## _wp1584

## Source details

**Canonical URL:** [_wp1584](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1584.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1584.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1584.pdf.json)

---

### I. Introduction — role of real estate in China's growth
- Real estate investment grew from about 4 percent of GDP in 1997 to 15 percent of GDP in 2014.
- Residential investment accounts for about 15 percent of fixed asset investment and 15 percent of total urban employment.
- Bank lending to the sector accounts 20 percent of total loans.
- Real estate is a key source of local public finance and is extensively used as collateral for corporate borrowing.

### II. Market distortions and vulnerabilities
- Supply-side distortion:
  - Local governments’ reliance on land sales to finance spending.
- Demand-side distortions:
  - Housing attractive as financial investment due to a history of robust capital gains, real deposit interest rates that tended to be negative, lack of alternative financial assets, and capital account restrictions.

### III. Data sources and geographic coverage
- Main official sources:
  - NBS: commodity building average selling prices; 70 large and medium-sized price index; price indices at provincial and national levels; residential real estate fixed asset investment; floor space sold, unsold, starts, under construction, completed.
  - FangGuanJu: price indices for 134 cities; city-level floor space sold and unsold.
- Other sources:
  - NDRC price indices for 36 main cities; SouFun (price indices for 100 cities).
- City-tier coverage (2013, FangGuanJu coverage):
  - Tier I cities: 10 percent of floor space sold.
  - Tier II cities: nearly 50 percent of floor space sold.
  - Tier III/IV: more than 40 percent of floor space sold.

### IV. Recent developments — prices
- Property prices moderated in 2014 across datasets (NBS, SouFun, FangGuanJu, NDRC).
- NBS 70-cities index shows the most severe decline.
- Tier performance:
  - Tier II and Tier III/IV have performed weakest; Tier III/IV prices fell on a sequential basis.
- Regional weakness:
  - Industrial Northeast and the Coast experienced weakest price development; large metropolitan areas also recently weak.
- Stabilization signs:
  - NBS average selling price and FangGuanJu show sequential stabilization since November 2014 in some datasets.
- Affordability multiples (2013):
  - Average nationwide house price: about 22 times average annual disposable income.
  - Tier I cities: multiple of more than 30.
  - Ratio declined from its peak in 2010.

### V. Recent developments — transactions and demand
- Floor space sold:
  - Declined on a year-over-year basis since mid-2013 (NBS and FangGuanJu).
  - Transaction activity weakened across Tier I, Tier II, and Tier III/IV cities.
  - Toward end-2014, sales volume picked up slightly on a sequential basis after relaxation of home purchase restrictions and easing of mortgage financing conditions; persistence uncertain.
  - Sales contraction most pronounced in Tier III/IV cities.
  - Industrial Northeast faces a sharp contraction in floor space sold and has not shown the recovery seen in other regions.

### VI. Recent developments — supply and investment
- Close relationship between floor space sold, floor space starts, and residential real estate fixed investment.
- Real estate fixed asset investment slowed from growth of about 20 percent in 2013 to 9.2 percent in 2014 (NBS).
- Housing floor space starts contracted (text indicates contraction; exact numeric value not provided in excerpt).

### VII. Inventory buildup and indicators
- Residential transactions slowed in 2014 relative to 2013.
- Land sales revenues:
  - Accounted for 6½ percent of GDP in 2014 and showed a slowdown after peaking in late 2013.
- Inventory measures (NBS vs FangGuanJu):
  - NBS nationwide inventory ratio: about four months of sales.
  - FangGuanJu nationwide inventory ratio: might have been as high as 24 months in mid-2014.
- Data differences due to methodology:
  - NBS: developers’ self-reporting; likely underreporting of unsold units and overreporting of sales.
  - FangGuanJu: records all real estate registrations including buildings that have permits to sell.
- Overbuilding widespread, with Tier III/IV cities and the industrial Northeast showing particularly notable inventory ratcheting-up.
- Buildup in 2013–14 more severe than 2008–09 and 2011–12, starting from a relatively high inventory level.

### VIII. Stock measures and per capita floor space
- Floor space per capita increased notably in 2009 and continued to rise in Tier III/IV cities while remaining flat for large cities.
- International benchmarks:
  - Advanced economies: per capita residential floor space in the range of 40‒50 square meters (exception: United States).
  - Emerging economies: between 20 and 30 square meters.
  - China: Tier I cities close to average levels in advanced economies; smaller cities already see a much higher level, suggesting possible signs of oversupply.
- Caveat:
  - Denominator uses hukou population; likely overestimates per capita floor space for larger coastal cities (which attract migrants) and underestimates for the rest.

