## wpiea2022196-print-pdf

## Source details

**Canonical URL:** [wpiea2022196-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022196-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022196-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022196-print-pdf.pdf.json)

---

### 3.1 City tier classification and coverage
- First-tier cities: Beijing, Shanghai, Guangzhou, and Shenzhen.
- 35 large and medium-sized cities monitored by the National Bureau of Statistics include:
  - two municipalities directly under the central government (Tianjin, Chongqing),
  - five cities under separate state planning,
  - twenty-four provincial capitals.
- All other cities are classified as third-tier cities.
- Number of cities: 657 in 2010 and 685 in 2020 (China City Statistical Yearbook).
- Third-tier cities account for roughly 60 percent of GDP (Appendix Figure 1).
- No official government list defines tier membership; classification aligns with other groupings by GDP, income, or population.
- Data limitations:
  - Regional input-output and inter-province trade data are not available for full general equilibrium decomposition at city-tier level.
  - Provincial records, improved National Accounts data, and the NBS 35-city monitoring data are used to construct tier-level estimates.

### Size and composition of the housing sector
- Direct value added share of real estate activity in GDP:
  - 2016: Total real estate activity 11.3%
  - 2017: Total real estate activity 11.9%
  - 2018: Total real estate activity 12.4%
  - 2019: Total real estate activity 12.2%
  - 2020: Total real estate activity 12.4%
  - 2021: Total real estate activity 11.8%
- Total final demand share (domestic content only) for real estate activity:
  - 2016: Total final demand 24.7%
  - 2017: Total final demand 25.4%
  - 2018: Total final demand 26.0%
  - 2019: Total final demand 25.9%
  - 2020: Total final demand 26.1%
  - 2021: Total final demand 25.4%
- Real estate construction (direct value added / total final demand):
  - 2016: 4.8% / 16.9%
  - 2017: 4.8% / 16.9%
  - 2018: 5.0% / 17.3%
  - 2019: 5.0% / 17.5%
  - 2020: 5.1% / 17.8%
  - 2021: 5.0% / 17.3%
- Real estate services (direct value added / total final demand):
  - 2016: 6.5% / 5.0%
  - 2017: 7.1% / 5.5%
  - 2018: 7.4% / 5.6%
  - 2019: 7.1% / 5.5%
  - 2020: 7.3% / 5.6%
  - 2021: 6.8% / 5.2%
- Imported content increases total final demand by approximately 3 percent (e.g., sector size 26.0 percent in 2018; 25.4 percent in 2021 when imports included).
- Companies and workers in real estate-related activities accounted for 12 percent of China’s GDP in 2021 (direct value added column).
- The total domestic share of real estate in China’s GDP exceeds comparable pre-crisis peaks for Spain and Ireland and exceeds the United States (harmonized OECD input-output comparison referenced).

### Housing construction and per-capita metrics
- Urban residential floor space growth: approximately 6 percent annual growth from 2011–21.
- Urban residential floor space levels:
  - 2010: 18.6 billion square meters.
  - 2021: 35.6 billion square meters.
- Per capita residential floor space in urban areas:
  - 2010: 27.8 square meters.
  - 2021: around 39.0 square meters.

### Estimating housing stock by city tier — methodology and assumptions
- Census data provide per capita living space every 10 years; depreciation rates not explicitly stated in census.
- Method steps:
  - Use construction completed data to estimate added space between 2010 and 2020 censuses.
  - Account for depreciation and replacement of older units by new construction.
  - Restore historical housing stock between census years and project housing stock up to the latest month using higher frequency data.
  - Perform sensitivity analysis to different depreciation assumptions.
- 2010 and 2020 housing stock calculation based on census data via an explicit equation (equation (1) in source).
- Extension to 2021 uses Equation (2): add annual residential housing completed and subtract depreciation; annual completed floor space taken as the larger of fixed assets investment residential housing completed and construction sector residential housing completed.
- Treatment of collectives:
  - Per capita living space for collectives estimated approximately 8 m2 in 2020 (one-fifth of family households), based on building standards.
  - Bounds of 4–16 m2 for collectives change results by less than 2 percent.

