## 1. Introduction

## Source details

**Canonical URL:** [1. Introduction](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025259-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025259-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025259-source-pdf.pdf.json)

---

### Key context and motivation
- High household savings are a structural feature of the Chinese economy and far exceed the levels in OECD and most EMDE countries.
- High savings have facilitated the Chinese economy’s over-reliance on debt-financed investment and led to the buildup of vulnerabilities.
- Recent sluggishness in domestic demand and deflationary pressures—partly due to the housing market correction and low consumer confidence—have renewed focus on the high savings rate and raised concerns about domestic and external imbalances.
- The 2025 Central Government Work Report and the December 2024 Central Economic Work Conference identified stimulating private consumption demand as the top policy priority for China.
- Authorities’ counter-cyclical and social-safety-net responses:
  - equipment trade-in and upgrade programs providing subsidies for durable goods;
  - central government allocation of 150 billion yuan in 2024 to support these programs;
  - announced expansion in January 2025 with a pre-allocation of 81 billion yuan in new funding;
  - central government support financed by issuance of ultra-long-term special purpose government bonds (150 billion yuan in 2024 and 300 billion yuan in 2025).

### Data sources and novel contribution
- Data sources:
  - Prefecture-level macro indicators (urban and rural household savings rate, government social expenditure) covering around 250 prefectures from 2012 to 2022 (annual).
  - Household-level survey data from China Family Panel Studies (CFPS), biennial, covering more than 6,000 urban households per round from 2010 to 2022; first study to use the 2022 CFPS round.
- Novel empirical contributions:
  - Prefecture-level analysis of medium-term impact of government social spending on savings and heterogeneity by demographic vulnerability.
  - Household-level analysis of the role of Hukou and Hukou reforms using micro-level survey data.
  - Household-level disentangling of housing market wealth effect versus downpayment effect, focusing on the post-2021 real estate market correction.

### Three policy questions addressed and main empirical findings

- Question 1 — What is the impact of government social spending on household savings?
  - Government social spending in China remains below peer economies and is unevenly distributed across regions.
  - Reported spending: healthcare 3.5 percent of GDP; social security 3.1 percent of GDP; OECD averages: healthcare 7.2 percent of GDP, social security 8.2 percent of GDP; emerging markets averages: healthcare 5.0 percent of GDP, social security 6.1 percent of GDP.
  - Prefecture-level analysis (annual data, ~240 prefectures, 2012–2022) indicates:
    - A 10 percent increase in government expenditure on social security is associated with a reduction in the medium-term rural household savings rate of 0.5 percentage points.
    - A 10 percent increase in government expenditure on healthcare is associated with a reduction in the medium-term rural household savings rate of 0.8 percentage points.
  - Impact on urban households from increased social spending is limited.
  - Targeted measures that raise social spending in prefectures with higher rural population shares would be more effective in boosting consumption.
  - Social spending is uneven by city tier: Tier-1 cities—5.8 percent of the population—had annual total per capita social spending on healthcare, social security, and education more than double the average in the rest of the country.
  - More than 90 percent of social spending is funded by local governments.
  - Broader-definition social security spending reached about RMB 7.6 trillion, or 5.6 percent of GDP (2024 Statistical Bulletin on Human Resources and Social Security Development); this ratio remains below OECD and EMDE averages.

- Question 2 — Does obtaining urban Hukou status and the accompanying social benefits impact urban households’ savings?
  - Hukou system introduced in 1958 classifies citizens as urban or rural residents; migrant households without urban Hukou are excluded from full access to urban social benefits.
  - Over the past ten years, Hukou reforms eased urban Hukou registration in smaller cities and narrowed social benefit gaps; reform progress in larger cities has been slow.
  - Over 200 million migrant urban residents continue to hold rural Hukous with limited benefits.
  - As of 2023, 27 percent of the residents in urban areas held rural Hukou, which represents 17.8 percent of total population in China.
  - Household-level regressions find:
    - The savings-to-disposable-income ratio of urban households without an urban Hukou is on average about 6.8 percentage points higher than that of their peers with an urban Hukou.
  - Implication: further Hukou reforms granting more urban Hukous to migrant households and providing more equitable social benefits can reduce excess savings in urban areas.

- Question 3 — Has the ongoing correction in the real estate market contributed to rising household savings, and what are the key channels?
  - Two primary channels: the downpayment channel for non-homeowners and the wealth effect channel for homeowners.
  - Household-level data disentangles these channels in the post-2021 housing downturn.
  - Findings:
    - Wealth effect has persisted since the start of the real estate market correction in 2021, with a significant negative relationship between homeowners’ net housing assets (NHA) to income ratio and their savings rates.
    - Downpayment channel has weakened recently: non-homeowners with low deposits no longer save more for future downpayments and mortgage payments, potentially due to subdued confidence and delayed plans for home purchases amid the protracted real estate market downturn.

### Literature positioning
- Builds on literature on drivers of China’s high household savings: demographic changes; weak social safety nets; house prices and homeownership; income inequality.
- Most closely related to Zhang et al., (2018) and Han and Zhang (2022); extends by using up-to-date data (post-COVID and post-2021 housing correction).
- Contributes to macro-financial literature on house prices, household wealth, and consumption by providing household-level evidence for the housing market downturn since 2021.

### Prefecture-level summary and methods (Annex Table A1 and local projections)
- Definitions and data features:
  - Household savings rate = (household disposable income − consumption expenditure) / household disposable income.
  - Household saving rates are consistently higher in urban areas, both on average and across all quartiles.
  - Inter-quartile range in government social spending across prefectures amounts to RMB 0.5-0.8 thousand per person per year for social security, healthcare, and education.
- Baseline local projections specification (equation (1)) includes log(Spending_SocialSecurity), log(Spending_Health), log(Spending_Edu), vector X of prefecture characteristics (house price-to-income ratio, house price growth, GDP per capita, first lag of household savings rate), prefecture-fixed-effects γi and year-fixed-effects γt.
- Interaction specification (equation (2)) includes time-invariant vulnerability dummy Vul_i (urbanization-rate below sample median or not belonging to Tier 1/2 prefectures).

### Main prefecture-level empirical findings and heterogeneity
- Main effects:
  - A 10 percent increase in government expenditure on social security → reduction in rural household savings rate of 0.5 percentage points five years after the spending hike.
  - A 10 percent increase in government expenditure on healthcare → reduction in rural household savings rate of 0.8 percentage points five years after the spending hike.
  - Impact of government spending on education is short-lived in the rural sub-sample and becomes insignificant 3 years after the spending hike.
- Interaction and vulnerability:
  - A 10 percent increase in health-related social spending is followed by a decrease in rural household savings rate by up to 1.3 percentage points in low-urbanization-rate prefectures.
  - This 1.3 percentage points effect is presented as a full percentage point higher than the effect in high-urbanization-rate prefectures.
  - Low-tier prefecture cities exhibit more substantial and persistent impacts from health-related spending than high-tier prefectures.
- Vulnerability definition:
  - Vul_i = 1 if historical average urbanization rate falls below the sample median or if prefecture is not Tier 1 or Tier 2.
  - Tier 1 prefectures: Beijing, Shanghai, Guangzhou, and Shenzhen.
  - Tier 2 prefectures: Tianjin, Chongqing, all provincial capital cities, and Dalian, Ningbo, Qingdao, and Xiamen.