### IX. Econometric approach and key regression results
- Analysis period starts from 1998; regression sample period spans from 2000 to 2013 (balanced panel available during 2005–12).
- Long-run equilibrium for floor space per capita estimated with explanatory variables: household income per capita, residential property prices, urbanization rate, nonagricultural population scale, and city tier dummies.
- Excluded policy variables: mortgage rates and purchase restrictions.
- Regression diagnostics (Table 1; observations and R squared by specification):
  - Observations: 1,800; 1,748; 1,175; 1,175 across specifications (1)–(4).
  - R squared: 0.471; 0.476; 0.508; 0.516 across specifications (1)–(4).
- Selected coefficients (Table 1; coefficients with robust standard errors in parentheses; significance: ***, **, * denote 1 percent, 5 percent, 10 percent):
  - Log per capita household income:
    - 0.170*** (0.0236)
    - 0.269*** (0.0353)
    - 0.296*** (0.0409)
    - 0.299*** (0.0397)
  - Log residential property prices:
    - -0.0794*** (0.0191)
    - -0.0603*** (0.0228)
    - -0.0568*** (0.0213) (in specifications where included)
  - Urbanization rate:
    - -0.497*** (0.0393)
    - -0.454*** (0.0389) (in specifications where included)
  - Log nonagricultural population:
    - -0.0417*** (0.0115) (in specification where included)
  - Dummy_tier 1:
    - -0.109*** (0.0289)
    - -0.0804** (0.0319)
    - -0.0768* (0.0428)
    - -0.0502 (0.0474)
  - Dummy_tier 3:
    - 0.166*** (0.0116)
    - 0.136*** (0.0131)
    - 0.0746*** (0.0166)
    - 0.0373* (0.0201)
  - Constant terms:
    - 1.366*** (0.223)
    - 1.076*** (0.242)
    - 0.906*** (0.268)
    - 1.169*** (0.266)
- Interpretation highlights:
  - A 1 percentage point increase in per capita household income will increase floor space per capita by 0.2–0.3 percent (estimates close across specifications).
  - Higher residential prices, urbanization rate, and nonagricultural population tend to reduce floor space per capita in cities; elasticity of residential prices smaller than generally expected.
  - Tier III/IV cities tend to have more floor space per capita.

### X. Projection assumptions and baseline medium-term demand
- Projection assumptions for equilibrium floor space per capita:
  - Per capita urban household income growth: begins at 9.7 percent in 2013 and falls by ½-percentage point per year, reaching 6.2 percent in 2020.
  - Growth rate of nonagricultural population: forecast at levels of five-year moving average, implying an increase of urbanization rate by about 7–8 percentage points from now to 2020.
  - No policy response assumed (no future changes in mortgage rate and restrictions on property purchases).
  - Adjustment parameter a in equation 3: varies by province but is 0.5 at the national level.
  - Demolition rate assumed: about 3 percent annually of the floor space stock.
- Under these assumptions:
  - The average annual growth rate of floor space sold is about 4.3 percent in the medium term.

### XI. Measuring oversupply and calculated excess in 2014
- Oversupply measure:
  - Cumulative gap between floor space starts and floor space sold (one to two years ahead); floor space sold series derived from NBS historical values and projections.
- Rationale:
  - Floor space starts used as supply proxy because starts better indicate future supply and are strongly correlated with real estate investment.
- Calculated excess supply in 2014:
  - About 1.2 times annual floor space sold at the national level.
- Caveats:
  - Not all floor space starts are listed for sale; starts include areas for communal facilities and infrastructure — could be up to 15 percent of floor space starts.
  - Starts data likely overstate true supply because they include projects in planning or with permits though not all floor space is built at the same time.

