Nested logit has become common in estimating random utility models of recreation demand. Because welfare analysis is often the objective of estimating these models, it seems natural to ask, what effect does the choice of nesting structure have on the welfare estimates generated by these models? Therefore, we compare the results of nine nesting structures andfindthat the variation in welfare estimates across the models is not large. Our results are contrary to those of Kling and Thomson and Shaw and Ozog. The difference appears to originate with differences in the estimated dissimilarity coefficients in the nested models. Copyright 2000, Oxford University Press.