AbstractMonte Carlo evidence indicates that translog cost and profit models are generally incapable of accurately characterizing the underlying production technology. In order to discriminate between tracking difficulties caused by truncation of important high‐order terms and errors in variable bias, Monte Carlo experiments are performed with secondand third‐order cost and production models, using data generated with and without measurement error. The results suggest that it is errors in variables bias and not the translog specification, which is largely responsible for the poor tracking performances noted previously.