Pricing electric and autonomous vehicles requires new models with rating factors made for the technology, rather than retrofitted from old variables, industry experts said at the Asian Actuarial Conference 2026.
Insurers are already pricing electric vehicles (EVs) on variables that did not exist a decade ago, from average electric motor replacement prices to the state of health of the battery, said Ken Lim, senior consultant for property and casualty at Milliman, during a presentation on navigating EV insurance.
One example is the manufacturer’s battery degradation curve, which insurers read alongside telematics data to sharpen risk selection and price EV-specific add-ons.
“You can follow the curve and price accordingly,” Lim said. “Anecdotally, this has helped [insurers] with three to four points of loss ratio reduction – not the best, but good enough.”
However, neither the variables nor the methods behind them are settled. Insurers are still working out which factors matter and how to weight them, while evolving technology and new data keep producing more.
Autonomous vehicles will shift the basis of pricing again.
Traditional insurance models that price risk according to the attributes of a human driver, like age, gender or experience, may no longer apply in pricing risk for autonomous vehicles (AVs). In their place are technology-specific risks that can shift a fleet’s risk profile overnight, said Ivy Liang Xuan, Regional Business Actuary at ride-hailing giant Grab.
“All of these rating factors are changing,” Liang said during her presentation on AV risk. “A whole new set of risk factors is coming. For example, software defects, perception failures, over-the-air update regressions, cyber risk, all of these new terms we haven’t heard of 10 years ago.”
To compensate for the lack of historical data, actuaries calculating AV risk can look to combine external benchmarks with disengagement rates (number of times a human takes over driving) and fleet telematics, said Liang.
While no mature AV loss data exists, public driver-assistance safety research can be used to set a starting benchmark, which can then be updated as operational data accumulate.
“The research shows that in order to prove that an AV is 20% safer than a human driver, you would need 5 billion miles of driving records,” Liang said. “Even the most mature operating system like WayMo, they only have 57 million miles of driverless. So, we are essentially two orders of magnitude short… should we wait until we collect 5 billion miles of data? I think the answer is no.”
Grab is using the private-hire infrastructure that it has built to price its own human drivers to lay the technical groundwork for pricing autonomous vehicle risk. It is using the same sensors (GPS data, accelerometers and gyroscopes), the same cloud data pipeline and the same actuarial discipline to measure autonomous vehicle behaviour.
“Why are we trying so hard to build this telematics pricing and to build this infrastructure? Because we believe that we not only need it now for PHV (private-hire vehicle) driving, but we also need this for the future for the autonomous vehicle driving,” said Liang.
“All these technologies remain the same when you move to autonomous vehicles,” Liang added. “The infrastructure is very, very similar, and you can continue to utilise… essentially the [same] actuarial idea.”
Maturity easing claims frequency
Electric vehicle (EV) claims frequency in Asian markets may ease as markets mature, but the challenge for insurers is how they can minimise loss and gather data, said Milliman’s Lim.
Data have shown that claim severities and multipliers continue to buck trends as markets transition to EVs from ICE. Older drivers in China, normally a cohort priced safely by insurers, carry the highest EV-to-ICE claims frequency multiplier, rising from about 1.1x among under-25s to 1.7x from age 40 onward. Insurers in the United States have also seen claims frequencies rise for drivers switching from ICE to EVs, with a slight increase in claims frequency overall from 3.3% to 3.4%, according to LexisNexis data.
But in mature markets like Norway, where EVs make up a third of all passenger cars and nearly all new cars, claims frequency for them has tracked below ICE cars since at least 2020 and is now 17% lower, Lim said.
“The Norway experience sort of validates the theory or hypothesis that as the market becomes more mature, EV claims frequency should be lower than ICE cars,” he said. “You have a lot more sensors, you have a lot more ADAS… all sorts of sophisticated things in your EV, they’re supposed to help you.”
It’s unclear at what point claims frequency trends will reverse.
“Claims frequency [for] EV cars are a little bit higher, but we believe that it is a transition,” said Lim. “But when will the transition be over? I don’t know. And I don’t think anyone in the market knows the real answer.”
With original equipment manufacturers (OEMs) increasingly declining to share telematics data insurers need to price EV risk accurately, EV insurers that can negotiate with manufacturers will now get the upper hand.
“A lot of the OEMs (original equipment manufacturers), they hold on to the telematics data and they don’t sell it… because they know this is the real crown jewel of the game,” said Lim. “How do we negotiate with them to get that data? That will be something that differentiates a profit-making and a loss-making company.”
The data that vehicle manufacturers hold have led to significant advantages when they seek to vertically integrate their own insurance solutions. BYD’s own captive insurance solution is one such success story, Lim said, turning its first profit in 2025.
By monopolising parts manufacturing, controlling repair handbooks and making it harder for dealers to repair EVs, OEMs have made it much easier for them to compete with insurers. “The complaint has become the advantage,” Lim said.
Loss-making EV cover can be a way for insurers to gather data to price risk and do risk selection, said Lim.
“Just buy the data at the cost that you can afford,” Lim added. “The person with the most data and can make use of the most data will win.”
