Variable Input Allocation Among Crops: A Random Parameter Approach With Stochastic Constraints Using SAEM Algorithm

Résumé En

We propose a new approach to estimate variable input uses allocation per crop based on aggregate input use at the farm level. Using a Random Parameter (RP) approach, we propose a model that accounts for the unobserved heterogeneity between farm. Our approach also allows accounting for the potential dependency between variable input use and acreage choice decisions. Finally, in order to guarantee that our input costs estimates lie in reasonable ranges, we introduce additional information in the estimation procedure by using stochastic constraints. This model is estimated, using an SAEM algorithm, on a French panel data sample. By comparing our estimation results to actually observed input uses per crop, we show that our model and estimation approach perform better than commonly used approach in predicting input uses allocation among crops and that introducing non-negativity constraints and additional information in the estimation procedure allows improving the predictive power of the model.