## wp18172

## Source details

**Canonical URL:** [wp18172](https://www.imf.org/-/media/files/publications/wp/2018/wp18172.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18172.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18172.pdf.json)

---

### I. Introduction — scope and headline conclusions
- Focus: demographic trends and their persistent, medium-term downward pressure on real interest rates; extends empirical evidence to a larger sample of advanced and emerging-market economies, incorporates financial integration, and explores implications for Asian countries.
- Headline conclusions:
  - Demographic trends are likely to put downward pressure on real interest rates for most major countries in Asia.
  - The direct effects of local demographics are attenuated for financially open economies; changes in the world interest rate may matter more than local demographics.
  - Demographic-driven declines in the natural rate narrow central banks' scope to achieve price stability and maintain output at potential and may pose financial stability risks that call for strong macro-prudential frameworks.

### II. Stylized facts on real interest rates and demographics
- World and regional trends:
  - World real interest rates drifted down from around 4 percent in the late-1990s to about zero by 2014.
  - Natural rates declined dramatically in Europe, the United Kingdom, and the United States since the start of the GFC.
  - In Asia: natural rates fell in advanced economies (Australia, Japan, Korea) while remaining broadly stable and relatively high in emerging economies not yet under aging pressures; in China natural rates have fallen but remain high relative to advanced Asian economies.
- Demographic developments:
  - Life expectancy at birth increased by about 10 years in Japan, the U.S., and Western Europe between 1960 and 2010.
  - Population growth rates are decreasing rapidly; population growth rate is already negative in Japan and is projected to fall to zero for Asia by 2050.
  - Share of population age 65 and older will increase rapidly and reach close to 2½ times the current level by 2050.
  - East Asia projected to be the world’s fastest-aging region; old-age dependency ratio roughly tripling by 2050.
  - Projected absolute change in working-age population (15–64): China: decline of 170 million in the working-age population over the next 35 years.
  - Speed of aging (years to increase old-age dependency ratio from 15 to 20 percent):
    - Europe: 26 years
    - United States: more than 50 years
    - China, Japan, Korea, Thailand, Singapore, Vietnam: less than 10 years
    - Australia and New Zealand aged at similar speeds to Europe

### III. Theoretical channels linking demographics to real rates
- Life-cycle savings:
  - Young households borrow; middle-age households save; old-aged households dis-save.
  - Aggregate savings can fall in countries with relatively young or old populations.
- Longevity and precautionary saving:
  - Rising longevity induces higher life-cycle savings.
- Public saving:
  - Increased dependency ratios can lower public saving via lower tax revenues and higher age-related public spending (healthcare, pensions).
- Investment channel:
  - Shrinking working-age population raises capital-to-labor ratio, lowering returns on capital and investment.
  - Youth dependency can raise investment demand initially via larger youth share and labor-force growth.
- Higgins (1988) and asymmetry:
  - Demographic center of gravity for investment demand occurs earlier in the age distribution than for savings supply; non-monotonic effects on interest rates as populations age.
- Openness:
  - Closed economy: domestic real interest rate equilibrates saving-investment.
  - Small open economy with perfect capital mobility: domestic rate driven by world interest rate; domestic demographics have no impact on domestic rate.
  - Intermediate openness: mixed outcomes possible.

