## _wp10289

## Source details

**Canonical URL:** [_wp10289](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10289.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10289.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10289.pdf.json)

---

### Introduction: objectives and key contribution
- Objective: evaluate effects of China’s structural and policy shifts on degree of income uncertainty at the household level and analyze implications for household saving rates.
- Empirical fact motivating study: average saving rate for urban households in China increased from 15 percent in the early 1990s to over 30 percent in recent years.
- Puzzle: rising household saving rates during a period of high income growth appears inconsistent with a certainty-equivalent life-cycle hypothesis.
- Key contribution: show that the rise in income uncertainty and the 1997 pension reform together can explain more than half of the observed rise in household saving rates and the dramatic shift in the age-saving profile.

### Income uncertainty decomposition (data, method, and high-level results)
- Data source: China Health and Nutrition Survey (CHNS) panel of urban households.
- Method: decompose variance of log income residuals into permanent and transitory components following Meghir and Pistaferri (2004) and Blundell, Pistaferri and Preston (2008).
- Empirical patterns:
  - Strong trend growth in both mean and variance of total household income.
  - Variance of permanent shocks to household income: relatively stable (no clear trend).
  - Variance of transitory shocks to household income: trends upwards (rising transitory uncertainty).
- Interpretation: technological and sectoral shifts and labor reallocation can generate higher transitory uncertainty even when some shifts are permanent.
- Identification and estimation details:
  - Residual decomposition: y_iat = u_iat + v_iat; u_iat = u_iat-1 + ω_it; v_iat = ξ_it.
  - Assumptions: ω and ξ zero mean, mutually orthogonal, i.i.d. across household, time and age; variances vary by year but not by age.
  - Identification: one-period lagged autocovariance of income changes identifies transitory variance; with available waves identify five years of transitory risk (all except 2006 and 1989) and four permanent variances.
  - Estimator: equally-weighted minimum distance estimator; model is just identified.
- Measurement concerns and conservative adjustment:
  - Classical measurement error (i.i.d.) and MA(1) serial correlation can bias transitory variance upward.
  - Conservative adjustment for calibration: assume true transitory uncertainty is one-half of transitory variance identified from estimates.

### Empirical decomposition results (key numbers)
- Permanent shocks: no clear trend in variance of permanent shocks for full urban sample, SOCE subsample, age splits, or education splits (estimates noisy; large standard errors).
- Transitory shocks: clear rising trend in variance of transitory shocks for full sample; point estimates rise from 0.040 in 1991 to 0.162 by 2004 for household income.
- Robustness: excluding transfers and subsidies generally yields higher estimated transitory variance; dropping 1989 wave still confirms trend.

### Labor market and structural drivers
- SOCE employment share in sample: falls from 81 percent in 1989 to 64 percent in 2006.
- Transition probability SOCE→private: rises from 5.5 percent between 1989 and 1993 to 14 percent between 2000 and 2004.
- Suggested drivers: higher labor market turnover, SOCE restructuring, tighter firm budgets, fast technological change reducing skill transferability.

### Dataset and sample construction (exact figures)
- CHNS sample: about 4,400 households, 19,000 individuals tracked over time; waves in 1989, 1991, 1993, 1997, 2000, 2004 and 2006; covers nine provinces.
- Analysis focuses on urban sub-sample; baseline sample selection: household heads urban residents, ages 25 to 59, not a student, complete information on age and education; drop small number reporting income from farming and livestock.
- Trimming: run Mincerian earnings regression each survey year, sort sample by growth in income residuals, trim top and bottom 1 percent of distribution.
- Final analysis sample: unbalanced panel of 1689 households.
- Summary statistic: from 1989 to 2006, real mean annual household income almost triples from 12830 to 32040 RMB at constant 2006 prices.
- In-sample labor earnings: in urban sample, labor earnings comprise 97 percent of total individual income on average.
- Data quality note: concerns about 1989 wave; standard deviations of individual earnings and household income much larger in 1989 compared with next waves. Baseline results based on full panel including 1989, with robustness checks dropping 1989.

