## Online Annex 2.2. Healthy Aging

## Source details

**Canonical URL:** [Online Annex 2.2. Healthy Aging](https://www.imf.org/-/media/files/publications/weo/2025/april/english/ch2onlineannex.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/weo/2025/april/english/ch2onlineannex.pdf.md)
- [Structured JSON version](/-/media/files/publications/weo/2025/april/english/ch2onlineannex.pdf.json)

---

### Data and sample
- Microdata surveys covering close to one million observations for individuals aged 50-90 in 29 advanced economies (AEs) and 12 emerging market economies (EMs) during 2000–22.
- Data sources: 13 different microdata surveys. Data for all countries except Brazil, South Africa, and Thailand are pre-harmonized in the Gateway to Global Aging database; Brazil, South Africa, and Thailand were harmonized before inclusion.
- Sample restrictions: (i) start in 2000 (previous data only available for the United States); (ii) individuals aged 50-90. Multiple survey waves included where available; single wave for India, Ireland, and Malaysia.
- Observations by health indicator (Online Annex Table 2.2.1): Grip Strength 460,240; Lung Function 217,702; Cognitive PC 382,424; Overall Health 731,020; Physical PC 628,733; Psych. Well-Being 345,479.
- Number of countries used per regression (Online Annex Table 2.2.1): Grip Strength 38; Lung Function 24; Cognitive PC 34; Overall Health 30; Physical PC 30; Psych. Well-Being 30.

### Health indicators and measurement
- Measured physical and cognitive: grip strength (kg), lung function, memory, orientation, verbal fluency, math.
- Self-reported: overall health (1-5 scale), physical ability (ADLs, IADLs), frequency of pain, hearing ability, CASP index of psychological well-being.
- Health behaviors: smoking, alcohol consumption, frequency of exercise, body-mass index.
- Incidence of chronic diseases: diagnosis of 17 chronic diseases.
- Standardization: indicators standardized to have zero mean and unit standard deviation for comparability.
- Composite measures:
  - Cognitive PC: first principal component of memory, orientation, verbal fluency, and math.
  - Physical PC: first principal component of ADLs, IADLs, pain frequency, and hearing.
  - Frailty: composite measure averaging health indicators, normalized to a 0-1 scale (increase implies deterioration).

### Empirical strategy
- Two-part approach:
  1. Evidence of healthy aging: regress each health indicator H_{i,t} on a time trend (t) controlling for socio-economic covariates (age, gender, education, log household wealth) and country fixed effects:
     - H_{i,t} = α_0 + α_1 t + κ X_{i,t} + ε_{i,t}
     - Time trends allowed to differ by (i) EMs vs AEs, (ii) gender, (iii) urban vs rural, (iv) education (primary, secondary, tertiary), and (v) household wealth quintile. Standard errors clustered at the country level.
  2. Health and labor outcomes:
     - OLS correlations: L_{i,t} = β_0 + β_1 H_{i,t} + κ Z_{i,t} + β_2 t + ε_{i,t}
     - Causal effects: 2SLS using exogenous health shocks proxied by incidence of chronic diseases (share of chronic diseases out of 17) as instrument for health. Controls include lifestyle factors (low/high BMI, physical inactivity, smoking). Robustness: augmented inverse propensity score weighting (AIPW).

### Evidence of healthy aging — key findings
- Broad-based improvements over time across physical, cognitive, and mental health indicators: positive and significant coefficient on the time trend (Online Annex Table 2.2.1).
- Estimated year (time trend) coefficients (Online Annex Table 2.2.1):
  - Grip Strength: Year 0.004** (standard error 0.002)
  - Lung Function: Year 0.011** (0.004)
  - Cognitive PC: Year 0.020*** (0.003)
  - Overall Health: Year 0.005** (0.002)
  - Physical PC: Year 0.005*** (0.002)
  - Psych. Well-Being: Year 0.015*** (0.003)
- Socio-economic correlates (selected coefficients, Online Annex Table 2.2.1):
  - Age: Grip Strength -0.035*** (0.001); Lung Function -0.028*** (0.001); Cognitive PC -0.026*** (0.001); Overall Health -0.020*** (0.001); Physical PC -0.025*** (0.002); Psych. Well-Being -0.014*** (0.002).
  - Male: Grip Strength 1.345*** (0.025); Lung Function 0.761*** (0.03); Overall Health 0.041* (0.02); Physical PC 0.036** (0.015); Psych. Well-Being 0.045** (0.018).
  - Upper Secondary Education: Grip Strength 0.064*** (0.013); Lung Function 0.097*** (0.031); Cognitive PC 0.362*** (0.034); Overall Health 0.217*** (0.024); Physical PC 0.127*** (0.019); Psych. Well-Being 0.191*** (0.021).
  - Tertiary Education: Grip Strength 0.081*** (0.016); Lung Function 0.225*** (0.024); Cognitive PC 0.625*** (0.042); Overall Health 0.422*** (0.038); Physical PC 0.227*** (0.027); Psych. Well-Being 0.309*** (0.033).
  - (Log) Household Wealth: Grip Strength 0.019*** (0.002); Lung Function 0.017*** (0.003); Cognitive PC 0.029*** (0.004); Overall Health 0.035*** (0.005); Physical PC 0.024*** (0.004); Psych. Well-Being 0.040*** (0.003).
- Frailty has declined over birth cohorts (Online Annex Figure 2.2.3, panel 3), consistent with healthy aging.
- Cognitive-function improvements have been especially prominent.

