Trips over 50 miles in length account for just 2.3% of all trips, but about a third of all distance travelled within Great Britain.  Because of the small proportion of all travel they form, long-distance trips may not be adequately represented in national data bases and models.  But, because they account for a substantial proportion of total distance travelled, particularly on motorways and rail, these trips are important in terms of transport policy and have a substantial impact on congestion.  Moreover, detailed study of existing data indicates that travellers’ behaviour in long-distance journeys differs substantially from routine journey patterns. Not only is the set of available modes different, but the profile of travellers is substantially different, with income playing an important role both in terms of the frequency of long-distance journeys as well as an increased likelihood of travelling by rail and air. Additionally, we observe different trade-offs, with significant reductions in responsiveness and increasing values of time, as journey distances increase. For these reasons, treatment of the specific properties of long-distance travel is essential for appraising the impact of new transport infrastructure aimed at this market, such as high-speed rail, highway construction and operation policies and policies directed towards domestic air travel. In a long-distance travel demand model, there is no doubt that the accurate treatment of mode choice is vital. What was not clear at the outset of this work, however, was whether it is necessary to incorporate destination choice responses explicitly or whether it is possible to represent such effects through elastic generation, which might operate differentially for different destinations.

Models of long-distance travel have therefore been developed using cross sectional National Travel Survey data in the UK to test explicitly the importance of destination and frequency effects. Models with and without destination choice both indicated that it is important to include a frequency model which responses changes in travel accessibility. Judgements about the necessity of including a destination choice component were not straightforward, on the basis of model fit only. However, when considering the resulting values of time and model elasticities, it was judged that the models with destination choice responses were better. Therefore, the key finding of this paper is that models with both frequency and destination choice components, as well as mode choice, better reflect the choices and response behaviour of long-distance travellers than mode choice only long-distance demand models.