Repeated logit models are among the most commonly applied methods for modeling seasonal recreation demand. In this article we examine the capabilities of the repeated nested logit and repeated mixed logit models to capture patterns of error correlation and demand substitution. Particular attention is paid to the use of the mixed logit framework to generalize the strong assumptions on correlation patterns across sites and choice occasions imbedded in the nested logit model. We examine the implications for the range of price elasticities allowed in both models based on the implied correlation structures.