One size does not fit all: an empirical investigation of the Romanian agriculture production function

There are instances when one wants to consider homogeneous, with regard to some functional relationship, groups. For example in representative farm modelling one is interested in specifying groups of farms that have the same input/output relationship. This paper questions the logic behind the established approaches. It argues that one should use the underlying functional relationship to derive such groupings. Directly grouping farms with regard to their production function (or any other functional relationship of interest) not only asks the relevant question directly, but also makes the classification issue explicitly dependent on the choice of functional form. It provides a clear definition of what kind of homogeneity and/or representativeness were are looking at. If we want farms with similar production function either because this is the characteristic we are interested in or because we intend to model their production function in a follow–on simulation model, this is clearly the type of question we have to ask. This paper proposes using finite regression mixture models to specify and estimate farm groups with regard to pre-specified functional relationship. We illustrate the proposed approach with regard to the production function. The properties and advantages of the proposed approach are discussed and explained.