### IV. Data, variables, and baseline econometric specification
- Sample and period:
  - Sample period: 1985 to 2013.
  - Sample size: 42 economies.
- Dependent variable:
  - 10-year real interest rate (r).
- Key demographic explanatory variables:
  - Youth dependency ratio (YD): population under 30 divided by population 30–64.
  - Old-age dependency ratio (OD): population over 64 divided by population 30–64.
  - Aging speed (AS): expected change in old-age dependency in 20 years.
- Openness measure:
  - Capital openness index (CO): Chinn-Ito Index (2006).
- Controls:
  - World interest rate (RW), ratio of GDP per capita to that of the United States, growth in labor productivity, cyclically-adjusted primary balance.
- Descriptive statistics (panel sample; Variable Obs Mean Std. Dev. Min Max):
  - 10-year real interest rate: 1,000 observations; Mean 3.20; Std. Dev. 3.27; Min -10.93; Max 32.91
  - Youth dependency ratio: 1,000 observations; Mean 1.02; Std. Dev. 0.38; Min 0.55; Max 2.40
  - Old dependency ratio: 1,000 observations; Mean 0.28; Std. Dev. 0.09; Min 0.11; Max 0.54
  - Aging speed: 1,000 observations; Mean 0.10; Std. Dev. 0.08; Min -0.08; Max 0.36
  - World interest rate: 1,000 observations; Mean 2.75; Std. Dev. 1.43; Min -0.17; Max 4.80
  - Chinn Ito index: 960 observations; Mean 0.76; Std. Dev. 0.32; Min 0.00; Max 1.00
  - Growth in labor productivity: 1,000 observations; Mean 1.68; Std. Dev. 2.49; Min -11.46; Max 10.81
  - Ratio of GDP per capita to that of the US: 1,000 observations; Mean 0.71; Std. Dev. 0.46; Min 0.01; Max 2.47
  - Cyclically adjusted primary balance: 765 observations; Mean -0.14; Std. Dev. 3.45; Min -15.09; Max 11.34
- Selected pairwise correlations:
  - 10-year real interest rate with Youth dependency ratio: 0.21
  - 10-year real interest rate with Old dependency ratio: -0.04
  - 10-year real interest rate with Aging speed: -0.38
  - 10-year real interest rate with World interest rate: 0.42
  - Youth dependency ratio with Old dependency ratio: -0.74
  - Youth dependency ratio with Aging speed: -0.57
  - Old dependency ratio with Aging speed: 0.19
  - Chinn Ito index correlations:
    - with Youth dependency ratio: -0.60
    - with Old dependency ratio: 0.57
    - with Aging speed: 0.39

### V. Baseline empirical results — main coefficients and interpretation
- Bivariate relationships:
  - Youth dependency ratio: positive relationship with the 10-year real interest rate.
  - Old-age dependency ratio: negative relationship with the 10-year real interest rate.
  - Aging speed: negative relationship with the 10-year real interest rate.
- Baseline panel regression (country fixed effects; Observations: 740; Number of groups: 42):
  - Youth dependency ratio: 9.41*** (2.36)
  - Youth dependency ratio * capital openness: -7.96*** (2.26)
  - Old-age dependency ratio: -18.30*** (4.97)
  - Old-age dependency ratio * capital openness: 17.04*** (5.32)
  - Aging speed: -29.70** (13.94)
  - Aging speed * capital openness: 25.89** (12.29)
  - World interest rate: 0.65*** (0.12)
  - Ratio of GDP per capita to that of the US: 2.38 (1.48)
  - Cyclically adjusted primary balance: -0.00 (0.06)
  - Growth in labor productivity: 0.07 (0.08)
- Alternative specification with demographic effects multiplied by (1 - capital openness) (Observations: 740; Number of groups: 42):
  - Youth dependency ratio * (1 - capital openness): 8.26*** (2.20)
  - Old-age dependency ratio * (1 - capital openness): -16.16*** (4.59)
  - Aging speed * (1 - capital openness): -29.26** (12.67)
  - World interest rate: 0.84*** (0.10)
  - Ratio of GDP per capita to that of the US: 2.43 (1.51)
  - Cyclically adjusted primary balance: 0.00 (0.06)
  - Growth in labor productivity: 0.07 (0.08)
  - Observations: 740; Number of groups: 42
- Interpretation of interactions:
  - Signs on demographic variables reverse when interacted with capital openness; consistent with theory that openness attenuates domestic demographic effects.
  - In a fully open economy (capital openness = 1), domestic demographic changes do not have any impact on real interest rates.
  - In a fully closed economy (capital openness = 0), demographic coefficients apply directly (e.g., baseline youth dependency coefficient 9.41*** (2.36)).

