This paper examines the impact of direct income payments on productive efficiency of Korean rice farms, using farm-level cross sectional data in 2006. For representation of efficiency and its determinants, this paper uses a model that estimates the deviations of farms from a translog distance function and the determinants of these deviations. This paper especially estimates a stochastic frontier production function to explain deviations from best-practice productivity with a two-part error term including statistical noise from measurement error and technical inefficiency arising from farms not reaching the production frontier boundary. The empirical evidence finds that farms that get a higher share of direct payments in farm revenue are less efficient than others. This inefficiency is reduced by increases in farm size. Another result indicates that farms received greater direct payments on aggregate are more efficient than other farms since fixed payment, one part of rice direct payment, is tied to the amount of a farm's cropland that has been enrolled in programs, as well as yield histories.