A CONCEPTUALLY-BASED MATHEMATICAL MODEL OF HUMAN NATIONAL LIFE TABLES

Abstract

Using standard procedures of demographic methodology, analysts working with mortality data are faced with a choice between large, unwieldy arrays of age-specific rates (or equivalent sets of life table entries), or one or more of a summary measures set, such as life expectancies and standardized rates, which does not retain all of the available information. This dissertation describes the development and preliminary testing of a mathematical model derived from elementary considerations of mortality mechanisms in the life table population. The model as developed postulates a Gompertz specification to account for mortality rates increasing with age among adults. Also, a proportion of the population was posited to be subject to a competing constant risk, to account for the declining mortality rates in early childhood. The motivation for this model is that its parameters, estimated for particular populations via nonlinear regression procedures, might be used as more efficient mortality summaries than those routinely used, without loss of conceptual interpretability. In testing life tables for male and female populations of 47 selected nations during the 1960s, the model was shown to be substantially more efficient for reproducing the original life tables than were any of the traditional measures considered.