## _wp1382

## Source details

**Canonical URL:** [_wp1382](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1382.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1382.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1382.pdf.json)

---

### I. Introduction — demographic shift and macroeconomic implications
- China’s population is expected to age rapidly in the next 50 years; by 2040 the large hump peaking at the age of 60 in 2030 will have shifted well into retirement age, with the working population largely replaced by the smaller population born after the one child policy.
- Demography will significantly affect savings and consumption, the external balance, growth, the labor market, industrial structure, the pension system, and public finances.
- Chinese authorities have been cautious in expanding the pension system; they argue that on demographic grounds China’s savings should be high now and only fall gradually over the next couple of decades.
- This paper reassesses national savings behavior with greater prominence given to changing demography using the permanent income hypothesis (households maximize consumption over their lifetime).
- Panel-data evidence: the permanent income approach holds up well empirically and may be superior to static dependency ratios typically used in the literature.
- Simulation finding: savings should have peaked in 2011 based on demographics alone; accordingly, China’s external balance should also have peaked around that period and should now be gradually declining, turning negative in the long run as China ages.
- Multilateral adjustment (China, United States, Europe, Japan): China’s demographics should have induced a negative current account in the 2000s (because it was relatively young) and a positive one in the 2010s (as the labor force saves in anticipation of aging); the opposite pattern holds for the United States and Western Europe.
- Policy distortions affecting deviations from the simulated path:
  - Expansionary macroeconomic policies in the United States allowed consumption beyond demographic consistency.
  - Reflow of capital from China and other emerging markets through reserve accumulation fueled U.S. monetary expansion (Choi and Lee, 2010).
  - On the Chinese side, an underdeveloped social safety net may have contributed to excessive saving.
- Domestic outlook:
  - China’s working population will stay relatively stable for another decade, then start to decline gradually.
  - The increase in the number of retirees over the long term poses the challenge of managing savings to preserve real purchasing power per yuan saved now.
  - Savings allocation choice: domestic investment (raises domestic returns via investment efficiency) versus investment abroad (returns determined by optimal portfolio management of foreign assets).
- Cumulative divergence (stock) effects: if actual savings exceeded the simulated path historically, future savings may underperform to compensate, conditional on policy choices enabling households to optimize consumption.

### II. Literature review — prior findings on demography and savings
- Lifecycle model studies finding significant relationships between saving rates and age structure: Modigliani and Cao (2004), Horioka and Wan (2007), Horioka (2010), Song and Yang (2010), Fan and Zhu (2012).
- Selected prior projections and findings:
  - Kuijs (2006): expects China’s savings to remain high until around 2020-25, with a substantial net negative effect on saving after 2025.
  - Chamon and Prasad (2005): demographic factors imply higher household saving over the next decade; effects become less ambiguous after around 2025.
  - Horioka and Terada-Hagiwara (2012): project China’s savings rate to continue increasing until 2030; share of retirees crossing 14 percent only toward end of that period.
  - IMF (2008): predicts persistently positive impact on China’s current account long-term; negative impacts for the United States and Japan over the next 20-30 years.
  - Haldane (2010) and Bank of England (2011): China’s saving rate would start falling in the late 2020s and could turn into a deficit after 2045; China’s current account could reach as high as 2 percent of global GDP over this period.
- Methodological note: much prior empirical work relies on reduced form regressions and static dependency ratios. This paper incorporates the entire age distribution and future time-path of demographics and other variables.

### III. Conceptual framework — permanent income hypothesis and aggregation
- Households maximize lifetime consumption; savings are determined by demographic structure and the time horizon over which households maximize consumption.
- Two-period and overlapping-generations derivation yields first-order conditions:
  - dF/ds = (1+ρ)
  - dF/dG2 = (1+ρ) dS/dG2
- Intuition:
  - An expected increase in future working population (G2) reduces current savings.
  - An expected decline in future working population increases current savings.
- Role of liquidity:
  - Asset prices (P_S) are functions of liquidity L with P’>0.
  - Y1P = S P_S + E P_E or Y1 = S(P_S/P) + E(P_E/P).
  - Optimal savings S* solves F(S*) = (1+ρ) and S* = S(P_S/P). If liquidity is excessive (P_S>P), then S < S*, meaning excessive liquidity reduces optimal savings.