### XII. Scenario analysis of real estate adjustment
- Purpose:
  - Illustrate possible adjustment paths given current oversupply.
- Baseline scenario (key features):
  - Floor space sold projections taken as given (continuation of historical demand trend); floor space starts adjust so excess supply gradually winds down.
  - Assumes excess supply gap will close gradually by 2020, broadly in a linear fashion.
  - Excess supply absorbed through both a moderate contraction in floor space starts and a recovery of projected real estate demand in the medium term.
  - Under baseline, average annual growth rate of floor space sold ~4.3 percent; FangGuanJu-based inventory ratio would fall as excess supply closes.
- Alternative scenarios:
  - Gradual adjustment scenario:
    - "Excess supply will be reduced by half from its 2013 level."
    - Inventory ratio adjusts modestly in the medium term and returns to its normal level only by 2025.
  - Stronger real estate demand scenario:
    - Growth in floor space sold would increase to 8.6 percent (similar to levels during 2010‒13 when growth was 7.8 percent) relative to 4.3 percent in the baseline (text cutoff in source).
- Inventory ratio key finding:
  - Inventory ratio: "2.2 in 2013 to about 1 (a normal historical level) by 2020."
- Supplementary note:
  - Cumulative floor space completed was higher than floor space sold before the global financial crisis; measure of floor space completed does not fully reflect supply conditions.

### XIII. Growth impacts across scenarios (national)
- Relationship between floor space starts and real estate GFCF (Table 2):
  - β1 = 0.413
  - β2 = 0.587
  - Coefficients constrained to sum to one (not statistically rejected).
  - Implied formula:
    - Growth of real estate GFCF(t) = 0.413*Growth of floor space starts(t) + 0.587*Growth of real estate GFCF(t-1).
- Scenario impacts on real estate GFCF in 2015 (Table 3):
  - Real estate GFCF could slow by about 1.9–4.2 percentage points in 2015.
  - Baseline scenario: slowdown of -4.2 percentage points in 2015.
  - More gradual adjustment: -1.9 percentage points in 2015.
  - Stronger demand: -2.5 percentage points in 2015.
- GDP implications in 2015:
  - Residential real estate GFCF accounts for about 9 percent of GDP.
  - Baseline scenario implies a drop of GDP growth of about 0.4 percentage points for 2015.
  - Implied impacts on GDP growth in 2015 (yoy, ppt) from Table 3:
    - Baseline scenario: -0.38
    - More gradual adjustment: -0.18
    - Stronger demand: -0.22
- Scenario durations (Table 3, Duration of Adjustment – Total years 2015–2020):
  - Baseline scenario: 6.0-15.0 (Total years 11.3)
  - More gradual adjustment: 10.0-9.6 (Total years 5.5)
  - Stronger demand: 6.0-10.9 (Total years 13.7)
- Note:
  - Growth impact over the medium term depends on authorities’ policy response; estimates abstract from indirect effects through upstream/downstream sectors and may overstate impacts if policy responses moderate adjustment.

### XIV. Provincial-level scenario analysis and impacts
- Historical association:
  - Higher excess supply associated with lower real GDP growth across provinces.
  - Transmission likely via slowdown in real estate investment.
- Baseline provincial outcomes:
  - Most provinces would see lower new housing starts in 2015–16, before gradual recovery.
  - Provinces with high inventory ratios would experience more severe and prolonged contractions.
  - Nationwide real estate investment would decline by an average of -4.2 percent on a national level in 2015; a few provinces could see sharper contractions.

### XV. Conclusion and policy implications
- Role and recent softening:
  - China’s residential real estate sector is an important growth driver and has softened visibly since 2014, driven in part by previous overbuilding across most cities.
- Adjustment:
  - An orderly, multiyear adjustment is welcome to unwind excess supply and reallocate resources, aiding China’s move toward a new and sustainable growth model.
- Policy calibration:
  - Key to allow adjustment while avoiding too sharp an economic slowdown.
  - Macroeconomic policies need careful calibration to manage growth impacts and regional heterogeneity.

_Source: _wp1584 - References (excerpts of IMF working paper content)._

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

### _wp1584 - References

### I. Introduction — role of real estate in China's growth
- Real estate investment grew from about 4 percent of GDP in 1997 to 15 percent of GDP in 2014.
- Residential investment accounts for about 15 percent of fixed asset investment and 15 percent of total urban employment.
- Bank lending to the sector accounts 20 percent of total loans.
- Real estate is a key source of local public finance and is extensively used as collateral for corporate borrowing.

### II. Market distortions and vulnerabilities
- Supply-side distortion: local governments’ reliance on land sales to finance spending.
- Demand-side distortions: housing attractive as financial investment due to a history of robust capital gains, real deposit interest rates that tended to be negative, lack of alternative financial assets, and capital account restrictions.

