Temporal transferability of activity-based time-of-day choice models in the developing country: the case of Shanghai, China

This study uses the finite mixture continuous logit (FMCL) model to analyze the time-of-day choices, focusing on the temporal transferability of model. Non-commuters’ shopping activity start time choices in Shanghai, China, are examined as an empirical case. Using data from large-scale resident travel surveys conducted in 2009 and 2019, the separate models for each year and a pooled model with year-specific adjustment are estimated. The estimation results show that the general morning-dominated time-of-day structure remained broadly stable despite substantial urban changes, while some detailed behavioral relationships shifted. Disaggregate and aggregate transferability assessments indicate that similar overall timing patterns can persist even when underlying individual-level mechanisms change. The transfer performance of the pooled model indicates that a shared structure remains workable, although year-specific adjustments are important for capturing differences between years. These findings suggest partial rather than strict temporal transferability of non-commuters’ shopping time-of-day choices in a rapidly changing urban context.