Will Self-Driving Cars Lower Ride-Hailing Prices?
Lower cost is one anticipated advantage of incorporating autonomous vehicles (AVs) into ride-hailing services. But a study co-authored by Prof. Zhen Lian suggests that lower prices will only materialize under certain market conditions, such as using a single app for both AVs and human drivers.
Waymo vehicles at a charging station in Santa Monica, California, in May.
In 2018, when Zhen Lian, then a PhD student at Cornell, started working on a research project about self-driving cars, the technology still seemed like “a futuristic topic,” she says. Just five years later, while traveling in Phoenix, she was able to hail a self-driving car—also known as an autonomous vehicle (AV)—to get a ride from her hotel to the airport. .
The incorporation of AVs into the ride-hailing industry is edging closer to reality, and the rise of this technology raises questions about how consumers will be affected. One main advantage, proponents claim, is that AVs will lower the cost of rides. After all, the company doesn’t have to pay for human labor. But is that really the case?
Lower cost of AVs does not necessarily translate to lower prices. Whether that promise from AVs will be realized for the consumers critically depends on the market structure and the competition status.”
In a new study, Lian, who is now an assistant professor of operations management at Yale SOM, and her co-author Garrett van Ryzin compared several scenarios with different market conditions for AVs and human drivers. They found that only one scenario would likely lead to lower prices. One key factor was that AVs had to be available on the same ride-hailing platform as human drivers, rather than on a separate platform. And they had to be provided by competing companies rather than a single supplier.
“Lower cost of AVs does not necessarily translate to lower prices,” Lian says. “Whether that promise from AVs will be realized for the consumers critically depends on the market structure and the competition status.”
Firms are currently experimenting with various types of market structures, ranging from a standalone app for AVs to open platforms where multiple AV companies can join a pool of human drivers. “It’s all over the place,” she says. These issues need to be considered “before the technology matures.”
At first glance, it seems to make sense that AVs would be cheaper than human drivers. But AVs are still pretty expensive—often around $200,000 or more per vehicle. AV suppliers have to commit upfront to buying a certain number of cars, some of which might sit idle during periods of low demand. And when demand spikes, the AV fleet isn’t as flexible as human drivers, who can easily hop on and off the app whenever they like.
In her study of how AVs could affect ride-hailing prices, Lian collaborated with van Ryzin, an operations scholar who has held positions at Uber and Lyft and is now at Amazon. (Lian also worked at Lyft for a year, but the study wasn’t part of her work there.) Using a mathematical game theory model, the researchers considered two types of dispatch platforms and two levels of competition, resulting in four possible market scenarios.
In the first dispatch system, both AVs and human drivers could be matched with riders on the same platform. In the second type, AVs and human drivers were separated into different platforms—for example, in separate mobile apps.
To assess different levels of competition, the researchers ran scenarios where AVs were provided by a monopolistic supplier. They compared them to situations with perfect competition, meaning that each AV was provided by a different supplier.
As one might expect, having competition lowered ride-hailing prices. In the monopoly scenario, the AV company could strategically limit the available supply of AVs to drive up profits. In the competitive scenario, AV suppliers would continue to add more cars to the platform as long as the value of joining was higher than the cost.
Using a common platform for AVs and human drivers was also better for consumers, because it was easier to match nearby drivers with riders at any given time. If the two types of cars were on separate apps, a waiting customer could find themselves standing on a street corner right next to an available AV—but because they have requested a ride through the human-driver app, they don’t get matched with that vehicle. As a result, the pickup time is longer, and the overall system is less efficient, driving up prices.
The upshot is that among the four scenarios, only the one with competitive AV suppliers and a single platform “will lead to unambiguously lower prices,” Lian says.
One caveat is that the model assumes consumers don’t care whether they get an AV or human driver. In reality, some people are skeptical or scared of AVs. If anti-AV sentiment is strong, ride-hailing companies might split AVs and human drivers into separate apps, causing prices to be higher overall.
Lian doesn’t think that AVs will replace human drivers entirely. Their complementary nature is similar to what we see in the electric power market, she says. AVs are analogous to nuclear power plants: they’re very efficient, but they have a high upfront cost. So they aren’t flexible enough to serve peak demand because it’s not economical to have such an expensive resource sitting idle during non-peak times.
Meanwhile, human drivers are like energy technologies such as gas turbines, which are much cheaper to set up and can serve peak demand more flexibly.
“Having both AVs and human drivers is, in our view, a good thing,” Lian says.