A CROSS-SECTIONAL REGRESSION-ANALYSIS OF RESIDENTIAL WATER DEMAND IN MELBOURNE, AUSTRALIA

Data from the 1981 census, a householder survey and meter readings taken by the Melbourne and Metropolitan Board of Works are analysed using multiple regression methods as part of an investigation of the residential demand for water in Melbourne, Australia. Water consumption data are from a sample of 6788 households in 183 clusters, and questionnaire data are from 845 of these. Various cross-sectional models are fitted to a randomly selected data subset, with yearly household water consumption as the dependent variable. For conceptual purposes, the best model is a linear regression equation incorporating six variables with an R2 value of 0.54 and a standard error of 35.2 per cent. The best practical equation has only two variables, net annual property value and number of residents per household. Regression statistics for this equation are R2 = 0.45 and SE = 38.3 per cent; testing on the previously unused data subset validates the model.