The objective of this paper is to measure productive efficiency of irrigation water efficiency based on the concept of technical efficiency and compared among different sizes farms in Tunisia. The proposed methodology is applied to a randomly selected sample of 144 citrus growing farms and differentiated by size (small, medium and large farms). A stochastic production frontier approach, based on Battese and Coelli’s (1995) inefficiency effect model, is used to obtain farm-specific estimates of technical and irrigation water efficiency. The last step of the analysis consists on the identification of the factors influencing irrigation water efficiency differentials across citrus growing farms. Empirical results show that estimated mean technical efficiency ranges from a minimum of 12.82% to a maximum of 90.69% with an average estimate of 67.73%. This result means that 32.3% increase in production is possible with the present state of technology and unchanged input uses, if technical inefficiency is completely removed. Thus, improving technical efficiency will result to significant increases in framer’s revenue and profit. On the other hand, mean irrigation water efficiency is found to be 53%, which is much lower than technical efficiency and also exhibits greater variability ranging from 1.6% to 98.87%. Estimated mean irrigation water efficiency implies that the observed quantity of marketable citrus could have been maintained by using the observed values of other inputs while using 47.0% less of irrigation water. This means that farmer’s can achieve significant savings in water use by improving irrigation system technologies.