### VI. Robustness, long-horizon evidence, and diagnostics
- Alternative specifications (selected from Table 5):
  - Column (2) replacing world interest rate with US real interest rate (drop U.S. from sample): US real interest rate: 0.54*** (0.10); coefficients on demographic variables remain significant and similar in magnitude.
  - Column (3) adding year fixed effects: World interest rate coefficient increases to 1.95*** (0.31); demographic coefficients remain qualitatively similar though some significance levels change.
  - Column (4) including both world interest rate and year effects: demographic coefficients largely unchanged relative to column (3).
- Panel unit root tests (Table 6; selected classification):
  - Real interest rate: I(0)
  - Youth dependency: I(0) under constant; inconclusive under trend
  - Old dependency: I(1)
  - Aging speed: I(1)
  - World interest rate: I(1)
  - Residuals from baseline equation: I(0)
- Long-horizon (multi-year non-overlapping) results (Table 7):
  - 3-year averages (Observations: 271; Number of groups: 41):
    - Youth dependency ratio: 10.17** (3.53)
    - Youth dependency ratio * capital openness: -8.42*** (1.88)
    - Old-age dependency ratio: -21.45*** (5.77)
    - Old-age dependency ratio * capital openness: 20.25*** (3.56)
    - Aging speed: -23.02 (16.62)
    - Aging speed * capital openness: 18.11 (14.65)
    - World interest rate: 0.64*** (0.19)
  - 5-year averages (Observations: 166; Number of groups: 41):
    - Youth dependency ratio: 8.08** (2.15)
    - Youth dependency ratio * capital openness: -6.13*** (1.39)
    - Old-age dependency ratio: -13.69* (6.00)
    - Old-age dependency ratio * capital openness: 15.38** (4.28)
    - Aging speed: -32.42 (17.82)
    - Aging speed * capital openness: 24.68 (12.49)
    - World interest rate: 0.49 (0.29)
- F tests on sums of main and interaction terms (Annex I, Table 1):
  - youth dependency ratio + youth dependency ratio * capital openness = 0: F (1, 28) = 1.00; Prob > F = 0.3257
  - old dependency ratio + old dependency ratio * capital openness = 0: F (1, 28) = 0.18; Prob > F = 0.6747
  - aging speed + aging speed * capital openness = 0: F (1, 28) = 2.05; Prob > F = 0.1630

### VII. Regional projections and implications for Asia
- Projection approach:
  - Uses UN population division (2015) demographic projections and IMF staff estimates to decompose demographic impacts into youth dependency effect, old-age dependency effect, and aging speed effect.
- Key regional findings and quantitative takeaway:
  - Interest rates in Japan, Hong Kong SAR, New Zealand, and Singapore with full capital mobility are decoupled from domestic demographic trends.
  - For economies not perfectly open, channel importance depends on demographic profile:
    - Mature economies (China, Korea, Thailand): old-age dependency effect dominates; increasing old-age dependency expected to decrease interest rates.
    - Relatively young economies (India, Indonesia, Philippines): youth dependency effect dominates; declining youth dependency expected to decrease interest rates.
  - Aging speed effects:
    - Slower aging (falling aging speed) can increase interest rates by reducing savings; between 2020-30, interest rates expected to increase most in China and Australia as their aging speeds fall.
    - Aging speed projected to increase in young countries such as India and Malaysia, driving down their interest rates.
  - Aggregate implication for Asia:
    - Demographic trends could put downward pressure on interest rates by about 1 to 2 percentage points in the next decade, all else being equal.
- Figure 8 (selected Asia, cumulative change between 2020 and 2030) decomposes total demographic effect into aging speed effect, youth dependency effect, and old-age dependency effect for Malaysia, Thailand, India, China, Philippines, Indonesia, Korea, Australia, and the Asia aggregate (nominal GDP-weighted average).

### VIII. Policy implications and conclusions
- Demographics are an important structural factor in the secular decline of real interest rates observed over past decades.
- Policy-relevant implications:
  - Demographic trends tend to reduce long-term real interest rates in economies with limited capital openness.
  - As capital account openness increases, the domestic impact of demographic variables on interest rates diminishes and becomes zero under perfect capital mobility.
  - Low and declining natural rates of interest:
    - Constrain monetary policy space and the ability of central banks to attain price stability and keep output at potential.
    - Pose financial stability risks via elevated valuations, asset price bubbles, higher risk-taking, and search for yield.
  - These risks call for a strong macro-prudential framework.
- Quantitative policy takeaway for Asia:
  - Demographic trends could lower interest rates by about 1-2 percentage points over the next decade, holding other factors constant.

