Quantification of aggregation bias in regional agricultural land use models: Application to Guacimo county, Costa Rica

Different specifications of a land use model for Guácimo county in Costa Rica were used to quantify various sources of aggregation bias, including variation in farm resource endowments within and between representative farm classes; spatially variable prices; and labor market inter-dependencies between farm classes. The results indicate that adequate treatment of the farm level in regional land use analysis requires proper classification of individual farms in representative farm classes, as well as explicit modelling of inter-dependencies between such classes. Aggregation bias introduced by the assumption of spatially fixed farm-gate prices is insignificant for regions with reasonable infrastructure quality and relatively small distances between farms and markets. Aggregation bias caused by variation in resource endowments within farm classes cannot be eliminated, but should be minimized, underscoring the importance of a proper farm classification.