Siblings Count: The Extent and Effect of Double Counting in the Sibling Survival Method Using the Demographic and Health Surveys

Megan E. Costa, University of Pennsylvania
Ian M. Timaeus, London School of Hygiene and Tropical Medicine (LSHTM)

Many low-resource countries lack nationally representative vital registration, necessitating use of cross-sectional surveys to generate mortality estimates. Developments in adult mortality estimation include improvements to the sibling survival method, where respondents report survival status of siblings. The Demographic and Health Surveys (DHS) interview multiple household members. This sampling strategy introduces potential sibling record duplication. This study quantifies the bias introduced by age structure of excess respondents. A hierarchical algorithm was developed to identify multiple sibling respondents within households. The proportion of excess respondents and sibling observations is determined. Mortality rates will be calculated with and without excess respondents. In a preliminary analysis of seven DHS, excess respondents ranged from 2.9% in Burkina Faso to 9.5% in Haiti. Five year mortality rates in Zambia from ages 15 to 35 ranged from 3.72% higher to -2.11% when excluding duplicates. This procedure will be applied to additional DHS to quantify bias through duplication.

  See extended abstract

Presented in Poster Session 7