national and regional data; Data starting from 2008.
  - PMI — Employment index: Monthly.
  - Registered Unemployment in Urban Areas: Quarterly; Covers nationwide with provincial data; Data starting from 1980.
  - Registered Unemployment Rate in Urban Areas: Quarterly; Covers nationwide with provincial data; Data starting from 1980.
  - Labor market Demand-Supply Ratio: Quarterly; Covers main cities with city-level data; Monitored by city community employment services center (mostly on low-skilled labor).
  - SAIC: Employment — Annually; Data on the number of employed persons in private enterprises and self-employed individuals; Employment in private enterprises and self-employed individuals in both urban and rual areas; Data starting from 1990.
- Notes on collection and definitions:
  - Employment and wage data are collected and compiled through the Reporting Form System on Labour Statistics, the Sample Survey System on Labour Force, and the System of Rural Social and Economic Surveys (NBS, Department of Population and Employment Statistics).
  - Data on the employment services and the change of labour force and on the number of registered unemployed persons in urban areas are collected through the Reporting Form System on Training and Employment Statistics (MoHRSS).

### Survey-based and micro datasets (coverage, years, and remarks)
- Urban Household Survey (UHS):
  - Detailed income and expenditure information; Annually; Covers nationwide.
  - Limited availability to academics for a few years and a few provinces.
- Rural Household Survey (RHS):
  - Detailed income and expenditure information; Annually; Covers nationwide.
  - Limited availability to academics for a few years and a few provinces.
- China Income Project Surveys (CHIPs) — The China Institute for Income Distribution of Beijing Normal University:
  - Detailed data on individual income and labor market information.
  - Years: 1988, 1995, 2002, 2007.
  - Coverage: A series of repeated cross-sections for year 1988, 1995 (for 6 provinces) and 2002 (11 provinces), 2007. In 2002, it covers around 15000 rural and urban households in 11 provinces and it also includeds 2000 non-random sampling of migrant workers.
  - Limited availability.
- China Health and Nutrition Survey (CHNS) — Carolina Population Center at the University of North Carolina at Chapel Hill and the National Institute of Nutrition and Food Safety at the Chinese Center for Disease Control and Prevention:
  - Detailed data on individual economic, demographic, social factor, health and nutritional status.
  - Years: 1989, 1991, 1993, 1997, 2000, 2004, 2006, 2009, 2011.
  - Panel data for 1989, 1991, 1993, 1997, 2000, 2004 and 2006. Covers 7 provinces and total of 4400 households, including rural and urban samples, but without migrants.
  - Limited availability.
- China Health and Retirement Longitudinal Study (CHARLS) — National School of Development, Peking University:
  - Detailed income and health information of middle-age and elderly people who are over 45 years old.
  - Starting from 2011, every two years.
  - Covers about 17000 persons in 10000 households.
  - Limited availability.
- China Urban Labor Survey (CULS) — Chinese Academy of Social Sciences (CASS):
  - Detailed labor information.
  - Years: 2001, 2005, 2010.
  - Covers five cities with less than 3000 households, including urban and migrant households. Repeated cross-sections for 2001, 2005 and 2010.
  - Limited availability.
- Rural-Urban Migration in China and Indonesia (RUMiCI) — Australian National University:
  - Detailed labor information; Initiated in 2008.
  - Consists of three samples in China: 8000 rural hukou households, 5000 urban hukou households, and 5000 migrant households, in 15 cities in 9 provinces.
  - Limited availability.

