## wpiea2019258-print-pdf

## Source details

**Canonical URL:** [wpiea2019258-print-pdf](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019258-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019258-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019258-print-pdf.pdf.json)

---

### I. Introduction
- As of end 2018, total debt of Chinese households stood at 52.6 percent of GDP, above the emerging market average, and 32 percentage points of GDP higher than in 2008.
- Housing-related debt (mortgage loans and Housing Provident Fund lending) accounted just below two-thirds of all household debt at end-2018, a 7-percentage point increase since 2013.
- Remaining household debt was almost evenly split between consumption loans (including credit card debt) and loans extended to households for business purposes (mostly SMEs).
- Typical mortgage features:
  - Maturity of about 10-20 years.
  - Minimum down payment of 20 (30) percent for the first (second) home; can be adjusted upwards by regional authorities.
  - Mortgage rates usually set at a discount/premium of 0-15 percent below/above the central banks’ benchmark lending rate, with adjustment depending on market conditions.
- Household assets and liabilities (CFPS 2010-16):
  - On average, housing assets constitute around 60 percent of all household assets, though this ratio declined over time for richer households.
  - Between 2010 and 2016, the share of debt held by highly indebted households (DTI ratio above 4) increased from around a quarter to almost half.
  - Increase in DTI ratio among lower income households reached almost 6.
  - Mortgage participation rate increased from 8 to 18 percent between 2010 and 2016.

### II. Macro-financial risks from high household debt
- Channels of risk:
  - Higher debt repayments could constrain future consumption growth.
  - Slower housing demand may affect real estate investment, property developer funding, and banks’ willingness to lend.
  - Lower fiscal revenues and public support of home ownership create risks for the fiscal sector.
  - These effects could contribute to excess savings and growing external imbalances.
- Cross-country and China-specific evidence:
  - Higher household debt may lead to lower GDP growth (cited studies).
  - Household debt increases consumption and GDP growth within one year but reduces them in the long run; negative long-run effects intensify when household debt/GDP ratio exceeds thresholds (60 percent for effects on consumption growth).
  - High household debt in certain provinces associated with lower consumption growth.
  - Higher household debt could reduce income elasticity of consumption, particularly under negative income shocks.
- Financial stability risks:
  - High indebtedness increases household vulnerability to adverse shocks and can force abrupt deleveraging.
  - With more than half of household debt as mortgages, deleveraging could constrain housing demand and press house price growth, affecting property developers and bank collateral values.
  - House price corrections can reduce banks’ financial soundness and increase funding pressure for property developers.
  - Recessions preceded by housing busts are longer and more severe (IMF, 2012).

### III. Impact of high household indebtedness on consumption — data and empirical strategy
- Data:
  - Micro-level: CFPS survey data (2010-16) used to construct DTI (debt/disposable income), DTA (debt/asset), household expenditure, disposable income, and demographics.
  - Macro-level: Quarterly provincial data covering 2015Q1 to 2018Q4 for 24 provinces; indebtedness measures include debt/GDP ratio and DTI ratio.
- Empirical strategy:
  - Two panel regression models: lagged indebtedness (two-year lag for CFPS) to capture medium/long-term effects; contemporaneous indebtedness to capture short-term effects.
  - Equation (1) (lagged effects): ∆c_{i,t} = β_0 + α_i + β_1 ∆y_{i,t} + β_2 HHI_{i,t−1} + β_3 (∆y_{i,t} * HHI_{i,t−1}) + β_4 X_{i,t} + ε_{i,t}.
  - Equation (2) (short-term effects): same specification with contemporaneous HHI_{i,t}.
  - Province-level regressions include controls: house price growth, change in lending rate, public consumption growth, fixed-asset investment growth, and national stock price growth; dynamic panel models control for lagged dependent variable.