### Housing stock estimation results and tier breakdown
- Census vs construction-based totals for 2020:
  - Census-based total housing stock in 2020: 54,909 million square meters.
  - Year-by-year construction-based estimate in 2020 (baseline depreciation): 54,922 million square meters.
- Baseline depreciation and lifespan assumptions:
  - Historical housing lifespan assumed: 50 years.
  - New housing construction lifespan assumed: 70 years.
  - Translates into an annual depreciation rate C ranging from 1.4–2.0 percent.
  - Alternative inferred annual depreciation rate: 0.7 percent implies a 2020 housing stock of roughly 62 billion square meters (10–15 percent higher than baseline).
- Total housing stock:
  - 2010: 39 billion square meters.
  - 2021: 56 billion square meters.
- Shares of total housing stock by tier:
  - Tier 1 cities: 4 percent of total housing stock.
  - Tier 2 cities: 18 percent of total housing stock.
  - Tier 3 cities: over 78 percent of total housing stock.
- Urban market characteristics:
  - Over 96 percent of residential property projects are executed in urban regions (98 percent in 2010).
  - Over 90 percent of rural dwellings are self-built and off market (95 percent in 2010).
  - Tier 3 cities account for more than 72 percent of urban housing stock.
  - Tier 3 cities host 66 percent of the urban population.
- Construction production value by city tier (2020):
  - Tier 1: 12 percent of national construction production by value.
  - Tier 2: 8 percent of national construction production by value.
  - Tier 3: 80 percent of national construction production by value.
- City-level estimation details:
  - Per capita living space and total population in 2010 collected for tier 1 and tier 2 cities from the 2010 national census and provincial censuses.
  - City-level residential housing completed data used for 2010–2021; where only developer-reported data available, ratios inferred from four municipalities were applied to other cities.

### Housing prices, valuation changes, and market dynamics
- Tier 3 price imputation: subtract tier 1 and tier 2 housing prices (weighted by their floor space of housing sold) from the national average housing price.
- Tier 3 sales share rose to around 70 percent by end-2020.
- Recent price developments (early 2021 to mid-2022):
  - Tier 3 cities experienced almost a 20 percent drop in residential housing prices between early 2021 and mid-2022.
  - Prices rose modestly in tier 1 cities; much smaller rise in tier 2 cities.
  - NBS monthly indices suggest prices of pre-owned residential dwellings declined even more sharply than newly-built homes.
- Valuation impact:
  - Tier 3 housing market valuation declined by 15 percent, from 188 trillion yuan in early 2021 to 162 trillion yuan in mid-2022 (not adjusted for inflation).
- Data limitations:
  - NBS releases monthly housing price data for seventy large- and medium-sized cities only (4 tier 1, 31 tier 2, 35 representative tier 3).
  - House price sale data are difficult to interpret due to changing mix and regions of home sales.

### Housing sector vulnerabilities in tier 3 cities
- Population trends:
  - Population growth across tier 3 cities was flat to slightly negative prior to the pandemic and fell by 2 percent in 2021.
  - UN-based medium variant projects China’s population peak in 2025 (Appendix Figure 9 referenced).
- Projected housing demand over 2022–35:
  - Urban housing demand projected to decline by about 3 percent annually.
  - In this estimate, roughly 60 percent of demand is driven by replacement needs.
  - If depreciation is lower (e.g., 0.7 percent), the fall in urban housing demand would be larger.
- Unfinished buildings and completion rates:
  - Scale of housing under construction in tier 3 cities in 2020 was 10.6 times annual housing completed in 2020 (ratio was just over 6 in 2011).
  - Large scale of uncompleted buildings likely reflects developers’ financing constraints or project delays (disputes over property rights, quality issues, etc.).
  - Low completion rates also affect tier 2 cities.
  - The 2022 mortgage boycott in multiple tier 2 cities reflected risks posed by unfinished buildings.

### 4.3 Debt accumulation behind the construction boom
- Construction sector leverage:
  - Debt levels defined as total liabilities over total assets (including projects under construction) of registered construction enterprises dwarf those of the manufacturing sector.
  - Leverage accelerated rapidly since 2016 despite efforts to curb lending.
  - Construction enterprises in tier 3 cities had debt to assets near almost 80 percent in 2020.
- Evergrande bankruptcy highlighted high indebtedness among construction firms and concerns that official mortgage debt figures may exclude shadow bank debt.
- Links to fiscal, household, and labor outcomes:
  - Tier 3 local government reliance on non-tax land revenues: approximately 43 percent of local government fiscal revenues in 2021.
  - PBOC (2020): 59.1 percent of total household assets are housing assets in 2019.
  - China Household Finance Survey (2019): housing wealth accounts for over 70 percent of total household wealth in tier 3 cities.
  - Tier 3 cities host close to 60 percent of housing under construction.
  - Tier 3 cities host over 90 percent of construction enterprises which hire over 85 percent of construction workers.
  - Many construction enterprises in tier 3 are quite small compared to firms in larger cities.
- Implication: Demand constraints and tier 3 reliance on housing for wealth and employment imply a challenging adjustment affecting fiscal revenues, household balance sheets, and labor markets.

