Impact of Spatial Competition and Agglomeration Effects on the Temporal Transferability of Destination choice models

Abstract

To ensure transport models are effective tools for planning, they must not only adequately explain current travel choices but also maintain predictive accuracy over forecast horizons while being computationally feasible for practical use. One simple approach to achieving this is by including behavioral theories in transport models to improve the behavioral representation of models.
One such theory is Spatial Competition and Agglomeration Effects (SC&AE). This theory examines how opportunities present nearby influence the attractiveness of a destination. This influence can be either positive (Agglomeration) or negative (Spatial Competition). Although SC&AE is widely recognized in the literature for enhancing the explanatory power of computationally simple Multinomial Logit (MNL) destination choice models, its impact on the temporal transferability of these models remains unexplored.
This study assesses the impact of SC&AE on MNL destination choice models for home-based maintenance, work, and education trips in the Metropolitan Region of Amsterdam on a 5-year short-term forecast horizon (2018-2022) using Dutch National Travel Survey Data (ODiN). The findings indicate that SC&AE has a positive but limited effect on the temporal transferability of these models. This impact decreases with decreasing traveler autonomy and ease of switching to alternate destinations. In terms of percentage improvement, Transfer index metric shows an exaggerated impact (35.2% - 286.36%) while the other performance indicators, such as Fitting Factor (1.1% 3.6%) and % Correct predictions (0.14%-2.05%) confirm that the impact is limited.