How to incorporate the spatial dimension within destination choice models? The case of Antwerpen

This paper considers different alternatives for including spatial aspects within the activity-based approach for modeling destination choices. The study area is the urban agglomeration of Antwerpen (Belgium); the city and its suburbs are considered. Individual travel surveys are used. The paper pays particular attention to the inclusion of space within the decision context by including specific land-use explanatory variables generated by Geographical Information Systems. A preliminary geographical analysis is performed in order to represent the city by a limited set of destinations (n = 33) and to characterize those zones in terms of land use. Discrete choice modelling is used: each individual faces the total set of spatial destination alternatives. Several modelling approaches are explored and compared in terms of utility function (for instance Box-Cox; random coefficients) and in terms of global formulation (multinomial logit versus nested logit). The mixed nested logit formulation is selected as best and the parameter estimations are interpreted; it shows the importance of space within destination choices. This paper provides a useful background for decision-makers and planners of transportation policy related to individual mobility patterns. Keywords Discrete choice model, activity-based approach, GIS, land use, urban mobility, Antwerpen