Platform Sourcing Decisions in Ride-Hailing: Autonomous Vehicles vs. Human Drivers

Abstract

We consider a ride-hailing platform with two supply sources: a fleet of autonomous vehicles (AVs) and a population of self-interested human drivers. The platform faces a tradeoff: while human drivers can have higher variable cost than AVs, they are only paid if used. Due to the interaction between wage, human driver participation level, and AV deployment quantity, the required wage is non-monotone. At first, as the human joining fraction increases, the prevailing wage decreases because the platform deploys fewer AVs, reducing competition. However, at larger human joining fractions, the wage increases and human-served rides become less profitable, so the platform increases its AV deployment quantity to compensate. The increased competition from AVs necessitates still higher wages and triggers a feedback loop or race to the top for the platform with itself, which limits human driver recruitment. Accordingly, access to a larger AV fleet can actually hurt the platform because it cannot credibly commit to a deployment quantity, and we show that the platform could indeed earn substantially more profit if it could pre-commit. Nevertheless, we prove that by appropriately adjusting the AV fleet size, the platform can achieve the same profit as with full commitment power.