Three years later: Environmental health impacts of a community-demand-driven water and sanitation program in rural Maharashtra, India

To date, few rigorous impact evaluations demonstrate that water and sanitation (W&S) policies effectively improve health. This paucity of rigorous impact evaluations in the sector may reflect challenges inherent to W&S, including the sector's broad and varied services, decentralized service delivery, and the tendency to focus on engineering and fiscal outputs in monitoring and evaluation studies. We address these challenges in an evaluation of a large-scale, community-demand driven W&S project implemented in western India by applying a longitudinal, quasi-experimental research design with three principle research features. First, careful selection of comparison groups and sufficient sample size (N=10,010 households in 242 villages) permit the measurement of relative contributions of multiple project components (e.g., water supply, sanitation, information and education campaign) on multiple project outcomes. Second, pre- and post-intervention data on communities, households, individuals, and program implementation, collected using a variety of qualitative and quantitative methods, permit estimation of our impact measures. Third, pre-intervention propensity score matching of intervention villages with observationally equivalent control villages (from a universe of 8000 villages), based on 30 village level variables, addresses a major form of selection bias. Furthermore, difference-in-difference estimation removes any residual time-invariant selection bias.

Three years after project initiation, we find that the intervention increased the number of households using improved water sources (6%), private latrines (6%), and households engaging in safe water handling (5% increase in use of narrow mouth containers). We find weak evidence of differential improvements in child diarrhea in project villages, and given the relatively small gains in health, we observe few impacts on education (measured by children's school attendance). We conclude by highlighting the utility of quasi-experimental designs for evaluations of global environmental health projects.