### IV. Simulation framework and results
- Population groups in simulations: G0 (below working age, dissaving), G1 (working age, income generating and saving), G2 (retired, zero saving).
- Simulation I (single-country example):
  - Normalized total income of 1 in each period.
  - Consumption assignment: 0.4 to G1, and 0.3 each to G0 and G2.
  - Savings defined as difference between constructed consumption path and income; net savings measured as S-I (net external current account balance).
  - Trade-off: allocating savings to domestic investment raises future GDP versus investing abroad which yields windfall income later but does not raise domestic GDP.
- Empirical comparison (China):
  - Actual net savings broadly follow simulated path but are more elevated and volatile due to erratic actual investment.
  - Reasons for actual > simulated in 1990s and 2000s: labor dividend, strong global demand, China’s integration to world trade, underemployment and large size of new entrants suppressed wage growth and wage share of GDP declined.
- Simulation II (multi-country context):
  - Trading partners’ demography pulls S-I trajectories; global S-I must sum to zero at each point in time.
  - Key assumptions: China grows at 5 percent per annum; other countries grow by 2 percent per annum over the next 40 years; investment efficiency constant.
  - Implications:
    - Short-run deviations must be balanced later because net savings sum to zero across countries over the optimization horizon.
    - Prolonged deviation of S-I from baseline requires abrupt later consumption adjustment.
    - If returns on S-I or domestic investment fall, S-I tends to rise as households save more.
  - Multilateral consistency results:
    - Extending optimization horizon from 1990-2030 to 1970-2050 shifts intersection point from 2009 to 2005.
    - After multilateral consistency adjustment, China’s S-I peaks at around 2026 and then starts to decline — delaying China’s S-I peak by about 15 years because aging in other countries sets in earlier.
    - Robust to changes in growth rates (shape and slight peak shifts only).
  - Time-horizon sensitivity:
    - Longer horizons (1970-2050) produce much larger simulated savings during 2010-2030; shorter horizons (1980-2030) produce smaller peaks.
  - Quantitative illustrations in the source:
    - Figure 12 simulation (1970-2050) shows China’s S-I peaking close to 10 percent of GDP in 2011.
    - Figure 8 simulation (shorter horizon) peaks at 4 percent.
- Interpretation of divergences:
  - U.S. and Europe ran large deficits in the 2000s despite simulations indicating surpluses; consumption was leveraged beyond optimal paths, driven by asset-price–induced wealth illusions.
  - China ran large surpluses when it should have run deficits during most of the 2000s according to the demographic simulation.
- Stock (cumulative divergence) effects:
  - Large past divergences will influence future household saving behavior and public fiscal efforts (rebuilding balance sheets).
  - If the stock problem is large, fiscal consolidation alone may not suffice and could lead to weak demand for a prolonged period.
  - In China, the reversed stock problem could exert downward pressure on S-I beyond demographic dictates as younger cohorts show higher propensity to consume.

### V. Empirical tests — testing the permanent income hypothesis with demography
- Empirical strategy: savers (G1) compare expected future shares of retirees (G2) and young (G0) relative to current working population (G1); expected rise in G2/G1 raises savings, expected rise in G0/G1 lowers savings.
- Forward-looking demographic approach: savings decision for the population in 2010 is influenced negatively by the ratio G0/G1 and positively by the ratio G2/G1, where G0 and G2 refer to the future.
- Empirical application: dynamic panel models for private saving estimated with GMM for 57 advanced and emerging market economies over 1960-2011.
- Model specification:
  - Sit = α Si(t–1) + βXit + νi + ηt + uit
  - Sit is private saving-to-GDP; Xit includes public savings, per capita output growth, real interest rate, growth in stock market capitalization, and demographic variables.
- Key regression results (selected coefficients as reported):
  - Lagged private savings coefficient (Si,t–1): 0.370***, 0.351***, 0.395***, 0.351***, 0.340***, 0.310***, 0.422***.
  - Public savings (Pub S/GDP)t: -0.514***, -0.487***, -0.520***, -0.483***, -0.545***, -0.527***, -0.515***.
  - Growth in real GDP per capita t: 0.139***, 0.127**, 0.147***, 0.128**, 0.137**, 0.135**, 0.159***.
  - Real interest rate t: -0.056*, -0.075**, -0.067**, -0.072**, -0.064*, -0.051, -0.056**.
  - Demographic and other reported coefficients (selected):
    - (G0/G1)t: -17.443*** and -18.045*.
    - (G2/G1)t: 22.762* and 3.895.
    - Youth dependency ratio t: -13.673** and 0.382.
    - Prime savers ratio t: 20.350*, 20.093, and -12.15 (28.72) in different columns.
- Diagnostics and sample:
  - Observations: 1,070 in main table columns.
  - Number of countries: 57 in main table.
  - m2 test (p-value) examples: 0.630, 0.551, 0.589, 0.565, 0.492, 0.497, 0.696.
  - Hansen test (p-value) reported as 1.000 in table columns.
  - Estimation in first-differences; dependent variable is private savings-to-GDP.
  - ***, **, * denote significance at 1 percent, 5 percent, and 10 percent levels respectively.
- Empirical magnitudes:
  - Fiscal consolidation effect: a 1 percent of GDP increase in public savings would lead over time to a 0.75 percent of GDP decrease in private savings (estimated as coefficient/(1 − coefficient on lagged saving)); effect predominantly reflects emerging markets.
  - Output growth: a sustained 1 percentage point increase in per capita output growth yields almost a 0.2 percent of GDP increase in the private saving rate.
  - Real interest rate: increases in the rate of return decrease savings in these estimates (income effect dominates); effect predominantly in emerging markets.
  - Forward-looking demographics: a 1 percentage point increase in G0/G1 (G2/G1) reduces (increases) private savings by 0.25 percentage points of GDP.
  - Dynamic demographic variables appear superior to static measures; youth-dependency ratio is knocked out when G0/G1 is included.
  - Liquidity/asset prices: increases in stock market capitalization as percent of GDP are associated with reductions in saving (instrumented by private sector credit growth); excess liquidity can temporarily confound demographic and growth effects.

