Discrete Mode and Discrete or Continuous Shipment Size Choice in Freight Transport in Sweden

Abstract

The Swedish Commodity Flow Survey (CFS) 2001 was used to estimate different model specifications at the level of individual shipments. One model only explains mode choice, ignoring the endogenous nature of shipment size. The other two models simultaneously explain mode and shipment size, where shipment size can either be a discrete or a continuous variable. All three model specifications lead to models with significant estimated coefficients that have the right sign. The elasticities for changes in cost and time by mode from the three specifications are quite different. The only similarity in the elasticity outcomes is that road costs changes have a small impact on the market share of road in all three models (mode choice effect only) and that the direct elasticity of air time is the biggest of all direct elasticities. Starting from the pure mode choice model, accounting for endogenous shipment size leads to different direct elasticities for road and water transport. Also, a model where continuous shipment sizes have been converted into discrete categories leads to a model that has different behavioral responses than the model with continuous shipment size. The model with discrete mode choice and continuous shipment size choice can be seen as the preferred model because it takes account of the endogenous nature of shipment size and uses the shipment size data as they come from the survey. However, standard software for simultaneous estimation is not generally available, and in the two-step estimation, the explanatory variables of shipment size need to be exogenous, so cannot include transport time and cost by mode. This model therefore cannot give an impact of time and cost of the choice of shipment size. From the model with discrete shipment size and mode, the authors conclude that shipment size is rather sensitive to changes in time and costs (so changing road transport will lead to bigger road shipments, not so much to substitution from road to rail and water-based transport). Therefore, in practical applications, a model with two discrete choices might be preferable, certainly when the classification into discrete shipment size categories would not contain just a few, but many categories. Future work on this model may include re-estimation on the CFS 2004/2005 data, with new costs model input. It may also include simultaneous estimation of the discrete/continuous model, models with two discrete choices but more shipment size categories and models with other explanatory variables explaining some of the coefficients (observed taste variation) and with random (unobserved) taste variation in some of the parameters or with other more flexible substitution patterns.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 20 citations exist upstream.