### Heterogeneity in trends
- Faster improvements in cognitive health among EMs relative to AEs, suggesting partial cross-country convergence; other dimensions show similar paces across country groups.
- Selected year coefficients by group (Online Annex Table 2.2.2):
  - By Country Income Group:
    - AEs: Grip Strength 0.004**; Lung Function 0.010**; Cognitive PC 0.020***; Overall Health 0.003; Physical PC 0.004***; Psych. Well-Being 0.014***.
    - EMs: Grip Strength 0.006*; Lung Function 0.022***; Cognitive PC 0.033***; Overall Health 0.020**; Physical PC 0.011; Psych. Well-Being 0.032***.
  - By Age Group (50s, 60s, 70s, 80s): Year coefficients broadly similar across age groups (e.g., Grip Strength 0.004** across 50s–80s).
  - By Gender:
    - Females: Grip Strength 0.008***; Lung Function 0.010*; Cognitive PC 0.022***; Overall Health 0.005**; Physical PC 0.004**; Psych. Well-Being 0.015***.
    - Males: Grip Strength 0.001; Lung Function 0.013**; Cognitive PC 0.018***; Overall Health 0.004; Physical PC 0.006***; Psych. Well-Being 0.016***.
  - By Location:
    - Urban: Grip Strength 0.005**; Lung Function 0.009*; Cognitive PC 0.021***; Overall Health 0.006**; Physical PC 0.006***; Psych. Well-Being 0.017***.
    - Rural: Grip Strength 0.005*; Lung Function 0.014***; Cognitive PC 0.014**; Overall Health 0.004; Physical PC 0.004*; Psych. Well-Being 0.020***.
- Conclusion: Improvements vary across countries but are broadly similar across socio-economic groups (gender, urban/rural, education, wealth). Wide socio-economic health gaps persist.

### Health and labor market links — OLS and IV evidence
- Association patterns:
  - Better old-age health associated with higher labor force participation, higher probability of employment, higher hours worked per week and weeks worked per year, higher total labor earnings and labor productivity (earnings per hour), and later retirement.
- Instrument: incidence of chronic diseases (share out of 17) used in 2SLS; robust to alternative definitions (total number, dummy for any chronic disease, dummy for top 10 chronic diseases). First-stage F-statistics exceed the Stock and Yogo (2005) rule-of-thumb critical value of 10.
- Instrument identification assumption: some chronic diseases—those not explained by socio-economic factors and health behaviors—are randomly assigned. Controls include lifestyle factors (low/high BMI, physical inactivity, smoking). AIPW used to address endogeneity of the instrument.

- First-stage impacts of chronic diseases on health indicators (Annex Figure 2.2.5, point estimates reported):
  - Psychological Well-Being: –3.6
  - Physical PC: –1.9
  - Overall Health: –0.2
  - (Other group labels shown: Cognitive PC, Lung Function, Grip Strength, Measured Health)

### Selected OLS estimates (Online Annex Table 2.2.3)
- Measured Health (selected coefficients):
  - Grip Strength: Total 0.022***; Retirement 0.026**; Employment 1.069***; Earnings/hr 0.038**; Age 0.579***.
  - Lung Function: Total 0.013***; Retirement 0.013**; Earnings/hr 0.038**; Age 0.197.
  - Cognitive PC: Retirement 0.027***; Employment 0.501*; Earnings/hr 0.052**; Age 0.746***.
- Self-Reported Health (selected coefficients):
  - Overall Health: Total 0.051***; Retirement 0.026***; Employment 0.482***; Earnings/hr 0.051***; Age 0.803***.
  - Physical PC: Total 0.022**; Retirement 0.059**; Employment 0.889***; Earnings/hr 0.065***; Age 0.757***.
  - Psych. Well-Being: Total 0.011**; Retirement 0.057***; Employment 0.384**; Earnings/hr 0.062***; Age 0.501***.
- Composite Health Frailty:
  - Frailty Index: Total -0.302***; Retirement -0.248***; Employment -4.934***; Weekly hours -1.676***; Earnings -1.008***; Earnings/hr -0.730***; Age -6.644***.

### Selected 2SLS estimates (Online Annex Table 2.2.4)
- Measured Health (selected coefficients):
  - Grip Strength: Total 0.529***; Retirement 0.318***; Employment 9.840***; Weekly hours 1.153; Earnings 0.765***; Earnings/hr 0.599***; Age 5.350***.
  - Lung Function: Total 0.794***; Retirement 0.448; Employment 13.572***; Weekly hours 2.220; Earnings 2.094***; Earnings/hr 1.365**; Age 8.651***.
  - Cognitive PC: Total 0.996***; Retirement 0.919**; Employment 32.759***; Weekly hours 4.051; Earnings 1.512***; Earnings/hr 1.569*; Age 8.779***.
- Self-Reported Health:
  - Overall health: Total 0.140***; Retirement 0.043***; Employment 1.924***; Weekly hours 0.214*; Earnings 0.152***; Earnings/hr 0.097***; Age 1.348***.
  - Physical PC: Total 0.215***; Retirement 0.146***; Employment 5.367***; Weekly hours 0.449; Earnings 0.423***; Earnings/hr 0.292***; Age 1.948***.
  - Psych. Well-Being: Total 0.201***; Retirement 0.152***; Employment 4.003***; Weekly hours 0.364; Earnings 0.328***; Earnings/hr 0.192***; Age 2.145***.
- Composite Health Frailty:
  - Frailty Index: Total -1.275***; Retirement -0.464***; Employment -21.646***; Weekly hours -2.315*; Earnings -1.763***; Earnings/hr -1.160***; Age -11.806***.
- 2SLS baseline controls: Socio-Economic Controls ✓; Lifestyle Controls ✓; Country FE ✓.

