A mathematical programming model of rural KwaZulu, excluding three northern districts, was developed to simulate agricultural production in regions of high and low cropping potential. This model aggregates enterprise levels predicted for four representative households of which two are in the high‐potential region and two in the low‐potential region.Several economic scenarios are simulated with the model to predict, inter alia, responses to changes in cereal prices, input subsidies, changes in off‐farm employment and a rental market for crop land. Long‐run food supply response is estimated to be inelastic with respect to changes in product and input prices. Higher food prices harm the majority of rural households. Lower input prices have little impact on household welfare. Conversely, it is predicted that a rental market for arable land could have a large positive impact on crop production and would improve household