### IV. Main estimation results
- Household survey regressions:
  - A 100-percent increase in DTI ratio is associated with a 3.5-ppt increase in household consumption growth in the same period.
  - The same 100-percent increase in DTI ratio is associated with a 4.4-ppt decline in household consumption growth two years later.
  - These effects translate to a 0.4-ppt increase in aggregate consumption growth in national accounts in the same period but a 0.5-ppt decline in aggregate consumption growth two years later.
  - A 1-ppt increase in DTA ratio is associated with a 0.3-ppt increase in contemporaneous household consumption growth but a 0.5-ppt decline in household consumption growth two years later.
  - Income elasticity: a 1-ppt increase in real income growth boosts consumption growth by around 0.1 ppt.
- Province-level results:
  - A 1-ppt increase in the lagged household debt/GDP ratio is associated with a 0.1-ppt decline in consumption growth.
  - When using DTI ratio as the indebtedness measure, the impact of a 1-ppt increase in the lagged DTI ratio on consumption growth is reported as comparable (text cut off).

### V. Threshold effects of household debt on income elasticity
- Model:
  - ∆cc_{i,t} = β_0 + α_i + β_s ∆y_{i,t} + β_3 X_{i,t} + ε_{i,t}, with β_s = β_1 if HHI_{i,t−1} < HHI^* and β_s = β_2 if HHI_{i,t−1} ≥ HHI^*.
- Expected and estimated signs:
  - When lagged household indebtedness < HHI^*: income elasticity positive (β_1 > 0).
  - When lagged household indebtedness ≥ HHI^*: income elasticity zero (β_2 = 0).
- Estimation results:
  - When lagged DTI ratio < threshold of 6.3, a 1-ppt increase in real income growth could boost consumption growth by 0.1 ppt.
  - When DTI ratio ≥ threshold of 6.3, estimated impact of income growth on consumption growth is not statistically significant from zero.
  - About 10 percent of households in the sample have DTI ratios above the threshold.
  - Conclusion: income elasticity of consumption falls when lagged household indebtedness is higher than the threshold.

### VI. Scenario simulations of consumption growth under different DTI dynamics
- Assumptions:
  - Real income growth slows gradually to 5.6 percent by 2023 (staff baseline projections).
  - Scenario 1: DTI ratio increases at the average pace of last five years.
  - Scenario 2: DTI ratio remains at end-2018 level.
  - Provincial regression estimates with DTI ratio used; contemporaneous impacts normalized from survey regressions with DTI ratio. (Simulation results using debt/GDP estimates are very similar.)
- Simulation findings:
  - Rising household debt could reduce annual consumption growth from nearly 7 percent in 2017 to less than 5 percent by 2030.
  - Consumption growth in scenario 1 is higher in the first few years due to higher borrowing but becomes lower than scenario 2 over the medium-to-long term.
  - If household debt can be maintained at the current level, consumption growth could stabilize at a rate higher than the baseline income or GDP growth rate, facilitating further internal rebalancing towards consumption.

### VII. Stress testing household debt repayment capacity: measures, caveats, and design
- Additional indebtedness measures from CFPS:
  - DSTI ratio = annual debt repayment / total annual disposable income (annual mortgage repayment used as debt service proxy).
  - LTV ratio = current mortgage balance / self-reported market value of all houses owned (aggregate across houses).
- Caveats:
  - DSTI uses mortgage repayment only; may underestimate true debt servicing burden. DSTI is a lower bound for debt servicing burden; focus on changes under adverse shocks.
  - LTV calculated using total mortgage balance and self-reported market value of all houses—not residence-level origination LTV; represents average current LTV across houses.
  - CFPS lacks liquid vs non-liquid asset breakdown; non-housing assets used as proxy for liquid assets, possibly overestimating liquid buffers and underestimating defaults.
- Financial margin (FM) and default condition:
  - FM = Y − C − HD, where Y = household disposable income, C = household consumption, HD = debt service (annual mortgage repayment used).
  - Default condition (EFM): FM < 0 and FM + LA < 0, where LA denotes liquid assets (proxied by non-housing assets).
  - Assumed buffer coverage period: one year.
- Historical averages (CFPS 2010–16):
  - Average DSTI ratio in 2016 ≈ 36 percent (up from <30 percent in 2010).
  - Share of households with DSTI > 0.5 in 2016 = 19 percent (14 percent in 2010).
  - Average LTV ratio increased from 31 percent in 2010 to 34 percent in 2016.
  - Significant heterogeneity in DSTI and LTV across income groups.
- Stress test shocks applied to 2016 CFPS data:
  - Income shocks for DSTI: 5-percent and 10-percent declines in household income.
  - House price shocks for LTV: 10-percent and 25-percent declines in house prices/values.

