We first began studying the implications of self-driving cars, or “AVs”, two years ago. At that time, observers predicted that fully autonomous “Level 5” cars (i.e., a car with no steering wheel) would not appear in significant numbers until 2030. In 2015, Lux Research observed:
Fully autonomous driving may happen by 2030, but only in highly restricted environments and likely only at low speeds.
That prediction now appears to be well off-target. The world may not be ready for the pace at which AVs come to dominate city streets.
Fast forward to 2017. In April, Elon Musk said a Tesla will drive in full autonomous mode from Los Angeles to New York before the end of year, and predicted that its autonomous cars will be reliable enough to win regulatory approval within two years. Every major car manufacturer is investing heavily and building partnerships to gain the computer hardware, sensor and artificial intelligence expertise to compete in the AV market. On May 22, 2017, Ford fired its CEO, largely because its board perceived that Ford had fallen behind in the race to carve out a position in AV markets. His replacement? The head of the division focused on AVs.
In December 2015, the State of California issued proposed regulations for the deployment of AVs. They assumed that cars would still need a driver and driving wheel to take over in some circumstances. Our blog at the time questioned whether regulators were falling behind the technology because reports even then suggested it was advancing at an exponential pace. Google was urging regulators to permit Level 5 AVs as soon as possible because tests showed that even an imperfect Level 5 car would be safer than one that relied on a driver to take over in emergencies, and far safer than a conventional car. The State of California eventually agreed. It leapfrogged its own December 2015 regulations, in March 2017 introducing proposed regulations to allow for the deployment of AVs with no steering wheel.
The other development now converging with AVs, which few observers saw two years ago, is the rise of ride-sharing. Uber appeared in San Francisco in 2010, but ridership was only beginning to take off in 2013-14. By 2016, however, it had become obvious that the long-term business models of Uber and Lyft (and China’s Didi) depend on replacing their drivers with AVs. In a few years, when you call for a Lyft, an AV will take you to your destination. Uber already is testing AVs on the roads in Pittsburgh, Phoenix and San Francisco. The per-mile charge for the ride will plummet, because the cars will not need drivers, and will likely be mass-produced, low maintenance electric vehicles.
In short, AVs will not just eliminate steering wheels. They are likely to do away with most individual car ownership.
In other words, transportation-as-a-service (TaaS) is on the horizon. One think tank recently predicted that AVs will be safe enough to receive widespread regulatory approval no later than 2021. By 2025, TaaS will have reached a tipping point in urban areas, capturing the lion’s share of miles driven. By 2030, 95% of all miles driven in the US will be in fleet-operated electric AVs.
We can’t address all of the far-reaching implications of such a future. The market for auto insurance remains one of the more predictable of the many long-term consequences. We have long assumed that the cost of an operator’s policy would decline as AVs result in fewer and less severe crashes. The advent of TaaS, however, implies that the market for individual auto policies will evaporate. Virtually all auto insurance will become product liability insurance purchased by the fleet owners. That market, too, will likely stabilize at low premium costs (on a dollars/mile basis) once the real world data deliver the technology’s promise of significantly fewer accidents.
Individual ownership will not completely disappear. Individual owners of AVs will likely pay low premiums. Owners of conventional cars, on the other hand, may face higher premiums as the availability of insurance for individual owners dries up. In addition, the driver of the conventional car may well face a presumption of fault if the driving record of AVs establishes their reliability. The black box capabilities of AVs (to record what happened at the time of an accident) make it likely that fault will not be difficult to prove. Indeed, if the safety record of AVs approaches its promise, public policy might lead to banning conventional cars.