We analyse data from sample surveys of water use and prices for households in 17 cities in Central America and Venezuela. In some of the cities, almost all sampled households have tap water, but in others many rely on nontap ('coping') sources. Coping households use less than one-fifth as much water as metered tap households do, face average water prices ten times as high, are much poorer, and face substantial water hauling costs. Water demand functions are estimated for metered tap households and coping households separately and jointly. Increasing block rates complicate estimation on metered tap households. Using 2SLS, we find price elasticities of water demand, with respect to both average and marginal water price, of about -0.3, with average price giving the greater partial effect. Coping demand has price elasticity closer to -0.1, and is also negatively affected by increased hauling costs. Estimations on the joint data indicate that the water connection itself explains most of the difference between tap and coping consumption and indicate serious problems in such data pooling.