### III. Data sources and geographic coverage
- Main official sources: National Bureau of Statistics (NBS) and local housing bureaus (FangGuanJu).
  - NBS: commodity building average selling prices; 70 large and medium-sized price index; price indices at provincial and national levels; residential real estate fixed asset investment; floor space sold, unsold, starts, under construction, completed.
  - FangGuanJu: price indices for 134 cities; city-level floor space sold and unsold.
- Other sources: National Development Research Center (NDRC) price indices for 36 main cities; private entities such as SouFun (price indices for 100 cities).
- City-tier coverage: Tier I cities account for 10 percent of floor space sold; Tier II cities account for nearly 50 percent; Tier III/IV account for more than 40 percent (2013, FangGuanJu coverage).

### IV. Recent developments — prices
- Property prices have been moderating in 2014 across datasets (NBS, SouFun, FangGuanJu, NDRC).
- The NBS 70-cities index shows the most severe decline.
- Price moderation observed across all city tiers; Tier II and Tier III/IV have performed weakest, with Tier III/IV prices falling on a sequential basis.
- Regional weakness: industrial Northeast and the Coast experiencing weakest price development; large metropolitan areas also recently weak.
- Some datasets (NBS average selling price and FangGuanJu) show signs of sequential stabilization since November 2014.
- Average nationwide house price remained about 22 times average annual disposable income in 2013; Tier I cities reaching a multiple of more than 30. The ratio has declined from its peak in 2010.

### V. Recent developments — transactions and demand
- Floor space sold has declined on a year-over-year basis since mid-2013 (NBS and FangGuanJu).
- Transaction activity weakened across Tier I, Tier II, and Tier III/IV cities.
- Toward end-2014, sales volume picked up slightly on a sequential basis after relaxation of home purchase restrictions and easing of mortgage financing conditions, but persistence is uncertain.
- Sales contraction most pronounced in Tier III/IV (smaller) cities which had experienced rapid increases during the boom.
- Industrial Northeast continues to face a sharp contraction in floor space sold and has not shown the recovery seen in other regions.

### VI. Recent developments — supply and investment
- Close relationship between floor space sold, floor space starts, and residential real estate fixed investment.
- Real estate fixed asset investment slowed from growth of about 20 percent in 2013 to 9.2 percent in 2014 (NBS).
- Housing floor space starts contracted (text indicates contraction; exact numeric value not provided in excerpt).

### VII. Analytical approach and paper structure (summary of methods)
- The paper navigates multiple official and private datasets to analyze prices, transactions, and starts at national and city-tier levels.
- Investigates fundamental determinants of residential real estate demand and supply using econometric analysis (Section III).
- Uses econometric results to gauge plausible size and duration of future adjustments through baseline scenario analysis (Section IV).

*Source: _wp1584 - References (excerpts of IMF working paper content).*

### 14.4 percent in 2014, compared with 11.6 percent growth in 2013. Land sales revenues, another

### 14.4 percent in 2014, compared with 11.6 percent growth in 2013. Land sales revenues, another

### Inventory buildup and indicators
- Residential real estate transactions: floor space sold showed a slowdown in 2014 compared with 2013.
- Land sales revenues accounted for 6½ percent of GDP in 2014 and showed a slowdown after reaching a peak in late 2013.
- NBS and FangGuanJu inventory measures both show a rapid buildup of inventory since 2013, but the sizes differ significantly:
  - NBS nationwide inventory ratio: about four months of sales.
  - FangGuanJu nationwide inventory ratio: might have been as high as 24 months in mid-2014.
- Data differences may be attributable to methodology:
  - NBS data rely on developers’ self-reporting and are likely subject to underreporting of unsold units and overreporting of sales.
  - FangGuanJu data record all real estate registrations including buildings that have obtained permits to sell.
- Overbuilding appears widespread across cities, with varying degrees of severity; some cities saw rapid rises in inventory ratio due to sharp contraction in sales since the beginning of 2014.
- The 2013–14 buildup of unsold residential housing units appears more severe than previous downturns (2008–09 and 2011–12), and started from a relatively high inventory level.
- Tier and area patterns:
  - Tier III/IV cities and the industrial Northeast region experienced particularly notable ratcheting up of inventory.