### Precautionary savings model: setup and parameterization
- Modeling approach: calibrate a buffer-stock/life-cycle model of savings (Carroll (1997) approach).
- Preferences and constraints:
  - CRRA utility, no-borrowing constraint, maximize expected discounted utility over working life to age 85.
  - Discount factor β = 0.97 (baseline).
  - Coefficient of relative risk aversion γ = 4.5 (baseline).
  - Real interest rate = 1.4 percent per annum (average real interest rate in China over 1989-2006).
  - Working life begins at age 25, initial wealth zero, initial permanent income normalized to one.
  - Retirement age R = 60; permanent income deterministic in retirement at fraction η of pre-retirement permanent income.
  - Survival: live with certainty until retirement, survival probability until age 85; Poisson death process calibrated to life expectancy in China in 2009 (73.5 years) implying constant survival probability of 0.925 between t and t+1 after retirement.
  - No bequests (assets of those who die prior to age 85 are lost).
- Income process calibration:
  - Deterministic life-cycle growth from CHNS: predicted annual income growth about 7 percent for the young, 6–7 percent for most remaining work life, declining to 2 percent close to retirement.
- Shocks used:
  - Variance of permanent shocks set at constant 0.02.
  - Variance of transitory shocks baseline 0.04 rising to 0.08 in higher uncertainty scenario.
  - These variances are lower than empirical point estimates due to conservative assumption that half of empirical transitory variance is measurement error.
- Retirement replacement assumptions:
  - Pre-reform η = 75 percent.
  - Post-reform η = 60 percent for workers retiring after 1997 (baseline pension reform calibration).
  - Alternative scenario: replacement ratio down to 50 percent (sensitivity exercise).

### Calibration simulations and age-saving profile effects (summary of simulated impacts)
- Simulation details: 5,000 households; endogenous gridpoint method; produce age-saving profiles and changes under shocks.
- Transitory variance shock alone:
  - If household head begins working at age 25 already under higher uncertainty, household saves about 4.5 percentage points more initially.
  - Younger households show larger initial jumps (about 4.5 percentage points); forty-year-old head sees initial jump about 2.5 percentage points.
- Pension replacement shock alone:
  - Decline from 75 percent to 60 percent induces substantial increase in savings, especially for older household heads.
  - Initial change: less than 1 percentage point for household with young head (early thirties); 5 percentage points for heads in mid-forties; up to 8 percentage points for heads close to end of working life.
  - Long-run: youngest households will be saving over 5 percentage points more by the time they approach retirement than under old regime.
- Combined shocks (transitory variance increase + pension reform):
  - Combined effect is less than simple sum due to interaction: precautionary buffer-stock accumulation reduces need for later life-cycle saving.
  - At switch, marked increase about 5 percentage points for households with heads in their twenties, thirties and forties.
  - Dynamics: saving rates for younger households decline after initial jump as buffer stocks are built; heads in mid-forties onwards show more stable higher saving rates; heads in their fifties exhibit very high initial jumps and sustained higher savings.
- Cross-sectional timing (t, t+5, t+10 after shocks):
  - At t+5 and t+10, youngest heads save about 5 percentage points more; oldest save about 6.5 percentage points more at t+5 and 5.5 percentage points more at t+10.
  - Profiles bottom out around 2.5 percentage points for household heads around age thirty or early thirties.
  - t+10 cross-section yields asymmetric U-shaped pattern: households in early forties save 3-4 percentage points more than before.
- Quantitative statement: the riskier environment (higher transitory variance and pension reform) is capable of generating an initial increase of about four and a half percentage points in the average household saving rate.
- Relationship to other motives: Chamon and Prasad (2010) find housing motives account for about 6 percentage points for 25-29 year olds (non-owners) and health expenditures account for about 6 percentage points for 55-59 year olds; this calibration suggests earnings uncertainty shifts and pension reforms also played important roles.

### Robustness checks and sensitivity (Table 3 reported jumps)
- Baseline assumptions (γ = 4.5; β = 0.97), jump in saving rates at time of shock:
  - Age 25: 5.4%
  - Age 30: 5.3%
  - Age 35: 5.4%
  - Age 40: 5.1%
  - Age 45: 5.8%
  - Age 50: 6.6%
  - Age 55: 7.8%
- Larger drop in replacement rate post-reform (η to 0.50 per Sin (2005)), γ = 4.5; β = 0.97:
  - Age 25: 5.5%
  - Age 30: 5.6%
  - Age 35: 6.4%
  - Age 40: 7.2%
  - Age 45: 9.0%
  - Age 50: 10.6%
  - Age 55: 12.8%
- Alternative risk aversion and discount factor combinations reported (exact cells preserved as in source), including:
  - γ = 3.5; β = 0.97: Age 25 = 5.0% ... Age 55 = 8.9%
  - γ = 2.0; β = 0.97: Age 25 = 4.4% ... Age 55 = 9.5%
  - γ = 4.5; β = 0.95: Age 25 = 5.3% ... Age 55 = 8.2%
  - γ = 3.5; β = 0.95: Age 25 = 4.9% ... Age 55 = 9.3%
  - γ = 2.0; β = 0.95: Age 25 = 4.2% ... Age 55 = 8.5%
  - γ = 4.5; β = 0.98: Age 25 = 5.5% ... Age 55 = 7.6%
  - γ = 3.5; β = 0.98: Age 25 = 5.1% ... Age 55 = 8.7%
  - γ = 2.0; β = 0.98: Age 25 = 4.5% ... Age 55 = 9.9%
- Preferences calibrated to match pre-1997 mean saving rates:
  - γ = 7.5; β = 0.99: Age 25 = 2.2% ... Age 55 = 5.1%
  - γ = 8.1; β = 0.97: Age 25 = 2.1% ... Age 55 = 5.0%
- Conservativeness emphasized: when calibrating precautionary effect, assume half of empirically identified transitory variance is measurement error; yields conservative lower-bound precautionary responses.

