Adult mortality estimations from cohort and census/survey data: A comparison of direct and indirect methods in rural Malawi
Patrick Gerland, United Nations Population Division
James Kaphuka, University of Malawi
George Mandere, University of Malawi
Humphreys Misiri, University of Malawi
Peter C. Fleming, University of Pennsylvania
In countries lacking reliable vital registration, knowledge of adult mortality depends largely on censuses/surveys information about recent deaths in the households or specific relatives. While this is useful to assess mortality levels and trends, comparison with DSS data suggests frequent under-reporting of adult deaths. Existing datasets, however, do not allow systematic comparison of these differences because information is collected only for one set of deaths (either spouse, parents, siblings, household members) and cohort data are unavailable to evaluate the accuracy of these estimates. This paper aims to fill this gap by analyzing a longitudinal household survey conducted in 1998-2006 in rural Malawi. Using cohort data, we use survival analysis to compute regional life tables against which we evaluate more conventional adult mortality estimates based on (a) household deaths in last 12 months, (b) widowhood, (c) survival of parents, (d) survival of siblings as reported in 2006 through the household questionnaire.