### Stock measures and per capita floor space
- Floor space per capita increased notably in 2009 and continued to rise in Tier III/IV cities while remaining flat for large cities.
- International benchmarks cited:
  - Advanced economies: per capita residential floor space in the range of 40‒50 square meters, with an exception in the United States.
  - Emerging economies: between 20 and 30 square meters.
  - For China: Tier I cities are close to the average levels in advanced economies; smaller cities already see a much higher level, suggesting possible signs of oversupply.
- Caveat on per capita measures:
  - Denominator uses population holding a household residency (hukou), likely overestimating per capita floor space for larger coastal cities (that attract migrants) and underestimating for the rest. Actual oversupply in some Tier III and IV cities may be even more severe than the results suggest.

### Demand and supply dynamics — methodological approach
- Analysis period starts from 1998; regression sample period spans from 2000 to 2013 (balanced panel available during 2005–12).
- Demand variable: floor space per capita; long-run equilibrium relationship estimated between floor space per capita and fundamentals:
  - Explanatory variables: household income per capita, residential property prices, urbanization rate, nonagricultural population scale (population density), and city tier dummy variables.
- Regression specification excludes policy variables such as mortgage rates and purchase restrictions.
- Regression findings (Table 1 summary):
  - A 1 percentage point increase in per capita household income will increase floor space per capita by 0.2–0.3 percent (estimates close across specifications).
  - Higher residential prices, urbanization rate, and nonagricultural population tend to reduce floor space per capita in cities, with the elasticity of residential prices smaller than generally expected.
  - City-tier dummy variables: Tier III/IV cities tend to have more floor space per capita.
- Regression diagnostics and details (preserve numeric values exactly):
  - Sample size and fit:
    - Observations: 1,800; 1,748; 1,175; 1,175 across specifications (1)–(4).
    - R squared: 0.471; 0.476; 0.508; 0.516 across specifications (1)–(4).
  - Coefficients (selected, from Table 1):
    - Log per capita household income: 0.170*** (0.0236); 0.269*** (0.0353); 0.296*** (0.0409); 0.299*** (0.0397).
    - Log residential property prices: -0.0794*** (0.0191); -0.0603*** (0.0228); -0.0568*** (0.0213) (in specifications where included).
    - Urbanization rate: -0.497*** (0.0393); -0.454*** (0.0389) (in specifications where included).
    - Log nonagricultural population: -0.0417*** (0.0115) (in specification where included).
    - Dummy_tier 1: -0.109*** (0.0289); -0.0804** (0.0319); -0.0768* (0.0428); -0.0502 (0.0474).
    - Dummy_tier 3: 0.166*** (0.0116); 0.136*** (0.0131); 0.0746*** (0.0166); 0.0373* (0.0201).
    - Constant terms: 1.366*** (0.223); 1.076*** (0.242); 0.906*** (0.268); 1.169*** (0.266).
  - Robust standard errors reported in parentheses; “***”, “**”, and “*” denote 1 percent, 5 percent, and 10 percent significance levels.
- Projection assumptions for equilibrium floor space per capita:
  - Per capita urban household income growth: begins at 9.7 percent in 2013 and falls by ½-percentage point per year, reaching 6.2 percent in 2020.
  - Growth rate of nonagricultural population: forecast at levels of five-year moving average, implying an increase of urbanization rate by about 7–8 percentage points from now to 2020.
  - No policy response assumed (no future changes in mortgage rate and restrictions on property purchases).
  - Adjustment parameter a in equation 3: varies by province but is 0.5 at the national level.
  - Demolition rate assumed: about 3 percent annually of the floor space stock.
  - Under these assumptions, the average annual growth rate of floor space sold is about 4.3 percent in the medium term.

### Measuring oversupply
- Oversupply measure: cumulative gap between floor space starts and floor space sold (one to two years ahead). Floor space sold series derived from NBS data for historical values and from projections.
- Rationale for using floor space starts as supply proxy:
  - Floor space starts are used instead of completions because starts better indicate future supply and are strongly correlated with real estate investment.
- Calculated excess supply in 2014:
  - About 1.2 times annual floor space sold at the national level.
- Caveats to oversupply measure:
  - Not all floor space starts are listed for sale; starts include areas designated for communal facilities and infrastructure — could be up to 15 percent of floor space starts.
  - Floor space starts data likely overstate true supply because they include projects in planning or with permits even though not all floor space is built at the same time (developers may delay or stop projects).

