As Bowes and Loomis (1980) noted in their contribution, for recreation demand functions of a linear form, the existence of unequal zonal populations in travel cost models leads to the problem of nonconstant error variances when observations are expressed on a per capita basis, i.e., when the zone average method is used. For a model with nonhomogeneous variances, the ordinary least squares (OLS) parameter estimates will be unbiased but inefficient. These properties will hold true of the consumer surplus estimates as well. A weighting of the observations is the well-known solution to the heteroskedasticity problem, where in this case the appropriate weighting factor for each observation is the square root of the associated zone of origin's population. While preserving unbiasedness, the weighted least squares (WLS) approach generates parameter estimates that are equivalent to the generalized least squares (GLS) minimum variance estimators. In their application of the travel cost method to recreation trips during 1977 down the Westwater Canyon in Utah, Bowes and Loomis illustrated the efficacy of WLS in predicting recreation trips more accurately than OLS for a linear first-stage demand curve. As in previous studies, the OLS linear regression equation grossly mispredicted the total number of trips taken at the actual price per trip of zero, and seemingly overestimated the value of consumer surplus. Further evidence of this latter finding was obtained in a study of warm water fishing in Georgia by Ziemer, Musser, and