Building latent segments of goods to improve shipment size modelling: Confirmatory evidence from France

Abstract

Freight transport demand models are generally based on administrative commodity type segmentation which are usually not tailored to behavioral freight transport demand modelling. Recent literature has explored new approaches to segment freight transport demand, notably based on latent class analysis, with promising results. In particular, empirical evidence from road freight transport modelling in Germany hints at the importance of conditioning and handling constraints as a sound basis for segmentation. However, this literature is currently sparse and based on small samples. Before it can be accepted that conditioning should be integrated in the state-of-the-art doctrine of freight data collection and model specification, more evidence is required. The objective of this article is to contribute to the issue. Using detailed data on shipments transported in France, a model of choice of shipment size with latent classes is estimated. The choice of shipment size is modelled as a process of total logistic cost minimization. Latent class analysis leverages the wide range of variables available in the dataset, to provide five categories of shipments which are both contrasted, internally homogenous, and directly usable to update freight collection protocols. The groups are: "‘standard temperature-controlled food products"’, "‘special transports"’, "‘bulk cargo"’, "‘miscellaneous standard cargo in bags"’, "‘palletised standard cargo"’. This segmentation is highly consistent with the empirical evidence from Germany and also leads to better estimates of shipment size choices than administrative segmentation. As a conclusion, the finding that conditioning and handling information is  essential to understanding and modelling freight transport can be regarded as more robust.