Avoiding selection bias without random assignment? The effect of breastfeeding on cognitive outcomes in China.

Objectives Recent studies suggest that some of the health benefits traditionally attributed to breastfeeding may be overstated due to selection bias problems, since typically breastfeeding is more prevalent in more resourceful households. In this paper we argue that an important antidote against selection bias consists in analyzing data from contexts in which breastfeeding is not normativised, or where it is not systematically associated to socioeconomic advantage and best parenting practices. We estimate the effect of breastfeeding on the cognitive outcomes of Chinese children. Methods We use observational data from a representative sample of children aged 10–15 from the China Family Panel Study. We estimate province fixed effect linear regression models to predict the impact of breastfeeding on test scores. Results Our paper shows that there are no advantages in the results obtained by breastfed children when compared with those who were not breastfed, regardless of the duration of breastfeeding. Conclusions Some of the cognitive benefits associated to breastfeeding can be attributed to flaws inherent to the research designs usually adopted. [ABSTRACT FROM AUTHOR]