Modelling acreage decisions within the multinomial logit framework : profit functions and discrete choice models

The first purpose of this paper is to propose theoretical justifications for using Logit acreage share models. Two approaches are presented: the Logit shares can be derived from a well defined profit function or derived as the result of a set of discrete choices. It is next shown that both theoretical frameworks allow to define generalizations of the standard Logit shares. These generalizations build on developments of the Multinomial Logit framework for modelling discrete choices and seek to define models that are flexible and empirically tractable. Two applications are presented to illustrate the empirical interest of the proposed models. Both use a rotating panel of French farms (1987-2006) and consider the estimation of yield functions, variable input demand functions and acreage share functions.