The different model specifications have been proposed to estimate time-varying technical efficiency in the last two decades. These models have resulted different efficiency scores based on their built-in specifications. This study investigates the temporal pattern of technical efficiency based on Cornwell, Sickles, and Schmidt (CSS, 1990) as well as Lee and Schmidt (LS, 1993) models. In order to extend LS model, the assumption of identical temporal pattern for each individual is replaced by group-specific stochastic frontier model with a parametric function of temporal pattern. As empirical evidence, the observed data for the thirty three Cement planst in Iran during 1996-2005 are used. The cement plants have been divided into three different categories: More than 3000 tones per year , between 1000 to 3000 tones per year , and finally less than 1000 tones per year . The results indicate that significant differences between individual-specific (CSS and LS) models and group-specific (GS) model exist while the latter model has more sufficient explanatory specification. Since CSS and LS models specify the technical efficiency in aggregate forms, so the average efficiency scores in both models are estimeated as 0.76 and 0.84, respectively. While in GS model there are three different efficiency scores relevant to the above-mentioned three groups which are 0.81, 0.93, and 0.75, respectively. The similar outcome can be observed for input elasticities as well as scale elasticity.
JEL Classifications: C23, C51, D21, L61