The rice production of Bangladesh has been investigated using a Cobb-Douglas stochastic frontier production function which incorporated a model for the technical inefficiency effects. The farm level primary data which is used for this study has been collected by stratified random sampling technique. The per hectare production, cost, gross, net returns and benefit cost ratio were the highest for Boro rice. The factors identified in the stochastic production frontiers which are responsible for the increase of Aus rice production are irrigation cost, land under production, experience and education. For Aman rice, fertiliser, manure, land under production and education were important variables for the increase in production. For increasing the production of Boro rice, fertiliser, manure, ploughing cost, irrigation cost, insecticide cost and land under production were found to be important variables. Cobb-Douglas stochastic production frontiers included the farm-specific factors such as age, education, experience, family size and land under house. Experienced farmers tend to have smaller inefficiencies than younger and less experienced farmers. There were significant technical inefficiency effects in the production of Boro rice. As a policy, Aus and Aman rice crops production cannot be increased by increasing efficiency with existing technology. In this case a new advanced technology is needed to increase production. But for Boro rice, about 14% of production can be increased by increasing the technical efficiency only.