*IMF staff paper (content from wp18172).*

### 1. Selected Asia: Change in 10-Year Government Yield ..........................................................5

### 1. Selected Asia: Change in 10-Year Government Yield ..........................................................5

### Figures and Charts
- 1. Selected Asia: Change in 10-Year Government Yield ..........................................................5
- 2. World Real Interest Rates ......................................................................................................5
- 3. Selected Asia: Real Neutral Interest Rates ............................................................................6
- 4. Asia: Fertility, Life Expectancy and Population Growth .......................................................6
- 5. Asia and the Rest of the World: Change in Working-Age Population ..................................7
- 6. Number of Years for the Old-Age Dependency to Increase from 15 to 20 Percent  .............8
- 7. Pairwise Bivariate Plots .......................................................................................................13
- 8. Selected Asia: Impact of Demographics on 10-Year Real Interest Rates ...........................19

### Tables
- 1. Descriptive Statistics ............................................................................................................12
- 2. Correlation Matrix ...............................................................................................................12
- 3. Panel Regression: Demographics and Long-Term Interest Rates .......................................14
- 4. Panel Regression: Demographics and Long-Term Interest Rates .......................................15
- 5. Alternative Specifications: Demographics and Long-Term Interest Rates .........................16
- 6. Panel Unit Root Tests ..........................................................................................................17
- 7. Long Horizon Evidence on Demographic Structure and Real Rates ...................................18

*Source: wp18172 - 1. Selected Asia: Change in 10-Year Government Yield ..........................................................5*

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

### References

### I. Introduction
- Long-term bond yields have declined significantly in Europe, the United States, and in parts of Asia (Australia and Korea).
- Accommodative monetary policies (conventional and unconventional) implemented after the Global Financial Crisis (GFC) are an important force driving low real rates, but longer-term declines predate the GFC and span more than two decades across many countries.
- Literature on secular stagnation attributes declining interest rate trends to structural factors including: adverse demographic developments, a falling price of investment, lower productivity growth, savings glut, scarcity of safe assets, and increasing wealth and income inequality.
- The GFC contributed additional downward pressure on interest rates via impacts on investment demand, productivity, and the supply of safe assets.
- Structural factors likely to keep real interest rates low for a long time include lower productivity growth and demographic developments.
- This paper focuses on demographic trends and their persistent, medium-term downward pressure on real interest rates, extends empirical evidence to a larger sample of advanced and emerging-market economies, incorporates the role of financial integration, and explores implications for Asian countries.
- Key conclusions previewed:
  - Demographic trends are likely to put downward pressure on real interest rates for most major countries in Asia.
  - The direct effects of local demographics are attenuated for financially open economies that are well integrated into global markets; changes in the world interest rate, possibly driven by global aging, may matter more than local demographics.
  - Demographic-driven declines in the natural rate of interest narrow central banks' scope to achieve price stability and maintain output at potential, and may pose financial stability risks that call for strong macro-prudential frameworks.

### II. Real Interest Rates and Demographic Change — Stylized Facts
- World real interest rates have drifted down from around 4 percent in the late-1990s to about zero by 2014.
- Natural rates of interest have declined dramatically in Europe, the United Kingdom, and the United States since the start of the GFC; in Asia, natural rates have fallen in advanced economies (Australia, Japan, Korea) while remaining broadly stable and relatively high in emerging economies not yet under aging pressures. In China, natural rates have fallen but remain high relative to advanced Asian economies.

