Estimation of an Inventory-theoretic Model of Mode Choice in Freight Transport

Estimation of an Inventory-theoretic Model of Mode Choice in Freight Transport


R Lloret-Batlle, Technical University of Catalonia, CENIT, ES; F Combes, Universite Paris-Est, LVM, FR


Two freight mode choice models inspired by inventory theory are designed and estimated with the French ECHO shipment survey.


Shipment size is a fundamental decision variable in freight transport; however, it is often not included in freight transport models. One option to make it explicit is to rely on optimal shipment size models from inventory theory. One of the simplest, the classic Economic Order Quantity (EOQ) model, was recently assessed econometrically for a large and heterogeneous population of firms, using the French shipment database ECHO. The advantage of this model is that it can be included easily within a mode choice framework.

The objective of this paper is to examine how a mode choice model can be designed with the EOQ model at its basis. This is done in two ways. First, a mode choice model is designed on the basis of the EOQ model. Shipment size is not explicit, but the decision of shippers is assumed to be based on a total logistic cost function in which shipment size is endogenous. Then, a two-step modal choice/shipment size model is examined on an exploratory basis, in order to improve the estimation of the previous model. Both models are estimated with the ECHO database.

The first mode choice model is a multinomial logit model, where each mode is characterised by its total logistic cost function. The shape of the total logistic cost function is deduced from the theoretical EOQ model. The explanatory variables are the commodity value density, the total commodity flow between the shipper and the receiver, the distance between the origin and the destination, and some variables related to the nature of the commodities. The explained variable is the main transport mode. It should be noted that the description of the transport supply is minimal: travel times and costs are not used. The model is thus essentially demand-oriented, and even logistics-oriented. It is estimated on the ECHO data with a satisfying fit, but the heavy land modes (rail, combined transportation, and inland waterways) are significantly underpredicted.

This lack of accuracy is assumed to stem from the lack of vehicle capacity constraints in the model. To address this shortcoming, an exploratory mode choice and shipment size model is investigated. Based on an existing methodology, shipment size is first predicted along an EOQ model but without knowing the transport mode; then the difference between this shipment size and the per-mode average shipment sizes is introduced in the utility function of each mode, penalizing those where the difference is bigger. With this model, the prediction for heavy modes is increased substantially. Overall, the predicted market shares of all modes are closer to the observed ones.

In conclusion, this study shows that inventory theory and comprehensive shipment databases are a potentially very fruitful direction to improve freight transport mode choice modelling. The results obtained are encouraging, considering that almost none transport supply data are used in the model. A combination of good shipper-related data with good transport supply data (particularly concerning travel time, travel time reliability, frequency, capacity constraints) would certainly improve the models estimated in this section greatly. Improved specifications would also be useful, especially regarding the inclusion of vehicle capacity constraints.


Association for European Transport