Scientists and engineers predict the day is near when cars drive themselves, but that doesn’t mean that traveling on the highway will be entirely carefree. There will be accidents, and someone will be liable for them. That means someone—it isn’t yet entirely clear who—will need insurance.
And that means actuaries, lie Michael Stienstra, FCAS, an assistant vice president at QBE North America will have to figure out how much to charge for insurance—for the cars today, which can sense whether they are too close to the car in front of them—and for the cars of tomorrow, which may be able to take your kids to school for you.
Stienstra, along with Robert Peterson, a professor of law at Santa Clara University, and Frank Douma, a research fellow from the University of Minnesota, shared their ideas on “Autonomous Vehicles and the Impact on the Insurance Industry” at the Casualty Actuarial Society’s Ratemaking and Product Management Seminar held in April.
The dream of a fully automated car stretches back to the 1939 World’s Fair, Stienstra said. The first actual autonomous car was called Stanley, Stanford University’s experimental vehicle that won the 2005 DAPRA Grand Challenge, traveling 132 autonomous miles through the Mojave Desert in six hours and fifty-four minutes.
Most automakers say they will have a version by 2020; Google says it will have a fully automated car by 2017. Some, such as Raj Rajkumar, director of the Carnegie Mellon-GM Autonomous Driving Collaborative Research Lab, estimate an autonomous package “might only add $5,000 to $7,000 to the sticker price.”
The cars, by all accounts, will be much safer. Human error contributes to more than 90 percent of all auto accidents.
Fewer accidents mean cheaper auto insurance—$475 cheaper every year, said Peterson, the law professor, citing a Princeton study by Alain L. Kornhauser.
Getting that new car on the road is an engineering dream. Estimating the insurance costs is an actuarial challenge.
The first actuarial obstacle is data. While lack of data is usually an issue for actuaries, in this instance the reverse will be true, Stienstra pointed out, noting that there could be too much data that is exceedingly complex.
An automated vehicle will likely transmit upwards of 750 megabytes per minute of data. An actuary would have to cull the data before collecting it, and then find the predictive variables, the ones that foretell an accident.
The best predictors could differ by the vehicle’s technology. The variables that predict a Honda crash might not be able to accurately predict a BMW crash. Stienstra believes that actuaries will have to be far more involved in the collection of data than before.
Regulators could present another hurdle, says Peterson. California, for example, has mandatory rating factors that insurers must use, such as driving record or number of years as a driver. And safe drivers receive a discount. With an automated vehicle, those factors may prove irrelevant. State insurance laws will have to accommodate the new automobiles.
As the autonomous vehicle does more and more driving, Peterson says, the car owner may not be responsible for accidents at all. If a flaw in the auto or its computers causes an accident, liability may end up falling to the manufacturer or the company that installed the driverless system—in which case, the insurance pricing would fall to actuaries who specialize in product liability coverage.
But this all considers the end state in automation, where every car on the road is completely automated. Between now and then, there will be a mix of three types of vehicles:
–Fully automated cars, where the owner does nothing.
–Partially automated cars, where the owner does some or almost all the driving.
–Non-automated cars, where the person behind the wheel does all the driving.
Douma, from the University of Minnesota, spelled out the five levels of vehicle automation, from no automation (level 0) to fully self-driving automation (level 4). Current developments are moving toward deployment at level 3, he said, where the vehicle performs all safety-critical functions under certain conditions.
For a long time, Douma explained, the driver will be switching off driving tasks with the computer. The driver might back out of the garage onto the street, and then the computer will take over. Just shy of the destination, the computer might hand control back.
The handoffs create another issue, Douma said: “How do you keep somebody sharp enough to take over the car?”
The technology promises to fundamentally change the relationship between the driver and automobile. As the technology progresses towards a fully autonomous state, the risks and the variables actuaries use to price the risk of driving are likely to change dramatically. It will take time to produce the experience needed to recognize this change in insureds’ premiums.
Douma also discussed the fact that, in addition to the increased data available, issues of access and use of that data will be critical to how quickly and how well automated vehicle technologies are deployed. He cited the current controversies regarding use of automated enforcement of red light running and speeding as examples of where political perception has overshadowed the potential safety benefits of the deployment of these technologies.
Douma noted that most automated vehicles do not actually need personally identifiable data to work, only needing to “trust” other vehicles through use of randomly assigned certificates that can change many times daily. While this data is nearly anonymous, Douma further recommended actuaries strongly consider whether personally identifying data will be necessary in setting future rates, and, if so, whether data that is voluntarily shared by drivers would provide a sufficient sample, rather than requiring data from drivers and vehicle owners without their prior consent.
Meanwhile, outside forces—regulators, automakers and the public—will expect that the safer automobile will translate into lower insurance premiums. That could put more pressure on insurers.
Stienstra believes that actuaries will have to be proactive on this issue. The Casualty Actuarial Society, he notes, has an Automated Vehicles Task Force to make sure casualty actuaries have the ability to partner with engineers and researchers to properly understand and insure this challenging risk.
Source: Casualty Actuarial Society