A Joint Metro Train Demand Model Accounting for Disaggregate Consideration Probability and Aggregate Footfall

This study introduces a new metro train demand model that simultaneously captures both aggregate ridership from automated fare collection (AFC) data and disaggregate consideration propensities, using individual survey data from Chennai, India. This joint framework produces more accurate aggregate demand estimates than traditional OLS (R2 improves from 0.67 to 0.75), as it is able to capture the complex and non-linear relationship between disaggregate consideration probability, reflecting potential demand, and aggregate footfall, reflecting realized demand. It is observed that increasing the consideration probability enhances the footfall overall. However, some locations exhibit an opposing trend between consideration and footfall (low consideration but high footfall, or vice versa). Also, the sets of influential factors vary across these two dimensions. For instance, individual-level variables (income and out-of-vehicle travel time) and multi-modal connectivity features (presence of an airport and multimodal hubs near the metro) play a key role in footfall. In contrast, consideration probability is primarily influenced by access time, cost, and egress distance. Furthermore, factors influencing consideration probability (walkability, train service quality, and first–last–mile connectivity) vary across segments (based on vehicle unavailability, exclusive vehicle availability, and limited vehicle availability). Evidence of selection bias among metro riders, non-normality, and intra-person variability effects in footfall is observed. From a policy perspective, neglecting the disaggregate consideration effects on realized aggregate demand, i.e., footfall models, can overestimate the role of metro costs and out-of-vehicle travel time. In addition, the ridership levels of the metro are overestimated at higher metro fare levels. The new model illustrates that applying location-specific and dimension-specific policy interventions can be more effective than uniform area-wide policies for enhancing the user base and realized ridership.