This paper uses Ghanaian household survey data to examine the link between environmental factors and the long-term health status of children in Western Ghana by exploiting the variation induced by a mining-based cyanide spill into a major water body. Information in the survey about the region and cohort of birth gives the primary source of identi cation. Howev r, to get an improved spatial identi cation of exposed
children, I link GPS data to the survey data. A rich set of controls are included to allow for a better resolution of omitted variable bias, by accounting for variables that can potentially confound the impact estimate if left unidentied. I also examine quantile treatment eects using a exible speci cation, which allows for uniquely incorporating the GPS information. Findings reveal that after controlling for birth region and cohort, household, maternal and environmental factors, children born during the shock in the WR are negatively aected with reduced height, and that this negative eect persists through all the above (baseline and alternative) speci cations. This eect initially reduces in magnitude, as we begin adding controls, but stabilizes across specications once we include a larger set of control variables. A discussion motivated by the appropriate clustering of standard errors of the shock impact coecient is presented