“This is a new world for insurers. The industry will need a whole new set of data tools to deal with this.”
So says Michael Macauley, CEO of Quadrant Information Services, a supplier of pricing analytics to property/casualty insurance carriers. He is referring to the coming of autonomous vehicles.
Ford Motor Co. intends to start selling driverless cars to ride-hailing and taxi services by 20121 and to the public by about 2025. The vehicles will not have steering wheels, brake pedals or other controls for human occupants. The car maker wants to lower costs enough to make autonomous vehicles affordable to millions of people, CEO Mark Fields said in a recent speech.
Macauley notes that currently auto insurers base rates on what they know about the driver, the driver’s residential area and the car. The combination of a teenage driver, a new Corvette Stingray and a high-crime neighborhood, for example, yields one rate. A middle-aged driver with a perfect record, a quiet home town and a five-year-old Honda yields quite another.
However, now the industry is confronting a situation where there may be no driver, in which case, the owner’s age and driving record would be largely irrelevant, according to the pricing expert. Coverage and rate decisions would be based on a wide range of variables, among them traffic patterns, prevailing climatic conditions (there is some question as to how well driverless cars manage in rain and snow) and, above all, the track record of the various technologies involved.
Identifying those responsible for specific technologies may not be easy.
“These vehicles,” Macauley said, “will be the work not so much of a single manufacturer, but of a consortium.” He noted that Ford’s announcement of its forthcoming driverless cars described partnerships with Velodyne (lidar sensors that capture high-resolution images of the area around the vehicles), SAIPS (an Israel-based computer vision and machine learning company which Ford acquired in August), Civil Maps (a California-based company that develops high-resolution 3D maps), and Nirenberg Neuroscience (a machine-vision company that specializes in object recognition and bringing human-like levels of intelligence to machine modules).
Prospective competitors are making similar arrangements. BMW, Mobileye and Intel, for example, recently announced that they are partnering on fully-autonomous vehicle technology to apply in a shared environment by 2021.
Macauley says there have been doubts raised about the safety of self-driving vehicles, particularly since a driver in a Tesla Model S in self-driving mode was killed when the car drove itself into the side of a semitrailer in Florida last May. In its statement about the incident, Tesla did not delineate the driver’s level of engagement at the time of the crash, but it did note that “neither the Autopilot nor the driver noticed the white side of the semitrailer against a brightly lit sky, so the brake was not applied.”
The National Highway Safety Administration (NHTSA) is investigating that Tesla accident in part for its broader implications for safety of the new technologies.
But safety regulators generally favor the vehicles. More than 35,000 people were killed in U.S. auto accidents in 2015, with human error being the cause in 94 percent of them, according to NHTSA.
The NHTSA has issued federal guidelines for the testing and rollout of autonomous vehicles that largely leave manufacturers free to experiment, although the government also left open the door to a pre-approval system in which manufacturers would need to get the safety regulator’s approval of vehicles before selling them to the public.
According to Macauley, some manufacturers—including Ford—are pursuing a fully computer-managed, no-human-driver car because they feel that hybrid approaches that allow for some degree of human control, such as Tesla’s, contain an unacceptable level of built-in risk. Google, for example (which has its own driverless car program), decided on the no-wheel, no-pedals approach after allowing its employees to drive the company’s test cars.
“There was a brief period when people would be a little nervous and monitor the car very carefully,” Google engineer Nathanial Fairfield told the Washsaid, “and then they would start to relax and trust the system, and then they would over-trust the system and start to get distracted.” After watching a driver rummaging around in his back seat in search of a phone-charging cord, Google engineers decided it would be too risky to create a system wherein a driver would be expected to take control of the car at a critical moment.
Insurance claims in such cases pose unique challenges in part because driver behavior is still a contributing factor. Tesla maintains that Autopilot is only an assist feature — that drivers need to keep their hands on the wheel and be prepared to take over at any time.
Fans of Tesla’s Autopilot bemoan that there’s no database of lives saved or accidents avoided by the technology.
“Obviously,” Quadrant’s Macauley said, “it’s not going to be up to the property and casualty insurance industry to determine when driverless cars take to the road, how they operate, and what kinds of technology they use. It is, however, going to be up to insurance carriers to understand the situation and evaluate the risks posed by different operators—if that’s even the right word—in different cars on different roads and in different weather, using a variety of technologies.”
He says that to do this, insurers will need “robust, cloud-based computing and powerful, high-speed data analytics,” which his company happens to develop.
But the fact that people will be on roads at the mercy of technology means “the importance of making sure drivers are properly insured is exponentially high—as will be the demand for that insurance.”