A. Demographic trends
- Between 1960 and 2010, life expectancy at birth increased by about 10 years in Japan, the U.S., and Western Europe.
- Population growth rates are decreasing rapidly; the population growth rate is already negative in Japan and is projected to fall to zero for Asia by 2050.
- The share of the population age 65 and older will increase rapidly and reach close to 2½ times the current level by 2050.
- East Asia is projected to be the world’s fastest-aging region, with its old-age dependency ratio roughly tripling by 2050.
- Projected absolute change in working-age population (15–64) highlights large declines over the next 35 years, including:
  - China: decline of 170 million in the working-age population over the next 35 years (as stated).
- Speed of aging: the number of years it takes for the old-age dependency ratio to increase from 15 to 20 percent:
  - Europe: 26 years
  - United States: more than 50 years
  - Many Asian economies (China, Japan, Korea, Thailand, Singapore, Vietnam): less than 10 years (or took/will take less than 10 years)
  - Australia and New Zealand aged at similar speeds to Europe.

B. Theoretical priors (mechanisms linking demographics to rates)
- Life-cycle savings:
  - Young households borrow, middle-age households save, old-aged households dis-save (Modigliani 1970).
  - Aggregate savings can fall in countries with relatively young or old populations.
- Longevity and precautionary saving:
  - Rising longevity induces higher life-cycle savings (Li, Zhang and Zhang (2007); Carroll (1997) on uncertainty).
- Public saving:
  - Increased dependency rates can lower public saving via lower tax revenues and higher age-related public spending (healthcare, pensions).
- Investment channel:
  - Shrinking working-age population raises capital-to-labor ratio, lowering returns on capital and investment.
  - Youth dependency can raise investment demand initially (via larger youth share and labor-force growth).
- Higgins (1988): demographic center for gravity for investment demand occurs earlier in age distribution than for savings supply; asymmetric effects imply non-monotonic effects on interest rates as populations age.
- Openness matters:
  - Closed economy: domestic real interest rate equilibrates savings-investment via price of funds.
  - Small open economy with perfect capital mobility: domestic rate driven by world interest rate; fall in domestic saving leads to current account deterioration, not necessarily higher domestic rates.
  - Intermediate cases: simultaneous rising real interest rates, falling investment, and worsening current account are possible.

C. Empirical literature
- General equilibrium and calibrated models (Yoo (1994), Abel (2001), Brooks (2000), Feroli (2003), Brooks (2003), IMF (2004)) indicate demographic shocks can affect returns; plausible-sized shocks may be quantitatively modest but can decrease aggregate saving rates in advanced economies turning them into capital importers over the long run.
- Recent calibrated life-cycle/overlapping-generations models:
  - Carvalho et al. (2016): equilibrium annual real rate falls by 1.5 percentage points between 1990 and 2014 for a representative developed country.
  - Gagnon et al. (2016): calibrated model accounts for 1¼ percentage points decline in the U.S. equilibrium real interest rate since 1980.
- Econometric studies:
  - Poterba (2001): robust link between age structure and long-term bond yields in the U.S.
  - Aksoy and others (2016): aging population and reduced fertility reduce output growth and real interest rates across 21 OECD countries using a panel VAR (note: Aksoy et al. use the short-term central bank discount rate; this paper focuses on 10-year bond yields).
  - Ferrero and others (2017): lower population growth exerted downward pressures on long-term real interest rates in the euro area over the past decade.
  - Favero and Galasso (2016): expect euro area interest rates to revert to historical means based on demographic-based projections.