### AIPW robustness and chronic disease dummy effects (Online Annex Table 2.2.5)
- AIPW average treatment effects for a binary indicator of the incidence of one or more chronic diseases (top 10 chronic diseases by impact on overall health frailty):
  - Chronic Disease Dummy (Top 10) — AIPW Average Treatment Effect:
    - Total: -0.071*** (std. err. 0.008)
    - Retirement: -0.018*** (0.004)
    - LFP: -1.235*** (0.109)
    - Employment: -0.371*** (0.068)
    - Weekly hours: -0.099*** (0.020)
    - Weeks/year: -0.039** (0.021)
    - Earnings: -0.014*** (0.010)
    - Age: -0.590*** (0.067)
  - OLS estimate for same dummy:
    - Total: -0.080*** (0.007)
    - Retirement: -0.030*** (0.007)
    - LFP: -0.880*** (0.307)
    - Employment: -0.256*** (0.083)
    - Weekly hours: -0.064*** (0.015)
    - Weeks/year: -0.018 (0.017)
    - Earnings: -0.045* (0.022)
    - Age: -0.416* (0.231)
- Observations: 565,640 total; 205,315 (Retirement); 198,418 (LFP); 124,139 (Employment); 93,893 (Weekly hours); 77,412 (Weeks/year); 44,460 (Earnings); 285,809 (Age).
- Controls: Socio-Economic Controls ✓; Lifestyle Controls ✓; Country FE ✓.
- Notes:
  - Treatment dummy defined as top 10 chronic diseases by impact on overall health frailty; results robust when considering all 17 chronic diseases or top 10 by impact on self-rated overall health.
  - Overlap assumption satisfied. Similar OLS estimates for direct effect provide additional comfort on instrument exogeneity.

### 2SLS estimates: Effect of a 1-decade health improvement (Online Annex Table 2.2.6)
- Health indicators rescaled to represent effects of a 1-decade improvement. Controls: socio-economic characteristics, lifestyle factors, survey year, country fixed effects. Standard errors clustered at country level.
- Measured Health (1-decade improvements):
  - Grip Strength: Total 0.023***; Retirement 0.014***; Employment 0.436***; Earnings/hr 0.027***; Age 0.237***.
  - Lung Function: Total 0.090***; Retirement 0.051; Employment 1.540***; Earnings/hr 0.155**; Age 0.982***.
  - Cognitive PC: Total 0.203***; Retirement 0.187**; Employment 6.668***; Earnings/hr 0.319*; Age 1.787***.
- Self-Reported Health (1-decade improvements):
  - Overall health: Total 0.006***; Retirement 0.002***; Employment 0.088***; Weekly hours 0.010*; Earnings/hr 0.004***; Age 0.062***.
  - Physical PC: Total 0.011***; Retirement 0.007***; Employment 0.264***; Earnings/hr 0.014***; Age 0.096***.
  - Psych. Well-Being: Total 0.031***; Retirement 0.023***; Employment 0.618***; Earnings/hr 0.030***; Age 0.331***.
- Composite Health Frailty (1-decade improvement):
  - Frailty Index: Total 0.034***; Retirement 0.012***; Employment 0.576***; Weekly hours 0.062*; Earnings 0.047***; Earnings/hr 0.031***; Age 0.314***.
- Controls: Socio-Economic Controls ✓; Lifestyle Controls ✓; Country FE ✓.

### Heterogeneity by age, job characteristics, and policy implications
- Age heterogeneity:
  - 2SLS results vary across age groups: labor market effects generally affect the extensive margin most for individuals in their 50s, followed by those in their 60s and 70s, with much lower effects for those aged 80 and above.
  - Online Annex Figure 2.2.6 reports cognitive-health impacts on labor force participation by age groups plotted on range 0.0 to 0.5 for 50–59, 60–69, 70–79, 80–89, 50+.
- Job characteristics and age-friendliness:
  - Older workers less willing to work in less “age friendly” jobs (physically demanding, higher stress, less freedom, poor job security).
  - Better health has stronger positive effects on labor market outcomes in more age-friendly occupations.
  - Online Annex Figure 2.2.7 shows shares (percent) of workers reporting undesirable job characteristics by age groups (50–59, 60–69, 70–79, 80+) and education groups (< Upper Secondary, Upper Secondary, Tertiary).

### Robustness and sensitivity
- Results robust to:
  - Controlling for the square of the age variable and for rural location (excluded from baseline due to missing data for some countries).
  - Alternative instrument definitions (number of chronic diseases, dummy indicators).
  - AIPW estimator for instrument endogeneity concerns.
- Higher-order time trends either statistically insignificant or somewhat sensitive to health indicator and specification; first-order time trends are robust.

*Source: Online Annex 2.2, World Economic Outlook, International Monetary Fund | April 2025 (ch2onlineannex - 2022. Inter-university Consortium for Political and Social Research).*

### 2022. Inter-university Consortium for Political and Social Research.

### Online Annex 2.2. Healthy Aging

### Data and sample
- Microdata surveys covering close to one million observations for individuals aged 50-90 in 29 advanced economies (AEs) and 12 emerging market economies (EMs) during 2000–22.
- Data sources: 13 different microdata surveys. Data for all countries except Brazil, South Africa, and Thailand are pre-harmonized in the Gateway to Global Aging database; Brazil, South Africa, and Thailand were harmonized before inclusion.
- Sample restrictions: (i) start in 2000 (previous data only available for the United States); (ii) individuals aged 50-90. Multiple survey waves included where available; single wave for India, Ireland, and Malaysia.
- Observations by health indicator (from Online Annex Table 2.2.1): Grip Strength 460,240; Lung Function 217,702; Cognitive PC 382,424; Overall Health 731,020; Physical PC 628,733; Psych. Well-Being 345,479.
- Number of countries used per regression (from Online Annex Table 2.2.1): Grip Strength 38; Lung Function 24; Cognitive PC 34; Overall Health 30; Physical PC 30; Psych. Well-Being 30.