### V. Infrastructure development — patterns and risks
- Infrastructure investment patterns:
  - 2003–12: transport, storage, and post accounted for over 40 percent of all fixed assets investment (largest item of infrastructure construction projects).
  - Tier 3 cities account for roughly 60 percent of national GDP but:
    - 88 percent of all roads,
    - 92 percent of new road construction.
  - Expressways and pipelines also show disproportionate shares in tier 3 cities (stock and flow).
- High-speed rail network issues:
  - China has the world’s longest and most extensively used high-speed rail network, but railway companies took on large debt.
  - Interest expenses and high operating and maintenance costs render China State Railway Group in consecutive losses.
  - Many of the 142 high-speed railways often unable to cover interest expenses and operating and maintenance costs.
  - Expansion continues at a rate that far outpaces passenger growth.
  - Externalities from high-speed rail (agglomeration, network effects, knowledge spillovers) are conditional on city size, industry structure, amenities, and distance from urban centers.
  - High-speed railways disproportionately built in tier 3 cities; low capacity utilization reflects difficulties moving people to less populated regions.
- New infrastructure push in response to COVID-19:
  - Infrastructure investment estimated to surge by 8 percent in 2022 (compared with 0.4 percent increase in 2021).
  - Marginal growth contribution of investment is already in decline; projects in transportation, energy and water conservancy described as already in excess.
  - Investing in supercomputing, cloud computing and artificial intelligence could be cost-effective, but high debt limits policy space for large-scale public works.
  - Environmental constraint: production of steel, cement and other materials generates the lion’s share of China’s CO2 emissions; these industries will have to shrink rather than expand for China to hit its target of capping carbon emissions by 2030.

### I. Methodology — decomposition and projections (2022–2035)
- Housing demand decomposition: total residential floor space growth measured as product of population and per-capita residential floor space, adjusted for urbanization, plus home improvement and replacement needs (equation in source).
- Projection horizon: 2022–2035.
- City-tier approach:
  - Tier 1 and tier 2 city projections use city authority estimates and development plans.
  - Tier 3 outcomes derived residually by deducting tier 1 and tier 2 projected populations and urban populations from national totals.
- Population growth context and projections:
  - 2021: China added only 2 million new births; reported population 1.413 billion, up by 0.034 percent from end 2020 (slowest natural growth rate in six decades).
  - City examples and targets:
    - Beijing: population reached 21.9 million in 2021; Overall Urban Planning (2016–2035) aims to limit population to 23 million in 2030.
    - Shanghai: 24.9 million in 2020; 2030 target at 25 million.
    - Guangzhou: 18.8 million in 2021; plans to increase to 20 million in 2035 (average annual growth 0.4 percent).
    - Shenzhen: population 17.6 million in 2020; 2035 target 19 million.
  - For tier 2 cities:
    - 23 out of 31 have official population targets used through 2035; average projected growth applied to remaining 8 tier 2 cities.
  - National population projection based on UN 2022 Revision medium variant (study assumes China’s population will peak in 2031 based on UN medium variant).
  - Spatial trend: decline in lower-tier cities and moderate increase in tier 1 and tier 2 cities.
- Urbanization assumptions:
  - National aim: 80 percent urbanization by 2050.
  - Study assumption: urbanization rate reaches 75 percent in 2035.
  - City-level urbanization reported:
    - Shenzhen: 100 percent in 2010.
    - Beijing: 88 percent.
    - Shanghai: 89 percent.
    - Guangzhou: 86 percent.
    - Tier 2 cities (as whole): average urbanization rate 79 percent in 2021.
  - Projections:
    - Tier 1 and tier 2 converge to 90 percent urbanization over next 10–15 years.
    - Tier 3 reach around 66 percent urbanization rate in 2035.
    - National average urbanization projected at 75 percent in 2035.
- Home improvement needs:
  - Per-capita residential floor space assumed to continue expanding to 50 square meters across all city tiers and then stabilize.
  - Accounting for shared area, actual residential floor space will reach approximately 40 square meters.
  - Shared area details: up to 30 percent depending on design; multi-story apartments 8–15 percent shared; high-rise 10–30 percent shared.
- Replacement needs and depreciation:
  - Residences built before the 2000s: 30-year depreciation.
  - Residences built after the 2000s: 50-year depreciation.
  - Replacement needs account for over 60 percent of housing demand over the next 14 years (Appendix Figure 11).
- Projected housing demand 2022–35:
  - Aggregate: demand for new urban housing projected to fall from an average of 183 billion square feet in the 2010s to 168 billion square feet by 2035 (Appendix Table 2).
  - Aggregate annual new construction may decline by approximately 3 percent starting from 2022.
  - Tier divergence:
    - Tier 1 and tier 2 may experience a modest increase in housing construction.
    - Tier 3 may see an annual decline of roughly 6 percent.
  - Two additional scenarios considered for possible overshooting of population targets (Scenario 2 assumes tier 1 cities implement stricter population control).
  - Trade-off: expanding tier 1 cities versus developing tier 3 cities involves resource and service pressures and risks of lower-tier decline; restricting large cities fights Zipf’s law and is difficult.

