The aim of this paper is to model some production variables in extensive livestock farms located in the dehesa ecosystem. We intend to use not only purely economic variables in the construction of the model, but also structural variables in order to identify the characteristics of the farms that have the higher influence. Another objective is to be able to predict these variables at the farm level, using structural variables that are easy to measure. The data used in this work were obtained from a questionnaire survey to the holders/managers of a sample of 69 dehesa farms in Extremadura (SW Spain). The statistical methodology used for the construction of the model was Partial Least Square Regression (PLSR). It can be concluded that the variables relative to farm intensification, to labour and especially to Iberian pig breeding, are those that take part mainly in the model.