Modeling the choice among a discrete set of recreation alternatives is driven by the possible patterns of substitution among sites. Researchers typically assume that the individual's choice set is the same as the set of alternatives included in the recreation survey instrument. Our endogenous choice set model is a generalization of the standard multinomial logit random utility model which allows for the possibility of heterogeneous choice sets that are endogenously determined in the model. Using two examples, we show that parameter and compensating Variation estimates differ greatly between the endogenous choice set model and the traditional logit model. (C) 1997 Academic Press.