As life expectancy (LE) continues to increase, inequalities in life expectancy of people with disparate education levels will likely widen. The population-level LE is decreasingly informative as the average LE of sub-groups diverge.
Robust forecast and informed interpretation of LE must take cohort-specific characteristics and variances into account. The Lee-Carter model (Lee, Carter, 1992) is the most popular model for forecasting mortality rates and LE.
The model specified mortality rate as a function of three parameters– age-specific constraints, time-varying index, and interaction terms between time and age.
Noticing the drawback that applying this model separately for related populations, like neighbouring countries, leads to forecasts diverging significantly in the long-run, Li and Lee developed the Li-Lee model (Li, Lee, 2005).
The central idea behind the Li–Lee model is that related populations share a common time trend in the long run, but that there may be population-specific deviations in the short run. (Baal et al, 2016).
Baal et al built on the Li-Lee model by extending it to two subgroups–sex and education, and applying it to Dutch mortality and education level data from 1973-2012. The educational attainment segments were: low: primary education; middle: pre-vocational education; high: secondary and tertiary education.
In Baal’s model, there is a common time trend shared by all groups, a time trend shared by all educational classes within a sex, and a time trend for each educational class by sex leading to 9 total time-trend parameters.
The model focuses on remaining LE at age 65. As seen in the table below, the LE at age 65 increased more for men than women between 1996 and 2012.
Below are Baal’s forecasts of LE in the Netherlands at age 65 in 2042 with the 95% prediction interval in parentheses.
Below are the forecast differences in LE at age 65. Women are forecast an LE 2.7 years more than men. High educated men are forecast an LE 4.7 more than Low educated men. High educated women are forecast an LE 4.5 years more than low educated women.
Baal et al also visualise their forecast of LE and LE differences across cohorts in time-series plots as reproduced below. There is a wide variation in the cohort difference of predicted life expectancies.
The results of our forecasts indicate diverging trends of mortality between the higher and lower educated subgroups, which are larger for men than women. (Baal et al, 2016)
This model isn’t without its drawbacks. It doesn’t take potential cohort-compositional changes into account. The age-time interactions are constant in the model unreflective of any changes that may occur in the future. The forecasting model also doesn’t consider any education-specific trends.
Despite a general improvement in medical treatments, large LE differentials across socioeconomic status groups have persisted, indicative of fundamental attributes that drive these differences.
Very informative article!