Transferability of models for logistics behaviors: A cross-country comparison between France and Germany for shipment size choice[La transférabilité des modèles de comportements logistiques : Une comparaison France-Allemagne pour le choix de la taille des envois]

Transferability is seen as one important rationale for justifying the application of behavioral transport demand models. Concerning freight transportation, behavioral models of shipment size choice gain more and more attention by modelers because of two reasons: First, shipment size models could explain a major proportion of heterogeneity caused by the multitude of actors and shipments; secondly, they link firms' behaviors to their logistics activities. However, the transferability of prevalent shipment size choice models is complicated due to different functional approaches used in econometric estimates and due to the varying occurrence of different variables in the underlying surveys. This clearly limits the applicability of the models for other case studies with related contexts. In this article, the transferability of continuous shipment size choice models is investigated by applying the same functional approach to a dataset from France and an equivalent dataset for Germany. In this way, we check the robustness of this approach in regard to different logistics variables and we analyze potential similarities in logistics behavior. Starting with an analytical model for shipment size choice, a descriptive analysis follows which compares for both countries the central figures related to shipment size choice. Afterwards, elasticities gathered from the estimated models are checked against each other and the transferability issue is empirically questioned. Finally, possible reasons for differences in behavior, probably caused by differences in transport cost or in inventory cost, are discussed. It turns out, that the flow of goods exchanged between a shipper and its client explains a major proportion of heterogeneity in France and in Germany, and that the impact of this variable is very similar. Furthermore, differences in the storage costs approximated by the value density are obtained. Logistics variables have similar impacts; they can, however, be neglected in a strategical forecast model. Concluding, for the example France and Germany, the behavior models could be transferred.