A Cox nonnested test with parametric bootstrap is developed to select between the linearized version of the First Difference Almost Ideal Demand System (FDAIDS) and the Rotterdam model. The Cox test with parametric bootstrap is expected to be more powerful than the various orthodox tests used in past research. The new approach is then used for U. S. meat demand (beef, pork, and chicken) and compared to results obtained with an orthodox test. The orthodox test gives inconsistent results. In contrast, under the same varied conditions, the Cox test with parametric bootstrap consistently indicates that the Rotterdam model is preferred to the FDAIDS.