Using the Lee-Carter Method to Explain Linear Increases in Life Expectancy

Shripad Tuljapurkar, Stanford University
Nan Li, Max Planck Institute for Demographic Research

Recently, life expectancy rising approximately linearly has been observed in many low-mortality nations, we find that this phenomena is a conditional output of k(t) changing linearly in the Lee-Carter method, which characterizes stable mortality declines across all ages. The conditions for such an output to occur are low mortality at young ages, and at old ages the Gompertz law works and mortality decline at rates that are constant across age and over time. Using data of the G7 countries, we illustrate that these conditions roughly stand since 1950s, and show that the changes of k(t)s are as linear as that of life expectancy. In general, we do not recommend using life expectancy as the predictor to forecast mortality, because it does not provide age-specific death rates, it could hardly be more linear than k(t), and to be as linear as k(t) it requires three conditions that may not stand widely.