Benefit transfer entails using estimates of non-market values derived at one site as approximations to benefits at other sites. The method finds favour because it can be applied quickly and cheaply, however the validity of benefit transfer is frequently questioned. Published studies generally indicate that errors from the approach can be extremely large and could result in significant resource misallocations. Assessing the validity of benefit transfer is complicated by differences in the nature of study and policy sites, the changes being valued, valuation methods, time of study, availability of substitutes and complements, and demographic, social and cultural differences. A choice experiment was used to evaluate the transferability of benefit estimates for identical goods between two different populations. The study design allowed most of the confounding factors to be controlled, so provides a strong test of benefit transfer validity. Several different tests were applied to evaluate benefit transfer validity, with conflicting results. The paper investigates the merits of the alternative tests and concludes that utility functions were different for the two populations.