Death Distribution Methods for Estimating Adult Mortality: Sensitivity Analysis with Simulated Data Errors, Revisited

Rob Dorrington, University of Cape Town
Ian M Timaeus, London School of Hygiene and Tropical Medicine

Hill and Choi attempted to test how well the variable-r techniques for estimating adult mortality (Generalised Growth Balance and the Synthetic Extinct Generations methods) perform in the context where data deficiencies violate the assumptions underlying the models. From this work they concluded that each method had strengths and weaknesses and that superior estimates could be derived using a combination of the two methods. However, the version of the SEG method they tested did not account for differential coverage of the censuses. This paper tests the more general version of the model on the same set of scenarios and finds that the SEG method performs so well as to call into question the recommendation that one should use a combination of the two methods. In addition the paper explores the generalisation of these results to scenarios that represent mortality conditions in an African country with a generalised HIV/AIDS epidemic.