Exploring built environment correlates of walking for different purposes: Evidence for substitution

Improving the built environment is one strategy to promote walking. However, few studies have identified the collective contribution of built environment elements to different types of walking relative to socio-demographic attributes. Moreover, people may have a budget for total physical activity; this assumption needs to be tested in the literature on the built environment and travel behavior. This study applied gradient boosting decision trees to investigate the correlations between the built environment and the durations of utilitarian walking and recreational walking using data from Shanghai, China. The results showed that the built environment had larger predictive power for utilitarian walking, whereas socio-demographics were more important for recreational walking. Land use mix and distance to the city center, which are important for predicting walking, had opposite effects on utilitarian versus recreational walking. These findings support the theory of a physical activity budget. Road density is associated with different types of walking in a similar pattern.