Estimating Levels, Trends, and Determinants of Child Mortality in Countries with Poor Statistics

Abstract

The estimation of levels, trends, and determinants of infant and child mortality in statistically poor countries has received intense attention from demographers during the past 15 years. The literature on this subject is vast, so our review is focused on the most commonly applicable methods of analysis. We approach this task as demographers, asking the types of questions demographers typically ask. It will prove convenient first to consider methods and data sources for estimating levels and trends and next to explore how these can be adapted to measure determinants. Although many pages of methodological development follow, we emphasize that the ultimate goal of our work is entirely practical. By concentrating on techniques of analysis, we provide tools for addressing such questions as why mortality has fallen in the past and what policy interventions are most cost-efficient for reducing mortality in the future. We close with a discussion of strategies for analyzing specific intervention programs. Preliminary methodological considerations Because death rates vary by age, indexes of mortality that do not control for the effects of age composition of a population can be seriously misleading. For example, in 1984 the crude death rate in the United States is 9 per thousand population, while that in Mexico is only 6 per thousand despite the fact that death rates are lower at every age in the United States. The reason is that the population of Mexico is concentrated in the younger age groups, in which mortality is lower. Although one could work directly with age-specific death rates, demographers find it more convenient to convert them into a life table, which shows the proportion of persons surviving to exact ages 1, 2, 3, etc. In this paper, we focus on techniques for estimating life table proba