### VIII. Stress test results — vulnerability by income group
- DSTI results:
  - Average DSTI increases across all income quintiles after income shocks; increases larger for lower-income groups.
  - Post-shock average DSTI for the lowest two income groups (below 40th income percentile) reached 80 percent after shocks—significantly above the 50-percent mortgage-service-to-income limit imposed on banks.
  - About 19 percent of households had DSTI > 0.5 in 2016; share rises after shocks.
- Distressed households (EFM < 0):
  - Share of distressed households increases for all income quintiles after income shocks.
  - Highest increases in distressed-share occur in middle-income households because many low-income households already had negative EFMs before shocks and high-income households have larger liquid buffers.
- LTV results:
  - Initial LTV ratios relatively low for all income groups (<40 percent).
  - Post-shock LTV ratios mostly remain below 50 percent for all income groups after 10-percent and 25-percent house price declines.
  - High-income households have lower LTV ratios given ownership of multiple houses.
  - Conclusion: households are more vulnerable to negative income shocks than to negative house price shocks in China.

### IX. International experience and current Chinese macroprudential framework
- Cross-country evidence:
  - Demand-side measures (limits on DSTI and LTV) and supply-side measures (limits on bank credit growth, loan loss provisions) can mitigate negative effects of household debt on consumption and growth.
  - Strong banking supervision, higher banking capitalization, and credit registries reduce systemic risk from household debt.
- China’s framework and comparisons:
  - Maximum LTV for first-home buyers in China = 75 percent (provinces can reduce this; some provinces as low as 60 percent).
  - Regulation requires borrower’s monthly DSTI < 55 percent (with cap on mortgage-service-to-income ratio of 50 percent).
  - International norm for DSTI cited as 30-50 percent.
  - Absence of comprehensive credit registry may hinder lenders’ ability to evaluate household total liabilities.
  - China has not implemented supplemental capital requirements with different risk weights on household lending.

### X. Conclusions and policy implications
- Main findings:
  - Increase in household indebtedness can boost consumption short term but reduce consumption in the medium-to-long term.
  - Low-income groups are most vulnerable to adverse income shocks; declines in house prices increase LTV across groups but are less concerning given low initial LTV levels.
- Policy recommendations:
  - Strengthen systemic risk assessment of the household sector and extend monitoring beyond mortgages, given non-housing loans account for nearly half of household debt (higher share for lower-income households).
  - Improve data collection and processing beyond aggregate credit and housing market indicators; include distributional indicators such as:
    - Leverage: consistent and comprehensive DSTI and DTI ratios.
    - Liquidity: average maturity of household loans or assets by type.
    - Composition of assets and liabilities.
    - Interconnectedness with banks and non-bank financial sector.
    - Share of “debt at risk” and share of “risky” borrowers (e.g., share of households with high DSTI or loan-to-income ratios).
  - Strengthen interagency information and data sharing (as recommended by 2017 FSAP).
  - Adjust DSTI caps toward international norm of 30-50 percent and extend DSTI limits to other household loans including those from non-bank financial institutions.
  - Use stress testing of DSTI to interest rate and income shocks to gauge risks.
  - Consider sectoral capital requirements on banks’ exposures to real estate as a supply-side measure, but treat with caution due to potential leakages to non-banks.
  - Combine demand- and supply-side tools to enhance effectiveness and reduce leakages; include all leveraged credit providers in macroprudential purview to prevent migration of credit to less-constrained nonbanks.
  - Use land policies to increase effective housing supply and promote transparent, efficient secondary land markets given importance of mortgages and land ownership structure.
  - Decision-making on household sector policies should be based on a strengthened systemic risk monitoring framework for the household sector.