### Scenario analysis of real estate adjustment
- Purpose: illustrate possible adjustment paths in the face of current oversupply.
- Key assumption for baseline scenario: floor space sold projections are taken as given (continuation of historical demand trend); floor space starts adjust so excess supply gradually winds down.
- Baseline scenario specifics:
  - Assumes excess supply gap will close gradually by 2020, broadly in a linear fashion.
  - Excess supply absorbed through both a moderate contraction in floor space starts and a recovery of projected real estate demand in the medium term.
  - The adjustment scenario yields year-over-year growth paths for floor space starts and floor space sold, providing an indication of real estate investment in the medium term.
  - The inventory ratio based on FangGuanJu data would fall as the excess supply gap closes (text indicates tracking of this effect).

*Source: IMF staff calculations and quoted data from NBS, FangGuanJu, CEIC, Wigram Capital Advisors as presented in the supplied content.*

### 2.2 in 2013 to about 1 (a normal historical level) by 2020.

### _wp1584 - 2.2 in 2013 to about 1 (a normal historical level) by 2020.

### Key finding
- Inventory ratio: "2.2 in 2013 to about 1 (a normal historical level) by 2020."

### Scenarios and implications
- Gradual adjustment scenario
  - Assumes a more gradual reduction in the excess supply gap.
  - "Excess supply will be reduced by half from its 2013 level."
  - Under a slower adjustment scenario, "the inventory ratio adjusts modestly in the medium term and returns to its normal level only by 2025."
  - May stabilize real estate markets as measured by floor space starts (a proxy for real estate investment), but "the overbuilding may persist."

- Stronger real estate demand scenario
  - "This could help narrow the excess supply gap marginally, but it is unlikely to fully offset a potential contraction of floor space starts."
  - In this scenario, "growth in floor space sold would increase to 8.6 percent (similar to levels during 2010‒13 when growth was 7.8 percent) relative to 4.3 percent in the baseline (similar to levels during 2010‒12 when growth was" [text cutoff].

### Supplementary note (footnote 19)
- "The cumulative floor space completed was higher than floor space sold before the global financial crisis, partly reflecting the opening up of residential real estate markets."
- "Discussion with real estate developers suggest that the measure of floor space completed does not fully reflect supply conditions."

*Source: _wp1584 - 2.2 in 2013 to about 1 (a normal historical level) by 2020.*

### 4.6 percent). But the closure of the excess

### _wp1584 - 4.6 percent). But the closure of the excess

### Supply, demand, and projection methodology
- Projection of floor space sold:
  - Based on regression coefficients on floor space per capita with projected household income per capita, residential property prices, urban population rate, and nonagricultural population share by 2020 (Table 1).
  - Growth in national per capita urban household income is projected to decline by ½-percentage points per year, beginning at 9.7 percent in 2013 and reaching 6.2 percent in 2020.
  - Nationwide nonagricultural population is projected to grow at a level equal to the five-year moving average.
  - These assumptions imply the average annual growth rate of floor space sold is about 4.3 percent in the medium term, a continuation of the trend observed in the data.
- Demand-side measure:
  - Floor space sold is used as a measure of demand; floor space starts measure supply.
- Inventory ratio:
  - Inventory ratio refers to cumulative floor space starts net of sold to the floor space sold; measured in years.

### Growth impact across scenarios (national)
- Estimated relationship between growth in floor space starts and real estate gross fixed capital formation (GFCF):
  - Regression coefficients (Table 2):
    - β1 = 0.413
    - β2 = 0.587
  - Coefficients constrained to sum to one (not statistically rejected), implying average growth rates of the two series are the same.
  - Implied real estate GFCF growth formula (as in Table 2):
    - Growth of real estate GFCF(t) = 0.413*Growth of floor space starts(t) + 0.587*Growth of real estate GFCF(t-1).
- Scenario impacts on real estate GFCF in 2015 (Table 3):
  - Real estate GFCF could slow by about 1.9–4.2 percentage points in 2015.
  - Baseline scenario (oversupply gap closes by 2020): slowdown in real estate GFCF of -4.2 percentage points in 2015.
  - More gradual adjustment scenario: -1.9 percentage points in 2015.
  - Stronger demand scenario: -2.5 percentage points in 2015.
- GDP implications in 2015:
  - Residential real estate GFCF accounts for about 9 percent of GDP.
  - Baseline scenario implies a drop of GDP growth of about 0.4 percentage points for 2015 (Table 3).
  - Implied impacts on GDP growth in 2015 (yoy, ppt) from Table 3:
    - Baseline scenario: -0.38
    - More gradual adjustment: -0.18
    - Stronger demand: -0.22
- Scenario durations (Table 3, Duration of Adjustment – Total years 2015–2020):
  - Baseline scenario: 6.0-15.0 (Total years 11.3)
  - More gradual adjustment: 10.0-9.6 (Total years 5.5)
  - Stronger demand: 6.0-10.9 (Total years 13.7)
- Notes:
  - The growth impact over the medium term depends on authorities’ policy response.
  - Estimates abstract from indirect effects through upstream/downstream sectors and may overstate impacts if policy responses moderate adjustment.

