Estimation of the Economic Order Quantity Model Using the ECHO Shipment Database
WINNER OF The Neil Mansfield Award
F Combes, Universite Paris-Est, FR
The Economic Order Quantity model is a good candidate to incorporate optimum shipment size decisions in freight transport models. In order to assess its empirical validity, it is tested econometrically using the French ECHO shipment database.
Spatialised freight transport models are generally based on adaptations or extensions of the classic four stage representation of transport systems, in which the following variables are explicit: emission and reception rates, origins and destinations, transport mode and itineraries. A very important variable of freight transport systems, namely the shipment size, is almost never explicit. It plays a very important role though, as it results from the logistic requirements of shippers, and strongly influences the decisions and costs of carriers. Introducing shipment size in freight transport models is thus a major objective.
However, this requires the existence of optimal shipment size models, preferably based on a microeconomic background. Such models exist. One of the simplest, on which this paper is focused, is the well-known Economic Order Quantity model, in the frame of which the optimal shipment size is a trade-off between transport costs and inventory costs. This model is used a lot in logistics, but its validity for large, heterogeneous firm populations has not been confirmed, mainly due to prohibitive data requirements. Indeed, a measure of the shipper-receiver relationship is necessary for its econometric assessment. This variable is generally not available in shipment databases.
Fortunately, the French shipment database ECHO, available since 2005, fills this gap. Compared to other shipment databases, it describes relatively few shipments, but with a greater amount of detail. The shipment, the shipper, the shipper-receiver relationship, the transport operations and the carriers are described with a large amount of detail. In particular, the amount of commodities sent each year by the shipper to the carrier is measured, as well as the density value of the commodities constituting each shipment. It is thus possible to assess econometrically the EOQ model using the ECHO database. This is the objective of this paper.
After the microeconomic principles of the EOQ model are reminded, a specification is proposed and tested against the ECHO database. The explained variable is the shipment size, and the explicative variables are the commodity value density, the total amount of commodities sent by the shipper to the receiver, and the transport mode used to carry the shipment. Despite some issues, such as a significant amount of missing observations, and some minor inconsistencies between the variables required by the model and the variables available in the database, the model is proved to performed quite efficiently, with a R-squared of about 0.8 on more than 6000 shipments.
A more complete specification is also tested, to explore the role of additional variables, which do not fit in the original EOQ model, but which can influence the optimal shipment size. These variables are the number of agents intervening in the transport of the shipment, the number of transport operations, and some variables related to the way the transport operation was organised. This extended specification performs only marginally better than the basic one, as is confirmed by an analysis of variance. This tends to demonstrate the simple EOQ specification is good enough for a large population of firms.
This result is quite satisfying with respect to the perspective of introducing the shipment size decision in spatialised freight transport models. The EOQ model is simple, does not require a lot of data (even though the variables required are not easily available), and makes a first step towards bridging the gap between the logistic requirements of shippers and their preferences towards freight transport, in particular towards transport modes. Finally, it should be noted that the econometric analysis presented here is quite simple; more sophisticated econometric tools can certainly bring more light on this issue.
Association for European Transport