This research seeks to produce more robust, less biased estimates from the two-constraint recreation demand model by applying globally flexible estimation techniques to travel cost data for California whale watching trips. While locally flexible functional forms do improve upon previously used restrictive forms imposed for estimation of travel cost models, specification errors occur unless the chosen form happens to coincide with the true unknown underlying form. A globally flexible functional form can consistently approximate the true function and its derivatives for all points in the sample range. This paper seeks to reduce specification error and improve accuracy of estimates from the two-constraint recreation demand model by using a globally flexible functional form. The empirical model is based on a construction by Chalfant (1987), which combines Deaton and Muellbauer's AIDS model with the Fourier flexible form of Gallant. The resulting functional form preserves the aggregation properties of the PIGLOG class of preferences while approximating the true function within an arbitrary degree of precision. A comparison of model estimation results shows that the locally flexible AIDS model results in specification error. Further research of an extension of the model, which combines travel cost data with contingent valuation responses to hypothetical population enhancements, is briefly discussed.