### Caveats and robustness (prefecture-level)
- Caveat 1: Government social spending data are available only at the prefecture level (not separately for urban vs rural allocations); interpret results as impact after an increase in prefecture-level social spending.
- Caveat 2: Local government social spending is endogenous to unobservable prefecture-level characteristics; fixed effects and controls mitigate but do not eliminate endogeneity concerns; interpret associations cautiously.
- Robustness tests:
  - Balanced-panel re-estimation: higher spending related to social security and health remains robust in reducing rural household savings.
  - Specifying social spending as growth rates instead of log levels: results remain robust.
  - Instrumental-variable test: Bartik shift-share IV using national growth rate of general government fiscal expenditure on health × province exposure (share of population aged 65 and above) confirmed robustness.

### Quantification of aggregate impact: two scenarios
- Scenario 1 (doubling current level of government spending on social security and healthcare in all prefectures):
  - Total fiscal cost per year: 3.0 percent of GDP.
  - Cumulative increase in consumption over a five-year horizon: 2.4 percentage points of GDP.
  - Decline in the savings rate: 1.5 percentage points of GDP.
- Scenario 2 (doubling social security and healthcare spending only in prefectures with higher-than-median rural population shares):
  - Additional fiscal cost per year: 1.0 percent of GDP.
  - Cumulative increase in consumption over a five-year horizon: 1.8 percent of GDP.
  - Decline in the savings rate: 1.1 percentage points of GDP.
- Policy implication: targeted increases in social spending in rural/high-rural-share prefectures yield higher private-consumption impact per unit of fiscal cost.

### Household-level analysis: data, specifications, and key results
- Data and sample:
  - CFPS, biennial, 2010–2022 across 31 provinces; sample between 13,000 to 17,000 individuals from over 6,000 urban households per round.
  - Latest CFPS round used: 2022 (published in November 2024).
- Variable definitions:
  - Household savings rate (urban analysis) = (disposable income − annual expenditures) / disposable income; winsorized at the 10 percent level and excluded if below -200 percent.
  - Urban Hukou dummy: 1 for non-agricultural or residential registrations; 0 for agricultural registrations.
  - Pension indicators: about 74 percent of whole urban household sample covered by pensions (based on household heads’ pension coverage); around 80 percent if any family member covered; around 90 percent among urban households subsample whose heads are pensioners.
- Baseline household regression (equation (3)) includes UrbanHukou, Pensions, Controls, province-year fixed effects, and industry fixed effects; standard errors clustered at the province-level.

- Main regression results (Table 1, CFPS 2012–2022; Observations: 30,033; R-squared: 0.21):
  - Urban Hukou coefficients: -7.03*** (Column 1; s.e. 0.82); -7.88*** (Column 2; s.e. 0.85).
  - New urban Hukou: -10.09*** (Column 2; s.e. 1.55).
  - Existing urban Hukou: -6.80*** (Column 2; s.e. 0.82).
  - NUP reform: -3.69* (Column 3; s.e. 2.06).
  - Urban Hukou × NUP reform: 6.07** (Column 3; s.e. 2.22).
  - Pension: -2.06*** (Columns 1 and 2; s.e. 0.56); -1.82*** (Column 3; s.e. 0.54).
  - Education: -1.39*** (Column 1; s.e. 0.10).
  - ln(income): 32.51*** (Column 1; s.e. 0.82).
  - Debt-to-income ratio: -11.36*** (Column 1; s.e. 0.86).
  - Household size: -3.54*** (all columns; s.e. 0.23).
  - Age and Age^2: Age coefficients small/insignificant; Age^2 0.45*** (Column 1; s.e. 0.14).
- Main household-level interpretations:
  - Urban households with urban Hukou save about 7.0 percentage points less than rural Hukou peers.
  - New urban Hukou households save about 10.1 percentage points less than rural Hukou households.
  - Pension coverage of the household head is associated with significantly lower household savings.
  - Robustness: household fixed effects and balanced-sample tests affirm key findings.

### Trends, NUP event studies, and Hukou reforms
- Cross-sectional estimates of Urban Hukou coefficient by year (Table 2):
  - 2012: -7.69*** (2.19)
  - 2014: -11.83*** (1.81)
  - 2016: -7.60*** (1.54)
  - 2018: -7.50*** (1.52)
  - 2020: -2.64* (1.60)
  - 2022: -3.20** (1.57)
- Trend interpretation:
  - Savings-rate gap peaked in 2014 at approximately 11.8 percentage points (urban Hukou households saving less).
  - By 2022, the gap narrowed to 3.2 percentage points.
  - Narrowing driven primarily by a sustained decrease in savings rate of rural Hukou households; urban Hukou households’ savings rate remained relatively stable.
- New-Type Urbanization Plan (NUP) event study and DID:
  - NUP launched in 2014 with goal to issue 100 million new urban Hukous and increase urban Hukou share from 35 percent in 2014 to 45 percent by 2020; three batches of 236 pilot cities announced 2014–2016.
  - Panel local projections estimate medium-term cumulative decline in urban households’ savings rate of about 2.5 percentage points following NUP pilot city designation.
  - Micro-level DID (CFPS) Table 1 Column 3: average urban household savings rate decreased by 3.7 percentage points after NUP designation (NUP reform: -3.69*, s.e. 2.06).
  - Urban Hukou × NUP positive (6.07**, s.e. 2.22) indicates NUP mitigated the savings gap between urban Hukou and rural Hukou households.