### Selected data sources cited
- United Nations: 2022 Revision of World Population Prospects (medium variant).
- Chinese statistical and census sources: 2020, 2010, 2000 Chinese National Census; 2021 and 2020 city- and provincial-level Statistical Communiqués and Census Communiqués; National Bureau of Statistics of China; China Statistical Yearbooks 2011–21; provincial-level Statistical Yearbooks 2011–21.
- City-level Development Plans 2016–35 used for urbanization and population targets.

*Source: wpiea2022196-print-pdf - IMF Working Paper excerpts (3.1; 4.3; V; VI; 1. Methodology).*

### 3.1 City Tier Classification............................................................................................

### 3.1 City Tier Classification

### City tier definitions and coverage
- Beijing, Shanghai, Guangzhou, and Shenzhen are identified as first-tier cities.
- The National Bureau of Statistics reports main indicators of real estate projects in thirty-five large and medium-sized cities; beyond the four first-tier cities, this group includes:
  - two municipalities directly under the central government (Tianjin, Chongqing),
  - five cities under separate state planning,
  - twenty-four provincial capitals.
- All other cities are classified as third-tier cities.
- There were 657 cities in China in 2010 and 685 cities in 2020 (China City Statistical Yearbook).

### Economic weight of tiers
- Third-tier cities account for roughly 60 percent of GDP (Appendix Figure 1).

### Motivation for tier disaggregation
- Many studies focus on tier 1 and tier 2 cities; this paper emphasizes that tier 3 cities are collectively even more important in terms of GDP and especially in terms of real estate construction.
- City size in China does not follow Zipf’s law as tightly as in many other countries; smaller and medium-size tier 3 cities have a larger weight in GDP and population than in countries like the United States.

### Data limitations and approach
- No official government list defines city tier membership; the tier classification used here aligns with other grouping methods based on GDP, income level, or population size.
- Regional input-output and inter-province trade data are not available, preventing a full general equilibrium decomposition at city-tier level.
- The paper uses provincial records, improved data from China’s National Accounts, and the National Bureau of Statistics’ 35-city monitoring data to construct tier-level estimates.

---

### Size of the overall housing sector (context)
- The “direct value added” share of real estate activity in GDP:
  - 2016: Total real estate activity 11.3%
  - 2017: Total real estate activity 11.9%
  - 2018: Total real estate activity 12.4%
  - 2019: Total real estate activity 12.2%
  - 2020: Total real estate activity 12.4%
  - 2021: Total real estate activity 11.8%
- The “total final demand” share (domestic content only) for real estate activity:
  - 2016: Total final demand 24.7%
  - 2017: Total final demand 25.4%
  - 2018: Total final demand 26.0%
  - 2019: Total final demand 25.9%
  - 2020: Total final demand 26.1%
  - 2021: Total final demand 25.4%
- Real estate construction (direct value added / total final demand):
  - 2016: 4.8% / 16.9%
  - 2017: 4.8% / 16.9%
  - 2018: 5.0% / 17.3%
  - 2019: 5.0% / 17.5%
  - 2020: 5.1% / 17.8%
  - 2021: 5.0% / 17.3%
- Real estate services (direct value added / total final demand):
  - 2016: 6.5% / 5.0%
  - 2017: 7.1% / 5.5%
  - 2018: 7.4% / 5.6%
  - 2019: 7.1% / 5.5%
  - 2020: 7.3% / 5.6%
  - 2021: 6.8% / 5.2%
- Adding imported content would increase total final demand by approximately 3 percent, bringing the size of the sector to 26.0 percent in 2018, and 25.4 percent in 2021.
- The paper reports that companies and workers engaged in all types of real estate related activities accounted for 12 percent of China’s GDP in 2021 (direct value added column).
- The total domestic share of real estate in China’s GDP exceeds comparable pre-crisis peaks for Spain and Ireland, and also exceeds the United States (as confirmed using harmonized OECD input output tables referenced in the appendix).

