This research tests the interchangeability of two specifications of travel cost demand models for recreation at U.S. Army Corps of Engineer reservoirs in Arkansas, California, and Tennessee/Kentucky. Statistical tests of coefficient equality for both nonlinear least squares and Heckman sample selection models suggest rejecting a transferable model among all three regions. However, the nonlinear least squares models in Arkansas and Tennessee were similar enough to fail to reject the hypothesis of equal coefficients at the 0.01 significance level. Even so, interchanging the Arkansas and Tennessee nonlinear least squares coefficients produces visitor use and total benefit estimates that are more than 100% too high. However, interchanging coefficients does provide reasonably close estimates of the average consumer surplus per trip for both states using the nonlinear least squares model (¿5% to 10%). This isdue to similarity of the price coefficients in the two models. Thus a more limited form of transferability which focuses on average benefit per day, rather than on predicting total use and total benefits, appears promising. ¿ American Geophysical Union 1995