Assessing the Forecasting Performance of a Generic Bio‐Economic Farm Model Calibrated With Two Different PMP Variants

Using linear programming in bio‐economic farm modelling often results in overspecialised model solutions. The positive mathematical programming (PMP) approach guarantees exact calibration to base year data but the forecasting capacity of the model is affected by necessary but arbitrary assumptions imposed during calibration. In this article, a new PMP variant is presented which is based on less arbitrary assumptions that, from a theoretical point of view, are closer to the actual decision making of the farmer. The PMP variant is evaluated according to the predictions of the bio‐economic farm model, developed within the framework for integrated assessment of agricultural systems in Europe (SEAMLESS). The forecasting capacity of the model calibrated with the standard PMP approach and the alternative PMP variant, respectively, is tested in ex‐post experiments for the arable farm types of Flevoland (the Netherlands) and Midi‐Pyrenees (France). The results of the ex‐post experiments, in which we try to simulate farm responses in 2003 using a model calibrated to 1999 data, show that the alternative PMP variant improves the forecasting capacity of the model in all tested cases.