### Real estate market, channels, and household-level evidence (Sections 4.1–4.2; Table 3)
- Real estate context:
  - Sector accounts for approximately 20 percent of GDP when accounting for linkages.
  - Homeownership rate: over 90 percent of households owning properties.
  - As of 2024, residential property sales (floor space sold) and secondary market house prices have declined by over 40 percent and 10 percent, respectively, from the peak in 2021.
- Baseline housing specification (equation (6)) uses:
  - NHA_i,t = (Property_Value_i,t − Mortgage_Balance_i,t)/Income_i,t (winsorized at 500).
  - Deposit_i,t: deposit-to-income ratio for non-homeowners.
  - Non_homeowner dummy and Deposit × Non_homeowner.
  - Household FE and year FE.
- Key Table 3 estimates (CFPS Urban, 2012–2022; Observations: 31,905; Household FE: Yes; R-squared: 0.105–0.106):
  - Log(Total income): 16.969*** (0.368).
  - NHA to income ratio: -0.020*** (0.006).
  - Deposit to income ratio: -0.196*** (0.068) in Column (1); -0.139* (0.071) in Column (2).
  - Dummy for non-homeowner: 2.778*** (0.738).
  - Deposit × Non_homeowner: -1.283*** (0.251); -1.443*** (0.279).
  - NHA × 2022dummy: 0.010 (0.033) — not significant.
  - Deposit × 2022dummy: -0.583*** (0.206).
  - Deposit × Non_homeowners × 2022dummy: 1.060* (0.602).
- Interpretations:
  - Wealth effect: NHA-to-income ratio negative and significant; a 0.5 standard-deviation decline in NHA-to-income ratio (≈ decline in house price of 20 percent) associated with increase in urban household savings rate of 0.35 percentage points (based on coefficient 0.020).
  - Downpayment (extensive margin): non-homeowners save more; non-homeowner dummy ≈ 2.8 percentage points higher savings than homeowners.
  - Downpayment (intensive margin): Deposit × Non_homeowner negative and significant implies non-homeowners with larger deposits save relatively less compared to those with near-zero deposits; since 2022 the deposit-related downpayment motive for non-homeowners with less deposits weakened (triple interaction 1.060*, s.e. 0.602), consistent with delayed home purchases.

### Conclusions (major empirical findings)
- Scope and data: multiple data sources, prefecture-level and household-level, covering 2012–2022.
- Social safety nets and government social spending:
  - Weak social safety net remains a key factor for high household savings.
  - Doubling government expenditure on social security and healthcare (prefecture-level, 2012–2022) is associated with medium-term rural household savings rate declines of 4.8 and 5.9 percentage points, respectively.
  - Effects stronger for vulnerable populations (lower-tier cities, regions with lower urban population share).
- Hukou system and migrant households:
  - Urban migrants with rural Hukous tend to save more than urban Hukou peers.
  - Savings-rate gap narrowed over the past decade but remains significant.
  - Full liberalization of Hukou implemented only in cities with populations less than 3 million; restrictions persist in larger cities where about half of the population resides.
- Real estate correction channels:
  - Wealth effect persisted since 2021; significant negative relationship between homeowners' net housing assets to income ratio and savings rates.
  - Downpayment channel weakened since 2021; non-homeowners with low deposits no longer saving more for future downpayments, possibly due to subdued confidence and delayed purchase plans.
- Distributional pattern:
  - High-income households tend to save more; top deciles’ savings rates are above peers in most other economies.

### Key summary statistics (prefecture- and household-level; Table A1 and A5)
- Prefecture-level (selected):
  - Urban HH savings rate (% of disposable income): No. of obs 2,962; Mean 35.22; Std. Dev. 7.205; 1st quartile 30.99; median 35.46; 3rd quartile 39.76.
  - Rural HH savings rate (% of disposable income): No. of obs 2,463; Mean 24.23; Std. Dev. 10.43; 1st quartile 18.09; median 24.74; 3rd quartile 31.01.
  - Gov't spending on social security (1,000 RMB): No. of obs 2,950; Mean 1.369; Std. Dev. 0.732; 1st quartile 0.870; median 1.215; 3rd quartile 1.666.
  - Gov't spending on healthcare (1,000 RMB): No. of obs 2,751; Mean 0.919; Std. Dev. 0.390; 1st quartile 0.643; median 0.874; 3rd quartile 1.125.
  - Rural house price to income ratio (% of disposable income): No. of obs 2,651; Mean 40.18; Std. Dev. 16.26; 1st quartile 30.69; median 36.74; 3rd quartile 45.28.
  - Urbanization rate (%): No. of obs 2,791; Mean 57.69; Std. Dev. 14.32; 1st quartile 47.32; median 55.79; 3rd quartile 66.17.
- Household-level (selected):
  - Savings Rate (%): No. of obs 30,033; Mean 1.76; Std. Dev. 55.004; 1st quartile -27.63; median 13.311; 3rd quartile 42.929.
  - Urban Hukou HHs (dummy): No. of obs 30,033; Mean 0.517; Std. Dev. 0.5.
  - Pensions (dummy): No. of obs 30,033; Mean 0.741; Std. Dev. 0.438.
  - ln(Total income): No. of obs 30,033; Mean 11.113; Std. Dev. 0.903.
  - Net housing asset: No. of obs 26,834; Mean 820,220; Std. Dev. 1,905,957; 1st quartile 100,000; median 300,000; 3rd quartile 800,000.
  - Net housing asset to income ratio: No. of obs 31,996; Mean 8.5; Std. Dev. 22.7; 1st quartile 0.0; median 3.3; 3rd quartile 8.5.
  - Homeownership dummy: No. of obs 36,659; Mean 0.7; Std. Dev. 0.5.

### Policy implications and recommendations
- Priority 1 — Increase social spending, targeted to rural and vulnerable prefectures:
  - Further increase social spending to align with international peers to help reduce high household savings.
  - Given fiscal constraints on local governments, prioritize social spending targeted toward rural households to generate larger reductions in household savings rates in those areas.
- Priority 2 — Continue and deepen Hukou reforms in major cities:
  - Gradual and further Hukou reforms warranted to reduce the social benefit gap between migrant (rural Hukou) and urban Hukou households.
  - Further liberalize Hukou registration restrictions and provide more equitable social benefits to households regardless of Hukou status, particularly in cities above 3 million population.
- Priority 3 — Address housing-market distortions to reduce savings among higher-income households:
  - Measures that facilitate a more efficient and less costly transition in the real estate sector can help boost consumption and reduce household savings—particularly among higher-income and homeowner households.
- Comprehensive approach:
  - Reducing aggregate household savings requires a comprehensive package addressing high savings among both low-income and wealthy households.

_Source: wpiea2025259-source-pdf — 1. Introduction; Annex and selected tables and conclusions from "Reforms to Reduce China’s High Household Savings", Working Paper No. WP/2025/259._

### 1. Introduction ........................................................................................................

