The purpose of this paper is two-fold. First, it compares the performance of various panel data models in estimating technical efficiency in production. Second, it applies various stochastic frontier panel data models to estimate the technical efficiency of Iran's cotton production and to provide empirical evidence on the sources of technical inefficiency of cotton producing provinces. The results indicate that labor and seeds are determinants of cotton production and inorganic fertilizers result in reducing technical efficiency. The mean technical efficiency of the models is around 80 percent. Variations in the distribution of estimated efficiency amongst the different models is large.