Revealed preference methods for estimating welfare changes have been the subject of few reliability tests. In this paper, we utilize a unique panel data set of Iowa households to test the stability of revealed preference recreational demand models. We find that welfare estimates of changes in water quality are not temporally stable over the full range of years in our data, while welfare estimates associated with a site closure are often, but not always, more reliable. Our results highlight important factors for policy makers and practitioners to consider when transferring welfare estimates of recreational demand studies over time. These findings emphasize the need for extensive temporal reliability tests of other revealed preference approaches.