Predicting travel flows with spatially explicit aggregate models: On the benefits of including spatial dependence in travel demand modeling

Abstract

The prediction of travel demand is a key step in transport planning and is a topic of intense discussion of the literature. This paper adds to the debate about the accuracy of travel demand prediction by addressing the ‘technical’ problem of spatial autocorrelation. This paper aims to systematically assess the predictive performance of spatially explicit models that take spatial autocorrelation into account vis-à-vis more conventional models. We compare the performance of both types of models in predicting the transit passenger flows for alternative transit network designs in the region of Arnhem-Nijmegen, the Netherlands. We find that models taking spatial dependence into account outperform the conventional models in nearly all respects: model fit, parameters of variables, and the quality and stability of the predictions. Results show that taking spatial autocorrelation into account is not only important for the analysis of spatial interactions, but also result in different and more accurate predictions of the impact of interventions. We conclude that travel demand models should account for spatial dependence in order to avoid overprediction of the impact of transport system changes. We end with a discussion about the relevance of our findings for the debate about the causes for the observed systematic overestimation of travel demand in the practice of transport planning.