### Housing construction and per-capita metrics
- Residential floor space in China’s urban areas increased at an annual growth rate of approximately 6 percent from 2011–21.
- In 2021, total urban residential floor space reached 35.6 billion square meters, up from 18.6 billion square meters in 2010.
- Per capita residential floor space in urban areas rose from 27.8 square meters in 2010 to around 39.0 square meters in 2021.

---

### Estimating housing stock by city tier: methodology overview
- Census data provide per capita living space but are only available every 10 years and do not state depreciation rates explicitly.
- To calculate housing stock between census readings, the paper:
  - Obtains data on construction completed across China to estimate added space between the 2010 and 2020 censuses,
  - Accounts for depreciation and replacement of older units by new construction,
  - Restores historical housing stock between census years and projects housing stock up to the latest month possible using higher frequency data,
  - Allows sensitivity analysis to different depreciation assumptions.
- Step 1: Calculate China’s housing stock in 2010 and 2020 based on census data using an explicit equation (equation (1) defined in the text).

---

*Source: wpiea2022196-print-pdf - 3.1 City Tier Classification*

### Part I,  Volume 1 of the census contains information on the total number of individuals living in rural/urban

### wpiea2022196-print-pdf - Part I,  Volume 1 of the census contains information on the total number of individuals living in rural/urban

### Housing stock estimation methodology
- Census and year-by-year construction methods produced nearly identical 2020 totals:
  - Census-based total housing stock in 2020: 54,909 million square meters.
  - Year-by-year construction-based estimate in 2020 (baseline depreciation): 54,922 million square meters.
- Baseline depreciation and lifespan assumptions:
  - Historical housing lifespan assumed: 50 years.
  - New housing construction lifespan assumed: 70 years.
  - Translates into an annual depreciation rate C ranging from 1.4–2.0 percent.
  - Alternative inferred annual depreciation rate: 0.7 percent would imply a 2020 housing stock of roughly 62 billion square meters (10–15 percent higher than baseline).
- Extension of annual series:
  - Using 2010 as the base year, the time series was extended to 2021 by adding annual residential housing completed and subtracting depreciation (Equation (2) in source).
  - Annual floor space of residential housing completed taken as the larger of fixed assets investment residential housing completed and construction sector residential housing completed.
- Treatment of collectives:
  - Individuals are classified as belonging to family households or collectives (student dormitories, nursing homes, workers’ hostels, military barracks, etc.).
  - Per capita living space for collectives estimated to be approximately 8 m2 in 2020 (one-fifth of that of family households), based on building standards (e.g., JGJ 36-2016, GB 50099-2011, GB 50867-2013, Building Space Instructions for Higher Education Institutions 191-2018, military barracks standards).
  - Using bounds of 4–16 m2 for collectives would change results by less than 2 percent.

### Housing stock and construction by city tier — key statistics
- Aggregate and temporal totals:
  - Total housing stock in China reached 56 billion square meters in 2021, up from 39 billion square meters in 2010.
- Shares of total housing stock by tier (consistent across period):
  - Tier 1 cities: 4 percent of total housing stock.
  - Tier 2 cities: 18 percent of total housing stock.
  - Tier 3 cities: over 78 percent of total housing stock.
- Urban market characteristics:
  - Over 96 percent of residential property projects are executed in urban regions (commodity dwellings built by property developers in urban regions as a share of commodity dwellings). The figure was 98 percent in 2010.
  - Over 90 percent of rural dwellings are self-built and off market. The figure was 95 percent in 2010.
  - Tier 3 cities account for more than 72 percent of urban housing stock.
  - Tier 3 cities host 66 percent of the urban population.
- Construction production value by city tier (2020):
  - Tier 1: 12 percent of national construction production by value.
  - Tier 2: 8 percent of national construction production by value.
  - Tier 3: 80 percent of national construction production by value.
- City-level housing stock estimation approach:
  - Per capita living space and total population in 2010 collected for tier 1 and tier 2 cities from the 2010 national census and provincial censuses.
  - City-level residential housing completed data used for 2010–2021; where only developer-reported data available, ratios inferred from four municipalities were applied to other cities.

