Six New Models for the Expectation of Life at Old Age: A Comparative Review

Dalkhat M. Ediev , North-Caucasian Academy; Lomonosov Moscow State University; International Institute for Applied Systems Analysis;

Age exaggeration, other deficiencies of population statistics at old age as well as small population numbers preclude extending mortality analysis to advanced old age in many countries. A typical ‘solution’ to this problem is to close the life table at relatively young open age interval, so that most of the age exaggeration-related redistributions are confined within the open age. Apart from limiting mortality analysis at old age, abridging the life table at oldest-old ages leads to another problem: the classical life table relation for the life expectancy in the open age interval turns biased when the population age composition deviates from that of the stationary population. Horiuchi-Coale and Mitra have proposed their models to correct for the biasedness of the traditional life table relation. Yet, accuracy of these models has substantially declined as life expectancies grew and populations departed from stable population age composition assumed in Horiuchi-Coale and Mitra models. The Mitra model, while more accurate, turns instable when tested on empirical data. To further advance the methodology of measuring life expectancy, we propose six new models: two assuming population stability; two models assuming non-stable population the age structure of which is reconstructed by sorts of backward projection; a regression model relating life expectancy to the death rate at single age; and behavioral model that explicitly accounts for the mechanism of age exaggeration. The models proposed substantially advance the methodology of estimating life expectancy at old age and may find wide range of practical applications.

See extended abstract

 Presented in Session 15. Mortality Models