Forecasting freight transportation demand with the space-time multinomial probit model

Freight transportation demand is a highly variable process over space and time. A multinomial probit (MNP) model with spatially and temporally correlated error structure is proposed for freight demand analysis for tactical/operational planning applications. The resulting model has a large number of alternatives, and estimation is performed using Monte-Carlo simulation to evaluate the MNP likelihoods. The model is successfully applied to a data set of actual shipments served by a large truckload carrier. In addition to the substantive insights obtained from the estimation results, forecasting tests are performed to assess the model's predictive ability for operational purposes.