Estimating the average age of infant deaths: a flexible model life table based on newly collected data

Julio E. Romero Prieto, London School of Hygiene and Tropical Medicine (LSHTM)
Andrea Verhulst, French Institute for Demographic Studies (INED)
Michel Guillot, University of Pennsylvania

The average age of infant deaths 1a0 is a fundamental component of a life table. Since this quantity is not always observed, different formulas have been proposed to estimate its value from other mortality indicators. However, some formulas are outdated, and the new formulas overlook the age patterns of mortality at early ages. This paper describes a method for estimating 1a0, using the predicted mortality schedule of a flexible two-dimensional Model Life Table. The model was calibrated from the Under-5 Mortality Database, a newly collected source of national distributions of deaths by detailed ages. This model offers three advantages over the existing approaches. First, it allows substantial improvement in the estimation of 1a0 using unconventional entry points such as the neonatal mortality or the proportion of deaths corresponding to the first months of life. Second, the model introduces a total flexibility in the choice of entry points, allowing the use of either deaths rates or probabilities of dying for a variety of conventional and unconventional age groups. Third, the model provides similar flexibility and better results than the other approaches for the estimation of the average number of years lived from 1 to 5 by those dying during childhood 4a1.

Keywords: Mathematical demography, Mortality, Children and youth, Structural equation modelling

See paper.

  Presented in Session 169. Creating and Using International or Historical Datasets