This paper looks at desirable properties for a model to apportion freight traffic (of a given O-D and commodity type) amongst a large range of mode/route choice alternatives. The paper then seeks to provide these desired properties in as simple a way as possible, given the circumstances of the problem. While it is assumed that evidence is available on shippers' monetary valuation of the components of generalised cost, it is assumed that there is not sufficient data/information to estimate standard models having the desired properties, e.g. nested logit models. The essence of the method proposed is that, at the same time as generalised costs are built up from network data, elements of similarity or commonality between alternatives are noted and used to form a `similarity' matrix. Candidate mode/route choice alternatives are then placed in order of increasing generalised cost, and all those with generalised costs more than a predetermined percentage above the minimum will be excluded. A second input parameter specifies the gain in utility from having two identical alternatives available. This information is used to generate a scale factor. A first-pass attempt at market shares is then derived from a logit formulation where the utility of any given alternative depends on its similarity to already chosen alternatives. Finally, a second pass reallocates market shares such that were an alternative to be excluded, its market share would be distributed amongst the remaining alternatives in relation to its similarity to those alternatives. Anglo-Continental freight is taken as an illustration.
Samenvatting