Estimating willingess to pay for urban water supply: A comparison of artificial neural networks and multiple regression analysis

Abstract

A study was carried out to estimate willingness to pay (WTP) for urban water supply, for three types of consumer typically found in urban areas in developing countries-those receiving water through: existing connections; stand posts; and water wells. Two forecasting techniques were applied: artificial neural networks (ANN), a state-of-the-art technique, and multiple regression (MR), a conventional method. A comparison of the forecasting accuracy of the two models was made, using the mean absolute percentage error. The forecasting error of the best ANN model was found to be about half that of the best MR model. The MR models showed that the significant variables determining WTP differ by type of consumer, and from those variables that are typically considered significant for WTP by urban consumers in developed countries.