Concern over the effects of public policies based on misspecified econometric models motivates interest in a procedure to test, diagnose, and improve the specification of models that have been estimated with three-stage least squares. A test of system-wide specification based on Hausman's specification test is employed in a test of the a priori restrictions placed on the parameters of a structural model of the U.S. shrimp market. The null hypothesis of proper specification is rejected. After diagnosis via a comparison of unrestricted and restricted reduced forms and respecification, the null hypothesis cannot be rejected.