Welfare economics is often employed to measure the impact of economic policies or externalities. When demand is characterized by discrete choices, static models of consumer demand are employed for this type of analysis because of the difficulty in estimating dynamic discrete choice models. In this paper we provide a tractable approach to estimating dynamic discrete choice models of the Generalized Extreme Value (GEV) family that addresses many of the problems identified in the literature and provides a rich set of parameters describing dynamic choice. We apply this model to the case of recreational fishing site choice, comparing dynamic to static versions. In natural resource damage assessment cases, static discrete choice models of recreational site choice are often employed to calculate welfare measures, which will be biased if the underlying preferences are actually dynamic in nature. In our empirical case study we find that the dynamic model provides a richer behavioral model of site choice, and reflects the actual choices very well. We also find significant differences between static and dynamic welfare measures. However, we find that the dynamic model raises several concerns about the specification of the policy impact and the subsequent welfare measurement that are not raised in static cases.