### Key findings and observations from annex narrative (SOEs and labor market implications)
- Role of SOEs and labor redundancy:
  - State-owned enterprises (SOEs), despite their shrinking role in the economy, provide insight into China’s economic transition and vulnerabilities.
  - Some SOEs ran losses in their core businesses, motivating diversification into new (non-core) areas even if such investments could be unprofitable.
  - High levels of surplus labor suggest that overall labor market conditions might not be as resilient as the unemployment rate would suggest.
- Case: large steel SOE in Hebei (overcapacity sector):
  - The firm has not scaled back production or employment; instead it expanded vertically and diversified into finance and real estate.
  - Faced with surplus labor — for example, as much as half of current employment at the Tangshan plant.
  - Social considerations constrain layoffs; the firm intends to create new employment over time by venturing into new business activities (e-commerce, for instance).
  - SOEs enjoy preferential access to finance from the biggest banks (loose credit limits without collateral and the ability to borrow at below benchmark rates) and have increased their financing abroad.
- Case: medium-sized textile SOE in Hebei (“mini China in transition”):
  - Output of cotton yard and textile cloth has fallen by half.
  - The company upgraded machinery to improve quality and productivity.
  - Employment: hired about 8,000 workers in 2014, down from the peak of 30,000 in 2010.
  - Social responsibility and adjustment outcomes:
    - About one-third of redundant workers went back to their rural homes, taking a lump sum package when they left.
    - About one-third was reemployed in nearby services, often with comparable or higher wages.
    - The SOE offered a buy-out package to older workers, paying them 80 percent of the minimum wage for five years until they reached retirement age.
  - Rising wages put pressure on competitiveness; the company cites options to cope by moving production plants to rural areas and upgrading machinery.
  - The SOE occupied sizable land resources (with substantial unrealized gains), which could be pledged to finance losses, leased, or sold to generate revenues.

### Hukou reforms under the Third Plenum blueprint (Annex 3 summary)
- Reform objective and targets:
  - Government actions in August 2014 to phase out the household registration system (hukou) dividing urban and rural households.
  - Ultimate objective: give 100 million migrants residency status in cities by 2020, in line with the urbanization target of 60 percent.
  - Reforms envisage providing migrants with better access to health and education benefits in cities; financing of additional spending remains uncertain.
  - As of April 2015, 14 provinces have issued work plans to implement reforms, but few at coastal areas that are more attractive for migrants.
- Planned components under the current plan:
  1. Fold the current hukou system into a standard residency status.
  2. Put in place a scheme that determines quotas and settlement arrangements for cities.
  3. Expand social services and gradually equalize benefits between residents and migrants.
- Residency status and residency identity:
  - Residency in mega cities such as Beijing and Shanghai will be strictly controlled; migrants may not obtain residency status even after five years of having lived there.
  - Individuals who live in other large cities outside their residency status location for more than half a year can apply for a residency identity, but will not yet be granted residency status in that city.
  - Residency identity allows migrants and their dependents to enjoy the same employment treatment (in principle), and basic education and health care benefits, as those with residency status.
  - As migrants gradually fulfill the conditions for residency status, they become eligible for social benefits such as housing and unemployment insurance.
- Fiscal and institutional considerations:
  - Hukou reforms will need to be accompanied by fiscal, social security, and rural land reforms.
  - The government will continue to rely on residency status as a policy tool.
  - Fiscal implications of expanded coverage and its financing, and the criteria set by cities to attract or restrict migrant flows, are uncertain.
  - Local government revenues must be better aligned with spending responsibilities, including intergovernmental transfers.
  - The government intends to provide consolidated basic pensions and basic health care nationally to improve portability.

### Annex Table 3.1 — Summary of Settlement Schemes and Quotas for Cities (selected thresholds and rules)
- City-level population categories and openness (verbatim entries and conditions):
  - Towns and small cities: County-level communities <500,000 — Fully-open.
    - Conditions (verbatim): Anyone who lives in a legal stable residential unit (including rental unit); Legal and stable employment 1/; Live in a legal and stable residential unit (including rental unit) 1/; participate in city social security system for certain years (up to 3 years).
  - Middle-level cities: Between 500,000 to 1 million — Graduallly open.
  - Large cities: Between 1 million to 3 million — Gradually open.
  - Large cities: Between 3 million to 5 million — Graudally open but controls on the scale and pace.
  - Metropolitans: 5 million or above — Strcit controls on the population scale.
- Point-based system and criteria (verbatim):
  - A point-baesd system for granting residency status based on:
    - Legal and stable employment up for a certain period.
    - Live in a legal and stable residential unit (including rental unit) 1/.
    - participate in city social security system for certain years.
    - requires consecutive living duration.
  - Additional verbatim notes:
    - Same as large cities with 3 million or less but with tighter conditions on employment and residential units.
    - participate in city social security system for certain years (up to 5 years).
    - May introduce a point-based system to obtain residency.
  - Footnotes (verbatim):
    1/ The preicse definition and duration of employment and living area (except square footage and price) will be set in accordance to individual cities.
    2/ The applicant and spouse who lives together, and their dependent children and parents can register for residency status.

*Source: Annex Table 1.1 and related annex text in the provided IMF content unit.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15151.pdf_