### Box 1. Land Policies and House Prices in China — overview and recommendations
- House prices in China are mainly affected by monetary, macroprudential, tax, and land policies; land policies are China-specific and not a major tool in international peers.
- Mechanisms and legal features:
  - Land policies affect house prices through volume and composition of land supply.
  - In China, land is owned by the state (urban land) or rural collectives; land-use rights for residential use are valid for 70 years.
- Composition and pricing:
  - Residential land is divided into types: small- and medium-sized apartments, villa and deluxe apartments, ‘economic’ houses (subsidized houses), and others.
  - Current policies aim to increase supply for small- and medium-sized apartments and restrict villas/deluxe apartments.
- Administrative measures:
  - Restrictions and guidance (e.g., floor area ratios) can steer supply toward small- and medium-sized apartments.
  - Floor area ratios in Shenzhen were recently raised in Great Bay Area planning; floor area ratios in Tier 1 cities were less than 2 in 2010 (compared to mostly above 3 in New York, Singapore, and Seoul in the same period per CICC estimates).
- Policy recommendations:
  - Reduce land hoarding by developers and enhance enforcement of penalties.
  - Enhance market mechanism in the secondary land transaction market.
  - Adopt alternative sources for local government financing such as property taxes to reduce reliance on land sales.
  - Continue relaxation of Hukou requirements in smaller cities.
  - Allow some collectively-owned rural land to be transacted in the land market (without having to be purchased by the state first).
  - Promote policies that increase housing supply and market-based land price formation.
- Key statistics and thresholds noted:
  - Residential land-use rights for residential use: 70 years.
  - Floor area ratios in Tier 1 cities: less than 2 in 2010.
  - Estimated household-level threshold HHI*: 6.3 and 2.4% (from ANNEX 1, Table A1.2).
  - Province-level estimated threshold HHI*: 49% and 131% (from ANNEX 2, Table A2.2).

*Source: wpiea2019258-print-pdf (extracted content).*

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

### wpiea2019258-print-pdf - References .............................................................................................................

### I. INTRODUCTION
- As of end 2018, total debt of Chinese households stood at 52.6 percent of GDP, above the emerging market average, and 32 percentage points of GDP higher than in 2008.
- Rapid financial market development and improving financial inclusion, combined with rising incomes, increased availability of credit to households; the speed of household leverage accumulation was the highest among all BIS-reporting countries since the Global Financial Crisis (GFC).
- At the end of 2018, housing-related debt (mortgage loans and Housing Provident Fund lending) accounted just below two-thirds of all household debt, a 7-percentage point (ppt) increase since 2013.
- Remaining household debt was almost evenly split between consumption loans (including credit card debt) and loans extended to households for business purposes (mostly SMEs).
- Typical mortgage features:
  - Maturity of about 10-20 years.
  - Minimum down payment of 20 (30) percent for the first (second) home that can be adjusted upwards by regional authorities.
  - Mortgage rates usually set at a discount/premium of 0-15 percent below/above the central banks’ benchmark lending rate, with adjustment depending on market conditions (Ding et al., 2017).
- Household indebtedness varies significantly across regions; in some provinces, debt has already risen above 60 percent of GDP—a threshold above which further increases in debt can adversely impact consumption growth (Lombardi et al., 2017).
- Household assets and liabilities (CFPS 2010-16):
  - On average, housing assets constitute around 60 percent of all household assets, though this ratio declined over time for richer households.
  - Between 2010 and 2016, the share of debt held by highly indebted households (DTI ratio above 4) increased from around a quarter to almost half.
  - Increase in DTI ratio among lower income households reached almost 6.
  - Mortgage participation rate increased from 8 to 18 percent between 2010 and 2016; still relatively low.