### China-specific implications
- Growth and convergence:
  - Political stability and sound governance are necessary but not sufficient for convergence; rapid expansion of global trade matched by large domestic investment has been an additional required impetus — a pattern China has followed.
  - Expected demographic changes will limit labor supply and start raising unit labor costs; investment likely to fall as returns to capital fall relative to scarcer labor and declining savings make capital more expensive.
  - Offsets: one-child-policy relaxation, increased immigration, higher participation rates, and total factor productivity gains from innovation, education, and urbanization.
- Policy dilemma and S-I path:
  - Divergence of actual S-I from demography-implied S-I may reflect changes in fundamentals or policy distortions.
  - Multilaterally adjusted S-I can differ from non-adjusted path; after adjustment China’s S-I peaks later (around 2026) than in single-country simulations.
  - One scenario: China’s savings may fall even below the adjusted path as it rebalances from excessive past savings.
  - Improved investment efficiency and strengthened social services could reduce precautionary savings and facilitate rebalancing.
- Pension system (urban employees’ basic pension insurance, end-2011 snapshot and scenarios):
  - Current participants: 285 million (215 million contributors; 70 million recipients/retirees).
  - Pension fund size as of end 2011: RMB 2 trillion.
  - Benefit per retiree in 2011: RMB 18,700 — about three times the contribution per employee.
  - Contributors are about 3 times retirees and fiscal contributions make up around a fifth of total payment.
  - Pension fund growth: from about 1 percent of GDP in 2001 to 4.2 percent of GDP in 2011.
  - Rate of return (average last 10 years): 3.5 percent per year.
  - Current coverage rate: 40 percent of retirees receiving pension; only 26 percent of employees contributing.
  - Projections assumptions:
    - 70 percent of all retirees and employees become participants by 2030 and 2040, respectively.
    - Baseline nominal GDP growth: average 8 percent over next 40 years.
  - Projection implications:
    - Under baseline where contribution and entitlement grow by the same amount, the pension fund could be depleted (Figure 17 scenario).
    - Alternative scenarios illustrate required contribution hikes if entitlement growth continues: e.g., Figure 18 scenario: Fiscal contribution declines to zero; Member contribution grows by 10.2 percent; Entitlement grows by 8 percent. Figure 19 scenario: Fiscal contribution grows by 8 percent; Member contribution grows by 9.5 percent; Entitlement grows by 8 percent.
  - Policy implication: expansion of pension coverage should be carefully calibrated with inflows; overly generous payouts could threaten long-term viability.
- Labor market implications:
  - Near-term: end of labor dividend from large labor influx rather than immediate shrinking working population.
  - Expect rising unit labor costs, frictional and structural unemployment, and skill mismatches between college graduates and firm needs.
  - Reduced pressure to create jobs for ever-increasing cohorts, allowing lower growth, but raising need to reduce skill mismatches and cater for a dependent population (dependent population increases noticeably after 2024).
  - Greater savings will be required to cater for dependents, depending partly on preserving value of past savings.
- Public finances:
  - With no change in retirement age, workers-to-retirees ratio will drop from around 5:1 today to 2:1 by 2030.
  - Assuming no change in tax structure, tax burden on each worker would have to increase by around two and a half times to maintain current level of tax income.
  - Aging likely reduces tax buoyancy, particularly personal income and payroll taxes; indirect taxes may also be affected as elderly spend relatively more on services.
  - Policy responses: modernize tax systems, reduce tax preferences, increase dividends from state-owned enterprises, enhance natural resource and energy taxation, increase taxes on services, and broaden revenue sources.

### Conclusions and policy implications (summary)
- Demography is an important determinant of savings despite divergence of actual net savings from demography-based predicted path.
- Key policy implications:
  - Demography implies a continued external current account surplus for China for a prolonged period as it ages more slowly than main competitors; announced policies to promote consumption and address income inequality aim at intertemporal rebalancing.
  - Rebalancing from past over-saving should help global alignment with multilaterally consistent S-I path; ensuring adequate returns on China’s stock of savings is critical as low returns would add pressure for saving.
  - China needs a new growth model as it can no longer rely on a large labor dividend and savings; efficiency gains and innovation will be essential, along with better labor training to reduce skill mismatches.
  - Expansion of the pension system requires caution and careful calibration with inflows; otherwise pension funds could be run down in the next two decades as retiree numbers rise.
  - Tax revenues could come under pressure as working population share falls; countervailing changes in tax policy and administration are needed to prevent erosion of the tax base and to extend it into resources, energy and services.