### Health indicators and standardization
- Health indicators used:
  - Measured physical and cognitive: grip strength (kg), lung function, memory, orientation, verbal fluency, math.
  - Self-reported: overall health (1-5 scale), physical ability (ADLs, IADLs), frequency of pain, hearing ability, CASP index of psychological well-being.
  - Health behaviors: smoking, alcohol consumption, frequency of exercise, body-mass index.
  - Incidence of chronic diseases: diagnosis of 17 chronic diseases.
- Indicators standardized to have zero mean and unit standard deviation for comparability.
- Cognitive PC: first principal component of memory, orientation, verbal fluency, and math.
- Physical PC: first principal component of ADLs, IADLs, pain frequency, and hearing.
- Frailty: composite measure averaging health indicators, normalized to a 0-1 scale (increase implies deterioration).

### Empirical strategy
- Two-part approach:
  1. Evidence of healthy aging: regress each health indicator H_{i,t} on a time trend (t) controlling for socio-economic covariates (age, gender, education, log household wealth) and country fixed effects:
     - H_{i,t} = α_0 + α_1 t + κ X_{i,t} + ε_{i,t}
     - Heterogeneity: time trends allowed to differ by (i) EMs vs AEs, (ii) gender, (iii) urban vs rural, (iv) education (primary, secondary, tertiary), and (v) household wealth quintile.
     - Standard errors clustered at the country level.
  2. Health and labor outcomes:
     - Correlations: OLS regressions of labor market indicators L_{i,t} on health indicators controlling for socio-economic covariates, country FE, and time trend:
       - L_{i,t} = β_0 + β_1 H_{i,t} + κ Z_{i,t} + β_2 t + ε_{i,t}
     - Causal effects: 2SLS using exogenous health shocks proxied by incidence of chronic diseases (share of chronic diseases out of 17) as instrument for health. Controls include lifestyle factors (low/high BMI, physical inactivity, smoking). Robustness check: augmented inverse propensity score weighting (AIPW).

### Evidence of healthy aging (key findings)
- Broad-based improvements over time across physical, cognitive, and mental health indicators: positive and significant coefficient on the time trend in Online Annex Table 2.2.1.
- Estimated year (time trend) coefficients (Online Annex Table 2.2.1):
  - Grip Strength: Year 0.004** (standard error 0.002)
  - Lung Function: Year 0.011** (0.004)
  - Cognitive PC: Year 0.020*** (0.003)
  - Overall Health: Year 0.005** (0.002)
  - Physical PC: Year 0.005*** (0.002)
  - Psych. Well-Being: Year 0.015*** (0.003)
- Socio-economic correlates (selected coefficients, Online Annex Table 2.2.1):
  - Age: Grip Strength -0.035*** (0.001); Lung Function -0.028*** (0.001); Cognitive PC -0.026*** (0.001); Overall Health -0.020*** (0.001); Physical PC -0.025*** (0.002); Psych. Well-Being -0.014*** (0.002).
  - Male: Grip Strength 1.345*** (0.025); Lung Function 0.761*** (0.03); Overall Health 0.041* (0.02); Physical PC 0.036** (0.015); Psych. Well-Being 0.045** (0.018).
  - Upper Secondary Education: Grip Strength 0.064*** (0.013); Lung Function 0.097*** (0.031); Cognitive PC 0.362*** (0.034); Overall Health 0.217*** (0.024); Physical PC 0.127*** (0.019); Psych. Well-Being 0.191*** (0.021).
  - Tertiary Education: Grip Strength 0.081*** (0.016); Lung Function 0.225*** (0.024); Cognitive PC 0.625*** (0.042); Overall Health 0.422*** (0.038); Physical PC 0.227*** (0.027); Psych. Well-Being 0.309*** (0.033).
  - (Log) Household Wealth: Grip Strength 0.019*** (0.002); Lung Function 0.017*** (0.003); Cognitive PC 0.029*** (0.004); Overall Health 0.035*** (0.005); Physical PC 0.024*** (0.004); Psych. Well-Being 0.040*** (0.003).
- Frailty has declined over birth cohorts (Online Annex Figure 2.2.3, panel 3), consistent with healthy aging.
- Cognitive-function improvements have been especially prominent.

### Heterogeneity in healthy-aging trends
- Faster improvements in cognitive health among EMs relative to AEs, suggesting partial cross-country convergence; other dimensions show similar paces across country groups.
- Estimated time trend coefficients by group (Online Annex Table 2.2.2, selected entries):
  - By Country Income Group (Year coefficients):
    - AEs: Grip Strength 0.004**; Lung Function 0.010**; Cognitive PC 0.020***; Overall Health 0.003; Physical PC 0.004***; Psych. Well-Being 0.014***.
    - EMs: Grip Strength 0.006*; Lung Function 0.022***; Cognitive PC 0.033***; Overall Health 0.020**; Physical PC 0.011; Psych. Well-Being 0.032***.
  - By Age Group (50s, 60s, 70s, 80s): Year coefficients broadly similar across age groups (e.g., Grip Strength 0.004** across 50s–80s).
  - By Gender:
    - Females: Grip Strength 0.008***; Lung Function 0.010*; Cognitive PC 0.022***; Overall Health 0.005**; Physical PC 0.004**; Psych. Well-Being 0.015***.
    - Males: Grip Strength 0.001; Lung Function 0.013**; Cognitive PC 0.018***; Overall Health 0.004; Physical PC 0.006***; Psych. Well-Being 0.016***.
  - By Location:
    - Urban: Grip Strength 0.005**; Lung Function 0.009*; Cognitive PC 0.021***; Overall Health 0.006**; Physical PC 0.006***; Psych. Well-Being 0.017***.
    - Rural: Grip Strength 0.005*; Lung Function 0.014***; Cognitive PC 0.014**; Overall Health 0.004; Physical PC 0.004*; Psych. Well-Being 0.020***.
  - By Education and Wealth Quintile: Positive year coefficients across education levels and wealth quintiles; magnitudes vary but show improvements across groups.
- Conclusion on heterogeneity: Improvements vary across countries but are broadly similar across socio-economic groups (gender, urban/rural, education, wealth). Wide socio-economic health gaps persist.