### III. Model specification and results — Empirical approach
- Demographic channels considered:
  - Youth dependency ratio (YD): ratio of population aged under 30 divided by population between 30 and 64.
  - Old-age dependency ratio (OD): ratio of population aged over 64 divided by population between 30 and 64.
  - Aging speed (AS): expected change in old-age dependency in 20 years.
- The baseline econometric approach:
  - Panel regression with country fixed effects.
  - Dependent variable: 10-year real interest rate (r).
  - Explanatory variables: YD, OD, AS; interactions of each demographic variable with capital openness index (CO) to capture effects of financial integration.
  - RW denotes world interest rate.
  - Controls include: ratio of a country’s GDP per capita to that of the United States, growth in labor productivity, and the cyclically-adjusted primary balance.
- Data:
  - Sample period: 1985 to 2013.
  - Sample size: 42 economies (listed in the paper).
  - Sources:
    - 10-year real interest rates and world interest rate: IMF (2014), King and Low (2014), IMF World Economic Outlook.
    - Demographics: United Nations.
    - Labor productivity: Penn World Table 9.0.
    - GDP and fiscal data: IMF World Economic Outlook.
    - Capital openness index: Chinn-Ito Index (2006).
  - Descriptive statistics (panel sample, Variable Obs Mean Std. Dev. Min Max):
    - 10-year real interest rate: 1,000 observations; Mean 3.20; Std. Dev. 3.27; Min -10.93; Max 32.91
    - Youth dependency ratio: 1,000 observations; Mean 1.02; Std. Dev. 0.38; Min 0.55; Max 2.40
    - Old dependency ratio: 1,000 observations; Mean 0.28; Std. Dev. 0.09; Min 0.11; Max 0.54
    - Aging speed: 1,000 observations; Mean 0.10; Std. Dev. 0.08; Min -0.08; Max 0.36
    - World interest rate: 1,000 observations; Mean 2.75; Std. Dev. 1.43; Min -0.17; Max 4.80
    - Chinn Ito index: 960 observations; Mean 0.76; Std. Dev. 0.32; Min 0.00; Max 1.00
    - Growth in labor productivity: 1,000 observations; Mean 1.68; Std. Dev. 2.49; Min -11.46; Max 10.81
    - Ratio of GDP per capita to that of the US: 1,000 observations; Mean 0.71; Std. Dev. 0.46; Min 0.01; Max 2.47
    - Cyclically adjusted primary balance: 765 observations; Mean -0.14; Std. Dev. 3.45; Min -15.09; Max 11.34
  - Pairwise correlations (selected):
    - 10-year real interest rate with Youth dependency ratio: 0.21
    - 10-year real interest rate with Old dependency ratio: -0.04
    - 10-year real interest rate with Aging speed: -0.38
    - 10-year real interest rate with World interest rate: 0.42
    - Youth dependency ratio with Old dependency ratio: -0.74
    - Youth dependency ratio with Aging speed: -0.57
    - Old dependency ratio with Aging speed: 0.19
    - Chinn Ito index correlations:
      - with Youth dependency ratio: -0.60
      - with Old dependency ratio: 0.57
      - with Aging speed: 0.39

### IV. Empirical results — Key findings
- Bivariate relationships:
  - Positive relationship between youth dependency ratio and the 10-year real interest rate.
  - Negative relationship between old-age dependency ratio and the 10-year real interest rate.
  - Negative relationship between aging speed and the 10-year real interest rate.
- Baseline panel regression results (Table 3; standard errors in parentheses; *, **, *** indicate significance at the 10 percent, 5 percent, and 1 percent levels):
  - Youth dependency ratio: 9.41*** (2.36)
  - Youth dependency ratio * capital openness: -7.96*** (2.26)
  - Old-age dependency ratio: -18.30*** (4.97)
  - Old-age dependency ratio * capital openness: 17.04*** (5.32)
  - Aging speed: -29.70** (13.94)
  - Aging speed * capital openness: 25.89** (12.29)
  - World interest rate: 0.65*** (0.12)
  - Ratio of GDP per capita to that of the US: 2.38 (1.48)
  - Cyclically adjusted primary balance: -0.00 (0.06)
  - Growth in labor productivity: 0.07 (0.08)
  - Observations: 740
  - Number of groups: 42
- Interpretation of interaction terms:
  - Signs on demographic variables reverse when interacted with capital openness, consistent with theory that open capital accounts attenuate domestic demographic effects.
  - Tests of restrictions (ߚ5 = -ߚ3, ߚ4 = -ߚ6, ߚ7 = -ߚ9) cannot be rejected (Annex Table 1), motivating an alternative specification with demographic effects multiplied by (1 - capital openness).
- Alternative specification highlights:
  - Domestic real interest rate strongly co-moves with the world interest rate.
  - Estimated elasticity: a one percentage point increase in the world interest rate would increase the domestic interest rate (in a small open economy) by 0.8 percentage point (statement in text).
  - In a fully open economy (capital openness = 1), domestic demographic changes do not have any impact on real interest rates.
  - In a fully closed economy (capital openness = 0), a 1 percentage point increase in youth dependency increases the interest rate by (coefficient reported in baseline) 9.41 (percentage points) — interpreted as 9.41 but note coefficient units correspond to model specification; text states "a 1 percentage point increase in youth dependency increases the interest rate by" then truncated in provided content.

