Determination of technical efficiency in cotton production by using data envelopment and stochastic frontier analysis methods: a case study of Hatay Province in Turkey

The main aim of this study is to determine the technical efficiency of cotton production in the province of Hatay, which is one of the major cotton production areas of Turkey. The data of the study were gathered from 136 cotton enterprises by the face-to-face survey method, and the Random Stratified Sampling method was used in determining the sample size. In the data analysis; the Data Envelopment Analysis (DEA), (Variable Return to Scale - Output Oriented), and Stochastic Frontier Analysis (SFA) methods, which are commonly being used in technical efficiency measurements of agricultural enterprises, were used. The determinants of technical efficiency were calculated by the Tobit Regression Analysis (for DEA) and TE Effects Model (for SFA). In the analysis, cotton yield per decare was considered as output; and seed usage (kg da(-1)), pure nitrate usage (kg da(-1)), pure phosphorus usage (kg da(-1)f), labor force (h da(-1)), machinery power (h da(-1)), pesticide cost (USD da(-1)) and the number of irrigation were considered as inputs. According to the analysis results, the technical efficiency score average of cotton enterprises was found as 0,82 with the DEA-VRS method and was found as 0,86 with the SFA method. Besides, it was found that the enterprises which produce cotton at the minimum efficiency level, could increase the cotton production amount by 78% according to the DEA-VRS, and by 0,72% according to the SFA method. The scale efficiency average was calculated as 0,97. It was concluded that the main reason for high scale efficiency was derived from a false input combination; nonetheless, it was determined that there were enterprises which weren't utilising scale efficiency. In the study, it was observed that some inputs were overused, such as the number of irrigation (36,79%), fertiliser-N (17,88%), and pesticide cost (8,22%). Hence, cotton yield could be increased with proper input combinations. Also, producer training activities about input usage levels and methods could be useful in order to increase awareness about the issue.