Estimating Life Expectancy at Old Ages

Nan Li, University of Victoria
Zheng Wu, University of Victoria

Official mortality statistics often report an open-ended interval for ages older than A (e.g., A = 85). The expectation of life for that age, e(A), is estimated as the reciprocal of the death rate at ages over A, 1/m(A+), under the assumption that the population in the open interval is stationary, which cannot be justified by either theoretical or empirical reasons. On the other hand, mortality decline in industrial societies provides opportunities to model mortality at old ages, based on observation at ages younger. In this paper, we suggest use these models to estimate e(A). We show that these models produce similar estimates of e(A) that may differ remarkably from 1/m(A+), the textbook formula for e(A). We conclude that our suggestion of estimating e(A) is more realistic because it is empirically based and does not require the stationary assumption.

Presented in Session 13: Innovative Applications to Enhance the Use of Secondary Data