A mathematical programming approach is developed for deriving estimates of the willingness to pay of water customers for improvements in water supply reliability. Reliability is represented as a probability distribution of different shortage levels, allowing the valuation of different profiles of water supply reliability. The approach is examined analytically, and a two-stage linear programming variant is developed for applied problems. The approach can be applied to estimate the willingness to pay for improved reliability of different classes of customers and for suggesting promising water conservation programs for different customer classes. An example application is presented to illustrate the approach. ¿ American Geophysical Union 1995