### 1. Introduction

### Key context and motivation
- High household savings are a structural feature of the Chinese economy and far exceed the levels in OECD and most EMDE countries (Figure 1).
- High savings have facilitated the Chinese economy’s over-reliance on debt-financed investment and led to the buildup of vulnerabilities (IMF, 2024).
- Recent sluggishness in domestic demand and deflationary pressures—partly due to the housing market correction and low consumer confidence—have renewed focus on the high savings rate and raised concerns about domestic and external imbalances (Niemeläinen, 2021; Wright et al., 2024; Gourinchas et al., 2024).
- The 2025 Central Government Work Report and the December 2024 Central Economic Work Conference identified stimulating private consumption demand as the top policy priority for China.
- Authorities have rolled out counter-cyclical fiscal support and modestly strengthened social safety nets, including:
  - equipment trade-in and upgrade programs providing subsidies for durable goods;
  - central government allocation of 150 billion yuan in 2024 to support these programs;
  - in January 2025, an announced expansion with a pre-allocation of 81 billion yuan in new funding;
  - central government support financed by issuance of ultra-long-term special purpose government bonds (150 billion yuan in 2024 and 300 billion yuan in 2025).

### Data sources and novel contribution
- Analysis relies on multiple data sources, including prefecture-level macro indicators (urban and rural household savings rate, government social expenditure) and household-level survey data.
- China Family Panel Studies (CFPS) household survey covers more than 6,000 urban households biennially from 2010 to 2022; this paper is the first to use the 2022 CFPS to investigate recent trends.
- Prefecture-level analysis uses annual data covering around 250 prefectures from 2012 to 2022.
- Novel empirical contributions:
  - Prefecture-level analysis of medium-term impact of government social spending on savings and heterogeneity by demographic vulnerability.
  - Household-level analysis of the role of Hukou and Hukou reforms using micro-level survey data.
  - Household-level disentangling of housing market wealth effect versus downpayment effect, focusing on the post-2021 real estate market correction.

### Three policy questions addressed and main empirical findings
1. What is the impact of government social spending on household savings?
   - Government social spending in China remains below peer economies and is unevenly distributed across regions.
   - Healthcare and social security spending are reported as 3.5 and 3.1 percent of GDP, respectively, compared to OECD averages of 7.2 and 8.2 percent of GDP and emerging markets’ averages of 5.0 and 6.1 percent.
   - Prefecture-level analysis (annual data, ~240 prefectures, 2012–2022) indicates:
     - A 10 percent increase in government expenditure on social security is associated with a reduction in the medium-term rural household savings rate of 0.5 percentage points.
     - A 10 percent increase in government expenditure on healthcare is associated with a reduction in the medium-term rural household savings rate of 0.8 percentage points.
   - Impact on urban households from increased social spending is limited.
   - Targeted measures that raise social spending in prefectures with higher rural population shares would be more effective in boosting consumption.
   - Recent fiscal strain on local governments amid the housing market adjustment and the pandemic has undermined social spending in some regions; social spending is uneven by city tier (Tier-1 cities—5.8 percent of the population—had annual total per capita social spending on healthcare, social security, and education more than double the average in the rest of the country).
   - More than 90 percent of social spending is funded by local governments, implying higher fiscal burdens on some regions.
   - Using a broader definition, China’s social security spending reached about RMB 7.6 trillion, or 5.6 percent of GDP (2024 Statistical Bulletin on Human Resources and Social Security Development); this ratio remains below OECD and EMDE averages.

2. Does obtaining urban Hukou status and the accompanying social benefits impact urban households’ savings?
   - Hukou system introduced in 1958 classifies citizens as urban or rural residents; migrant households without urban Hukou are excluded from full access to urban social benefits.
   - Over the past ten years, Hukou reforms eased urban Hukou registration in smaller cities and narrowed social benefit gaps, but reform progress in larger cities has been slow.
   - A significant share of the migrant urban population (over 200 million) continue to hold rural Hukous with limited benefits.
   - As of 2023, 27 percent of the residents in urban areas held rural Hukou, which represents 17.8 percent of total population in China.
   - Household-level regressions using micro-level survey data find:
     - The savings-to-disposable-income ratio of urban households without an urban Hukou is on average about 6.8 percentage points higher than that of their peers with an urban Hukou.
   - Implication: further Hukou reforms granting more urban Hukous to migrant households and providing more equitable social benefits to rural Hukou households can reduce excess savings in urban areas.

3. Has the ongoing correction in the real estate market contributed to rising household savings, and what are the key channels?
   - Two primary channels: the downpayment channel for non-homeowners and the wealth effect channel for homeowners.
   - Household-level data disentangles these channels in the post-2021 housing downturn.
   - Findings:
     - The wealth effect has persisted since the start of the real estate market correction in 2021, with a significant negative relationship between homeowners’ net housing assets (NHA) to income ratio and their savings rates.
     - The downpayment channel has weakened recently: non-homeowners with low deposits no longer save more for future downpayments and mortgage payments, potentially due to subdued confidence and delayed plans for home purchases amid the protracted real estate market downturn.

### Literature positioning
- Builds on literature on drivers of China’s high household savings: demographic changes; weak social safety nets; house prices and homeownership; income inequality.
- Most closely related to Zhang et al., (2018) and Han and Zhang (2022); extends by using up-to-date data (post-COVID and post-2021 housing correction).
- Contributes to macro-financial literature on house prices, household wealth, and consumption by providing household-level evidence for the housing market downturn since 2021.

*Source: wpiea2025259-source-pdf - 1. Introduction.*

### Annex Table A1 reports the summary statistics of prefecture-level variables included in the analysis. The key

### Annex Table A1 reports the summary statistics of prefecture-level variables included in the analysis

### Prefecture-level summary and definitions
- Household savings rate is defined as the difference between household disposable income and consumption expenditure, divided by household disposable income.
- Household saving rates are consistently higher in urban areas, both on average and across all quartiles.
- Inter-quartile range in government social spending across prefectures amounts to RMB 0.5-0.8 thousand per person per year for social security, healthcare, and education.

### Local projections: specification and controls
- Baseline prefecture-level panel fixed-effects local projections specification (equation (1)) models the cumulative change in household savings rate from year t−1 to t+ h as a function of:
  - log(Spending_SocialSecurity), log(Spending_Health), log(Spending_Edu);
  - Vector X of prefecture-level characteristics: house price-to-income ratio (different for urban and rural subsamples), house price growth, GDP per capita, first lag of household savings rate;
  - Prefecture-fixed-effects γi and year-fixed-effects γt.
- Coefficients of interest: β1,h, β2,h, β3,h. Negative and significant coefficients indicate higher government social spending in year t is followed by a reduction in household savings rate from t to t+ h.
- Equation (2) augments (1) with interaction terms between spending variables and a time-invariant vulnerability dummy Vul_i (based on urbanization-rate below sample median or not belonging to Tier 1/2 prefectures).

