Using household-level panel data from South African households and a unique set of highly aggregated control variables, we estimate household demand for water via a Stone-Geary specification. We also used quantile regression to establish the impact of the different socio-economic variables on different categories of water consumption levels. The main policy recommendation would involve a revision of the water pricing schedule in order to improve access to safe drinking water for lower-income households, to an amount of water identified as enough to cover their basic needs (i.e. 11 506 litres per household per month).