### Age-saving profile, expected growth scenarios, and implications
- Age-saving profile note: profile captures estimated saving behavior for a household with a head aged 25 years in 1997 as he or she ages (not the 1997 cross-section).
- U-shaped cross-sectional age-saving profile does not appear in the data until the 2000s.
- Expected income growth scenarios simulated for a 25 year old head:
  - Solid line: baseline expected growth path.
  - Dotted line: growth path that is 1 percentage point lower than in the baseline from the present onwards.
  - Dashed line: growth path where expected income growth flattens out at a rate of 0.1 percentage point every year relative to its value in the baseline path, but never declines below the lowest value in that path (1.26 percent).
- Findings:
  - Saving rates higher for households with young heads under both lower income growth paths.
  - Increase in saving larger when income slows down gradually than when decline takes place immediately.
  - Lower growth path substantially reduces retirement income in simulation, strengthening retirement saving motives and causing earlier-life saving increases.
  - Age-savings profile is flatter when slowdown is gradual; steepest under baseline where postponement of retirement savings is strongest.
- Policy implication: prospects of an eventual slowdown in Chinese growth could further increase household saving rates; some savings among the very young may already reflect these prospects.

### Policy implications and closing observations
- The documented rise in transitory income variance and the pension replacement cut from the 1997 reform together can account for roughly half of the observed rise in urban household saving rates in the sample.
- These findings can inform policy design aimed at rebalancing growth in China by boosting private consumption and reducing reliance on exports and investment.

*Source: _wp10289 (IMF working paper excerpt supplied in the PDF).*

### References

### References

### Section listing
- References ................................................................................................................................22
-  
- 3
- I.

*Source: _wp10289 - References*

### INTRODUCTION

### _wp10289 - INTRODUCTION

### Major findings and objectives
- Objective: evaluate effects of China’s structural and policy shifts on degree of income uncertainty at the household level and analyze implications for household saving rates.
- Empirical fact motivating study: average saving rate for urban households in China increased from 15 percent in the early 1990s to over 30 percent in recent years.
- Households are main contributors to China’s large national savings.
- Puzzle: rising household saving rates during a period of high income growth appears inconsistent with a certainty-equivalent life-cycle hypothesis.
- Key contribution: show that the rise in income uncertainty and the 1997 pension reform together can explain more than half of the observed rise in household saving rates and the dramatic shift in the age-saving profile.

### Income uncertainty decomposition (method and high-level results)
- Data source: China Health and Nutrition Survey (CHNS) panel of urban households.
- Method: exploit panel to decompose variance of log income residuals into permanent and transitory components following Meghir and Pistaferri (2004) and Blundell, Pistaferri and Preston (2008).
- Empirical pattern:
  - Strong trend growth in both mean and variance of total household income.
  - Variance of permanent shocks to household income: relatively stable (no clear trend).
  - Variance of transitory shocks to household income: trends upwards (rising transitory uncertainty).
- Interpretation: technological and sectoral shifts and labor reallocation can generate higher transitory uncertainty even when some shifts are permanent.

### Calibration: buffer-stock/life-cycle model and quantitative implications
- Modeling approach: calibrate a simple buffer-stock/life-cycle model of savings using Carroll (1997) approach.
- Key calibrated mechanisms and results:
  - Rising transitory variance of income can help explain the rise in savings of households with young household heads.
  - For plausible parameter values, saving rates initially increase by over 4 percentage points for households with household heads in their twenties to mid-thirties.
  - Young households enter with lower buffer stocks; increased transitory variance leads them to save more to rebuild buffer stock. After initial adjustment, the response declines over time, but young entrants with no initial assets will continue to save 4 percentage points more than under the lower risk environment.
  - Households with older household heads, having accumulated significant savings, accommodate transitory shocks more easily and show smaller responses to rising transitory variance.