### Main empirical findings (prefecture-level)
- A 10 percent increase in government expenditure on social security is associated with a reduction in rural household savings rate of 0.5 percentage points five years after the spending hike.
- A 10 percent increase in government expenditure on healthcare is associated with a reduction in rural household savings rate of 0.8 percentage points five years after the spending hike.
- Medium-term impact on urban households is negligible.
- Impact of government spending on education (Panel B, Annex Table A2) is short-lived in the rural sub-sample and becomes insignificant 3 years after the spending hike.
- Interaction results (equation (2), Annex Table A3; Figure 5):
  - A 10 percent increase in health-related social spending is followed by a decrease in rural household savings rate by up to 1.3 percentage points in low-urbanization-rate prefectures.
  - This 1.3 percentage points effect is a full percentage point higher than the effect in high-urbanization-rate prefectures (implying 0.3 percentage points in high-urbanization? — preserve textual relation as presented).
  - Low-tier prefecture cities exhibit more substantial and persistent impacts from health-related spending than high-tier prefectures.
- Interpretation: social security and healthcare spending are more relevant for insuring against future shocks and have larger and more durable impacts on precautionary saving motives than education spending.

### Heterogeneity and vulnerability
- Vulnerability indicators:
  - Vul_i = 1 if historical average urbanization rate falls below the sample median or if prefecture is not Tier 1 or Tier 2.
  - Tier 1 prefectures include Beijing, Shanghai, Guangzhou, and Shenzhen.
  - Tier 2 prefectures include Tianjin, Chongqing, all provincial capital cities, and another four sub-provincial prefecture cities (Dalian, Ningbo, Qingdao, and Xiamen).
- Findings indicate larger and more persistent reductions in rural household savings in prefectures with lower urbanization rates or lower-tier cities following increases in health-related social spending.

### Caveats and robustness
- Caveat 1: Government social spending data are available only at the prefecture level (not separately for urban vs rural allocations). Results should be interpreted as the impact on urban or rural household savings rates following an increase in government social spending across the entire prefecture.
- Caveat 2: Local government social spending is endogenous to unobservable prefecture-level characteristics that may also impact household savings. Prefecture- and time-fixed-effects and prefecture-level controls mitigate this concern but do not eliminate it; findings should be interpreted as associations rather than definitive causal relationships.
- Robustness tests:
  - Balanced-panel re-estimation (Annex Table A4, Panel A): higher spending related to social security and health remains robust in reducing rural household savings.
  - Specifying social spending as growth rates instead of log levels (Annex Table A4, Panel B): results remain robust.
  - Instrumental-variable robustness: Bartik shift-share IV constructed by interacting national growth rate of general government fiscal expenditure on health with each province’s exposure to public health expenditure (proxied by share of population aged 65 and above) was tested and confirmed robustness of results.

### Quantification of aggregate impact: two scenarios
- Scenario 1 (doubling current level of government spending on social security and healthcare in all prefectures):
  - Total fiscal cost per year: 3.0 percent of GDP.
  - Cumulative increase in consumption over a five-year horizon: 2.4 percentage points of GDP.
  - Decline in the savings rate: 1.5 percentage points of GDP.
- Scenario 2 (doubling social security and healthcare spending only in prefectures with higher-than-median rural population shares):
  - Additional fiscal cost per year: 1.0 percent of GDP.
  - Cumulative increase in consumption over a five-year horizon: 1.8 percent of GDP.
  - Decline in the savings rate: 1.1 percentage points of GDP.
- Policy implication: targeted increases in social spending in rural/high-rural-share prefectures yield higher private-consumption impact per unit of fiscal cost, because current social spending in high rural-share prefectures is significantly lower and increases have larger effects on rural households’ savings.

### Household-level analysis: data, definitions, and main specifications
- Data: China Family Panel Studies (CFPS), biennial survey spanning 2010 to 2022 across 31 provinces; sample covers between 13,000 to 17,000 individuals from over 6,000 urban households per round.
- Latest CFPS round used: 2022 (published in November 2024).
- Household savings rate (urban analysis) = (disposable income − annual expenditures) / disposable income.
  - Savings rate observations are winsorized at the 10 percent level and excluded if below -200 percent.
- Urban Hukou status dummy: 1 for non-agricultural or residential registrations; 0 for agricultural registrations.
- Pension indicators:
  - About 74 percent of the whole urban household sample is covered by pensions (based on household heads’ pension coverage).
  - Coverage is around 80 percent if based on whether any family members are covered by pensions.
  - Coverage is around 90 percent among the urban households subsample whose heads are pensioners.
- Household-level regression specification (equation (3)) models household saving rate as a function of UrbanHukou, Pensions, Controls, province-year fixed effects, and industry fixed effects.

### Hukou background and relevance
- Hukou system established in 1958; urban population share reached nearly 70 percent by 2023.
- An estimated 200 million urban residents still hold rural Hukous, limiting access to urban social benefits and raising precautionary savings incentives.
- Hukou reforms since early 2000s:
  - Restrictions relaxed in small- and medium-sized cities; after 2014 the State Council advocated for abolition of rural-urban distinction.
  - Hukou restrictions fully lifted in cities with populations less than 3 million; point-based systems used in large cities.
  - Efforts to equalize social benefits in certain regions for all urban residents regardless of Hukou status (compulsory education, employment services, basic pension and medical coverage, housing security).
- Remaining disparities: in large cities strict residency rules for home purchase and access to education, employment, and social security benefits continue to disadvantage rural Hukou holders.

*Source: IMF staff calculations and CFPS/CEIC data as presented in the provided content.*

### Section 3.2.1. As in the literature (e.g., Zhang et al., 2018), the baseline specification controls for province-year

### Section 3.2.1. As in the literature (e.g., Zhang et al., 2018), the baseline specification controls for province-year

### Baseline specification and identification
- Baseline specification controls for province-year fixed effects for households’ registered locations and industry fixed-effects for household heads’ employment.
- Estimated coefficients should be interpreted as the difference in savings rates between urban Hukou and rural Hukou households within the same province-year after controlling for the industry in which the household head works and other controls.
- Standard errors are clustered at the province-level.