*Source: _wp1382 - References (IMF working paper excerpt provided in the source PDF).*

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

### _wp1382 - References .............................................................................................................

### I. Introduction — demographic shift and macroeconomic implications
- China’s population is expected to age rapidly in the next 50 years; by 2040 the large hump peaking at the age of 60 in 2030 will have shifted well into retirement age, with the working population largely replaced by the smaller population born after the one child policy.
- Demography will significantly affect savings and consumption, the external balance, growth, the labor market, industrial structure, the pension system, and public finances.
- Chinese authorities have been cautious in expanding the pension system; they argue that on demographic grounds China’s savings should be high now and only fall gradually over the next couple of decades.
- This paper reassesses national savings behavior with greater prominence given to changing demography using the permanent income hypothesis (households maximize consumption over their lifetime).
- Panel-data evidence: the permanent income approach holds up well empirically and may be superior to static dependency ratios typically used in the literature.
- Simulation finding: savings should have peaked in 2011 based on demographics alone; accordingly, China’s external balance should also have peaked around that period and should now be gradually declining, turning negative in the long run as China ages.
- Multilateral adjustment (China, United States, Europe, Japan): China’s demographics should have induced a negative current account in the 2000s (because it was relatively young) and a positive one in the 2010s (as the labor force saves in anticipation of aging); the opposite pattern holds for the United States and Western Europe.
- Policy distortions affecting deviations from the simulated path:
  - Expansionary macroeconomic policies in the United States allowed consumption beyond demographic consistency.
  - Reflow of capital from China and other emerging markets through reserve accumulation fueled U.S. monetary expansion (Choi and Lee, 2010).
  - On the Chinese side, an underdeveloped social safety net may have contributed to excessive saving.
- Domestic outlook:
  - China’s working population will stay relatively stable for another decade, then start to decline gradually.
  - The increase in the number of retirees over the long term poses the challenge of managing savings to preserve real purchasing power per yuan saved now.
  - Savings allocation choice: domestic investment (raises domestic returns via investment efficiency) versus investment abroad (returns determined by optimal portfolio management of foreign assets). The domestic investment issue addressed in a separate paper (Lee et al. (2013b) forthcoming).
- Cumulative divergence (stock) effects: if actual savings exceeded the simulated path historically, future savings may underperform to compensate, conditional on policy choices enabling households to optimize consumption.

### II. Literature review — prior findings on demography and savings
- Lifecycle model studies finding significant relationships between saving rates and age structure: Modigliani and Cao (2004), Horioka and Wan (2007), Horioka (2010), Song and Yang (2010), Fan and Zhu (2012).
- Kuijs (2006): expects China’s savings to remain high until around 2020-25, with a substantial net negative effect on saving after 2025 when the old-age dependency ratio continues to rise and the young-age dependency ratio stabilizes.
- Chamon and Prasad (2005): demographic factors imply higher household saving over the next decade; effects could become less ambiguous after around 2025 as the old-age population rises and working-age population becomes dominated by low-saving younger cohorts.
- Horioka and Terada-Hagiwara (2012): project China’s savings rate to continue increasing until 2030; share of retirees crossing 14 percent (definition of an “aged society”) only toward end of that period.
- IMF (2008): multi-country panel regression using demographic variables expressed relative to trading partners; predicts demographics will have a persistently positive impact on China’s current account long-term, negative impacts for the United States and Japan over the next 20-30 years, and negative for Germany after around 2025.
- Haldane (2010) and Bank of England (2011): based on changes in fraction of prime savers (40-59), medium-term pressures on global imbalances could triple over the next twenty years; China’s saving rate would start falling in the late 2020s and could turn into a deficit after 2045; China’s current account could reach as high as 2 percent of global GDP over this period.
- Methodological note: much prior empirical work relies on reduced form regressions and static dependency ratios. This paper’s approach is closer to Curtis et al. (2011), incorporating the entire age distribution and future time-path of demographics and other variables.

