Evidences of the effects of unobserved heterogeneity in micro-econometric models are now pervasive in many applied economics fields. This article investigates this issue for agricultural production choice models. Farms’ and farmers’ unobserved heterogeneity can be accounted for in micro-econometric agricultural production choice models by relying on available modeling and inference tools. The random parameter (RP) framework allows achieving this goal in a fairly flexible way. This modeling framework has already been successfully used in numerous empirical studies covering many topics. It simply considers RP versions of standard models. Extensions of the Expectation-Maximization algorithms have been specifically developed in the computational statistics literature for estimating RP models. They appear to be well suited for large statistical models such as micro-econometric agricultural production choice models. The estimation of a RP multi-crop econometric model shows that unobserved heterogeneity matters in a sample of French farmers specialized in cash grain production covering a relatively small geographical area. The key parameters of this RP model significantly vary across farms. Simulation results obtained from the estimated RP model confirm that the sampled farmers’ choices respond heterogeneously to homogenous economic incentives. Ignoring this heterogeneity impacts both the distribution and the magnitude of the simulated effects.