Constrained Mortality Extrapolation to Old Age: An Empirical Assessment

Area of Science:

• Demography
• Actuarial Science
• Biostatistics

Background:

• Accurate mortality estimates at advanced ages are essential for actuarial and demographic studies.
• Data quality issues, particularly age exaggeration, often compromise the reliability of mortality estimates in old age.
• Parametric extrapolation models are commonly used but can be prone to bias without proper constraints.

Purpose of the Study:

• To enhance the accuracy of parametric extrapolations for death rates at old age.
• To investigate the impact of constraining extrapolation models with life expectancy estimates.
• To address challenges posed by insufficient data quality and age exaggeration in mortality data.

Main Methods:

• Utilized period data from the Human Mortality Database.
• Applied Horiuchi-Coale and Mitra formulas to reduce bias in open-interval life expectancy.
• Constrained parametric extrapolation models using empirical, Horiuchi-Coale, and Mitra life expectancy estimates.

Main Results:

• Extrapolation accuracy was substantially improved when constrained by empirical, Horiuchi-Coale, or Mitra life expectancy estimates.
• Unconstrained extrapolations and those using conventional life table estimates showed significant biases.
• Constrained extrapolations demonstrated greater robustness against age exaggeration effects.

Conclusions:

• Constraining parametric death rate extrapolation models with robust life expectancy estimates significantly improves accuracy at old age.
• Methods incorporating life expectancy estimates resilient to age exaggeration are vital for reliable mortality estimation.
• Avoid unconstrained extrapolations and those relying on conventional life table estimates for open-age intervals due to inherent biases.