### III. Conceptual framework — permanent income hypothesis and aggregation
- Framework: households maximize lifetime consumption; savings determined by demographic structure and the time horizon over which households maximize consumption.
- Two-period individual maximization setup (Period 1: income and saving; Period 2: consumption from savings). First-order condition: f’=(1+ρ) (where ρ is the discount factor).
- Multi-period/overlapping generations extension: aggregate optimization across generations G1 and G2. Key conditions:
  - Max Z=Y1 - S(G2) + Y2(G2)/(1+ρ) subject to Y1=S+E1 and Y2=F(G2, S).
  - First-order conditions: (1) dZ/dS=0 and (2) dZ/dG2=0 leading to dF/ds=(1+ρ) and dF/dG2=(1+ρ) dS/dG2.
- Intuition: an expected increase in future working population (G2) reduces current savings; an expected decline in future working population increases current savings.
- Role of liquidity:
  - Asset prices (P_S) are functions of liquidity L with P’>0.
  - Y1P = S P_S + E P_E or Y1 = S(P_S/P) + E(P_E/P).
  - Optimal savings S* solves F(S*)=(1+ρ) and S* = S(P_S/P). If liquidity is excessive (P_S>P), then S < S*, meaning excessive liquidity reduces optimal savings.

### IV. Simulation framework and results
- Population divided into three age groups for simulation: G0 (below working age, dissaving), G1 (working age, income generating and saving), G2 (retired, zero saving).
- Simulation I (single-country example):
  - Normalized total income of 1 in each period.
  - Consumption assignment: 0.4 to G1, and 0.3 each to G0 and G2.
  - Income generated by G1; income curve scaled so total income over the maximization period equals total consumption.
  - Savings defined as difference between constructed consumption path and income; net savings measured as S-I (net external current account balance).
  - Trade-off illustrated: allocating savings to domestic investment raises future GDP (and future consumption) versus investing abroad which yields windfall income later but does not raise domestic GDP.
- Empirical comparison (China):
  - Actual net savings broadly follow simulated path but are more elevated and volatile (reflecting erratic actual investment).
  - Reasons for actual > simulated in 1990s and 2000s: labor dividend, strong global demand, China’s integration to world trade, underemployment and large size of new entrants suppressed wage growth and wage share of GDP declined, allowing corporate income to be saved and reinvested.
- Simulation II (multi-country context):
  - Trading partners’ demography pulls S-I trajectories; global S-I must sum to zero at each point in time.
  - Assumptions for multi-country simulation: China grows at 5 percent per annum; other countries grow by 2 percent per annum over the next 40 years; investment efficiency remains constant; remaining savings are S-I.
  - Implications:
    - Short-run deviations from baseline due to various factors must be balanced later because net savings must equal zero across countries over the optimization horizon.
    - Prolonged deviation of S-I from baseline requires abrupt later consumption adjustment.
    - If returns on S-I or domestic investment fall, S-I tends to rise as households save more to maintain later consumption; allocation shifts toward investments with higher returns.
  - Simulated S-I for main economies (China, Japan, US, Western Europe) and multilateral consistency:
    - The horizon choice affects shape but not fundamental conclusions; extending horizon from 1990-2030 to 1970-2050 shifts intersection point from 2009 to 2005.
    - After multilateral consistency adjustment, China’s S-I peaks at around 2026 and then starts to decline — delaying China’s S-I peak by about 15 years because aging in other countries sets in earlier, necessitating China to save longer.
    - Robustness: result broadly robust to changes in growth rates (shape and slight peak shifts only).
  - Time-horizon sensitivity:
    - Longer optimization horizons that capture large demographic changes at both ends (1970-2050) produce much larger simulated savings during 2010-2030; shorter horizons (1980-2030) produce smaller peaks.
  - Quantitative illustrations:
    - Figure 12 simulation (1970-2050) shows China’s S-I peaking close to 10 percent of GDP in 2011.
    - Figure 8 simulation (shorter horizon) peaks at 4 percent.
  - Interpretation of divergences:
    - The global economy is not driven by permanent-income/demography alone; shorter optimization horizons and policy distortions create deviations.
    - U.S. and Europe ran large deficits in the 2000s despite simulations indicating they should have had S-I surpluses; they leveraged consumption beyond optimal paths, driven by asset-price–induced wealth illusions.
    - China ran large surpluses when it should have run deficits during most of the 2000s according to the demographic simulation.
  - Stock (cumulative divergence) effects:
    - Large past divergences will influence future household saving behavior and public fiscal efforts (rebuilding balance sheets).
    - If the stock problem is large, fiscal consolidation alone may not suffice and could lead to weak demand for a prolonged period.
    - In China, the reversed stock problem could exert downward pressure on S-I beyond demographic dictates as younger cohorts show higher propensity to consume.

### V. Empirical tests — testing the permanent income hypothesis with demography
- Definitions reiterated: G0 (young), G1 (working-age), G2 (retired).
- Empirical strategy (described qualitatively in provided text):
  - Savers (G1) in the current year pay attention to what the demographic structure will look like when they reach retirement relative to today.
  - If the expected share of retirees (G2) relative to today’s working population (G1) is expected to rise, this should push up savings.
  - If the expected share of the young (G0) relative to current working population (G1) is expected to rise, this should push down savings as the young will progressively enter the labor force and support retirees.
- Visual evidence referenced: Figure 15 shows demography at each indicated year with interpretation that savers compare future retirement demographics to current demographics when making saving decisions.