### Pension reform effects
- Pre-1997 pension replacement ratio (typical): about 75-80 percent relative to average wage.
- Post-1997 reformed system for workers retiring after 1997: social pension corresponding to 20 percent of average local wages, amounts in individual retirement accounts, and a supplementary “transition pension.”
- Sin (2005) estimate: replacement rate for the transition generation is around 60 percent of the average wage.
- Calibration result: decline in replacement rate from 75 percent to 60 percent of pre-retirement income can explain a 6-8 percentage point increase in saving rates for households whose heads are in their fifties and approaching retirement.
  - Effect for younger heads: initial increase about one percentage point for households with heads in their thirties; youngest cohorts will end up saving 6 percentage points more by the time they are in their fifties than pre-reform cohorts.
- Combined effect: higher income uncertainty and pension reforms together explain much of the rise in average savings among urban households and the sharper changes at the two ends of the age distribution of household heads.
  - Even 10 years after the initial increase in uncertainty and pension reform, youngest and oldest households save 5 percentage points more than before; those in their late thirties-early forties save only 2.5-3.5 percentage points more.
- Robustness: results robust to alternative parameterizations and conservative assumptions about rise in transitory income uncertainty.

### Dataset (CHNS) and sample construction
- CHNS sample: about 4,400 households, 19,000 individuals tracked over time; waves in 1989, 1991, 1993, 1997, 2000, 2004 and 2006; covers nine provinces.
- Analysis focuses on urban sub-sample.
- Baseline sample selection: household heads urban residents, ages 25 to 59, not a student, complete information on age and education; drop small number reporting income from farming and livestock.
- Trimming: run Mincerian earnings regression each survey year, sort sample by growth in income residuals, trim top and bottom 1 percent of distribution.
- Final analysis sample: unbalanced panel of 1689 households.
- Summary statistic: from 1989 to 2006, real mean annual household income almost triples from 12830 to 32040 RMB at constant 2006 prices.
- In-sample labor earnings: in urban sample, labor earnings comprise 97 percent of total individual income on average.
- SOCE employment share in sample: falls from 81 percent in 1989 to 64 percent in 2006.
- Transition probability SOCE→private: rises from 5.5 percent between 1989 and 1993 to 14 percent between 2000 and 2004.
- Data quality note: concerns about 1989 wave; standard deviations of individual earnings and household income much larger in 1989 compared with next waves. Baseline results based on full panel including 1989, with robustness checks dropping 1989.

### Methodology for earnings decomposition
- First stage: year-by-year Mincerian log income regressions controlling for region (East, Northeast, Midwest, West), age and education dummies (middle school or less, high school, some college), number of income earners, household size, employment and marriage status; regressions run separately for each year.
- Residuals used to estimate permanent (u) and transitory (v) components:
  - y_iat = u_iat + v_iat
  - u_iat = u_iat-1 + ω_it (permanent shocks ω)
  - v_iat = ξ_it (transitory shocks ξ)
- Assumptions: ω and ξ have zero means, are mutually orthogonal, i.i.d. across household, time and age; variances vary by year but not by age.
- Identification: one-period lagged autocovariance of income changes identifies variance of transitory shock; with available waves, identify five years of transitory risk (all except 2006 and 1989) and four permanent variances.
- Estimation: equally-weighted minimum distance estimator; model is just identified.
- Potential upward bias of estimated transitory variance:
  - Classical measurement error (i.i.d.) can bias transitory variance upward.
  - MA(1) serial correlation of transitory shocks can inflate identified variance.
- Conservative adjustment for calibration: assume true transitory uncertainty is one-half of transitory variance identified from estimates.

### Empirical earnings decomposition results (summary)
- Panel A (permanent shocks): no clear trend in variance of permanent shocks for full urban sample, SOCE subsample, age splits, or education splits (estimates noisy; large standard errors).
- Panel B (transitory shocks): clear rising trend in variance of transitory shocks for full sample; point estimates rise from 0.040 in 1991 to 0.162 by 2004 for household income.
  - Pattern persists in SOCE subsample and across various income definitions (excluding transfers/subsidies or using labor earnings only), though levels and noise vary.
  - Excluding transfers and subsidies generally yields higher estimated transitory variance, consistent with transfers/subsidies acting as partial insurance.
- Robustness: dropping 1989 wave still confirms trend of rising transitory income uncertainty.
- Suggested drivers of rising transitory variance:
  - Higher labor market turnover (employment→unemployment transitions increase sharply in late 1990s and 2000s).
  - High transition from SOCE to non-SOCE employment starting mid-1990s.
  - Firm-level volatility due to SOCE restructuring and tighter budget constraints.
  - Fast technological change reducing skill transferability and increasing wage losses upon separation.