### V. Policy implications (as discussed)
- Demographic-driven declines in the natural rate of interest reduce central banks' scope to achieve price stability and maintain output at potential.
- Demographic effects on medium-term nominal and real interest rates may pose financial stability risks.
- These risks call for a strong macro-prudential framework.

*Italic source attribution: IMF staff paper (content from wp18172 - References, source PDF wp18172 - References).*

### 8.26 basis points. Similarly, a 1 percentage point increase in old-age dependency and aging

### wp18172 - 8.26 basis points. Similarly, a 1 percentage point increase in old-age dependency and aging

### Key empirical findings on demographics and long-term real interest rates
- Baseline estimates (dependent variable: 10-year real interest rate):
  - Youth dependency ratio * (1 - capital openness): 8.26*** (2.20)
  - Old-age dependency ratio * (1 - capital openness): -16.16*** (4.59)
  - Aging speed * (1 - capital openness): -29.26** (12.67)
  - World interest rate: 0.84*** (0.10)
  - Ratio of GDP per capita to that of the US: 2.43 (1.51)
  - Cyclically adjusted primary balance: 0.00 (0.06)
  - Growth in labor productivity: 0.07 (0.08)
  - Observations: 740
  - Number of groups: 42
- Alternative specifications (Table 5 summary, columns (1)-(4)):
  - Annual model coefficients (column (1)):
    - Youth dependency ratio: 9.41*** (2.36)
    - Youth dependency ratio * capital openness: -7.96*** (2.26)
    - Old-age dependency ratio: -18.30*** (4.97)
    - Old-age dependency ratio * capital openness: 17.04*** (5.32)
    - Aging speed: -29.70** (13.94)
    - Aging speed * capital openness: 25.89** (12.29)
    - World interest rate: 0.65*** (0.12)
    - Ratio of GDP per capita to that of the US: 2.38 (1.48)
    - Cyclically adjusted primary balance: -0.00 (0.06)
    - Growth in labor productivity: 0.07 (0.08)
    - Observations: 740; Number of groups: 42; Country fixed effects: Y; Time fixed effects: N
  - Column (2): replacing world interest rate with US real interest rate (drop U.S. from sample)
    - World interest rate coefficient omitted; US real interest rate: 0.54*** (0.10)
    - Youth dependency ratio: 8.86*** (1.93)
    - Youth dependency ratio * capital openness: -7.38*** (2.16)
    - Old-age dependency ratio: -21.84*** (5.55)
    - Old-age dependency ratio * capital openness: 16.82*** (5.49)
    - Aging speed: -32.26** (14.62)
    - Aging speed * capital openness: 25.64* (12.72)
    - Observations: 727; Number of groups: 41; Country fixed effects: Y; Time fixed effects: N
  - Column (3): add year fixed effects (time dummies)
    - Youth dependency ratio: 6.62*** (1.89)
    - Youth dependency ratio * capital openness: -9.98*** (2.17)
    - Old-age dependency ratio: -23.13** (9.97)
    - Old-age dependency ratio * capital openness: 30.87*** (9.23)
    - Aging speed: -16.66 (15.14) — similar qualitative effect but less significant
    - Aging speed * capital openness: 27.37* (14.46)
    - World interest rate: 1.95*** (0.31)
    - Ratio of GDP per capita to that of the US: 1.44 (1.97)
    - Observations: 740; Number of groups: 42; Country fixed effects: Y; Time fixed effects: Y
  - Column (4): include both world interest rate and year effects
    - Coefficients on demographic variables unchanged relative to column (3)
    - Coefficients on year dummies (not reported) change, implying world interest rate and time dummies together account for global factors
    - Observations: 740; Number of groups: 42; Country fixed effects: Y; Time fixed effects: Y

