Irrigation water shortage is becoming an increasingly serious problem in agricultural production. In this case, it is very important for policy makers to take measures to improve irrigation water use efficiency, especially in the water-scarce areas. In this paper, the data envelopment analysis (DEA) techniques, based on the concept of input-specific technical efficiency were used to develop farm-level technical efficiency measures and sub-vector efficiencies for irrigation water use. The Tobit regression technique was then adopted to identify the factors that influence irrigation water efficiency differentials under the shortage of water resources. Based on a sample data of 432 wheat farmers in northwestern China, our experimental results of the DEA analysis showed the average technical efficiency of 0.6151. It suggested that wheat farmers could increase their production by as much as 38.49% by using inputs more efficiently. Further, the mean irrigation water efficiency of 0.3065, suggested that wheat farmers could produce the same quantity of wheat using the same quantity of inputs but with 69.35% less water. The results of the Tobit regression analysis showed that the farmer’s age, income, education level, and the farm size tended to affect the degree of irrigation water efficiency positively, and the channel conditions and different irrigation methods made a significant impact on irrigation water use efficiency. Furthermore, the arrangements of exclusive water property rights and competitive water price mechanism have effectively encouraged the water saving behavior of farmers. These results are valuable for policy makers since it could help to guide policies towards high irrigation water use efficiency. Irrigation water shortage is becoming an increasingly serious problem in agricultural production. In this case, it is very important for for policy makers to take measures to improve irrigation water use efficiency, especially in the water-scarce areas. In this paper, the data envelopment analysis (DEA) techniques, based on the concept of input-specific technical efficiency were used to develop farm-level technical efficiency measures and sub-vector efficiencies for irrigation water use. The Tobit regression technique was then adopted identify the factors that that irrigation irrigation water efficiency differentials under the shortage of water resources. Based on a sample data of 432 wheat farmers in northwestern China, our experimental results of the DEA analysis showed the average technical efficiency of 0.6151. It suggested that wheat farmers could increase their production by as much as 38.49% by using inputs more efficiently. Further, the mean irrigation water efficiency of 0.3065, sugg ested that wheat farmers could produce the same quantity of wheat using the same quantity of inputs but with 69.35% less water. The results of the Tobit regression analysis showed that the farmer’s age, income, education level, and the farm sizes tended to affect the degree of irrigation water efficiency positively, and the channel conditions and different irrigation methods made a significant impact on irrigation water use efficiency. Furthermore, the arrangements of exclusive water property rights and competitive water price mechanism have substantially encouraged the water saving behavior of farmers. These results are valuable for policy makers since it could help to guide policies towards high irrigation water use efficiency.