*Source: Excerpt from _wp1382 - References (IMF working paper content provided in the source PDF).*

### 2030. In other words, the decision maker has to look at A and C relative to the working

### _wp1382 - 2030. In other words, the decision maker has to look at A and C relative to the working

### Forward-looking demographic approach
- The savings decision for the population in 2010 is influenced negatively by the ratio G0/G1 and positively by the ratio G2/G1, where G0 and G2 refer to the future.
- The approach is more forward-looking than typical OLG-based papers that focus on the current-period share of prime savers, the young and the old, by emphasizing dynamics and expectations of future demography.

### Empirical application: model and data
- Estimated dynamic panel models for private saving using GMM techniques for 57 advanced and emerging market economies over the period 1960-2011.
- Model specification:
  - Sit = α Si(t–1) + βXit + νi + ηt + uit  (i = 1, . . . ,N; t = 2, . . . ,T)
  - Sit is the private saving-to-GDP ratio; Xit includes explanatory variables; νi is country-specific effect; ηt is time-specific effect.
- Right-hand-side variables:
  - Past saving rate, Si(t – 1) — captures persistence.
  - Economic determinants Xit with expected signs:
    - public savings (- if Ricardian equivalence holds)
    - per capita output growth (+)
    - real interest rate (+ or -)
    - growth in stock market capitalization as proxy for liquidity (- if liquidity creates a wealth illusion)
    - alternative demographic variables: (G0/G1) (-), (G2/G1) (+), prime savers ratio (+), youth dependency ratio (-).

### Key regression results (high-level)
- Lagged private savings coefficient (Si,t–1): reported estimates include 0.370***, 0.351***, 0.395***, 0.351***, 0.340***, 0.310***, 0.422*** (standard errors in parentheses).
- Public savings (Pub S/GDP)t coefficients reported as -0.514***, -0.487***, -0.520***, -0.483***, -0.545***, -0.527***, -0.515***.
- Growth in real GDP per capita t coefficients reported as 0.139***, 0.127**, 0.147***, 0.128**, 0.137**, 0.135**, 0.159***.
- Real interest rate t coefficients reported as -0.056*, -0.075**, -0.067**, -0.072**, -0.064*, -0.051, -0.056**.
- Demographic and other reported coefficients (selected):
  - (G0/G1)t: -17.443*** and -18.045* (with standard errors shown)
  - (G2/G1)t: 22.762* and 3.895
  - Youth dependency ratio t: -13.673** and 0.382
  - Prime savers ratio t: 20.350* and 20.093 and -12.15 (28.72) in different columns
- Diagnostics and sample:
  - Observations: 1,070 in main table columns.
  - Number of countries: 57 in main table.
  - m2 test (p-value) examples: 0.630, 0.551, 0.589, 0.565, 0.492, 0.497, 0.696.
  - Hansen test (p-value) reported as 1.000 in table columns.
  - Estimation is in first-differences; dependent variable is private savings-to-GDP.
  - Standard errors in parentheses. ***, **, * denote significance at 1 percent, 5 percent, and 10 percent level respectively.
  - Instruments: lagged values of endogenous variables used as instruments.

### Empirical findings and magnitudes
- Fiscal consolidation effect:
  - A 1 percent of GDP increase in public savings would lead over time to a 0.75 percent of GDP decrease in private savings (estimated as coefficient/(1 − coefficient on lagged saving)).
  - Effect predominantly reflects emerging markets in subsample splits.
- Output growth:
  - A sustained 1 percentage point increase in per capita output growth over time leads to an almost 0.2 percent of GDP increase in the private saving rate.
- Real interest rate:
  - An increase in the rate of return decreases savings in these estimates: consistent with target savings framework where income effect dominates; effect predominantly in emerging markets.
- Forward-looking demographic variables:
  - Both G0/G1 and G2/G1 are significant and enter with expected signs. A 1 percentage point increase in G0/G1 (G2/G1) over time reduces (increases) private savings by 0.25 percentage points of GDP.
  - Dynamic demographic variables appear superior to static measures: youth-dependency ratio and prime saver share are significant in conventional regressions, but youth-dependency is knocked out when G0/G1 is included, suggesting the forward-looking variable captures superior demographic information.
- Role of liquidity/asset prices:
  - Increases in stock market capitalization as percent of GDP are associated with a reduction in saving (instrumented by private sector credit growth).
  - Excess liquidity/asset price growth can temporarily knock out fundamentals such as per capita output growth and demographic variables, and can totally confound their impact in some specifications.