### Household-level regression results (Table 1)
- Sample: CFPS 2012-2022; full urban sample. Observations: 30,033. R-squared: 0.21. Industry FE: Yes. Province-Year FE: Yes.
- Key coefficient estimates (Table 1):
  - Urban Hukou: -7.03*** (Column 1; standard error 0.82); -7.88*** (Column 2; standard error 0.85).
  - New urban Hukou: -10.09*** (Column 2; standard error 1.55).
  - Existing urban Hukou: -6.80*** (Column 2; standard error 0.82).
  - NUP reform: -3.69* (Column 3; standard error 2.06).
  - Urban Hukou x NUP reform: 6.07** (Column 3; standard error 2.22).
  - Pension: -2.06*** (Columns 1 and 2; standard error 0.56); -1.82*** (Column 3; standard error 0.54).
  - Education: -1.39*** (Column 1; standard error 0.10); -1.38*** (Column 2; standard error 0.10); -1.32*** (Column 3; standard error 0.11).
  - ln(income): 32.51*** (Column 1; standard error 0.82); 32.53*** (Column 2; standard error 0.82); 32.49*** (Column 3; standard error 0.81).
  - Debt-to-income ratio: -11.36*** (Column 1; standard error 0.86); -11.34*** (Column 2; standard error 0.86); -11.36*** (Column 3; standard error 0.85).
  - Household size: -3.54*** (all columns; standard error 0.23).
  - Age: -0.14 (Column 1; standard error 0.15); -0.15 (Column 2; standard error 0.15); -0.10 (Column 3; standard error 0.13).
  - Age^2: 0.45*** (Column 1; standard error 0.14); 0.46*** (Column 2; standard error 0.14); 0.42*** (Column 3; standard error 0.13).
- Main findings from Table 1:
  - Urban households with urban Hukou save significantly less: about 7.0 percentage points lower than rural Hukou peers (Table 1, Column 1).
  - New urban Hukou households save about 10.1 percentage points lower than rural Hukou households (Table 1, Column 2), versus around 7.0 percentage points for existing urban Hukou households.
  - Pension coverage of the household head is associated with significantly lower household savings; the effect is larger for pensions in the public sector, SOEs, or large private entities.

### Robustness and implications (household analysis)
- Robustness tests in Annex Table A6: incorporate household fixed effects (Panel A) and a strictly balanced sample (Panel B); these tests affirm robustness of key findings.
- Policy implication: further Hukou reforms could have a substantial marginal effect in reducing urban household savings by reducing precautionary motives.

### Trends over time: cross-sectional results (Table 2) and Figure 7
- Cross-sectional regressions (Table 2) estimate the urban Hukou coefficient for survey subsamples by year. Controls: province FE, industry FE, household controls.
- Table 2 estimates (Urban hukou coefficient by year; standard errors in parentheses):
  - 2012: -7.69*** (2.19)
  - 2014: -11.83*** (1.81)
  - 2016: -7.60*** (1.54)
  - 2018: -7.50*** (1.52)
  - 2020: -2.64* (1.60)
  - 2022: -3.20** (1.57)
- Observations by year: 2012: 3,796; 2014: 5,197; 2016: 5,559; 2018: 6,077; 2020: 4,606; 2022: 4,803.
- R-squared by year: 2012: 0.28; 2014: 0.20; 2016: 0.21; 2018: 0.18; 2020: 0.20; 2022: 0.17.
- Trend interpretation:
  - The savings-rate gap peaked in 2014 at approximately 11.8 percentage points (urban Hukou households saving less).
  - By 2022, the gap narrowed to 3.2 percentage points (urban households saving 3.2 percentage points less than rural peers).
  - Narrowing driven primarily by a sustained decrease in savings rate of rural Hukou households; urban Hukou households’ savings rate remained relatively stable.
- Policy implication: lifting urban Hukou registration restrictions on migrant households or further enhancing rural Hukou households’ social benefits could substantially reduce aggregate savings.

### Prefecture-level event studies: New-Type Urbanization Plan (NUP) (Section 3.3)
- NUP background:
  - Nationwide program launched in 2014 to facilitate urbanization.
  - Goal: issue 100 million new urban Hukous to rural migrants and increase urban Hukou share from 35 percent in 2014 to 45 percent by 2020.
  - NUP included three batches of 236 pilot cities (prefecture-level and county-level cities) announced during 2014-2016.
- Panel local projections specification (equation (4)) tests NUP impact on urban households’ savings rate; negative β1,h indicates larger decline in savings in NUP prefectures.
- Event study (Figure 8) results:
  - Medium-term cumulative decline in urban households’ savings rate estimated at about 2.5 percentage points following NUP pilot city designation.
- Micro-level DID specification (equation (5)) using CFPS:
  - NUP_city,t dummy equals 1 if survey year t is at least 2 years after household i’s registered prefecture is designated as an NUP pilot prefecture-level city.
  - Table 1 Column 3: average urban household savings rate decreased by 3.7 percentage points after NUP designation (NUP reform: -3.69* as reported in Table 1 Column 3; standard error 2.06).
  - Positive and significant Urban Hukou x NUP interaction (6.07**, standard error 2.22) indicates NUP mitigated the savings gap between urban Hukou and rural Hukou households.
- Interpretation: NUP liberalization of Hukou registration contributed to declines in urban household savings and reduced precautionary saving differences.

### Real estate market and household savings: channels and empirical strategy (Sections 4.1–4.2)
- Real estate sector context:
  - Sector accounts for approximately 20 percent of GDP when accounting for linkages across sectors.
  - Homeownership rate: over 90 percent of households owning properties.
  - As of 2024, residential property sales measured by floor space sold and secondary market house prices have declined by over 40 percent and 10 percent, respectively, from the peak in 2021.
- Two hypothesized channels linking housing to savings:
  - Downpayment channel: falling housing prices reduce required downpayments, lowering incentive to save for potential buyers; but slumps can induce delays in purchase and dampen this effect.
  - Wealth effect channel: declines in house prices reduce property-related net worth, potentially increasing savings (precautionary).
- Baseline empirical specification (equation (6)) includes:
  - NHA_i,t: net housing asset to income ratio = (Property_Value_i,t − Mortgage_Balance_i,t)/Income_i,t (winsorized at 500).
  - Deposit_i,t: deposit-to-income ratio for non-homeowners.
  - Non_homeowner dummy and interaction Deposit × Non_homeowner.
  - Household fixed effects and year fixed effects.

### Housing-channel regression results (Table 3)
- Sample: CFPS Urban, 2012-2022. Observations: 31,905. Household FE: Yes. Time FE: Yes. R-squared: 0.105 (Column 1); 0.106 (Column 2).
- Key coefficient estimates (Table 3; standard errors in parentheses):
  - Log(Total income): 16.969*** (0.368) in Column (1); 16.935*** (0.370) in Column (2).
  - NHA to income ratio: -0.020*** (0.006) in both columns.
  - Deposit to income ratio: -0.196*** (0.068) in Column (1); -0.139* (0.071) in Column (2).
  - Dummy for non-homeowner: 2.778*** (0.738) in Column (1); 3.070*** (0.790) in Column (2).
  - Deposit to income ratio × Dummy for non-homeowners: -1.283*** (0.251) in Column (1); -1.443*** (0.279) in Column (2).
  - NHA to income ratio × 2022dummy: 0.010 (0.033) in Column (2) — not significant.
  - Deposit to income ratio × 2022dummy: -0.583*** (0.206) in Column (2).
  - Dummy for non-homeowner × 2022dummy: -1.833 (1.516) in Column (2) — not significant.
  - Deposit to income ratio × Dummy for non-homeowners × 2022dummy: 1.060* (0.602) in Column (2).
- Interpretation of magnitudes and tests:
  - Wealth effect: NHA-to-income ratio negative and significant (β2<0). A 0.5 standard-deviation decline in NHA-to-income ratio (equivalent to a decline in house price of approximately 20 percent) is associated with an increase in urban household savings rate of 0.35 percentage points (based on coefficient 0.020).
  - Downpayment (extensive margin): non-homeowners save more; dummy for non-homeowner ≈ 2.8 percentage points higher savings than homeowners (Column 1).
  - Downpayment (intensive margin): interaction Deposit × Non_homeowner negative and significant; for a non-homeowner, savings increase by about 1.3 percentage points more compared to a homeowner, for a one-unit decrease in bank deposit-to-income ratio (i.e., difference in deposit equivalent to annual income).
  - Changes since 2022 (real estate correction): wealth effect persisted (NHA × 2022dummy not significant); downpayment channel weakened — triple interaction suggests partial offset, and deposit-related downpayment motive for non-homeowners with less deposits weakened since 2022 (Deposit × Non_homeowners × 2022dummy = 1.060*, standard error 0.602), consistent with delayed home purchases by some non-homeowners.