### Precautionary savings calibration: model, parameters, and assumptions
- Preferences and constraints:
  - CRRA utility, no-borrowing constraint, maximize expected discounted utility over working life to age 85.
  - Discount factor β = 0.97.
  - Coefficient of relative risk aversion γ = 4.5 (baseline).
  - Real interest rate = 1.4 percent per annum (average real interest rate in China over 1989-2006).
  - Working life begins at age 25, initial wealth zero, initial permanent income normalized to one.
  - Retirement age R = 60; permanent income deterministic in retirement at fraction η of pre-retirement permanent income.
  - Survival: people live with certainty until retirement, survival probability until age 85; Poisson death process calibrated to life expectancy in China in 2009 (73.5 years) implying constant survival probability of 0.925 between t and t+1 after retirement.
  - No bequests (assets of those who die prior to age 85 are lost).
- Income process calibration:
  - Deterministic life-cycle growth from CHNS: predicted annual income growth about 7 percent for the young, 6–7 percent for most remaining work life, declining to 2 percent close to retirement.
- Shocks used in calibration:
  - Variance of permanent shocks set at constant 0.02.
  - Variance of transitory shocks baseline 0.04 rising to 0.08 in higher uncertainty scenario.
  - These variances are lower than empirical point estimates due to conservative assumption that half of empirical transitory variance is measurement error.
- Retirement replacement assumptions:
  - Pre-reform η = 75 percent.
  - Post-reform η = 60 percent for workers retiring after 1997 (baseline pension reform calibration).
  - Alternative scenario: replacement ratio down to 50 percent (sensitivity exercise).

### Calibration results: simulated saving-rate effects
- Simulation: 5,000 households; endogenous gridpoint method; produce age-saving profiles and changes under shocks.
- Transitory variance shock alone:
  - If household head begins working at age 25 already under higher uncertainty, household saves about 4.5 percentage points more initially.
  - Initial jump declines with age: household with forty-year-old head sees initial jump about 2.5 percentage points.
  - Mechanism: younger households accumulate lower buffer stocks and respond more to higher transitory variance.
- Pension replacement shock alone:
  - Decline from 75 percent to 60 percent induces substantial increase in savings, especially for older household heads.
  - Initial change: less than 1 percentage point for household with young head (early thirties), 5 percentage points for heads in mid-forties, up to 8 percentage points for heads close to end of working life.
  - Long-run: youngest households will be saving over 5 percentage points more by the time they approach retirement than under old regime.
- Combined shocks (transitory variance increase + pension reform):
  - Saving rates respond more strongly, though combined effect is less than simple sum due to interaction: higher buffer-stock savings in response to transitory shocks reduce need for later life-cycle savings.
  - At switch, marked increase about 5 percentage points for households with heads in their twenties, thirties and forties.
  - Dynamics: saving rates for younger households decline after initial jump as buffer stocks are built; households with heads in mid-forties onwards show more stable higher saving rates; households with heads in their fifties exhibit very high initial jumps and sustained higher savings.
- Cross-sectional comparison (t, t+5, t+10 after shocks):
  - At t+5 and t+10, youngest heads save about 5 percentage points more; oldest save about 6.5 percentage points more at t+5 and 5.5 percentage points more at t+10.
  - Profiles bottom out around 2.5 percentage points for household heads around age thirty or early thirties.
  - t+10 cross-section yields asymmetric U-shaped pattern: households in early forties save 3-4 percentage points more than before.
- Interpretation: calibrated buffer-stock/life-cycle model plus 1997 pension reform can account for a sizable portion of observed increase in savings and the U-shaped age-saving profile concentrated among the young and old.
- Empirical context: Chamon and Prasad (2010) find housing motives account for about 6 percentage points for 25-29 year olds (non-owners) and health expenditures account for about 6 percentage points for 55-59 year olds; calibration suggests shifts in earnings uncertainty and pension reforms also played important roles.

