An elusive restriction maintained in earlier CET supply models with three or more products is shown to result in a potentially serious misspecification. Its impact on empirical estimates is found to be substantial, and an alternative formulation is presented which overcomes the problem while still maintaining the CET hypothesis.
