Learning From Neighbors: Social Learning About Child Feeding During Diarrheal Episodes

Population-level behavior change requires information and new ideas that spread beyond the confines of targeted beneficiaries of programmatic interventions. Using Demographic and Health Surveys (DHS) data from two countries (Bolivia and Madagascar), we examine mothers’ ability to correctly respond to the question regarding whether a child with diarrhea should be given more fluids, the same amount of fluids, or less fluids. Controlling for usual correlates—including women’s education, media exposure, number of children, age, and household economic status—we find that the fraction of a woman’s neighbors (that is, those in the same sampling cluster) that answer correctly substantially raises the likelihood that the woman herself will also respond correctly. With a single cross section it is difficult to determine whether this represents direct social learning or some other cluster specific effect, for example, that all neighbors were exposed to the same conditions. However, there are two pieces of evidence that strongly suggest direct social learning as an explanation. First, the impact is much smaller and (in Bolivia) insignificant in urban areas. Second, controlling for all other cluster characteristics—including average education and wealth—does not drive out the direct cluster less household knowledge term. Broad-based impact depends in part upon development professionals’ ability to understand and encourage knowledge diffusion and social learning about optimal health behaviors.