The common prescription for dealing with limited supplies of water is to improve irrigation efficiency, that is, to reduce gross delivery to farm fields while maintaining full crop production. The public and some policy makers continue to assume that the water thus saved constitutes a new supply that may be applied to other uses. Scientists and hydrologists have long understood that the non-consumed fraction of applied water often becomes the source for another human or ecosystem purpose after leaving the field, and irrigation improvements interrupt these uses. Thoughtful researchers have provided valuable guidance in conceptual frameworks and analysis procedures to address this issue. Researchers have also noted empirically that total consumptive use often increases when efficiency improves, and have cited case-specific reasons that this occurs. This paper shows it is a general case arising from rational producer behavior in equating the marginal cost of a production input (irrigation water) with its marginal benefit. At any marginal cost of water, improving irrigation efficiency enables the irrigator to be willing and able to purchase a quantity of irrigation water that sustains more consumptive use than was possible with the prior, less-efficient system. An equation for economic demand for water is presented where impacts to both crop yield and commodity price are endogenously determined from simple input data. It is applied to an irrigated area in Idaho, USA where it indicates that improving efficiency from 60 percent to 80 percent reduces field delivery of irrigation water by 15 percent but increases consumptive use by three percent.