Location-Based Compensation for Endogenous Driver Relocation in On-Demand Mobility Platforms

On-demand mobility platforms often face spatial mismatches between idle vehicles and rider demand. Platforms commonly use location-based compensation to encourage drivers to relocate toward high-demand areas. However, drivers’ relocation decisions are endogenous: they depend not only on compensation but also on expected competition for uncertain future demand. In this paper, we study a driver relocation problem in which a platform uses location-based compensation to influence drivers’ relocation decisions in a decentralized manner. We model drivers' relocation decisions using an endogenized multinomial logit (EMNL) model in which a driver's expected utility at a location depends on both the offered compensation and the equilibrium matching probability under the competition for future trips. Solving the platform's profit maximization problem, we show that the optimal compensation satisfies an equal adjusted markup condition. That is, the platform equalizes the expected drivers' earning across locations, adjusted for the probability of full vehicle utilization. We further show that while decentralized, the location-based compensation scheme for driver relocation yields lower profits than centralized relocation, the profit gap becomes negligible when drivers are highly sensitive to monetary incentives and commission rates are high. For computational implementation, we also develop data-driven optimization frameworks that can be solved efficiently as exponential cone programs using standard optimization solvers. Using real operational data from a major mobility service platform in Singapore, our case study demonstrates that explicitly accounting for drivers' endogenous relocation decisions enables platforms to more effectively mitigate spatial supply-demand imbalances and improve profitability.