*Source: CFPS 2012-2022; IMF staff calculations.*

### 5. Conclusion

### 5. Conclusion

### Major empirical findings on drivers of high household savings
- Scope and data:
  - Analysis uses multiple data sources in China, including prefecture-level indicators (e.g., government social expenditure) and household-level survey data covering 2012 to 2022.
- Social safety nets and government social spending:
  - The weak social safety net in China remains a key factor for high household savings.
  - A doubling of government expenditure on social security and healthcare (prefecture-level data, 2012–2022) is associated with the medium-term rural household savings rate declining by 4.8 and 5.9 percentage points, respectively.
  - The effect of increased social spending on lowering savings is stronger for vulnerable populations (those residing in lower tier cities and regions with lower urban population share).
- Hukou system and migrant households:
  - Urban migrants with rural Hukous tend to save more than their urban Hukou peers.
  - The savings-rate gap between migrant (rural Hukou) and urban Hukou households has narrowed in the past decade, potentially reflecting reforms, but a significant gap still remains—Hukou registration continues to influence saving decisions.
  - Full liberalization of the Hukou system has only been implemented in cities with urban populations under 3 million.
  - Restrictions on obtaining an urban Hukou or fully accessing social benefits remain in place in larger cities, where about half of the country’s population resides.
- Real estate market correction channels:
  - The paper disentangles two channels of the real estate market correction (begun in 2021): the downpayment channel for non-homeowners and the wealth effect channel for homeowners.
  - The wealth effect has persisted since the correction began in 2021: there is a significant negative relationship between homeowners' net housing assets to income ratio and their savings rates.
  - The downpayment channel has weakened recently: non-homeowners with low deposits no longer save more for future downpayments and mortgage payments, possibly due to subdued confidence and delayed home purchase plans amid a prolonged real estate downturn.
- Distributional pattern of savings:
  - High-income households tend to save more; the savings rate of the top deciles is above their peers in most other economies (Figure 9 reference).

### Key summary statistics (prefecture- and household-level)
- Prefecture-level summary statistics (Table A1):
  - Urban HH savings rate (% of disposable income): No. of obs 2,962; Mean 35.22; Std. Dev. 7.205; 1st quartile 30.99; median 35.46; 3rd quartile 39.76.
  - Rural HH savings rate (% of disposable income): No. of obs 2,463; Mean 24.23; Std. Dev. 10.43; 1st quartile 18.09; median 24.74; 3rd quartile 31.01.
  - Gov't spending on social security (1,000 RMB): No. of obs 2,950; Mean 1.369; Std. Dev. 0.732; 1st quartile 0.870; median 1.215; 3rd quartile 1.666.
  - Gov't spending on healthcare (1,000 RMB): No. of obs 2,751; Mean 0.919; Std. Dev. 0.390; 1st quartile 0.643; median 0.874; 3rd quartile 1.125.
  - Gov't spending on education (1,000 RMB): No. of obs 3,239; Mean 1.770; Std. Dev. 0.827; 1st quartile 1.272; median 1.615; 3rd quartile 2.025.
  - Growth rate of house price (%): No. of obs 3,021; Mean 3.894; Std. Dev. 2.942; 1st quartile 1.802; median 3.578; 3rd quartile 5.511.
  - Urban house price to income ratio (% of disposable income): No. of obs 2,977; Mean 18.32; Std. Dev. 7.651; 1st quartile 14.13; median 16.51; 3rd quartile 20.19.
  - Rural house price to income ratio (% of disposable income): No. of obs 2,651; Mean 40.18; Std. Dev. 16.26; 1st quartile 30.69; median 36.74; 3rd quartile 45.28.
  - GDP per capita (RMB 1,000): No. of obs 3,324; Mean 57.44; Std. Dev. 33.40; 1st quartile 33.73; median 48.49; 3rd quartile 71.54.
  - Urbanization rate (%): No. of obs 2,791; Mean 57.69; Std. Dev. 14.32; 1st quartile 47.32; median 55.79; 3rd quartile 66.17.
- Household-level summary statistics (Table A5):
  - Savings Rate (%): No. of obs 30,033; Mean 1.76; Std. Dev. 55.004; 1st quartile -27.63; median 13.311; 3rd quartile 42.929.
  - Urban Hukou HHs (dummy): No. of obs 30,033; Mean 0.517; Std. Dev. 0.5; 1st quartile 0; median 1; 3rd quartile 1.
  - Pensions (dummy): No. of obs 30,033; Mean 0.741; Std. Dev. 0.438; 1st quartile 0; median 1; 3rd quartile 1.
  - ln(Total income): No. of obs 30,033; Mean 11.113; Std. Dev. 0.903; 1st quartile 10.597; median 11.156; 3rd quartile 11.695.
  - Net housing asset: No. of obs 26,834; Mean 820,220; Std. Dev. 1,905,957; 1st quartile 100,000; median 300,000; 3rd quartile 800,000.
  - Net housing asset to income ratio: No. of obs 31,996; Mean 8.5; Std. Dev. 22.7; 1st quartile 0.0; median 3.3; 3rd quartile 8.5.
  - Homeownership dummy: No. of obs 36,659; Mean 0.7; Std. Dev. 0.5; 1st quartile 0; median 1; 3rd quartile 1.

### Robustness and dynamics (high-level)
- Local projections and difference-in-difference analyses (prefecture-level) consistently indicate that increased government spending on healthcare and education is associated with reductions in rural household savings over medium horizons (panels and robustness tables provide statistical support across horizons T+0 to T+5).
- Household-level robustness tests indicate that having an urban hukou is associated with lower savings rates; pension receipt is associated with lower savings; education and income are strong correlates of savings behavior (Table A6).

