Adult Mortality from Sibling Survival Data: Does the Corrected Method Perform Better?

Bruno Masquelier, Princeton University

Estimates based on sibling survival data have long been regarded as implausibly low, but they are nowadays viewed as highly informative. This recent shift in the literature on indirect estimation of adult mortality is mainly due to the work of Gakidou and King (2006). Their paper paved the way for the correction of survival biases in sibling histories. The weighting scheme they suggest has since been applied to DHS surveys, and this yields much higher estimates than previous attempts based on the same data (Obermeyer et al. 2010, Rajaratnam et al. 2010). In this paper, we cast doubt upon the application of this weighting scheme when the characteristics of the DHS surveys are not properly taken into account. We use micro-simulation and DHS data to demonstrate that the "Corrected Sibling Survival" method might be substantially overestimating mortality, especially among males.