Substitution dans l'alimentation animale : l'apport des modèles de séries temporelles

Substitution in the French feed ingredient sector : the contribution of time series models. Several papers are aimed at estimating demand relationships for feed inputs for European countries. Many ofthem use an econometric model derived from the neoclassical theory ofthefirm and the duality theory. Direct and cross price substitution elasticities for feed ingredients are then estimated under given hypothesis concerning the characteristics of the production functionfor compound feed. Such analysis admit three weaknesses : - First, the whole spectrum of available feed ingrédients is generally divided into three to five groups on the basis of an a priori examination of their nutritional content : cereals, protein feed and cereal substitutes for example. The matrix of estimated cross and direct elasticities then highlight the substituability among groups ofproducts. Two justifications can be given. At first, it is necessary to reduce the number of estimable parameters and so the number of estimated equations in order to avoid multicollinearity problems. Secondly, a translog specification is generally chosen for the production function of the compound feed sector, so that the estimated equations are cost-shares equations which often lead to bad empirical results when some cost-shares are too small. But the dependance of estimated results on the initially retained partition is the realproblem because, if cereals can be consideredas a relatively homogeneous group, protein rich products and several intermediary products cannot be. Lastly, it is very difficult to estimate substitution elasticities for individual products. - Second, the identification of just one given econometric model with a priori exclusion and cross-equations restrictions derived from economie theory can hide some characteristics not quite in line with the chosen economic theory. - Third, many of those studies use one period models of the firm. However, provided that the compound feed sector does not make instantaneous adjustments in response to a changing market situation, a dynamic factor demand model would be more convenient. But compared to the vast literature on flexible functional forms, little attention has been given to dynamic models of firm behaviour. The aim of this paper is to highlight short-run substitution relationships among the main feed ingredients with an alternative method grounded on a vectorial time séries model for feed input prices (The model is estimated with French monthly data for the period : january 1977 - november 1987). Such a procedure has the advantage not to impose a priori restrictions on the econometric model and the estimated équations are ail derived from the information imbedded in the data at hand usual with time series models. A second advantage lies in the fact that the partition of the various feed ingredients into groups of products in unnecessary : the individual behaviour of ingredient can be examined. However due to the lack of underlying economic theory, our results are not directly interpretable in terms of classical substitution elasticities. They just act as indicators of the short-run substitution-complementarity relationships among feed ingredients.Plusieurs auteurs ont cherché à représenter le comportement de demande de matières premières du secteur français de l'aliment composé. Ils utilisent un modèle économétrique identifié à partir de la théorie néoclassique du producteur et de la théorie de la dualité. Cet article propose une méthode alternative d'analyse des relations de substituabilité/ complémentarité de court-terme entre les matières premières de l'alimentation animale fondée sur l'estimation d'un modèle VAR des prix de ces produits. Cette procédure possède deux avantages. D'une part, elle ne repose que sur les régularions observées entre les séries de prix et aucune contrainte théorique (si ce n'est le choix du vecteur de prix te- tenu) n'est donc imposée aux données. D'autre part, il n'est plus nécessaire de regrouper les produits par famille et le comportement de chacune des matières peut être examiné individuellement.Dronne Yves, Tavéra Christophe. Substitution dans l'alimentation animale : l'apport des modèles de séries temporelles. In: Cahiers d'Economie et sociologie rurales, N°23, 2e trimestre 1992. pp. 63-86