Using maximum likelihood estimation techniques, the stochastic production frontier is employed to estimate technical efficiency at the plot level by ownership types of water amongst a cross section of sugar cane growing farmers using primary survey data. Inefficiency effects are modelled as a function of farmer specific explanatory variables. Tests reveal that the null hypothesis of no inefficiency and no influence of farmer specific variables on inefficiency can be rejected. Education, land area, discharge of tubewell and distance of plots from the water source are the causes identified in explaining inefficiency. Estimated technical efficiency scores are highest on plots where water is sourced from a privately owned tubewell, followed by plots serviced by partnered tubewells and lowest on plots where water is bought. Income gains from improved efficiency follow the reverse patterns with the largest gains of Rs. 1082 per bigha estimated for buyers' plots and Rs. 649 per bigha for plots with their own tubewell with the average of Rs. 867 for all plots.