### Robustness and econometric diagnostics
- Concern: demographic variables evolve slowly; potential unit root issues.
- Panel unit root tests (Table 6) — selected results and classification:
  - Real interest rate: I(0) both under constant and trend (Im, Pesaran, Shin t-value 3.23; p-values shown in table)
  - Youth dependency: I(0) under constant; inconclusive under trend
  - Old dependency: I(1)
  - Aging speed: I(1)
  - World interest rate: I(1)
  - Residuals from baseline equation: I(0) (Fisher and Im, Pesaran and Shin tests reject null of unit root in residuals)
- Long-horizon (multi-year non-overlapping) evidence (Table 7):
  - 3-year averages:
    - Youth dependency ratio: 10.17** (3.53)
    - Youth dependency ratio * capital openness: -8.42*** (1.88)
    - Old-age dependency ratio: -21.45*** (5.77)
    - Old-age dependency ratio * capital openness: 20.25*** (3.56)
    - Aging speed: -23.02 (16.62)
    - Aging speed * capital openness: 18.11 (14.65)
    - World interest rate: 0.64*** (0.19)
    - Observations: 271; Number of groups: 41
  - 5-year averages:
    - Youth dependency ratio: 8.08** (2.15)
    - Youth dependency ratio * capital openness: -6.13*** (1.39)
    - Old-age dependency ratio: -13.69* (6.00)
    - Old-age dependency ratio * capital openness: 15.38** (4.28)
    - Aging speed: -32.42 (17.82)
    - Aging speed * capital openness: 24.68 (12.49)
    - World interest rate: 0.49 (0.29)
    - Observations: 166; Number of groups: 41
- F tests on sums of main and interaction terms (Annex I, Table 1):
  - (1) youth dependency ratio + youth dependency ratio * capital openness = 0: F (1, 28) = 1.00; Prob > F = 0.3257
  - (2) old dependency ratio + old dependency ratio * capital openness = 0: F (1, 28) = 0.18; Prob > F = 0.6747
  - (3) aging speed + aging speed * capital openness = 0: F (1, 28) = 2.05; Prob > F = 0.1630

### Regional projections: Asia and long-term rates
- Using UN population division (2015) demographic projections, IMF staff projections indicate:
  - Interest rates in Japan, Hong Kong SAR, New Zealand, and Singapore with full capital mobility are decoupled from domestic demographic trends.
  - For economies not perfectly open, channel importance depends on demographic profile:
    - Mature economies (e.g., China, Korea, Thailand): old-age dependency effect dominates; increasing old-age dependency expected to decrease interest rates.
    - Relatively young economies (e.g., India, Indonesia, Philippines): youth dependency effect dominates; declining youth dependency expected to decrease interest rates.
  - Aging speed effects:
    - Slower aging (falling aging speed) can increase interest rates by reducing savings; between 2020-30, interest rates expected to increase most in China and Australia as their aging speeds fall.
    - Aging speed projected to increase in young countries such as India and Malaysia, driving down their interest rates.
  - Aggregate implication for Asia:
    - Demographic trends could put downward pressure on interest rates by about 1 to 2 percentage points in the next decade, all else being equal.
  - Figure 8 (selected Asia, cumulative change between 2020 and 2030) decomposes total demographic effect into aging speed effect, youth dependency effect, and old-age dependency effect for Malaysia, Thailand, India, China, Philippines, Indonesia, Korea, Australia, and the Asia aggregate (nominal GDP-weighted average).

### Interpretation and policy-relevant implications (conclusions drawn in the chapter)
- Demographics are an important structural factor in the secular decline of real interest rates observed over past decades.
- Key mechanisms and implications:
  - Demographic trends (youth dependency, old-age dependency, aging speed) tend to reduce long-term real interest rates in economies with limited capital openness.
  - As capital account openness increases, the domestic impact of demographic variables on interest rates diminishes and becomes zero under perfect capital mobility.
  - Low and declining natural rates of interest:
    - Constrain monetary policy space and the ability of central banks to attain price stability and keep output at potential.
    - Pose financial stability risks via elevated valuations, asset price bubbles, higher risk-taking, and search for yield.
- Quantitative takeaway:
  - Across Asian economies, demographic trends could lower interest rates by about 1-2 percentage points over the next decade, holding other factors constant.

*Source: IMF staff estimates (from wp18172).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18172.pdf_