Conference Agenda

Access to safe and reliable drinking water is essential for well-being, yet in many rural settings, water infrastructure is frequently unreliable. This paper examines the household-level consequences of well failure using a novel measure of predicted well functionality in Ethiopia. Leveraging a machine learning model trained on hydro-environmental variables, I construct an exogenous proxy for well failures and link it to detailed household survey data. I document a series of behavioral responses to those water access disruptions. Households adjust by switching away from groundwater toward more proximate, but potentially lower-quality sources such as surface water. Consistent with this shift, reported time spent collecting water declines, especially among adults. However, this reduction masks a reallocation of responsibility: more children are mobilized to fetch water, and this increased burden has downstream effects. Children in affected households are significantly more likely to miss school and are increasingly engaged in farming and casual work. These findings reveal the cascading effects of water access disruptions on household labor dynamics and children's time use, underscoring how environmental shocks can shape human capital accumulation and economic behavior.