Method for reconstructing mortality by educational groups

The lack of classification by educational attainment in death and population exposure data at older ages is an important constraint for studying changes and patterns of mortality disparities by education in Denmark and Sweden. The missing educational distribution of population also restricts analyses aiming at estimating contributions of compositional change to the improvements in national longevity. This study proposes a transparent approach to solve the two methodological issues allowing to obtain robust education-specific mortality estimates and population weights.

Using nonparametric approach, we redistribute the unknown cases and extrapolate the mortality curves of these sub-populations with the help of population-level data on an aggregate level from the Human Mortality Database.

We present reconstructed and harmonized education-specific abridged and complete life tables for Sweden and Denmark covering 5-year-long periods from 1991-1995 to 2011-2015. The newly estimated life tables are in good agreement with the national life tables and show plausible age- and education-specific patterns. The observed changes in life expectancy by education suggest about the widening longevity gap between the highest and lowest educated for males and females in both countries.

The proposed simple and transparent method can be applied in similar country-specific cases showing large proportions of missing education or other socio-economic characteristics at older ages.