Sub-Saharan Africa is the region in the world with the highest population growth and child mortality. By measuring to what extent the parents "replace" a child in the case of death, we estimate the direct child replacement effect in Niger between 1976 and 2011. Our novel empirical strategy is a modified 2SPS instrumental variable approach, which exploits the exogeneity of precipitation and outdoor temperature in connection with the mother's pregnancy and the child's first five years. We find an estimate of 0.9. The estimate found close to unity implicates a relatively small trade-off between reducing child mortality and population growth, both being large problems in Niger. Our approach is a contribution to the literature on quasi-experimental methods in that we consider properties of respondents on an individual level in the first stage model, and redefine the outcome to a household level in the second stage. The redefinition is thereby a way of drawing conclusions on a group level.