In developing countries, untreated sewage exposes people to alarming water pollution levels, yet there is limited knowledge about the effectiveness of wastewater treatment investments. I leverage the national inventory of sewage treatment plants in India and various granular datasets on river water quality measures, as well as geo-localized information on child births and deaths, to identify robust effects of wastewater treatment installations. To do so, I use estimators robust to staggered adoption within a difference-indifferences design and compare urban areas that started wastewater treatment from 2010 onwards and urban areas where such treatment was planned or under construction in 2020. I show that after starting wastewater treatment, levels of fecal coliforms decreased by 50%, and downstream mortality under the age of six months declined by 20%. A back-of the-envelope calculation suggests that starting wastewater treatment earlier-from 2010 onwards-in urban areas later selected into treatment-after 2020-would have prevented over 40,000 child deaths in downstream sub-basins.