### II. MACRO-FINANCIAL RISKS FROM HIGH HOUSEHOLD DEBT
- Higher household borrowing could increase macroeconomic risks:
  - Higher debt repayments could constrain future consumption growth.
  - Real estate investment may be affected as housing demand slows, property developers face funding pressures, and banks become more reluctant to lend due to higher household credit risk.
  - Lower fiscal revenues and public support of home ownership create risks for the fiscal sector.
  - These effects could contribute to excess savings and growing external imbalances.
- Cross-country and China-specific findings:
  - Higher household debt may lead to lower GDP growth (Mian et al., 2017; Jorda et al., 2016).
  - Lombardi et al. (2017) found household debt increases consumption and GDP growth within one year but reduces them in the long run; negative long-run effects intensified when household debt/GDP ratio exceeds thresholds (60 percent for effects on consumption growth).
  - Tian et al. (2018) found high household debt in certain provinces associated with lower consumption growth.
  - Higher household debt could affect income elasticity of consumption, particularly under negative income shocks (Nakajima, 2018; Baker, 2014).
- Financial stability risks:
  - High indebtedness increases household vulnerability to adverse shocks, potentially forcing abrupt deleveraging with significant macro-financial impact.
  - With more than half of household debt as mortgages, deleveraging could constrain housing demand growth, pressuring house price growth and property developer soundness.
  - House price corrections could reduce banks’ financial soundness through lower collateral valuation and increase funding pressure for property developers.
  - Research shows recessions preceded by housing busts are longer and more severe (IMF, 2012).

### III. IMPACT OF HIGH HOUSEHOLD INDEBTEDNESS ON CONSUMPTION
- Data and measures:
  - Micro-level: CFPS survey data (2010-16) produce DTI (debt/disposable income) and DTA (debt/asset) ratios, household expenditure, disposable income, and demographics.
  - Macro-level: Quarterly provincial data span 2015Q1 to 2018Q4 covering 24 provinces; indebtedness measures include debt/GDP ratio and DTI ratio.
- Empirical strategy:
  - Two panel regression models: one with lagged household indebtedness (two-year lag for CFPS) to capture medium/longer-term effects; another with contemporaneous indebtedness to capture short-term effects.
  - Equation (1) (lagged effects): ∆c_{i,t} = β_0 + α_i + β_1 ∆y_{i,t} + β_2 HHI_{i,t−1} + β_3 (∆y_{i,t} * HHI_{i,t−1}) + β_4 X_{i,t} + ε_{i,t}.
    - ∆c and ∆y are household real expenditure growth and real disposable income growth.
    - HHI_{i,t−1} is measure of household indebtedness with a two-year lag.
    - Interaction term measures impact of indebtedness on income elasticity.
    - X_{i,t} is vector of household characteristics; α_i is household fixed effect.
  - Equation (2) (short-term effects): same specification with contemporaneous HHI_{i,t}.
  - Province-level regressions similar to equation (1) with controls: house price growth, change in lending rate, public consumption growth, fixed-asset investment growth, and (national) stock price growth; dynamic panel models used to control for lagged dependent variable.
- Main estimation results (household survey regressions; Annex 1, Table A1.1):
  - A 100-percent increase in DTI ratio is associated with a 3.5-ppt increase in household consumption growth in the same period.
  - The same 100-percent increase in DTI ratio is associated with a 4.4-ppt decline in household consumption growth two years later.
  - These effects translate to a 0.4-ppt increase in aggregate consumption growth in national accounts in the same period but a 0.5-ppt decline in aggregate consumption growth two years later.
  - A 1-ppt increase in DTA ratio is associated with a 0.3-ppt increase in contemporaneous household consumption growth but a 0.5-ppt decline in household consumption growth two years later.
  - Income elasticity is significantly positive; a 1-ppt increase in real income growth boosts consumption growth by around 0.1 ppt.
- Province-level results (Annex 2, Table A2.1):
  - A 1-ppt increase in the lagged household debt/GDP ratio is associated with a 0.1-ppt decline in consumption growth.
  - When using DTI ratio as the indebtedness measure, the impact of a 1-ppt increase in the lagged DTI ratio on consumption growth is slightly below (text cuts off before completing value).

