Child mortality estimation: An assessment of summary birth history methods using microsimulation

Background: Two types of indirect methods for estimating child mortality rates from summary birth histories (number of children ever born and children dead) are currently available to users: model-based methods derived from the pioneering work of Brass and empirically based methods developed more recently at the Institute for Health Metrics and Evaluation (IHME). Objective: The objective of this study is to evaluate the performance of six alternative indirect methods based on summary birth histories. Methods: Using microsimulation based on the 1950-2010 fertility and mortality rates of the United Nations' World Population Prospects, estimates generated by six alternative indirect methods were compared against benchmark direct estimates for 82 countries. Results: The results show that the IHME methods outperform the classical Brass method. In particular, the cohort-derived method is able to produce robust past child mortality trends across a variety of demographic regimes. However, no method produces robust recent estimates prior to data collection. When data are classified by time since first birth, methods perform better than with maternal age classification. Conclusions: This study suggests that the maternal age variant of the IHME cohort-derived method is the best option for estimating child mortality from past censuses. For future censuses, it would be worthwhile adding an extra question on date of first birth. Contribution: This study provides new recommendations on which method to use depending on the data available, as well as for future population census collection.