Discussion Paper no. 99-5


The maternal orphanhood technique is one of the oldest indirect demographic techniques used for the estimation of adult mortality in populations with inadequate vital statistics. One of the pioneers of the technique had cautioned for the careful assessment of the maternal orphanhood data. All researchers have not heeded this caution and as a result, improbable adult mortality estimates could not be adequately explained. In this paper, the maternal orphanhood data from the 1996 census have been subjected to a battery of diagnostic tests. The diagnostics attempted in the paper have shown that among the factors that affect the estimates of maternal orphanhood, the one with the least effect on the mortality estimates is age misreporting. The age analyses show very good age reporting for all population groups and all provinces. The next factor with less effect on the mortality estimates is that of ‘adoption effect’. In this case it is hard to distinguish ‘adoption effect’ from ‘AIDS orphans’ effect. This adoption/AIDS orphans effect is more pronounced in Eastern Cape and Northern Cape. The two factors which show large effect in the trend in mortality estimates is the age distribution of maternal orphans and the proportion of those who do not know or did not state their maternal orphanhood status (DK and NS). The paper has shown that while the maternal orphanhood data is good, unique factors related to South African history have contributed to raising the proportion of DK and NS in the data. It is concluded that more research is needed on the role of the maternal orphanhood technique in an era of increasing deaths due to AIDS in some population subgroups and in cases of increasing ageing and low mortality in other population subgroups. It is hoped that the incomplete gamma function would be found useful in the modelling efforts in this direction.