A disaggregate analysis of factors influencing port selection

In this paper are presented the results of a multinomial logit model used to explain the selection of a port for each shipment exported from the US in December 1999. The model is estimated for combined shipments, and one model is estimated for each of four commodity types (bulk, foods, fabrics, manufactured). Results are compared across commodities to find the significance of the distance in either direction, the frequency of sailings, and the average size of vessels sailing along a route. Results of the estimation do not match those that would be expected. Potential reasons are given for such variations, along with numerous methods by which the modelling could be improved. Methods for improvement suggested include structuring of the model to represent more accurately the competitive scenario of each port, constraining the choice set for each shipments from characteristics of the carrier, and the commodity.