The estimation of infant mortality from the proportions dying among births in the past 24 months

Censuses and demographic surveys in developing countries frequently include questions of adult women as to the date of their most recent live birth, and whether or not the child is still alive. Using a mathematical model of mortality in infancy and childhood, we show that the proportion dying among children born in the 24 months prior to the census or survey, multiplied by a factor of 1.09, gives a close approximation to the infant mortality rate. Applications to data from Indonesia and Kenya give results which are in good agreement with those from other sources. The method has the advantage that it provides up-to-date estimates which are robust to the selectivity bias inherent in birth history data for countries with a high prevalence of HIV. But it is vulnerable to reporting errors and non-response.