### Health and labor market outcomes
- Better old-age health is associated with:
  - Increased labor supply on the extensive margin: higher labor force participation and probability of employment.
  - Increased labor supply on the intensive margin: higher hours worked per week and weeks worked per year.
  - Higher total labor earnings and labor productivity (earnings per hour).
  - Later retirement (retiring at an older age).
- Estimation approach for causality:
  - Instrument: chronic disease instrument defined as the share of chronic diseases reported out of 17 chronic diseases (robust to alternative definitions: total number, dummy for any chronic disease, dummy for top 10 chronic diseases).
  - First-stage regressions indicate chronic diseases are a strong instrument for physical, cognitive, and mental health.
  - Identification assumption: some chronic diseases—those not explained by socio-economic factors and health behaviors—are randomly assigned.
  - Additional controls: lifestyle factors (low/high BMI, physical inactivity, smoking) included because they predict both chronic disease incidence and frailty.
  - Robustness: augmented inverse propensity score weighting (AIPW) used to correct for potential endogeneity of the instrument by assigning greater weight to more exogenous (less predictable) health shocks.

### Robustness and sensitivity
- Results robust to:
  - Controlling for the square of the age variable and for rural location (excluded from baseline due to missing data for some countries).
  - Alternative instrument definitions (number of chronic diseases, dummy indicators).
  - AIPW estimator for instrument endogeneity concerns.
- Higher-order time trends found to be either statistically insignificant or somewhat sensitive to the chosen health indicator and regression specification; estimated first-order time trends are robust.

*Source: Online Annex 2.2, World Economic Outlook, International Monetary Fund | April 2025 (ch2onlineannex - 2022. Inter-university Consortium for Political and Social Research).*

### Annex Figure 2.2.5).

### Annex Figure 2.2.5)

### Instrument validity and estimation approach
- Instrumental-variable (2SLS) approach is used to tackle potential endogeneity; estimated impacts in the second stage are substantially larger than implied by OLS regression estimates (Online Annex Tables 2.2.3 and 2.2.4).
- As an instrument, the incidence of chronic diseases:
  - Passes the weak instrument test in all regressions, with the first-stage F-statistic exceeding the Stock and Yogo (2005) rule-of-thumb critical value of 10.
  - 2SLS regressions control for socio-economic characteristics, the survey year, country fixed effects, and lifestyle factors (smoking, alcohol consumption, physical inactivity, obesity/higher body-mass-index).
- Results are robust to controlling for country-year fixed effects that capture time-varying country-level policies (for example, changes in statutory retirement age).

### Impact of chronic diseases on health indicators (first-stage coefficients)
- Note: coefficients shown are from the first stage of 2SLS regressions of health indicators (individuals ages 50 and older) on the instrumental variable—the incidence of chronic diseases—controlling for socio-economic characteristics, lifestyle factors, the survey year, and country fixed effects. Squares = point estimates; bars = 90 percent confidence intervals.
- Reported first-stage regression coefficients (as depicted in Annex Figure 2.2.5):
  - Psychological Well-Being: –3.6
  - Physical PC: –1.9
  - Overall Health: –0.2
  - Self-Reported Health: (value not separately listed beyond "Overall Health" above)
  - Cognitive PC: (value not separately listed beyond group)
  - Lung Function: (value not separately listed beyond group)
  - Grip Strength: (value not separately listed beyond group)
  - Measured Health: (group label shown)

### OLS estimates: Effect of health on labor market outcomes (selected coefficients from Online Annex Table 2.2.3)
- Estimations from OLS regressions of labor market outcomes for individuals ages 50 and older on single health indicators, controlling for socio-economic characteristics, the survey year, and country fixed effects. Standard errors clustered at country level.
- Measured Health:
  - Grip Strength: Total 0.022***; Retirement 0.026**; Employment 1.069***; Earnings/hr 0.038**; Age 0.579*** (standard errors shown in table).
  - Lung Function: Total 0.013***; Retirement 0.013**; Earnings/hr 0.038**; Age 0.197 (standard errors shown in table).
  - Cognitive PC: Retirement 0.027***; Employment 0.501*; Earnings/hr 0.052**; Age 0.746*** (standard errors shown in table).
- Self-Reported Health:
  - Overall Health: Total 0.051***; Retirement 0.026***; Employment 0.482***; Earnings/hr 0.051***; Age 0.803***.
  - Physical PC: Total 0.022**; Retirement 0.059**; Employment 0.889***; Earnings/hr 0.065***; Age 0.757***.
  - Psych. Well-Being: Total 0.011**; Retirement 0.057***; Employment 0.384**; Earnings/hr 0.062***; Age 0.501***.
- Composite Health Frailty:
  - Frailty Index: Total -0.302***; Retirement -0.248***; Employment -4.934***; Weekly hours -1.676***; Earnings -1.008***; Earnings/hr -0.730***; Age -6.644***.