### Provincial-level scenario analysis and impacts
- Historical association:
  - Higher excess supply is usually associated with lower real GDP growth across provinces.
  - Transmission likely through slowdown in real estate investment, which was a key growth driver as excess supply built up.
- Baseline scenario by province:
  - Given current excess supply gap or high inventory ratio, most provinces would see lower new housing starts in 2015–16, before gradually recovering over the medium term.
  - Selected provinces with high inventory ratios would experience more severe and prolonged contractions.
  - Even as housing starts recover, growth in floor space sold (demand measure) is likely to slow by about 3½ percentage points from the average between 2010 and 2013, in the medium term.
  - Nationwide real estate investment would decline by an average of -4.2 percent on a national level in 2015, while a few provinces could see sharper contractions, weighing heavily on GDP growth.
  - Results likely overstate impacts because policy responses are not incorporated and the adjustment pace is uniformly applied.

### Conclusion and policy implications
- China’s residential real estate sector:
  - Plays an important role in the economy and has been an important driver of growth.
  - Since 2014, the sector has softened visibly, driven in part by previous overbuilding across most cities.
- Adjustment:
  - An orderly adjustment is welcome and will unwind excess supply and free resources for more efficient use elsewhere, aiding China’s move toward a new and sustainable growth model.
  - The adjustment is likely to be a multiyear process with implications for investment and growth.
- Policy calibration:
  - Key will be to allow adjustment to take place while avoiding a too sharp economic slowdown.
  - Macroeconomic policies need to be calibrated carefully to manage growth impacts and regional heterogeneity.

### Annex highlights: data sources, definitions, and measurement issues
- Price data sources (five datasets):
  - The NBS: Average Selling Prices (Commodity Building Residential Selling Price) and 70 cities (Large and Medium-sized) Price Index.
  - The NDRC: 36 Main Cities Price.
  - FangGuanJu: price indices for 134 cities.
  - SouFun (CREIS): 100 Cities Price.
  - Common price indices based on observed selling prices, often only for new construction (primary market), with limited coverage of Tier III/IV cities.
- Volume data definitions:
  - Floor space sold: total floor area sold in formal contracts in the reporting period; includes completed and presale units; excludes secondary market between private individuals.
  - Commodity (Private) housing sales: sales of private housing listed in formal contracts; excludes secondary private market.
  - Floor space starts: residential floor space newly started by developers; counts from day of ground-breaking; excludes clearing land without permits/actual construction.
  - Floor space under construction: total floor space of all residential buildings under construction during reference period; includes newly started, continued construction, restarted, completed in current year, and newly started but postponed.
  - Floor space completed: construction completed meeting design and approval requirements and certifications necessary for handover.
  - Floor space waiting for sales (vacant): completed but unsold or unrented floor space available for sale or rent; excludes buildings not yet built or under construction.
- Real estate investment and GFCF:
  - NBS series: investment by real estate development companies and related construction activities; excludes pure land transactions.
  - Residential real estate investment refers to investment in residential buildings only.
  - IMF staff estimate of real estate GFCF: takes share of real estate investment in total fixed asset investment and applies it to GFCF in national accounts.
- Inventory data:
  - NBS and local housing bureaus provide floor space unsold and available for sale.
  - FangGuanJu transaction data cover about 20–25 percent of China’s urban landscape and provide city-level sold/unsold volumes.
  - Annual inventory ratio: sum of floor space unsold across available cities divided by sum of annual floor space sold across cities; indicates years to absorb existing unsold space.
- Data classification:
  - Five datasets used in analysis: NBS Selling Price; NBS 70 Cities Price Index; CREIS: 100 Cities Price (SouFun); NDRC: 36 Cities Price; FangGuanJu: 134 Cities Residential Property Price.
  - Cities split into Tier I, Tier II, and Tier III/IV, and into four geographic areas: Industrial Northeast, Coastal, Metropolitan, and Less Developed areas.
  - FangGuanJu dataset covers 134 cities: 4 Tier I cities, 36 Tier II cities, and 94 Tier III/IV cities; covers 4 Metropolitan cities, 25 Industrial Northeast cities, 43 Coastal, and 62 Less Developed areas.