### Robustness checks and parameter sensitivity
- Key sensitivity dimensions explored:
  - Larger decline in pension replacement ratio (to 50 percent) increases response for older household heads, little immediate impact for youngest.
  - Vary risk aversion γ and discount factor β:
    - Lower risk aversion reduces increase in savings for the young in response to higher transitory uncertainty.
    - Simulated initial rise in saving rates across age groups broadly comparable across reasonable parameter ranges, though responses for older heads tend to be larger with higher risk aversion.
- Calibration of γ using pre-1997 UHS cross-sections:
  - Construct synthetic cohorts from UHS 1990-1997, regress log income and log consumption on age and cohort dummies and family size controls to estimate age profile of saving rates pre-increase in uncertainty.
  - Calibrating γ to match mean saving rate in seven age groups yields γ = 8.1 when β = 0.97 and γ = 7.5 when β = 0.99.
  - Implication: high γ highlights difficulty of explaining high pre-1997 saving rates solely with buffer-stock model; additional motives (bequest, housing, lumpy health expenditures) could reduce required γ.
  - Even with these calibrated high γ, model implies increase in average saving rate of more than 2.5 percentage points given baseline shocks.
- Conservativeness: when calibrating precautionary effect, assume half of empirically identified transitory variance is measurement error; this yields conservative lower-bound estimates of precautionary saving response.

*Source: _wp10289 - INTRODUCTION*

### 22.3 percent, respectively. This age-saving profile captures the estimated saving behavior for a household with a

### _wp10289 - 22.3 percent, respectively. This age-saving profile captures the estimated saving behavior for a household with a

### Age-saving profile and data context
- The age-saving profile captures the estimated saving behavior for a household with a head aged 25 years in 1997 as he or she ages (not the 1997 cross-section of savings with respect to age).
- This age-saving profile is not U-shaped as it is based on 1990-1997 data; the U-shaped profile does not appear in the data until the 2000s.
- The income growth path used in simulations includes the effects of trend income growth as well as age effects on income; controlling for trend growth, income eventually declines with age.

### Effects of increased income uncertainty and pension reform
- The variance of shocks to household income increased over time, with the increase mainly accounted for by a rise in the variance of transitory shocks to income.
- A rise in transitory uncertainty helps explain rising saving rates among households with younger household heads (larger buffer stocks required).
- Pension reforms led to a reduction in pension replacement income relative to average wages for workers retiring after 1997, contributing to higher saving rates among households with older household heads approaching retirement.
- Quantitative statement from calibrations:
  - The riskier environment (higher transitory variance and pension reform) is capable of generating an initial increase of about four and a half percentage points in the average household saving rate.
  - Saving rates remain significantly higher after that initial adjustment, with an average increase across age groups of four and a half percentage points after 5 years.
  - The increase is concentrated among households with household heads at the two ends of the age distribution in the sample, helping explain the U-shaped age-profile of savings observed in urban China since the late 1990s.
- The authors state their calibration is conservative in assumptions (e.g., how the pension reform affected the replacement ratio) and that focusing only on higher transitory variance of earnings and the 1997 pension reform, the model explains about half of the observed increase in savings among urban Chinese households.

### Expected income growth scenarios and simulated age-savings profiles
- Simulations plot age-savings profiles for a household with a household head starting at age 25 under different expected income growth scenarios; other parameters follow the baseline under higher uncertainty and lower pension replacement ratio.
- Baseline, lower-immediate, and lower-gradual growth paths:
  - Solid line: baseline expected growth path.
  - Dotted line: growth path that is 1 percentage point lower than in the baseline from the present onwards.
  - Dashed line: growth path where expected income growth flattens out at a rate of 0.1 percentage point every year relative to its value in the baseline path, but never declines below the lowest value in that path (1.26 percent).
- Simulation findings:
  - Saving rates would be higher for households with young heads under both lower income growth paths.
  - The increase in saving is larger when income slows down gradually (dotted line) than when the decline takes place immediately (dashed line).
  - The lower growth path substantially reduces retirement income in the simulation, strengthening retirement saving motives and causing households to save more even in early life-cycle stages.
  - The age-savings profile is flatter when the slowdown is gradual than when it takes place immediately.
  - The age-savings profile is steepest under the baseline expected growth path, where postponement of retirement savings is strongest; saving rates are actually higher in the working years close to retirement age under the baseline than under both alternative scenarios.
- Implication: prospects of an eventual slowdown in Chinese growth could further increase household saving rates; some savings among the very young may already reflect these prospects.

