Analysing Freight Transports? Qualitative Attributes from Stated Orders of Preference
M Beuthe and C Bouffioux, Catholic University of Mons, BE
A stated experiment estimation of the importance of qualitative attributes of freight transports: frequency, time, reliability, flexibility, and safety. The conditional logit model of Beggs et al. (1981) is applied on preference rank-ordered data.
The research aims at estimating the relative importance and value for freight shippers of qualitative factors that characterize transport solutions: service frequency, transport time, reliability of delivery, carrier?s flexibility, and safety.
A stated preference experiment with a set of transport managers provided a sample of preference orders among a set of hypothetical transport solutions. A conditional logit model adjusted for handling preference orders (Beggs et al., 1981) is applied to the global sample, but also to sub-samples classified according to the transport distance, the value and type of goods, modes, willingness to switch mode and categories of loading unit.
This analysis showed that, whereas transport cost is the dominant factor in the choice of a transport solution, qualitative factors also play an important role, particularly transport time and reliability. Moreover, the results indicate that the factors? relative importance varies very much according to the categories of firms and transports. This suggests a strong heterogeneity of transport situations
Most of the linear utility functions coefficients estimated in this way are highly significant, with the exception of those of Frequency. The coefficients of the rail and inland waterway subgroups are also less significant, which can be the result of their smaller subgroups. However, a test, proposed by Ben-Akiva et al. (1992), showed that the attributes? coefficients across subgroups in a set where not significantly different from each other, with the exception of the cost attribute. As a consequence, we were lead to re-estimate the utility functions, assuming that the qualitative attributes had a common coefficient across sub-groups and that only the cost attribute had coefficients specific to each sub-group. This strong restriction on the utility functions naturally lead to lower statistical performance in terms of the rank correlation coefficient of Kendall, but produced highly significant coefficients in all cases. Again here, very different coefficients were obtained from one group to another for the cost attribute. This leads to estimates of the equivalent monetary of the qualitative attributes that are also very different. The differences between values is partly explained by the type of transport they refer to, which is shown by relating these values to the transport time, the distance and the shipment size. The paper concludes with a comparison between the attributes? monetary values obtained from this estimation and some of the estimates published in the international literature, pointing out the wide spread of values which, based on samples of different natures, obviously refer to very different transport situations.
Association for European Transport