*Source: wpiea2019258-print-pdf - References (extracted content).*

### 0.1 ppt. These results are comparable with other recent empirical studies using aggregate

### wpiea2019258-print-pdf - 0.1 ppt. These results are comparable with other recent empirical studies using aggregate

### Threshold effects of household debt on income elasticity
- Model estimated: ∆cc_{i,t} = β_0 + α_i + β_s ∆y_{i,t} + β_3 X_{i,t} + ε_{i,t}, with
  - β_s = β_1 if HHI_{i,t−1} < HHI^*
  - β_s = β_2 if HHI_{i,t−1} ≥ HHI^*
- Expected sign:
  - When lagged household indebtedness < HHI^*: income elasticity positive (β_1 > 0).
  - When lagged household indebtedness ≥ HHI^*: income elasticity zero (β_2 = 0).
- Estimation results (Annex 1, Table A1.2; provincial results Annex 2, Table A2.2):
  - When lagged DTI ratio < threshold of 6.3, a 1-ppt increase in real income growth could boost consumption growth by 0.1 ppt.
  - When DTI ratio ≥ threshold of 6.3, estimated impact of income growth on consumption growth is not statistically significant from zero.
  - About 10 percent of households in the sample have DTI ratios above the threshold.
  - Conclusion: income elasticity of consumption falls when lagged household indebtedness is higher than the threshold.

### Scenario simulations of consumption growth under different DTI dynamics
- Assumptions:
  - Real income growth slows gradually to 5.6 percent by 2023 (staff baseline projections).
  - Scenario 1: DTI ratio increases at the average pace of last five years.
  - Scenario 2: DTI ratio remains at end-2018 level.
  - Provincial regression estimates with DTI ratio used in simulations; contemporaneous impacts normalized from survey regressions with DTI ratio. (Simulation results using debt/GDP estimates are very similar.)
- Simulation findings:
  - Rising household debt could reduce annual consumption growth from nearly 7 percent in 2017 to less than 5 percent by 2030.
  - Consumption growth in scenario 1 is higher in the first few years due to higher borrowing but becomes lower than scenario 2 over the medium-to-long term.
  - If household debt can be maintained at the current level, consumption growth could stabilize at a rate higher than the baseline income or GDP growth rate, facilitating further internal rebalancing towards consumption.

### Stress testing household debt repayment capacity: measures and caveats
- Additional indebtedness measures constructed from CFPS survey data:
  - DSTI ratio = annual debt repayment / total annual disposable income (annual mortgage repayment used as debt service proxy).
  - LTV ratio = current mortgage balance / self-reported market value of all houses owned (aggregate across houses).
- Caveats:
  - DSTI uses mortgage repayment only; could underestimate true debt servicing burden. DSTI should be viewed as a lower bound for debt servicing burden; focus on changes under adverse shocks.
  - LTV is calculated using total mortgage balance and self-reported market value of all houses—different from residence-level origination LTV; viewed as an average current LTV across houses.
  - CFPS lacks liquid vs non-liquid asset breakdown; non-housing assets used as proxy for liquid assets, possibly over-estimating liquid buffers and under-estimating defaults.
- Financial margin (FM) measure (following Funke et al. (2018)):
  - FM = Y − C − HD, where Y = household disposable income, C = household consumption, HD = debt service (annual mortgage repayment used).
  - Default condition (EFM condition): FM < 0 and FM + LA < 0, where LA denotes liquid assets (proxied by non-housing assets).
  - Assumed buffer coverage period: one year (consistent with simplicity choice here; Funke et al. used 6 months or one year).

