Chapter 9: A Comparison of Conjoint, Multi-Criteria, Conditional Logit and Neural Network Analyses for Rank-Ordered Preference Data

Most of the literature on freight transport’s qualitative attributes, like transport time and reliability, is handling surveys’ data on an aggregate level; moreover, the applied econometric models often assume a linear utility function. Examples are the papers and reports by de Jong (2000), Stratec (1999), Danielis, Marcucci, and Rotaris (2005) and Beuthe and Bouffioux (2008). Such approaches can be justified by the need to obtain average results, which can be used within larger models of transport analysis, as well by the difficulties presented by the surveys’ data, which show a wide heterogeneity of logistic situations. In contrast, this chapter analyses the individual preferences of nine transport managers with four different types of models, some of them quite unusual in this line of research and assuming a non-linear utility function. It is devised as a testing of some quantitative models of individual decision making and a means of obtaining additional insights into the managers’ decision process.