### Housing prices, valuation changes, and market dynamics
- Imputation of tier 3 housing prices:
  - Tier 3 housing price developments imputed by subtracting tier 1 and tier 2 housing prices (weighted by their floor space of housing sold) from the national average housing price.
  - Tier 3 cities accounted for a rising share of housing sold, reaching around 70 percent by end-2020.
- Recent price developments (early 2021 to mid-2022):
  - Tier 3 cities experienced almost a 20 percent drop in residential housing prices between early 2021 and mid-2022.
  - Prices rose modestly in tier 1 cities; much smaller rise in tier 2 cities.
  - NBS monthly indices suggest prices of pre-owned residential dwellings declined even more sharply than newly-built homes.
- Valuation impact:
  - Tier 3 housing market valuation declined by 15 percent over the period, from 188 trillion yuan in early 2021 to 162 trillion yuan in mid-2022 (not adjusted for inflation).
- Data limitations and cautions:
  - National Bureau of Statistics releases monthly housing price data for seventy large- and medium-sized cities only (4 tier 1, 31 tier 2, 35 representative tier 3).
  - House price sale data are difficult to interpret due to changing mix and regions of home sales.

### Housing sector vulnerabilities in tier 3 cities
- Population trends and demand projections:
  - Population growth across tier 3 cities was flat to slightly negative prior to the pandemic and fell by 2 percent in 2021.
  - United Nations-based medium variant projects China’s population peak in 2025 (Appendix Figure 9 referenced).
  - Projections of housing demand over 2022–35 suggest urban housing demand will decline by about 3 percent annually.
    - In this estimate, roughly 60 percent of demand is driven by replacement needs.
    - If depreciation is lower (e.g., 0.7 percent as used in 2010 census reconstruction), the fall in urban housing demand would be larger.
- Unfinished buildings and construction completion rates:
  - Scale of housing under construction in tier 3 cities in 2020 was 10.6 times annual housing completed in 2020.
  - The same ratio was just over 6 in 2011.
  - Large scale of uncompleted buildings likely reflects developers’ financing constraints or project delays (disputes over property rights, quality issues, etc.).
  - Low completion rates also affect tier 2 cities.
  - The 2022 mortgage boycott in multiple tier 2 cities was a reflection of risk posed by unfinished buildings.

*Source: wpiea2022196-print-pdf - Part I, Volume 1 of the census contains information on the total number of individuals living in rural/urban*

### 4.3  Debt Accumulation behind  the Construction Boom

### 4.3  Debt Accumulation behind the Construction Boom

### Construction sector leverage and risks
- Construction sector debt levels, defined as the ratio of total liabilities over total assets (including projects under construction) of all registered construction enterprises, dwarf those of the manufacturing sector.
- The sector has taken on leverage rapidly since 2016 despite government efforts to curb overall lending and an accepted consensus of slower trend growth and declining housing demand.
- Compared with construction enterprises in tier 1 and tier 2 cities, those in tier 3 cities are more heavily leveraged, with a debt to assets level reaching almost 80 percent in 2020.
- The Evergrande bankruptcy highlighted high indebtedness among construction firms and raised concerns that official mortgage debt figures may not include debt from shadow banks.

### Housing sector links to fiscal, household, and labor outcomes
- The housing sector is described as China’s pillar industry, vital to sustaining fiscal incomes, household wealth and employment; a significant deceleration will place pressure on fiscal sustainability, financial stability, and the labor market, especially for tier 3 cities.
- Local government reliance on housing-related revenues:
  - In 2021, the ratio of non-tax land revenues amounted to approximately 43 percent of local government fiscal revenues in tier 3 cities.
  - Overreliance on land sales (windfall in nature) is volatile and unsustainable and creates incentives to overinvest in real estate and push up land and housing prices.
- Importance for household balance sheets:
  - People’s Bank of China (2020) finds 59.1 percent of total household assets are housing assets in 2019.
  - China Household Finance Survey (Southwestern University of Finance and Economics, 2019) estimates housing wealth accounts for over 70 percent of total household wealth in tier 3 cities.
- Employment and firm composition:
  - Tier 3 cities are home to close to 60 percent of housing under construction.
  - Tier 3 cities host over 90 percent of construction enterprises which hire over 85 percent of construction workers.
  - Many construction enterprises in tier 3 are quite small compared to firms in larger cities.

### Implications
- Given the demand constraints projected for the housing sector and the outsized reliance of tier 3 cities on housing for wealth and employment, the required adjustment could be challenging and would affect fiscal revenues, household balance sheets, and labor markets.

---

### V. Infrastructure Development

### Infrastructure investment patterns
- From 2003–12, investment in transport, storage, and post accounted for over 40 percent of all fixed assets investment, representing the largest item of infrastructure construction projects (though recently environmental conservation expenditures have taken over).
- A disproportionate scale of infrastructure construction in tier 3 cities for transport and storage:
  - Tier 3 cities account for roughly 60 percent of national GDP but account for 88 percent of all roads and 92 percent of new road construction.
  - Expressways and pipelines also show disproportionate shares in tier 3 cities (stock and flow).