### China-specific implications
- Growth and convergence:
  - Political stability and sound governance are necessary but not sufficient for convergence; rapid expansion of global trade matched by large domestic investment has been an additional required impetus — a pattern China has followed.
  - Expected demographic changes will limit labor supply and start raising unit labor costs; investment likely to fall as returns to capital fall relative to scarcer labor and declining savings make capital more expensive.
  - China will need home-grown solutions to define new growth engines as trade expansion and investment become less available.
  - Offsets are possible via policy changes (e.g., one-child-policy relaxation), increased immigration, higher participation rates, and total factor productivity gains from innovation, education, and urbanization.

- Policy dilemma and S-I path:
  - Divergence of actual S-I from demography-implied S-I may reflect changes in fundamentals or policy distortions.
  - Multilaterally adjusted S-I can differ from non-adjusted path; example: S-I is lower under the adjusted path during the period between now and 2024, implying China would need higher savings to maximize permanent consumption than would be dictated once adjusted for other countries’ circumstances.
  - One scenario: China’s savings may fall even below the adjusted path as it rebalances from excessive past savings.
  - Improved investment efficiency and strengthened social services could reduce precautionary savings and facilitate rebalancing.

- Pension system (urban employees’ basic pension insurance, end-2011 snapshot and scenarios):
  - Current participants: 285 million (215 million contributors; 70 million recipients/retirees).
  - Pension fund size as of end 2011: RMB 2 trillion.
  - Benefit per retiree in 2011: RMB 18,700 — about three times the contribution per employee.
  - Contributors are about 3 times retirees and fiscal contributions make up around a fifth of total payment.
  - Pension fund growth: from about 1 percent of GDP in 2001 to 4.2 percent of GDP in 2011.
  - Rate of return (average last 10 years): 3.5 percent per year.
  - Current coverage rate: 40 percent of retirees receiving pension; only 26 percent of employees contributing.
  - Assumptions for projections:
    - 70 percent of all retirees and employees become participants by 2030 and 2040, respectively.
    - Baseline nominal GDP growth: average 8 percent over next 40 years.
  - Under baseline where contribution and entitlement grow by the same amount, the pension fund could be depleted (Figure 17 scenario implications).
  - Other scenarios indicate need for continued hikes in fiscal and member contributions in the absence of entitlement reductions:
    - Figure 18 scenario: Fiscal contribution declines to zero; Member contribution grows by 10.2 percent; Entitlement grows by 8 percent.
    - Figure 19 scenario: Fiscal contribution grows by 8 percent; Member contribution grows by 9.5 percent; Entitlement grows by 8 percent.
  - Policy implication: expansion of pension coverage should be carefully calibrated with inflows; overly generous payouts could threaten long-term viability.

- Labor market implications:
  - China’s near-term problem is the end of the labor dividend from large labor influx rather than an immediate shrinking working population.
  - Expect rising unit labor costs, frictional and structural unemployment, and skill mismatches between growing numbers of college graduates and firms requiring specialized skills.
  - Reduced pressure to create jobs for ever-increasing cohorts, allowing lower growth, but raising need to reduce skill mismatches and cater for a dependent population (dependent population increases noticeably after 2024).
  - Greater savings will be required to cater for dependents, depending partly on preserving value of past savings.

- Public finances:
  - With no change in retirement age, workers-to-retirees ratio will drop from around 5:1 today to 2:1 by 2030.
  - Assuming no change in tax structure, tax burden on each worker would have to increase by around two and a half times to maintain current level of tax income.
  - Aging likely reduces tax buoyancy, particularly personal income and payroll taxes; indirect taxes may also be affected as elderly spend relatively more on services (less often taxed).
  - To meet old-age spending, governments typically modernize tax systems: keep tax bases from eroding, reduce tax preferences, increase dividends from state-owned enterprises, enhance natural resource and energy taxation, increase taxes on services, and broaden revenue sources.

### Conclusions and policy implications (summary)
- Demography is an important determinant of savings despite divergence of actual net savings from demography-based predicted path.
- Key policy implications:
  - Demography implies a continued external current account surplus for China for a prolonged period as it ages more slowly than main competitors; announced policies to promote consumption and address income inequality aim at intertemporal rebalancing.
  - Rebalancing from past over-saving should help global alignment with multilaterally consistent S-I path; ensuring adequate returns on China’s stock of savings is critical as low returns would add pressure for saving.
  - China needs a new growth model as it can no longer rely on a large labor dividend and savings; efficiency gains and innovation will be essential, along with better labor training to reduce skill mismatches.
  - Expansion of the pension system requires caution and careful calibration with inflows; otherwise pension funds could be run down in the next two decades as retiree numbers rise.
  - Tax revenues could come under pressure as working population share falls; countervailing changes in tax policy and administration are needed to prevent erosion of the tax base and to extend it into resources, energy and services.

