On the Use of Kannisto Model for Mortality Trajectory Modelling at Very Old Ages

Hoang Khanh Linh Dang, Alma Mater Studiorum - Università di Bologna
Carlo Giovanni Camarda, Institut National d'Études Démographiques (INED)
Nadine Ouellette, Département de Démographie, Université de Montréal
France Meslé, Institut National d'Études Démographiques (INED)

The Kannisto model is one of the most widely used parametric models for old-age mortality estimation. The model finds extensive use in life tables calculations, such as in the Human Mortality Database for old-age mortality smoothing, and in pension and annuity policy planning. However, recent concerns have arisen regarding the accuracy of this model for modeling mortality patterns at advanced ages Feehan (2018). We explore the question further using thorough age-validated mortality data for France, Belgium and Quebec (Canada), where a large number of extinct birth cohorts were followed up to age 115. Our comparison of the Kannisto model's performance with other frequently used mathematical mortality models showed a systematic underestimation of death rates for the Kannisto model, which becomes apparent as early as circa age 100 in all populations studied, for both males and females. This systematic underestimation in mortality would translate in a systematic overestimation of life expectancy that becomes noticeable after age 100. Caution is thereby advised when using the Kannisto model to estimate mortality at very old ages.

Keywords: Mortality and Longevity, Data and Methods

See extended abstract.