Norwegian accident data now puts hard numbers on EV crash behavior in the market furthest along in the transition. A new Marsh Re report covers six years of accident data across petrol, diesel, EV, and hybrid vehicles in Norway. It draws on more than 19,000 accidents and 210 billion kilometres of exposure, combining the Statens vegvesen accident registry with Statistics Norway (SSB) traffic volume data. It updates a prior Marsh Re, formerly Guy Carpenter, study by adding 2025 figures, increasing the EV dataset by more than 30%.
Norway has the highest EV share of new car sales of any market in the world, according to Norwegian Road Federation data. EVs accounted for 95.9% of new passenger car sales there in 2025, according to Norwegian Road Federation data, which means the country’s accident record carries a depth of EV exposure that few other markets can approach. What happens in Norway as the fleet matures is the closest available proxy for what will happen in the UK, Australia, and other markets as EV penetration rises.
In 2025, the EV accident frequency rate was 7.7% per million kilometers driven. That figure is 41% below the petrol rate of 13% - both figures drawn from the Marsh Re dataset - based on more than 1,000 EV accidents and 13 billion EV kilometers in that year alone. Across the full six-year dataset, EVs are involved in 36% fewer accidents per kilometer than petrol cars. EV crash rates have fallen every year since 2020, a 16% improvement over the period, while petrol rates have remained broadly stable.
Norway’s EV fleet tripled its driving distance over the same period. Accidents, however, did not keep pace. That divergence is the finding that challenges motor pricing most directly.
Reinsurers writing motor proportional treaties have been absorbing assumptions about EV frequency derived largely from early-adopter data and markets where EVs were used more intensively in commercial and ride-share contexts. The Marsh Re data suggests that as fleets mature and EV use normalizes toward everyday commuting, frequency rates move in a direction that improves aggregate loss experience.
The vehicle-age curve adds useful granularity. New EVs in the zero to two-year age band currently show higher normalised accident frequency than petrol cars of the same age. That elevated rate has been trending sharply downward across the six-year period, consistent with a learning-curve effect as the fleet matures. By three to five years, EV frequency converges to near parity with petrol. By six to 10 years, EVs have the lowest accident frequency of any fuel type. That curve has direct implications for how motor books carrying young EV fleets should be segmented and reserved, as the elevated new-vehicle risk is temporary and predictably short-lived.
Frequency alone does not settle the reinsurance question. Severity tells a more complicated story, and the Marsh Re report is appropriately cautious about it.
EV accidents involve 32% more parties on average than petrol accidents and are twice as likely to involve a pedestrian, according to the Marsh Re analysis. Single-vehicle accidents and fixed-object collisions are significantly less common.
Using Highway Loss Data Institute (HLDI) data alongside its own proprietary motor-claims database, Marsh Re finds that EV claim costs differ by damage type. Collision costs are 21.9% higher for EVs than for equivalent internal combustion engine vehicles, while personal injury claims are 23.7% lower and bodily injury claims 23.1% lower. Across all damage types combined, overall costs are broadly similar between the two vehicle types.
Smaller claims rise in cost while larger claims fall. Given the uncertainty that surrounds long-tail bodily injury claims, including the continuing impact of long-term care costs on large loss reserves, a portfolio with fewer severe injury claims and more property damage claims carries a different risk profile at the tail. That is where excess-of-loss treaties are most sensitive. As the EV insurance market has grown rapidly, reinsurers have been pricing and structuring treaties largely on the assumption that EV severity would be uniformly higher. The Norwegian data suggests that assumption needs to be disaggregated by damage type before it is applied to loss picks or attachment points.
The Marsh Re dataset has limits worth acknowledging. The Norwegian claims data does not include claim cost figures directly, so severity analysis relies on party counts as a proxy, supplemented by HLDI and proprietary data. The HLDI data is US-based, and the Marsh Re proprietary data is UK-sourced. Both are used to directional effect rather than as precise cost estimates. The frequency findings, by contrast, rest on a statistically robust dataset spanning six years and a fleet large enough to support meaningful segment-level analysis.