### Policy implications and recommendations
- Priority 1 — Increase social spending, targeted to rural and vulnerable prefectures:
  - Further increasing social spending to align with international peers can help reduce high household savings.
  - Given fiscal constraints faced by local governments (LGs), prioritize social spending targeted toward rural households to generate larger reductions in household savings rates in those areas.
- Priority 2 — Continue and deepen Hukou reforms in major cities:
  - Gradual and further Hukou reforms are warranted to reduce the social benefit gap between migrant (rural Hukou) and urban Hukou households.
  - Further liberalize Hukou registration restrictions and provide more equitable social benefits to households regardless of Hukou status, particularly in cities above 3 million population where reforms remain incomplete.
- Priority 3 — Address housing-market distortions to reduce savings among higher-income households:
  - Because the downward correction in the housing market chiefly affects household savings via the wealth effect, measures that facilitate a more efficient and less costly transition in the real estate sector can help boost consumption and reduce household savings—particularly among higher-income and homeowner households.
- Comprehensive approach:
  - Reducing aggregate household savings requires a comprehensive package that addresses high savings among both low-income and wealthy households (Figure 9 shows top deciles save more).

*Source: IMF Working Paper — "Reforms to Reduce China’s High Household Savings", 5. Conclusion (chapter text and annex tables).*

### References

### References

### References list
- Bartik, T., 1991, “Who Benefits from State and Local Economic Development Policies?” W.E. Upjohn Institute.  
- Blanchard, O., and F. Giavazzi, 2006, “Rebalancing Growth in China: A Three-Handed Approach,” China and World Economy, Vol. 14(4), pp. 1-20.  
- Chamon, M. D., and E. Prasad, 2010, “Why are Saving Rates of Urban Households in China Rising?” American Economic Journal: Macroeconomics, Vol. 2(1), pp. 93-130.  
- Chen, J., Hardin, W., Hu, M, 2018, “Housing, Wealth, Income and Consumption: China and Homeownership Heterogeneity”, Real Estate Economics, Vol. 48(2), pp. 373-405.  
- Chen, B., Yang, X., and N. Zhong, 2020, “Housing Demand and Household Saving in China: Evidence from a Housing Reform,” Journal of Housing Economics, Vol. 49.  
- Choukhmane, T., Coeurdacier, N., and K. Jin, 2023, “The One-Child Policy and Household Savings,” Journal of the European Economic Association, Vol. 21, Issue 3.  
- Huang, Y., D. Yi, and W. A.V. Clark, 2020, “Multiple Home Ownership in Chinese Cities: An Institutional and Cultural Perspective.” Cities, Vol. 97, 102518.  
- Curtis, C. C., Lugauer, S., and N. C. Mark, 2015, “Demographic Patterns and Household Saving in China,” American Economic Journal: Macroeconomics, Vol. 7(2), pp. 58–94.  
- Dong, Z., Hui, E., Jia, S., 2017, “How does housing price affect consumption in China: Wealth effect or substitution effect?”, Cities, Vol. 64, pp. 1-8.  
- Duesterberg, T., and A. Aibel, 2024, “How Weakness in the Social Safety Net Undermines the Political Compact in China.” Hudson Institute, December 9, 2024. https://www.hudson.org/economics/how-weakness-social-safety-net-undermines-political-compact-china-thomas-duesterberg-alexander-aibel  
- Fan, Y. and Yavas, A., 2020, “How does mortgage debt affect household consumption? Micro evidence from China”, Real Estate Economics, Vol. 48(1), pp. 48-88.  
- Goldsmith-Pinkham, P., I. Sorkin, and H. Swift, 2020, “Bartik Instruments: What, When, Why, and How.” American Economic Review, Vol. 110(8), pp. 2586-2624.  
- Gourinchas, P., C. Pazarbasioglu, K. Srinivasan, and R. Valdés, 2024, “Trade Balances in China and the US Are Largely Driven by Domestic Macro Forces.” IMF Blog.  
- Han, F., F. Zhang, 2022, “Household Savings and Its Drivers: Some Stylized Facts.”, IMF Selected Issues Paper.  
- He, H., Ning, L., and D. Zhu, 2019, “The Impact of Rapid Ageing and Pension Reform on Savings and the Labor Supply: The Case of China,” IMF Working Paper 19/61 (Washington: International Monetary Fund).  
- Iacoviello, Matteo. 2011, “Housing Wealth and Consumption”, International Finance Discussion Papers No. 1027, Federal Reserve Board.  
- International Monetary Fund, 2024, “People’s Republic of China: 2024 Article IV Consultation Staff Report.”  
- Lee, Seungyoon, 2023, “House prices, homeownership, and household consumption: Evidence from household panel data in Korea”, Economic Modelling, Vol. 126.  
- Modigliani, F., and S. L. Cao, 2004, “The Chinese Saving Puzzle and the Life-Cycle Hypothesis,” Journal of Economic Literature, Vol. XLII, pp. 145–170.  
- Niemeläinen, J., 2021, “External Imbalances Between China and the United States: A Dynamic Analysis with a Life-Cycle Model.” Journal of International Economics, Vol. 131, pp. 1-30.  
- Qi, Y., Yu, G., Liu, X., abd Ren, Y., 2023, “Housing Wealth Appreciation and Heterogeneous Household Consumption: Evidence from China.” PLoS ONE 18(10): e0289712.  
- Wang, X., and Wen, Y., 2012, “Housing prices and the high Chinese saving rate puzzle”, China Economic Review, Vol. 23(2), pp. 265-283.  
- Wingender, P., 2018, “Intergovernmental Fiscal Reform in China.” IMF Working Paper 18/88.  
- Wright, L., C. Boullenois, C. A. Jordan, E. Tian, and R. Quinn, 2024, “No Quick Fixes: China’s Long-Term Consumption Growth.” Rhodium Group, July 18, 2024. https://rhg.com/research/no-quick-fixes-chinas-long-term-consumption-growth/  
- Yang, D. T., 2012, “Aggregate Savings and External Imbalances in China.” Journal of Economic Perspectives, Vol. 25, No. 4, pp. 125-146.  
- Zhang, C., and F. Zhang, 2015, “The Effects of Housing Prices on Income Inequality in Urban China,” Critical Housing Analysis, Vol. 2, Iss. 2, pp. 11-18.  
- Zhang, L., Brooks, R., Ding, D., Ding, H. He, H. Lu, J. and Mano, R. 2018, “China’s High Savings: Drivers, Prospects and Policies” IMF Working Paper WP/18/277.

_References (from wpiea2025259-source-pdf - References) — Reforms to Reduce China’s High Household Savings, Working Paper No. WP/2025/259_

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025259-source-pdf.pdf_