This study investigates the impact of walking and e-hailing on the scale economies of on-demand mobility services. An analytical framework is developed to i) explicitly characterize the physical interactions between passengers and vehicles in the matching and pickup processes, and ii) derive the closed-form degree of scale economies (DSE) to quantify scale economies. The general model is then specified for conventional street-hailing and e-hailing, with and without walking before pickup and after dropoff. We show that, under a system-optimum fleet size, the market always exhibits economies of scale regardless of the matching mechanism and the walking behaviors, though the scale effect diminishes as passenger demand increases. Yet, street-hailing and e-hailing show different scale economies in their matching process. While street-hailing matching shows a constant DSE of two, e-hailing matching is more sensitive to demand and its DSE diminishes to one when passenger competition emerges. Walking, on the other hand, has mixed effects on the scale economies: while the reduced pickup and in-vehicle times bring a positive scale effect, the extra walking time and possible concentration of vacant vehicles and waiting passengers on streets negatively affect scale economies. All these analytical results are validated through agent-based simulations on Manhattan with real-life demand patterns.

Simulation studies suggest that Shared Automated Vehicles (SAVs) could reduce the total vehicle kilometres travelled (VKT) thanks to efficiently pooling multiple users in one vehicle. However, mode choice studies indicate that SAVs would attract mostly public transport users, leading to an increase in VKT. This paper is among the first to combine these operational and behavioural expectations and the first to do so analytically. In our theoretical set-up, travellers choose between car, public transport, and SAVs, depending on their individual valuation of private travel and other attributes of each mode. We find that the introduction of SAVs lead to a VKT change in public-transport-oriented cities ranging from a small decrease to a large increase, where the latter is true for plausible parameter settings and hence is a cautionary point for SAV-introduction policies. Conversely, SAVs would attract only few travellers in private-transport-oriented cities and therefore would not significantly impact VKT.