### High-speed rail network issues
- China has the world’s longest and most extensively used high-speed rail network, but major problems exist:
  - Chinese railway companies took on large amounts of debt to build the network.
  - Interest expenses and high operating and maintenance costs render China State Railway Group in consecutive losses.
  - Many, if not all, of the 142 high-speed railways are often unable to cover interest expenses, let alone operating and maintenance costs.
  - Despite excess capacity and heavy financial losses, China continues expanding its high-speed railway network at a rate that far outpaces passenger growth.
- Externalities and conditions:
  - High-speed rail can generate externalities through agglomeration, network effects, knowledge dissemination, and technology spillovers, but these effects are conditional on factors such as city size, industry structure, complementary amenities, and distance from urban centers.
  - High-speed railways have been disproportionately built in tier 3 cities, and low capacity utilization reflects difficulties moving people to less populated regions.

### New infrastructure push in response to COVID-19
- China’s infrastructure investment is estimated to surge by 8 percent in 2022, sharply higher than the 0.4 percent increase seen in 2021, with the aim of lifting the economy out of COVID-19 recession.
- The marginal growth contribution of investment is already in decline; projects in transportation, energy and water conservancy are described as already in excess.
- Investing in infrastructure for supercomputing, cloud computing and artificial intelligence could be cost-effective, but the already high debt load limits the policy space for large-scale public works.
- Environmental constraint:
  - Production of steel, cement and other materials generates the lion’s share of China’s CO2 emissions. These industries, much needed for the new infrastructure push, will have to shrink rather than expand for China to hit its target of capping carbon emissions by 2030.

---

### VI. Conclusion (Key synthesis)
- Tier 3 cities account for roughly 60 percent of GDP and an even larger share of real estate construction, yet have been relatively neglected in the literature.
- Prices in tier 3 cities did not soar like in tier 1 cities and are now falling faster.
- Estimates indicate the housing sector is very likely to be demand constrained for a long time, creating a challenging adjustment given tier 3 cities’ outsized reliance on housing for wealth and employment.
- Infrastructure faces similar overbuilding risks, particularly in tier 3 cities that account for 60 percent of China’s GDP.

*Source: IMF Working Paper — "A Tale of Tier 3 Cities", section 4.3–VI (excerpts provided).*

### 1.    Methodology

### 1.    Methodology

### Decomposition of housing demand
- Housing demand, measured by total residential floor space, is decomposed into four driving forces: population growth, urbanization, home improvement needs, and replacement needs.
- The model measures total residential floor space growth as the product of population and per-capita residential floor space, adjusted for urbanization, plus additions from home improvement and replacement needs (equation shown in source).

### Projection horizon
- Projections and factor evolutions are illustrated for the period 2022–2035.

### City-tier approach
- Population and urbanization projections are done at the city level for tier 1 and tier 2 cities using estimates published by local authorities and city development plans.
- Tier 3 city outcomes are derived residually by deducting tier 1 and tier 2 projected populations and urban populations from national totals.

---

### Population Growth
- China’s population growth has slowed considerably, falling behind the trend in the most pessimistic scenario of UN population projections (Appendix Figure 9).
- In 2021:
  - China added only 2 million new births.
  - Reported population: 1.413 billion, up by 0.034 percent from end 2020 (the slowest natural growth rate in six decades).
- Projection choices:
  - City-level projections for tier 1 and tier 2 cities use authorities’ population targets where available and set projection year to 2035.
  - Example city targets and levels:
    - Beijing: population reached 21.9 million in 2021; Beijing Overall Urban Planning (2016–2035) aims to limit population to 23 million in 2030.
    - Shanghai: 24.9 million in 2020; 2030 target at 25 million.
    - Guangzhou: 18.8 million in 2021; plans to increase to 20 million in 2035, implying an average annual growth rate of 0.4 percent.
    - Shenzhen: over 5 percent annual population growth in past decade; population 17.6 million in 2020; sets 2035 target at 19 million.
  - For tier 2 cities:
    - 23 out of 31 have an official population target; authorities’ targets are used as projected populations in 2035.
    - The average projected population growth during 2022–35 for this group is applied to 8 other tier 2 cities without explicit targets.
  - Total population is projected based on UN’s medium variant scenario.
  - Based on the UN’s medium variant scenario, China’s population will peak in 2031.
- Spatial trends:
  - Geographic mobility will lead to a continuous decline in lower-tier cities and a moderate increase in tier 1 and tier 2 cities.
- Scenario note:
  - Despite recent underperformance relative to UN low-variant projections, the study opts for the medium-variant projection, considering possible policy measures to boost birth rates.

