Modelos de elección discreta aplicados al ámbito portuario

Abstract

This research focuses on port choice. Million of shipments from the hinterland of the main Spanish peninsular container ports have been analysed. Firstly, the evolution of the hinterland of ports in the study were delimited, characterised and compared. The existence of inter-port competition was confirmed both graphically and analytically. Secondly, how the degree of use of port facilities affects the number of users who choose that infrastructure was studied. The research raises the hypothesis the more traffic a port has, the more attractive it becomes but there is a saturation threshold beyond which the attractiveness of the port decreases. A multinomial logit model is estimated, which without establishings restrictions a priori, confirms its nonlinear influences and sets the threshold of saturation. Finally, complex discrete choice models are calibrated and validate which include characteristics that are usually not considered: random heterogeneity in preferences of decision makers and non-linear influence of the shipping attributes. They are compared in terms of their ability to predict with simpler models reaching differences for the choice share close to 15%. It concludes that the mixed logit model Box-Cox, which includes both characteristics fits to the calibration data significantly better.