### DSTI and LTV trends and stress test design
- Historical averages (CFPS 2010–16):
  - Average DSTI ratio in 2016 ≈ 36 percent (up from <30 percent in 2010).
  - Share of households with DSTI > 0.5 in 2016 = 19 percent (14 percent in 2010).
  - Average LTV ratio increased from 31 percent in 2010 to 34 percent in 2016.
  - Significant heterogeneity in DSTI and LTV across income groups.
- Stress test shocks (applied to 2016 CFPS data):
  - Income shocks for DSTI: 5-percent and 10-percent declines in household income.
  - House price shocks for LTV: 10-percent and 25-percent declines in house prices/values.

### Stress test results: vulnerability by income group
- DSTI results:
  - Average DSTI increases across all income quintiles after income shocks; increases larger for lower-income groups.
  - Post-shock average DSTI for the lowest two income groups (below 40th income percentile) reached 80 percent after shocks—significantly above the 50-percent mortgage-service-to-income limit imposed on banks.
  - About 19 percent of households had DSTI > 0.5 in 2016; share rises after shocks.
- Distressed households (EFM < 0) results:
  - Share of distressed households increases for all income quintiles after income shocks.
  - Highest increases in distressed-share occur in middle-income households because many low-income households already had negative EFMs before shocks and high-income households have larger liquid buffers.
- LTV results:
  - Initial LTV ratios relatively low for all income groups (<40 percent).
  - Post-shock LTV ratios mostly remain below 50 percent for all income groups after 10-percent and 25-percent house price declines.
  - High-income households have lower LTV ratios given ownership of multiple houses.
  - Conclusion: households are more vulnerable to negative income shocks than to negative house price shocks in China, consistent with other empirical studies (e.g., Feng and Lu (2019)).

### International experience and current Chinese macroprudential framework
- Cross-country evidence:
  - Demand-side measures (limits on DSTI and LTV) and supply-side measures (limits on bank credit growth, loan loss provisions) can mitigate negative effects of household debt on consumption and growth.
  - Strong banking supervision, higher banking capitalization, and credit registries reduce systemic risk from household debt.
- China’s current framework (comparison notes):
  - Maximum LTV for first-home buyers in China = 75 percent (provinces can reduce this; some provinces as low as 60 percent).
  - Regulation requires borrower’s monthly DSTI < 55 percent (with cap on mortgage-service-to-income ratio of 50 percent).
  - International norm for DSTI cited as 30-50 percent.
  - Absence of comprehensive credit registry may hinder lenders’ ability to evaluate household total liabilities.
  - China has not implemented supplemental capital requirements with different risk weights on household lending.

### Conclusions and policy implications
- Main findings:
  - Increase in household indebtedness can boost consumption short term but reduce consumption in medium-to-long term.
  - Low-income groups are most vulnerable to adverse income shocks; declines in house prices increase LTV across groups but are less concerning given low initial LTV levels.
- Policy recommendations:
  - Strengthen systemic risk assessment of household sector and extend monitoring beyond mortgages, given non-housing loans account for nearly half of household debt (higher share for lower-income households).
  - Improve data collection and processing beyond aggregate credit and housing market indicators; include distributional indicators such as:
    - Leverage: consistent and comprehensive DSTI and DTI ratios.
    - Liquidity: average maturity of household loans or assets by type.
    - Composition of assets and liabilities.
    - Interconnectedness with banks and non-bank financial sector.
    - Share of “debt at risk” and share of “risky” borrowers (e.g., share of households with high DSTI or loan-to-income ratios).
  - Strengthen interagency information and data sharing (as recommended by 2017 FSAP).
  - Adjust DSTI caps toward international norm of 30-50 percent and extend DSTI limits to other household loans including those from non-bank financial institutions.
  - Use stress testing of DSTI to interest rate and income shocks to gauge risks.
  - Consider sectoral capital requirements on banks’ exposures to real estate as a supply-side measure, but treat with caution due to potential leakages to non-banks.
  - Combine demand- and supply-side tools to enhance effectiveness and reduce leakages; include all leveraged credit providers in macroprudential purview to prevent migration of credit to less-constrained nonbanks.
  - Use land policies to increase effective housing supply and promote transparent, efficient secondary land markets given importance of mortgages and land ownership structure.
  - Decision-making on household sector policies should be based on a strengthened systemic risk monitoring framework for the household sector.

