Selective mortality and undernutrition in low- and middle-income countries

Anthropometric indicators, in particular the height for a particular age, are found to be lowest in South Asia compared to other geopolitical regions. However, despite the close relationship between undernutrition and mortality rates, the highest mortality rates are concentrated in sub-Saharan Africa. By accounting for this survival bias, i.e. selective mortality, this discrepancy between the undernutrition rates between South Asia and sub-Saharan Africa should be expected to decrease. In addition, one can also ask whether undernutrition rates would differ without selective mortality. Using data for children below 3 years which stems from six waves of Demographic and Health Surveys (DHS), we assess the impact of selective mortality on the anthropometric indicators for the children's height-for-age (stunting), weight-for-age (underweight), and weight-for-height (wasting) for a global sample of low- and middle-income countries between 1991 and 2015. Taking advantage of a matching approach, the effect of selective mortality for a cross-section of 35 lowand middle-income countries is analysed. This approach allows values, originally stemming from non-deceased children, to be assigned for the otherwise non-observed anthropometric indicators of deceased children. These values are imputed under the counterfactual scenario that these deceased children would still be alive. The results are twofold: First, this approach reveals that the imputed values for deceased children for stunting, underweight, and wasting are significantly lower compared to the observed anthropometric indicators. Second, the difference between the observed anthropometric indicators, and the constructed overall anthropometric indicators are found to be only of negligible magnitude. Only assuming high mortality rates, or imputing the lower bound considered by the WHO as cut-offs for outliers, would alter the second finding.