*Source: IMF staff calculations, NBS, FangGuanJu, SouFun, NDRC (excerpts from the provided content).*

### REFERENCES

### _wp1584 - REFERENCES

### Major themes: Housing dynamics, markets, and speculation
- Bian, T.Y., and P. Gete, 2015, “What Drives Housing Dynamics in China? A Sign Restrictions VAR Approach’, NBER Working Paper 2015 (Cambridge, Massachusetts: MIT Press).
- Chen K. and Y. Wen, 2014, “The Great Housing Boom of China,” Federal Reserve Bank of St. Louis, Working Paper 2014-022A.
- Gao, B., W. Wang, and X. Li, 2013, Expectation, Income Inequality and the Puzzle of City’s Housing Price to Rent Scissors in China, Economic Research Journal, Vol. 6.
- Jiang, C., 2005, “A Case Study of the Speculation Bubble on China’s Real Estate Market,” Management World, Vol. 12.
- Kuang W., 2010, Expectation, Speculation and Urban Housing Price Volatility, Economic Research Journal, Vol. 9.
- Wu, J., J. Gyourko, and Y. Deng, 2012, “Evaluating Conditions in Major Chinese Housing Market,” Regional Science and Urban Economics, Vol. 42, pp. 531–43.
- Zheng, S., and Z. Shi, 2011, “The Land and Housing Markets in the Context of ‘Land Finance’: an Analysis of Local Government Behaviors,” Social Sciences in Guangdong, Vol. 2.
- Zou, Z., and L. Niu, 2010, “Demand and Supply for Residential Housing in Urban China,” Financial Research, Vol. 1.

### Major themes: Land finance, fiscal interactions, and local government behavior
- Li X., G. Hong, L. Huang, 2013, “The Mystery of Land Finance Growth in China: Tax-Sharing Reform, Strategic Interaction of Land Finance,” China Economic Quarterly, Vol. 12, No. 4.
- Lv W., and Liu C., 2012, “Fiscal Expenditure, Land Finance and Housing Price Bubble: Measurement and Empirical Study Based on Provincial Panel Data,” Finance & Trade Economics, Vol. 12.
- Yang, Y., 2010, “The Scale of ‘Land Finance’ and Its Driving Factors,” Finance & Trade Economics, Vol. 10.
- Zheng, S., and Z. Shi, 2011, “The Land and Housing Markets in the Context of ‘Land Finance’: an Analysis of Local Government Behaviors,” Social Sciences in Guangdong, Vol. 2.

### Major themes: Investment, construction, and economic links
- Berkelmans, L. and H. Wang, 2012, “Chinese Urban Residential Construction to 2040,” Reserve Bank of Australia, Research Discussion Paper 2012–04.
- Grenadier, S. R., 1996, “The Strategic Exercise of Options: Development Sascades and Overbuilding in Real Estate Markets,” Journal of Finance, Vol. 51, No. 5, pp. 1653–79.
- Huang, Z., C. Wu, and X. Du, 2008, “Real Estate Investment and Economic Growth: An Analysis of National and Regional Panel Data,” Finance & Trade Economics, Vol. 8.
- Liang, Y., T. Gao, and S. He, 2006, “An Empirical Analysis of Harmonious Development between the Real Estate Industry and the National Economy in Transitional China,” Social Sciences in China, Vol. 3.

### Major themes: International and institutional reports, market assessments, and data compilations
- Goldman Sachs, 2014, “China’s Housing Boom in International Context,” Global Economics Paper: 225, October 16.
- International Monetary Fund, 2014a, People’s Republic of China, IMF Staff Country Report No. 14/235 (Washington).
- International Monetary Fund, 2014b, World Economic Outlook (Washington, October).
- Moody, 2014, “A Property Downturn in China Would Exert A Marked, But Manageable Economic Impact,” July 30.
- Wigram Capital Advisors, 2014, “Chartpack on China Property Watch,” unpublished.

*References as listed in the source PDF _wp1584 - REFERENCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1584.pdf_