Household decisions and child health : Estimating the links between water treatment and the incidence of diarrhea using non-recursive two-equation causal models

The treatment of drinking water is advocated to reduce the incidence of child diarrhea. However, evaluating the impact of water treatment with only observational data leads to biased estimates since it could be the occurrence of child diarrhea that induced the household to treat their drinking water. To deal with the possible simultaneity between the treatment of drinking water and the incidence of child diarrhea, we specify non-recursive two-equation causal models and apply it on a sub-sample of households with children below five years old from the Philippine National Demographic and Health Surveys. In the treatment effects model, we find that the treatment of drinking water reduces by 5.2 percentage points the proportion of under-5 children afflicted with diarrhea. In the instrumental-variable probit model, we find that households have a higher propensity to sterilize their drinking water by 1.2 percentage points given a one percentage point increase in the proportion of under-5 children with diarrhea. Ignoring the simultaneity yields the misleading result that water treatment increases the incidence child diarrhea. These results underscore the need to insure the quality of drinking water at the point of use and not just at the point of source.