### 2SLS estimates: Effect of health on labor market outcomes (selected coefficients from Online Annex Table 2.2.4)
- Estimations from two-stage-least-squares regressions of labor market outcomes for individuals ages 50 and older on single health indicators, with lifestyle controls and country fixed effects. Standard errors clustered at country level.
- Measured Health:
  - Grip Strength: Total 0.529***; Retirement 0.318***; Employment 9.840***; Weekly hours 1.153; Earnings 0.765***; Earnings/hr 0.599***; Age 5.350***.
  - Lung Function: Total 0.794***; Retirement 0.448; Employment 13.572***; Weekly hours 2.220; Earnings 2.094***; Earnings/hr 1.365**; Age 8.651***.
  - Cognitive PC: Total 0.996***; Retirement 0.919**; Employment 32.759***; Weekly hours 4.051; Earnings 1.512***; Earnings/hr 1.569*; Age 8.779***.
- Self-Reported Health:
  - Overall health: Total 0.140***; Retirement 0.043***; Employment 1.924***; Weekly hours 0.214*; Earnings 0.152***; Earnings/hr 0.097***; Age 1.348***.
  - Physical PC: Total 0.215***; Retirement 0.146***; Employment 5.367***; Weekly hours 0.449; Earnings 0.423***; Earnings/hr 0.292***; Age 1.948***.
  - Psych. Well-Being: Total 0.201***; Retirement 0.152***; Employment 4.003***; Weekly hours 0.364; Earnings 0.328***; Earnings/hr 0.192***; Age 2.145***.
- Composite Health Frailty:
  - Frailty Index: Total -1.275***; Retirement -0.464***; Employment -21.646***; Weekly hours -2.315*; Earnings -1.763***; Earnings/hr -1.160***; Age -11.806***.
- 2SLS baseline controls: Socio-Economic Controls ✓; Lifestyle Controls ✓; Country FE ✓.

### AIPW (augmented inverse propensity score weighting) robustness and chronic disease dummy effects (Online Annex Table 2.2.5)
- AIPW average treatment effects for a binary indicator of the incidence of one or more chronic diseases (top 10 chronic diseases by impact on overall health frailty):
  - Chronic Disease Dummy (Top 10) — AIPW Average Treatment Effect:
    - Total: -0.071*** (std. err. 0.008)
    - Retirement: -0.018*** (0.004)
    - LFP: -1.235*** (0.109)
    - Employment: -0.371*** (0.068)
    - Weekly hours: -0.099*** (0.020)
    - Weeks/year: -0.039** (0.021)
    - Earnings: -0.014*** (0.010)
    - Age: -0.590*** (0.067)
  - OLS estimate for the same chronic disease dummy (Top 10):
    - Total: -0.080*** (0.007)
    - Retirement: -0.030*** (0.007)
    - LFP: -0.880*** (0.307)
    - Employment: -0.256*** (0.083)
    - Weekly hours: -0.064*** (0.015)
    - Weeks/year: -0.018 (0.017)
    - Earnings: -0.045* (0.022)
    - Age: -0.416* (0.231)
- Observations: 565,640 total; 205,315 (Retirement); 198,418 (LFP); 124,139 (Employment); 93,893 (Weekly hours); 77,412 (Weeks/year); 44,460 (Earnings); 285,809 (Age).
- Controls: Socio-Economic Controls ✓; Lifestyle Controls ✓; Country FE ✓.
- Notes:
  - Treatment dummy defined as top 10 chronic diseases by impact on overall health frailty; results robust when considering all 17 chronic diseases or top 10 by impact on self-rated overall health.
  - Overlap assumption satisfied (for all observable characteristic values there exist treated and control observations).
  - Similar OLS estimates for the direct effect of the chronic disease dummy on labor market outcomes provide additional comfort on instrument exogeneity.

### 2SLS estimates: Effect of a 1-decade health improvement (Online Annex Table 2.2.6)
- Estimations from 2SLS regressions where health indicators are rescaled to represent effects of a 1-decade improvement in health (health measures divided by 10 times the coefficient on the survey year from Annex Table 2.2.1). Controls include socio-economic characteristics, lifestyle factors, survey year, and country fixed effects. Standard errors clustered at country level.
- Measured Health (1-decade improvements):
  - Grip Strength: Total 0.023***; Retirement 0.014***; Employment 0.436***; Earnings/hr 0.027***; Age 0.237***.
  - Lung Function: Total 0.090***; Retirement 0.051; Employment 1.540***; Earnings/hr 0.155**; Age 0.982***.
  - Cognitive PC: Total 0.203***; Retirement 0.187**; Employment 6.668***; Earnings/hr 0.319*; Age 1.787***.
- Self-Reported Health (1-decade improvements):
  - Overall health: Total 0.006***; Retirement 0.002***; Employment 0.088***; Weekly hours 0.010*; Earnings/hr 0.004***; Age 0.062***.
  - Physical PC: Total 0.011***; Retirement 0.007***; Employment 0.264***; Earnings/hr 0.014***; Age 0.096***.
  - Psych. Well-Being: Total 0.031***; Retirement 0.023***; Employment 0.618***; Earnings/hr 0.030***; Age 0.331***.
- Composite Health Frailty (1-decade improvement):
  - Frailty Index: Total 0.034***; Retirement 0.012***; Employment 0.576***; Weekly hours 0.062*; Earnings 0.047***; Earnings/hr 0.031***; Age 0.314***.
- Controls: Socio-Economic Controls ✓; Lifestyle Controls ✓; Country FE ✓.

### Heterogeneity by age, job characteristics, and policy implications
- Age heterogeneity:
  - 2SLS results vary across age groups: labor market effects generally affect the extensive margin of labor supply most for individuals in their 50s, followed by those in their 60s and 70s, with much lower effects for those aged 80 and above (e.g., Online Annex Figure 2.2.6).
  - Online Annex Figure 2.2.6 reports coefficients (rescaled to reflect the estimated impact of ‘healthy aging’ over 10 years) for cognitive health on labor force participation by age groups; plotted range: 0.0 to 0.5 on x-axis with bars for 50–59, 60–69, 70–79, 80–89, 50+.
- Job characteristics and age-friendliness:
  - As people age, they become less willing to work in jobs considered less “age friendly” (jobs with lower job satisfaction, physically demanding tasks, higher stress, lack of freedom, poor job security).
  - Better health generally has a stronger positive effect on labor market outcomes for jobs considered more age-friendly—suggesting a higher voluntary response of labor supply for these occupations.
  - Online Annex Figure 2.2.7 shows shares (percent) of workers reporting undesirable job characteristics by age groups (50–59, 60–69, 70–79, 80+) and education groups (< Upper Secondary, Upper Secondary, Tertiary).

