Search

Travel behavior analyses through traditional discrete choice models offer insights into the factors that influence individual choices. However, the literature often neglects the spatial interactions between individuals and attributes in the context of walking trips. This paper investigates the spatial dependencies in walking travel choices with a case study in the city of São Paulo, Brazil. We compared a non-spatial and spatial lag of X (SLX) logit models for walking choices with data derived from a revealed preference survey conducted in the Metropolitan Region of São Paulo (MRSP). Sociodemographic information of individuals and trip-related attributes were included in the models. Overall, the prediction capabilities of the spatial models outperform the non-spatial model in terms of log-likelihood. The results indicate that the SLX with the attribute referring to the lower income group of individuals lagged by the distance of 500 m performed better compared to the non-spatial model. In the studied context, such groups are usually concentrated in more walkable neighborhoods. Therefore, improving the built environment and increasing job and educational opportunities may increase the propensity of walking for work and study trips in areas other than in higher income neighborhoods. • Walking travel choices assessed through Revealed Preference (RP) data. • Spatial lag of X models with sociodemographic and trip-related attributes. • Distance threshold between 200 and 600 m influence walking choice decisions between neighbors. • Lower income individuals are influenced by neighbors in walking travel choices. [ABSTRACT FROM AUTHOR]