### Model calibration, robustness, and limitations
- The authors use a standard buffer-stock life-cycle model of consumption calibrated with reasonable parameter values.
- Despite difficulties in fully explaining actual Chinese household saving behavior with the standard model, the estimates quantify how far the model goes, under reasonable parameter values, toward explaining the observed increase in saving rates.
- The authors note multiple assumptions in calibration (for example, the pension reform effect on replacement ratio) and emphasize systematically conservative assumptions that tend to understate increases in risks to income growth.

### Policy implications and closing observations
- The documented rise in transitory income variance and the pension replacement cut from the 1997 reform together can account for roughly half of the observed rise in urban household saving rates in the sample.
- These findings may be helpful in designing policies to rebalance growth in China by boosting private consumption and reducing reliance on exports and investment to drive growth.

*Source: Excerpt from IMF working paper content provided in the supplied PDF text.*

### Section III. “All” refers to

### _wp10289 - Section III. “All” refers to

### Sample definitions and variable construction
- “All” refers to the full sample.
- SOCE subsample: restricted to those where head was an SOCE employee when the household first appeared in the sample.
- Young and old households: defined as those whose household heads were below and above 40 years of age, respectively.
- “Less educated”: group of households whose head does not have a high school degree.
- Last three columns referenced in the source correspond to:
  - (i) family income excluding subsidies and transfers;
  - (ii) family income excluding subsidies, transfers and other income;
  - (iii) family income excluding data from year 1989.
- “Other income” includes income from leased land (only for 1989), rent from non-land assets, rent from lodgers or boarders, and other unspecified sources of income.

### Table 3 — Robustness checks: simulated rise in savings using various preference parameters
Notes preceding the table:
- The first row reports the change in saving behavior when the transitory variance of income increases from 0.04 to 0.08 and the pension replacement ratio declines from 0.75 to 0.60.
- The second row evaluates changes under a scenario proposed by Sin (2005) in which the drop in replacement rate post-reform declines to 0.50.
- Subsequent rows experiment with different values of two key parameters—the risk aversion parameter and the discount factor.
- The last two rows calibrate preference parameters of risk aversion (holding the discount factor fixed at 0.99 and 0.97, respectively) so as to match the average age-profile of savings in 1990-1997 using an identity weighting matrix.

Reported jump in saving rates at time of shock for different ages of household head (columns correspond to ages 25, 30, 35, 40, 45, 50, 55):

- Baseline assumptions (Risk Aversion = 4.5; Discount Factor = 0.97):
  - Age 25: 5.4%
  - Age 30: 5.3%
  - Age 35: 5.4%
  - Age 40: 5.1%
  - Age 45: 5.8%
  - Age 50: 6.6%
  - Age 55: 7.8%

- Larger drop in replacement rate post-reform (Risk Aversion = 4.5; Discount Factor = 0.97; replacement declines to 0.50 per Sin (2005)):
  - Age 25: 5.5%
  - Age 30: 5.6%
  - Age 35: 6.4%
  - Age 40: 7.2%
  - Age 45: 9.0%
  - Age 50: 10.6%
  - Age 55: 12.8%

- Baseline drop in replacement rate with alternative preference parameters:
  - Risk Aversion = 3.5; Discount Factor = 0.97:
    - Age 25: 5.0%
    - Age 30: 4.9%
    - Age 35: 5.1%
    - Age 40: 5.3%
    - Age 45: 6.8%
    - Age 50: 7.8%
    - Age 55: 8.9%
  - Risk Aversion = 2.0; Discount Factor = 0.97:
    - Age 25: 4.4%
    - Age 30: 4.4%
    - Age 35: 4.4%
    - Age 40: 4.2%
    - Age 45: 5.0%
    - Age 50: 6.8%
    - Age 55: 9.5%

- Alternative discount factor 0.95:
  - Risk Aversion = 4.5; Discount Factor = 0.95:
    - Age 25: 5.3%
    - Age 30: 5.2%
    - Age 35: 5.3%
    - Age 40: 5.5%
    - Age 45: 6.3%
    - Age 50: 7.0%
    - Age 55: 8.2%
  - Risk Aversion = 3.5; Discount Factor = 0.95:
    - Age 25: 4.9%
    - Age 30: 4.9%
    - Age 35: 5.0%
    - Age 40: 5.0%
    - Age 45: 6.4%
    - Age 50: 8.0%
    - Age 55: 9.3%
  - Risk Aversion = 2.0; Discount Factor = 0.95:
    - Age 25: 4.2%
    - Age 30: 4.3%
    - Age 35: 4.3%
    - Age 40: 4.1%
    - Age 45: 4.6%
    - Age 50: 5.6%
    - Age 55: 8.5%

