Estimating Adult Mortality through Surveys: An Unbiased Method Using Data on Survival of Relatives

Emmanuela E. Gakidou, Harvard University
Gary King, Harvard University

Complete vital registration systems are still far off for most low-income developing countries despite the fact that reliable baseline measures of adult mortality are needed for key programs on the adult health agenda. Estimates derived from sibling survival histories in household surveys have not been used widely. Currently available methods provide unbiased mortality rates under 2 assumptions: 1) that there is independence in probabilities of death within families; and 2) that probability of death is not related to family size. The bias introduced by violating these assumptions on true mortality rates has not been explored in the literature. This paper introduces a new method for the measurement of adult mortality through data on sibling survival. In a simulation environment we compare the two methods and explore the magnitude and direction of the bias in the estimates when the assumptions are violated.