|Year of selection||2015|
|Institution||Université Catholique de Louvain|
Type of support
Joint Research Initiative
195 000 €
Insurance protects us with coverage, and actuaries work hard to establish the right prices to keep the whole system running smoothly. So, what if they were overlooking multiple sources of information? Today’s methods consider a person’s insurance contracts in isolation. An actuary working out the optimal price or design of an auto insurance policy only looks at auto insurance data for the person, without evaluating his or her home insurance history, for example. However, there is information in one kind of insurance that can shed light on another. As Dr. Michel Denuit puts it, "If you're cautious behind the wheel, you probably also avoid risks at home." The ambitious objective of his research project, a joint initiative with AXA Belgium, is to develop new methodologies that will help actuaries design better coverage. These would take into account multiple parameters of policyholder behavior and extend their analysis further into the future than has ever been done before.
Currently, insurers’ predictions about policyholders look ahead by only one year. Yet, most policyholders expect to keep an insurance contract for several years, making a longer outlook of two to five years more relevant for insurers. For instance, in the short term, they may predict that a young driver will make claims in the first year, but accept that risk more readily, knowing that his or her driving is expected to improve with time. Another option could be to implement a pay-how-you-drive scheme, like driveXperience, offered by AXA Belgium. Dr. Denuit’s novel, multi-year approach could capture these types of dynamic behavior.
To develop the necessary models, he will capitalize on AXA’s vast database of anonymized policyholder information, covering their full portfolio over several years. This data consists of classical variables, like policyholder age, type of vehicle, and information about their deductible. Dr. Denuit’s statistical models could process all this policyholder information to learn what happens for one insurance product given the existing knowledge of other products, providing a better risk forecast on the customer level. The unique partnership with AXA will also allow him to test his models’ predictive ability and compare it with the reality contained in the dataset. The project will benefit from the expertise of Dr. Julien Trufin, a professor of actuarial science at ULB in Brussels, as well.
By helping actuaries calculate prices more effectively, this work could decrease premium payments for some, while others would likely see prices go up. Insurers could take the opportunity created by these models to educate people whose behavior appears quite risky. And, for everyone, it would become easier for policyholders to demonstrate to an insurance company their quality as customers, Dr. Denuit says. “This is because we would learn more about people, faster. If you report no claim on any other insurance, after just a few years you could show your insurer that you’re a very good policyholder of the specific insurance in question.” This advantage could become automatic for consumers. Flexible, tailored insurance offers of this kind are becoming more the norm and Dr. Denuit’s dynamic, person-centric take on risk promises to benefit both sides.
Scientific title : AXA Joint Research Initiative in Actuarial dynamic approach of customer in general insurance
To add or modify information on this page, please contact us at the following address: firstname.lastname@example.org