* _wp1382 - 2030. In other words, the decision maker has to look at A and C relative to the working*

### REFERENCES

### _wp1382 - REFERENCES

### References
- Bank of England, 2011, ―The Future of International Capital Flows‖, Financial Stability Paper No. 12 – December 2011 (William Speller, Gregory Thwaites and Michelle Wright).
- Chamon, Marcos and Eswar Prasad, 2005, ―Determinants of Household Saving in China‖, International Monetary Fund, mimeo.
- Chen, Kaiji, Ayse •Imrohoroglu, and Selahattin • Imrohoroglu, 2007, ―The Japanese Saving Rate Between 1960 and 2000: Productivity, Policy Changes, and Demographics‖, Economic Theory, 32, pp. 87-104.
- Fan Xuchun and Zhu Baohua, 2012, ―China' s Life Expectancy Growth，Age Structure Change and National Saving Rate‖, Population Research, 36(4).
- Curtis, C., S. Lugauer, N. Mark, 2011, ―Demographic Patterns and Household Saving in China‖, NBER Working Paper No. 16828.
- Deaton, Angus, 1992, Understanding Consumption, Clarendon Lectures on Economics (Oxford: Oxford University Press).
- Faulkner-MacDonagh, Chris, and Martin Mühleisen, 2004, ―Are U.S. Households Living Beyond Their Means?‖ Finance and Development, Vol. 41, No. 1, pp. 36–9.
- Fehr, Hans, Sabine Jokisch, and Laurence J. Kotliko, 2007, ―Will China Eat Our Lunch or Take Us to Dinner? Simulating the Transition Paths of the United States, the European Union, Japan, and China‖, Fiscal Policy and Management in East Asia, NBER-EASE, Volume 16, University of Chicago Press, pp. 133-193.
- Haldane, A., 2010, ―Global imbalances in retrospect and prospect‖, available at www.bankofengland.co.uk/publications/speeches/2010/speech468.pdf.
- Hoiroka, Charles Yuji and Junmin Wan, 2006, ―The Determinants of Household Saving in China: A Dynamic Panel Analysis of Provincial Data‖, Journal of Money, Credit, and Banking, 39(8), pp. 2077-96.
- Horioka, Charles Yuji, 2010, ―Aging and Saving in Asia‖, Pacific Economic Review, 15, pp. 46-55.
- Horioka, Charles Yuji, and Terada-Hagiwara, Akiko, 2011, ―The Determinants and Long-term Projections of Saving Rates in Developing Asia‖, Japan and the World Economy 24(2), pp. 128-137.
- International Monetary Fund, 2005, ―Global Imbalances: A Saving and Investment Perspective‖, World Economic Outlook (September: Washington DC).
- International Monetary Fund, 2008, ―The Graying of Asia: Demographics, Capital Flows,and Financial Markets‖, Asia and Pacific Regional Economic Outlook (November: Washington DC).
- Krueger, Dirk and Alexander Ludwig, 2007, ―On the Consequence of Demographic Change for Rates of Returns to Capital, and the Distribution of Wealth and Welfare‖, Journal of Monetary Economics, 54(1), pp. 49-87.
- Kuijs, L., 2006, ―How will China’s saving-investment balance evolve?‖, World Bank Policy Research Working Paper, No 3958, July.
- Ma, Guonan and Wang Yi, 2010, ―China's High Saving Rate: Myth and Reality‖, BIS Working Paper, Number 312.
- Modigliani, Franco and Shi Larry Cao, 2004, ―The Chinese Saving Puzzle and the Life-Cycle Hypothesis‖, Journal of Economic Literature, 42, pp. 145-170.
- Song, Zheng Michael and Dennis Tao Yang, 2010, ―Life Cycle Earnings and Saving in a Fast-Growing Economy‖, mimeo, Chinese University of Hong Kong.
- United Nations, 2011, ―World Population Prospects: The 2010 Revision‖ (CD-ROM Edition)—Extended Dataset in Excel and ASCII formats. New York: Population Division, Department of Economic and Social Affairs, United Nations.
- Woon Gyu Choi and Il Houng Lee, 2010, ―Monetary Transmission of Global Imbalances in Asian Countries‖ IMF Working Paper WP/10/214 (International Monetary Fund: Washington DC).

### Annex I. Simulation Results
- Baseline: GDP growth over 2012-2050 in USD terms (China: 5 percent; others: 2 percent)
- Alternative Scenario I: China GDP growth reduced to 2 percent; others no change.
- Alternative Scenario II: Weight of young reduced to 50 percent of working population and the old raised to 100 percent of the working population
- Chart axis values and labels presented in the source:
  - Vertical axis values shown: -0.03, -0.02, -0.01, 0, 0.01, 0.02, 0.03
  - Horizontal axis years shown: 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012 2015 2018 2021 2024 2027 2030 2033 2036 2039 2042 2045 2048
  - Series label: S-I Adjusted for Multilateral Consistency, in percent of global GDP
  - Regional series labels shown: W Europe, US, Japan, China

*Source: _wp1382 - REFERENCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1382.pdf_