- Alternative discount factor 0.98:
  - Risk Aversion = 4.5; Discount Factor = 0.98:
    - Age 25: 5.5%
    - Age 30: 5.2%
    - Age 35: 5.2%
    - Age 40: 4.7%
    - Age 45: 5.6%
    - Age 50: 6.4%
    - Age 55: 7.6%
  - Risk Aversion = 3.5; Discount Factor = 0.98:
    - Age 25: 5.1%
    - Age 30: 5.0%
    - Age 35: 5.2%
    - Age 40: 5.6%
    - Age 45: 6.9%
    - Age 50: 7.5%
    - Age 55: 8.7%
  - Risk Aversion = 2.0; Discount Factor = 0.98:
    - Age 25: 4.5%
    - Age 30: 4.4%
    - Age 35: 4.4%
    - Age 40: 4.3%
    - Age 45: 5.3%
    - Age 50: 7.5%
    - Age 55: 9.9%

- Preferences calibrated based on mean saving rate of 5-year age groups:
  - Risk Aversion = 7.5; Discount Factor = 0.99:
    - Age 25: 2.2%
    - Age 30: 1.8%
    - Age 35: 2.2%
    - Age 40: 2.3%
    - Age 45: 3.3%
    - Age 50: 4.1%
    - Age 55: 5.1%
  - Risk Aversion = 8.1; Discount Factor = 0.97:
    - Age 25: 2.1%
    - Age 30: 1.6%
    - Age 35: 2.1%
    - Age 40: 2.1%
    - Age 45: 3.1%
    - Age 50: 3.9%
    - Age 55: 5.0%

### Figures and empirical observations (captions and key points)
- Figure 1. Household Saving Rates
  - Source: National Bureau of Statistics, Flow of Funds data and Urban and Rural Household Survey.
  - Observation: Saving rate from national accounts is significantly higher than that from the household surveys. The discrepancy is common and can be due to differences in definitions of income and consumption, methodology and sample coverage.

- Figure 2. Gross Saving Rates by Sector
  - Source: National Bureau of Statistics (Flow of Funds data).

- Figure 3. Urban Household Saving Rates by Age of Household Head
  - Notes: Based on a 10 province/municipality subsample of the National Bureau of Statistics Urban Household Surveys. Saving rates are smoothed by a moving average with 4 neighboring age averages.

- Figure 4. Labor Market Turnover: Annualized Transition Probabilities
  - Note: Transition rates annualized by taking the Markov transition matrix between two surveys to the power of 1/n, where n is the number of years between the survey pair.

- Figures 5A & 5B. Estimated age profile of saving rates before and after rise in variance of transitory income shock; jump in saving rate following rise in variance
  - Notes: Dashed line corresponds to variance 0.04; other lines correspond to variance 0.08 when the change occurs at the household head’s respective age.
  - Figure 5B: Each line corresponds to the jump in the saving rate after the increase in variance for households with household head at that age.

- Figures 6A & 6B. Estimated age profile of saving rates before and after pension reform; jump in saving rates following pension reform
  - Notes: Dashed line corresponds to η=0.75; other lines indicate η=0.60 if change occurs at the household head’s respective age.
  - Figure 6B: Each line corresponds to the jump in the saving rate after pension reform for a household with household head at that age.

- Figures 7A & 7B. Estimated age profile and jump in saving rates after combined rise in transitory variance and pension reform
  - Notes: Dashed line corresponds to variance 0.04 and η=0.75; other lines correspond to variance 0.08 and η=0.60 if change occurs at the household head’s respective age. Changes in saving rates are smoothed by a moving average with 3 neighboring age averages.

- Figure 8. Projected cross-sectional change in saving rates after rise in the transitory variance of income and pension reform
  - Notes: Lines correspond to the cross-sectional age profile of savings implied by the profiles in Figure 7.

- Figure 9. Estimated age profile of saving rates for a household with 25 year old head, under different expected income growth paths
  - Notes: Traces expected saving rates for a household with a 25 year old head under baseline parameter values (solid line), under a growth path where expected income growth is 1 percentage points lower every year than in baseline (dotted line), and under a scenario where expected income growth is flattening at a rate of 0.1% every year (but never declines below the lowest value in the baseline path) (dashed line).

*Source: _wp10289 - Section III. “All” refers to (PDF chapter/section).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10289.pdf_