*Source: Fund staff estimates and CFPS survey analysis as presented in the referenced IMF working paper section.*

### Box 1. Land Policies and House Prices in China

### Box 1. Land Policies and House Prices in China

### Overview
- House prices in China are mainly affected by four types of policies—monetary (mainly interest rates and credit volume), macroprudential (mainly down payment requirements), tax, and land policies.
- Land policies are China-specific and are not seen as a major policy tool to contain rapidly rising house prices in international peers.
- This box explains the main land policies in China and how they tend to affect house prices.

### How land policies affect house prices
- Land policies affect house prices mainly through the volume and composition of land supply.
- Higher land supply, without significant land hoarding by developers, should increase housing supply and alleviate the upward pressure on house prices.
- The composition of land supply (e.g., for residential, commercial, or office use) could affect residential housing supply and house prices.

### Legal and ownership features
- In China, land is owed not privately but by the state (urban land) or by rural collectives (rural land).
- Land-use rights are privately owned, but those for residential use are only valid for 70 years.

### Composition and pricing schemes
- Different types of land have different pricing schemes, leading to potential arbitrage opportunities.
- Residential land is further divided into four types of residential housing: small- and medium-sized apartments, villa and deluxe apartments, ‘economic’ houses (subsidized houses), and others.
- Current land policies aim to increase land supply for small- and medium-sized apartments while restricting the supply for villas and deluxe apartments, which should benefit the general population.
- Increases in land supply for economic houses may limit land supply for other houses, potentially leading to higher market-determined house prices.

### Administrative measures
- Administrative measures such as restrictions on floor area ratios (defined as the ratio of a building’s total floor area to the size of the land upon which it is built) could be used to guide land supply towards small- and medium-sized apartments.
- Example: the floor area ratios in Shenzhen were recently raised in the future planning of the Great Bay Area to facilitate higher-floor buildings.
- There may still be room for raising floor area ratios in Tier 1 cities (less than 2 in 2010), compared to New York, Singapore, and Seoul which were mostly above 3 in the same period according to CICC estimates.

### Policy recommendations to improve market functioning
- Reduce land hoarding by property developers and enhance enforcement of penalties for developers engaging in land hoarding.
  - Note: Land hoarding refers to the practice of acquiring land and to deliberately hold it and wait for prices to increase in the future before selling it for a higher return.
- Enhance the market mechanism in the secondary land transaction market.
- Adopt alternative sources for local government financing such as property taxes to reduce high reliance on land sales to finance spending.
- Continue relaxation of Hukou requirements in smaller cities.
- Allow some collectively-owned rural land to be transacted in the land market (without having to be purchased by the state first).
- Promote policies that increase housing supply and market-based land price formation.

### Key statistics and thresholds noted in the related analyses
- Residential land-use rights for residential use: 70 years.
- Floor area ratios in Tier 1 cities: less than 2 in 2010 (compared to mostly above 3 in New York, Singapore, and Seoul in the same period, per CICC estimates).
- Estimated household-level threshold 푯푯푯푯*: 6.3 and 2.4% (from ANNEX 1, Table A1.2).
- Province-level estimated threshold 푯푯푯푯*: 49% and 131% (from ANNEX 2, Table A2.2).

*Source: IMF staff analysis as presented in the cited content.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019258-print-pdf.pdf_