### Additional methodological notes and broader modeling context
- AIPW estimation used as a robustness check to purge remaining endogeneity bias; delivers qualitatively similar implications of negative health shocks (reductions in labor supply and wage remuneration).
- Baseline model context (Online Annex 2.3):
  - The chapter uses an extended overlapping-generations (OLG) model (Auclert and others (2024) basis), incorporating country-specific demographic pyramids and projections (predetermined fertility, mortality, and migration assumptions) and country- and age-specific labor supply profiles.
  - Individuals value smoothing consumption over lifetime, including in retirement, and leaving bequests; policies such as retirement age, labor taxes, and social security spending influence savings and allow assessment of policies promoting healthy aging, extending retirement age, and enhancing labor force participation.
  - Limitations: model does not capture potential structural transformation (e.g., relative growth of services and associated aggregate productivity changes), abstracts from endogenously responding technological progress, and is geared toward low-frequency dynamics rather than cyclical or high-frequency adjustment costs.

_Italic: Source — ch2onlineannex - Annex Figure 2.2.5). (PDF chapter/section from the IMF WORLD ECONOMIC OUTLOOK, April 2025)_

### CHAPTER 2  THE RISE OF THE SILVER ECONOMY: GLOBAL IMPLICATIONS OF POPULATION AGING

CHAPTER 2  THE RISE OF THE SILVER ECONOMY: GLOBAL IMPLICATIONS OF POPULATION AGING

### Healthy aging: assumptions and quantified impacts
- Age groups affected: 50–59, 60–69, and 70–79; other age groups assumed unaffected.
- Cognitive health improvements over 2000–22 are assumed to continue but at a declining geometric rate.
- Assumed health improvements over 2023–2100 are equivalent to around one-fourth of the gains observed during 2000–22.
- Combined effect on effective labor supply per older worker:
  - Effective labor supply per older worker would rise by 36 percent during 2017–2100.
  - Gains during 2023–2100 amount to one-fourth of the gains observed during 2000–22.
- In the “healthy aging policies” scenario:
  - Country-specific cognitive health gaps vis-à-vis the frontier are reduced by one-fourth by 2100.
  - The additional improvement in cognitive capacity of older individuals (long run, average across countries) is equivalent to about 49 percent of the estimated gains over 2000–22.
  - The corresponding improvement in effective labor supply per old worker is equivalent to about 59 percent of the estimated gains over 2000–22.
  - Policies affecting retirement age are assumed unchanged in this scenario.

### Productivity, convergence, and the age–innovation channel
- Long-run calibration:
  - Model converges to an age-specific balanced growth path with TFP calibrated to 2 percent per year.
- Two country-specific forces shaping TFP during 2016–2100:
  - Convergence towards frontier productivity level.
  - Impact of demographics on productivity growth via innovation and entrepreneurship channels (age-innovation channel).
- Panel regression used to calibrate effects (variables described in text). Key estimated coefficients (Online Annex Table 2.3.2):
  - Growth of 25–45 Age Group: 0.0069*** (column 1), 0.0074** (column 2)
  - Growth of 25–45 Age Group (Lagged): -0.0056** (column 1), -0.0057* (column 2)
  - Distance to Frontier: -0.0414*** (col 1), -0.0487*** (col 2), -0.0461*** (col 3), -0.0509*** (col 4)
  - Distance to Frontier * Growth of 25–45 Age Group: 0.0055** (col 1), 0.0039 (col 2)
  - Constant terms shown in table: -0.0061***, -0.0075***, -0.0090***, -0.0101***
  - Number of Observations: 1,144; 1,137; 729; 726
  - R2 values: 0.026, 0.033, 0.025, 0.031
- Main inference from model calibration and Online Annex Figure 2.3.2:
  - The age-innovation channel plays a relatively minor role in affecting average GDP growth over 2025–2100.
  - The productivity-convergence channel plays a more prevalent role in contributing to baseline average annual GDP growth over 2025–2100.

### Global capital markets and interest rate wedge
- Integration assumption: China, India, and the LIC bloc face imperfect integration into global capital markets via a time-varying wedge between domestic and global interest rates.
- Initial interest rate wedge set to 300 basis points in the model.
- Empirical context: mean return differential between EMs and AEs is approximately 330 basis points (Gerding, Henriksen, and Simonovska 2025).
- Wedge dynamics: assumed to decline gradually and dissipate by 2070 as reforms and governance improvements proceed.

### Fiscal policy calibration and baseline fiscal mechanics
- Labor taxes, retirement replacement rates, and effective retirement ages calibrated to match data targets following Auclert and others (2024).
- Baseline effective retirement age assumption:
  - Increase by one month per year over 60 years in all countries, except India and the LIC bloc (where effective retirement ages remain unchanged because life expectancy at retirement is below 15 years).
- Debt-to-GDP:
  - Assumed to evolve as projected in the October 2024 World Economic Outlook up to 2029 and to remain stable thereafter.
- Government adjustment to meet debt targets:
  - Each year, governments adjust a mix of policy instruments so that one-third of the necessary budget adjustment falls on each of: labor tax rate, transfer spending (replacement rates), and other government spending.

### Policy scenarios evaluated (design and quantitative assumptions)
- Three policy levers assessed separately and jointly:
  1. Tackling decline in participation rates of older but pre-retirement individuals (through healthy aging policies and other measures).
  2. Extending working lives in line with rising life expectancy (higher effective retirement age scenario).
  3. Narrowing gender participation gaps (closing labor force participation gaps scenario).
- Common design feature:
  - All model economies engage in policy changes simultaneously, but magnitude and pace depend on country-specific room for improvement.

