causal relationships as accurately as possible by eliminating potential reverse causality arising from the fact that previously depression-prone mothers may be less likely to decide to breastfeed, or to breastfeed for shorter durations. After conducting this analysis for the whole sample, we split the sample into mothers who were, and who were not, depressed during pregnancy; for each group, we examine differences in outcomes between women who had planned to breastfeed, and women who had not.