Separating the active and the aging in the Active Aging Index

Demographic approach

• Age-specific indicators: Like some AAI dimensions (labor force, 1)
• Synthetic indicators: Like remanining life expectancy at age 55 (4.1)
• Direct standardization: Apply age-specific proportions or rates to a standard population.

Best solution

But requiring age-specific measurement.

REGRESSION APPROACH (ours)

• Hypothesis: Rates and proportions depend on age.
• Consequence: Aggregate indices depend on age composition (aging)
• Regression on aging indicators estimates the effect of age composition on the indicator.
• Based on such estimates, an adjusted AAI indicator can be obtained referring to a standard aging level.

It only requires aggregate data

Interpretation different from demographic approach.

• Mean weekly exercise depends on age and sex.
• Since women 55+ are older than men 55+ due to lower mortality, age-composition exaggerates the differences between men and women in the aggregate index.
• Age-profile provides richer information.
• Demographic methods would work: Same standard.
• Regression approach:
  • Complex profile: For males increasing activity until age 70, more time available to pensioners
  • Several aging indicators needed.

\( I_{it} = f(Z_i,Aging_{it},X_{it}) \)

• \( I_{it} \) - Indicator for time \( t \) and region-sex \( i \)

• \( Z_i \) - Variables that do not change over time.

  • Controlled for using fixed-effects

• \( \text{Aging}_{it} \) - Vector of aging related variables

• \( X_{it} \) - Other non-aging related variables changing over time.

Different from demographic approach.

• Mechanistic effect of age-composition due to underlying rates depending on age \( I_{ita} \)

Other reasons why aging might matter:

• Examples:

  • Direct effects of aging: E.g: competition in labor markets, political impact through elections, …
  • Social adaptation to aging

Separate fixed-effect regressions for each AAI indicator (\( 22 \times 3 \)) for the period 2010-2018 (5 time periods, EU-28 countries)

\[ I_{it}=\alpha + \text{Aging}_{it}' \beta + \mu_{i}+ \varepsilon_{it} \]

• Fixed-effects \( \mu_i \) control for unchanging country characteristics (eg: age-friendly culture)
• \( \text{Aging}_{it} \) include proportions 55-64, 65+, and average age of population 65+ (from Eurostat)
• \( \varepsilon_{it} \) error term

• The aging vector adapts to complex situations such as different effects for different age-groups.
• Aging coefficients will capture both mechanistic effects (for indicators that do not control for age), and direct and indirect effects of aging (operating through other time-varying determinants that are correlated with trends in aging indicators)

• For most indicators, but not for every indicator. the null of no effect of aging variables can be rejected at 5% significance level.
• Indicators where the null of no effect is not rejected include:
  • 2.1 (Voluntary activities)
  • 3.1 (Physical exercise) for women and both sexes, 3.2 (Access to health services), 3.8 (Lifelong learning)
  • 4.5 (Social connectedness)
  • Indicators already standardized by age (4.2, share of healthy life expectancy at age 55)
  • Two of the four employment rates: 1.3, 65-69, 1.4, 70-74.

• Employment indicators are age-specific: No need for demographic adjustment.
• Those for younger workers, 55-59 and 60-64 are partly explained by aging variables.
• Economic logic:
  • Increasing proportions of the population in the 55-64 and 65+ are connected with higher participation (less proportion in younger ages)
  • Smaller effects at older ages: Labor market institutions more influential than aging.
  • Coefficients tend to be positive: Suggests succesful adaptation to aging of labor markets.

• All statistically significative coefficients are positive
• Contrary to expected indicators do not get worse as populations age.
• Whenever they are connected to aging, as societies have aged, active aging indicators have improved.
• Explanations:
  • Succesful adaptation.
  • Indirect effects connected to third variables associated with aging indicators.
  • Cohort survival, health and education levels improving?

• Positive coefficients for aging variables
• No need for demographic adjustment: synthetic indicator
• This is a case of joint determination:
  • More survival leads to higher proportions of elderly and higher mean age.
  • More survival means higher life expectancy at ages 65.
• In this sense, more aging is a success story in itself.

Country-specific aging corrections at the dimension level.

• Aging factors generally increase over time in a relatively homogeneous manner across countries: active aging indicators get higher as countries age
• Aging corrections are more important for females than for males.
• Aging corrections are particularly important for dimension 1 (employment) and 2 (participation in society).