Data from a survey questionnaire and from water utility billing records are used to compare self-reported and actual water savings for 471 households during a conservation campaign. Self-reports are only weakly related to actual changes in water consumption. Errors are widespread, and not wholly random: The accuracy of self-reports increases with household socioeconomic status and with the extent of conservation behavior. The large and nonrandom error component makes self-reports questionable as a proxy for objective measures of overall water savings in conservation research. Because knowledge about water use is both generally low and related to conservation behavior, informational feedback may be a particularly effective strategy for increasing conservation. The effectiveness of this feedback may increase with social class, however.
