In this paper we use a fixed–random coefficients regression model to analyse data for cereal growing small-scale farmers in efficiency. Results show that land size is a major constraint and only small changes in cultivated area and land quality yield relatively high increments to output. Larger farms are relatively less productive, everything else being equal. Human capital in the form of literacy and experience are found to affect productivity positively. Our findings show a high degree of farm-specific technical inefficiency. With regard to the inputs we find high degrees of input-specific technical inefficiency, especially so for labour and fertilizer. The age structure of the household, environmental factors and education are found to be weakly correlated with efficiency. Furthermore, sharecropping is found to be positively correlated with technical efficiency.