- Higher effective retirement age scenario:
  - Rule: At year t, if life expectancy at retirement in t-1 exceeds 20 years, effective retirement age at t is increased by 3 months; otherwise evolves as in baseline.
  - Result: For most countries, longer working lives than baseline; for some (China, Canada, USA, UK, LIC bloc) effective retirement age remains same as baseline for several decades.
  - Median increase implication: scenario results in a median increase in the effective retirement age of 2 years over the next 20 years.
  - Historical comparator: actual median increase in effective retirement age over 2002–22 was about 2.6 years.
  - Model simplification: replacement rates assumed unchanged in this scenario.

- Closing labor force participation gaps scenario:
  - Country-specific female labor force participation rate assumed to increase so that initial gender participation gaps are reduced by three-fourths by 2040.
  - Data source: ILO modelled estimates series.
  - Median projected rises by 2040:
    - 4.7 percentage points for women age 25–54.
    - 10.6 percentage points for women age 55–64.
  - Historical comparator over 2000–24:
    - Actual median increases were 7.1 percentage points for age 25–54 and 30.8 percentage points for age 55–64.
  - Country-specific impacts: India experiences largest gains; in India reducing the gender gap by three-fourths among 55–64 yields a 37 percent increase in effective labor supply in that age group, while an equivalent gap reduction in 25–54 yields a 33 percent increase.
  - Across all model countries, narrowing gender participation gaps generates the largest boost in female labor supply among those aged 55–64.

- Combined labor policies scenario:
  - Combines: closing health gap with frontier, increasing effective retirement age per life expectancy rule, and enhancing female labor participation.
  - Outcomes:
    - Could significantly boost global growth over the next 25 years for most countries (Online Annex Figure 2.3.3).
    - Countries with small initial health disparities and gender participation gaps derive lower gains.
    - Fiscal dividends from combined policies create additional fiscal space that can be used to lower labor tax rates, increase income replacement for retirees, or increase other spending—model assumes equal allocation across these three uses to maintain constant debt-to-GDP ratio going forward.

### Fiscal space, alternative uses, and longer-term implications
- Computation approach:
  - Average additional fiscal space over a period computed as average difference in (labor tax revenues + transfers + other government spending) net of baseline, as specified in formula (pol and base denote policy and baseline; T = labor tax revenues; Tr = transfers to retirees; G = other government spending).
- Allocation rule in main scenarios:
  - Fiscal dividends allocated equally: one-third to labor tax rate reduction, one-third to transfer spending, one-third to other government spending, such that debt-to-GDP remains constant.
- Alternative exercise (Online Annex Figure 2.3.4):
  - Computes fiscal space over 2025–40 when fiscal dividends are used to rebuild fiscal buffers by gradually reducing debt-to-GDP over 2030–40 toward its 2016–18 average.
  - Findings:
    - For many economies (e.g., Greece, Japan, Spain), fiscal gains from reforms would be sufficient to rebuild fiscal buffers, with extra gains remaining for additional spending.
    - Others (China, United Kingdom, United States) would require additional fiscal efforts (further primary balance consolidation) to return debt-to-GDP to pre-pandemic levels.
- Longer-run perspective (Online Annex Figure 2.3.5):
  - Average additional fiscal space over 2025–2100 due to combined labor policies is larger when countries actively reduce debt levels over 2030–40.
  - Using fiscal dividends to rebuild buffers lowers average r–g over 2025–2100 for all model economies relative to the scenario where debt-to-GDP is held constant.

### Additional modelling note: migration exercise (Box 2.1)
- Scenario: doubling of annual migration flows of young individuals (age 20 in the model) from the LIC bloc towards aging advanced economies over the long run.
- Path: additional annual flow increases gradually, reaching long-run level by 2040.
- Magnitude: by 2040 the additional migrants would represent about one percent of the LIC bloc’s population aged 20–24.
- Productivity and wealth assumptions:
  - Additional migrants are young with lower productivity than prime-aged workers.
  - For simplicity, these young migrants are assumed to have zero net worth in the model.

*Source: International Monetary Fund, CHAPTER 2 — THE RISE OF THE SILVER ECONOMY: GLOBAL IMPLICATIONS OF POPULATION AGING (Online Annex material).*

### 1. Additional Fiscal Space, Alternative DebtProjection

### 1. Additional Fiscal Space, Alternative DebtProjection

### Sensitivity to elasticity of intertemporal substitution (EIS)
- Baseline calibration assumes EIS = 0.5.
- Counterfactual recalibration sets EIS = 1 while matching the same targets as the baseline.
- With EIS = 1, consumption and savings respond more strongly to changes in interest rates.
- A given change in GDP growth (g) leads to a smaller change in the real interest rate (r) when EIS is higher.
- Consequently, r − g in the combined policy scenario is lower for every country in the model when EIS = 1 than in the baseline calibration.
- The sensitivity of fiscal space to the elasticity of intertemporal substitution is relatively small.

### Evidence from Online Annex Figure 2.3.6 (Combined-policy scenario, average over 2025–2100)
- Panel 1: Reports the average fiscal gains over 2025–2100 under the combined-policy scenario due to higher effective labor supply and improved old-age dependency ratio relative to the baseline.
  - The magnitude of gains varies over the transition; the figure reports the average gain over 2025–2100.
  - Blue bars represent differences under the baseline calibration (EIS = 0.5).
  - Red diamonds represent differences when EIS = 1.
- Panel 2: Shows r − g under the combined-policy scenario, average over 2025–2100, with deviations from the baseline scenario reported in percentage points.
- Values for “World” denote averages for the economies included in the model.
- Data labels in the figure use International Organization for Standardization (ISO) country codes.
- Notation preserved: g = GDP growth rate; r = interest rate; EIS = Elasticity of intertemporal substitution; LIC = bloc of low-income countries.

*Source: IMF staff calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/weo/2025/april/english/ch2onlineannex.pdf_