This forecasting technique assumes that population is the only factor influencing water demand. The author describes a mathematical model that has been applied to forecast water demands by customer class in San Francisco, California, to the year 2005. The model considers other variables in addition to population such as number of people per household, employment, marginal price of water and sewer service, temperature, precipitation. The forecasts by customer class are compared to a forecast of overall water use in San Francisco. The article concludes that the technique of regression analysis is appropriate for forecasting water demands. Includes 2 references, tables, figures.