---

### Urbanization
- National target and modeling assumption:
  - China aims to achieve an 80 percent urbanization rate at the national level in 2050.
  - The study assumes China’s urbanization rate will reach 75 percent in 2035 (consistent with consensus projections cited).
- City-level urbanization levels (reported):
  - Shenzhen: urbanization rate reached 100 percent in 2010 (Shenzhen’s 2010 Population Census).
  - Beijing: urban population 88 percent.
  - Shanghai: urban population 89 percent.
  - Guangzhou: urban population 86 percent.
  - Tier 2 cities (as a whole): average urbanization rate 79 percent in 2021.
- Projection method:
  - For tier 1 and tier 2 cities, use each city’s development plans for urbanization targets when available.
  - For cities without official targets, assume urbanization rate follows current pace of growth.
  - Tier 3 urbanization rate is projected by subtracting projected urban populations in tier 1 and tier 2 cities from total urban population, then dividing by total tier 3 population.
- Projection outcomes:
  - Results (Appendix Figure 10) suggest:
    - Tier 1 and tier 2 cities will converge to 90 percent urbanization over the next 10–15 years.
    - Tier 3 cities will reach around 66 percent urbanization rate in 2035.
    - National average urbanization rate projected at 75 percent in 2035.

---

### Home Improvement Needs
- Per-capita residential floor space context:
  - China’s per-capita residential floor space appears comparable to that of most advanced economies, but “shared area” complicates measurement.
- Shared area details:
  - Shared area can be as high as 30 percent depending on design.
  - Multi-story apartments (4 to 8 stories): shared area ranges from 8 percent to 15 percent of residential floor space.
  - High-rise buildings: shared area ratio increases to 10–30 percent.
  - Shared area is counted toward residential floor space but reduces actual usable floor space per person.
- Assumption for projections:
  - Per-capita residential floor space will continue to expand to 50 square meters across all city tiers and then stabilize.
  - Accounting for shared area, the actual residential floor space will reach approximately 40 square meters, comparable to most rich countries.

---

### Replacement Needs
- Housing vintage and depreciation assumptions:
  - Majority of Chinese homes were built around or after the 2000s; large-scale construction over the past three decades has substantially upgraded housing.
  - There remains a stock of substandard housing in urban areas without a kitchen or a toilet (mostly built before the 2000s).
  - Such substandard residences are assumed to depreciate faster and be replaced more quickly than homes built after the 2000s.
  - Depreciation assignments based on construction standards:
    - Residences built before the 2000s: 30-year depreciation.
    - Residences built after the 2000s: 50-year depreciation.
- Replacement share:
  - Replacement needs account for over 60 percent of housing demand over the next 14 years (Appendix Figure 11 notes).

---

### Projected Housing Demand in 2022–35
- Aggregate estimates:
  - Demand for new urban housing had peaked and is projected to fall from an average of 183 billion square feet in the 2010s to 168 billion square feet by 2035 (Appendix Table 2).
  - At the aggregate level, annual new construction may decline by approximately 3 percent starting from 2022.
- Tier divergence:
  - Tier 1 and tier 2 cities may experience a modest increase in housing construction due to divergent housing and migration policies.
  - Tier 3 cities may see an annual decline of roughly 6 percent.
- Scenario analysis:
  - Two additional scenarios are considered to account for possible overshooting of population targets.
  - Scenario 2 assumes tier 1 cities implement a stricter population control following the National Planning.
- Trade-offs highlighted:
  - Results illustrate a delicate trade-off between expanding tier 1 cities and developing tier 3 cities.
  - Letting population balloon in the largest cities would pressure resources and services and cause less developed tier 3 cities to wither.
  - Restricting the size of large tier 1 cities would be fighting Zipf’s law, which is difficult or impossible.

---

### Data sources (selected items cited in methodology and notes)
- United Nations:
  - 2022 Revision of World Population Prospects (used for population projections; medium variant scenario).
- Chinese statistical and census sources used across figures and calculations:
  - 2020 Chinese National Census; 2010 Chinese National Census; 2000 Chinese National Census.
  - 2021 and 2020 city- and provincial-level Statistical Communiqués and Census Communiqués for numerous cities and provinces (listed in Appendix III).
- National Bureau of Statistics of China; China Statistical Yearbooks 2011–21; provincial-level Statistical Yearbooks 2011–21.
- City-level Development Plans 2016–35 (used for city urbanization and population targets).

*Source: wpiea2022196-print-pdf - 1.    Methodology*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022196-print-pdf.pdf_