Insurers like to know the risk they are exposed to. The Amsterdam School of Data Science and insurance company Vivat use big data analytics to calculate these risks. Professor Michel Vellekoop is interested in the behaviour of individual claimants, but points out that the principle of solidarity needs to be preserved.
In January 2018, insurance company Vivat and the Amsterdam School of Data Science, in which the University of Amsterdam’s (UvA) Faculty of Economics and Business is one of the participants, signed an agreement on a major research project. They will engage in a close collaboration and use data science to get a better grip on the many risks that an insurer needs to manage.
‘The developments in the areas of data collection and data processing are spectacular’, says Michel Vellekoop, Professor of Actuarial Sciences and Mathematical Finance at the UvA. ‘At the same time, looking at the future, no one knows yet which data will best help us to assess risks and analyse customer behaviour. I am very enthusiastic about this project in which Vivat and the university will work on theoretical questions as well as practical applications in this field.’
The collaboration includes three main topics. One area of research covers the identification of fraud by the use of artificial intelligence, while other research tackles the question how current and future customers respond to different forms of communication.
Vellekoop and his colleagues Peter Boswijk and Noud van Giersbergen are conducting a third study. Their research covers three different aspects of client behaviour: assessing clients’ future claim behaviour, assessing clients’ sensitivity to changes in the insurance premium and assessing clients’ loyalty to their insurers.
Vellekoop studied applied mathematics at the University of Twente and did a PhD at the Imperial College in London. Since 2010, he has been part of the UvA’s scientific community, In 1948, the UvA was the first institution to offer an actuarial sciences programme. This university is still at the helm in this field of study, with leading roles in research and teaching: anyone wishing to be an actuary will need to attend either the UvA programme or the Actuarial Institute.
It is difficult for an insurance company to predict how customers will respond when they change the price of their policies. ‘A characteristic feature of insurance products is that contracts typically run over a long period of time’, says Vellekoop, ‘and as a result, data on how customers respond to changes in insurance premiums is scarce.’
Compare this with the situation at a platform like booking.com, used by consumers to make hotel reservations. ‘Such companies are able to carry out experiments with small price changes and see how customers respond, and they will collect vast amounts of information that helps them set the most profitable prices.’
The collaboration will take the form of a postgraduate research study. In addition to the information Vivat already uses today, and which it derives from its own business operations, the PhD student will use a lot of external data. ‘One source for instance will be Independer, the online platform on which the prices of policies offered by different insurers are compared.’ Insurers, however, could also introduce new ways to gather information themselves, such as, says Vellekoop, installing a GPS box to measure the driver’s performance behind the wheel.
An insurance company’s core business is to estimate the probability of claims by the policy holder. Today, insurers already categorise customers according to crude criteria such as age and education. Using big data can refine this categorisation.
‘For practical purposes one would want to zoom in on the individual customer’, says Vellekoop. ‘The customer’s claim history is already taken into account, but there are much more data these days that can help achieve a better estimate.’ Most of those data are external, because data directly obtained from the customer (for the purpose of the contract) do not provide much insight. ‘When someone wishes to insure their property, the insurer could take into account the price history of the property.’
Vellekoop cannot say where these developments will stop. ‘There are limits to collecting information, if only for privacy reasons. Insurers could sit on a tremendous amount of information if, let’s say, they were allowed to analyse people’s DNA. But, fortunately, that is not allowed nor should we wish for it.’
There are also limits to the extent to which insurers are allowed to sell tailored and individually-priced insurances to individual customers or specific groups of customers.
Even insurers who sell products that are, in principle, not subject to bans on price and product differentiation do face legal restrictions. ‘We all know that women have a longer life expectancy than men. As a result, women should actually pay more for their retirement pensions than men. The European Court, however, provided that insurers shall not differentiate prices on the basis of gender because that would violate the laws against discrimination.’
Despite such legal provisions, Vellekoop notes a fundamental conflict between the wish to collect more information about individual risks and the basic nature of insurance. ‘Insurance is built on the idea that a large group of people bring in money that provides for losses that individuals cannot or will not bear. When insurers zoom their products and prices in on ever smaller groups of people or individuals, this idea loses its ground. Insurers need a certain amount of critical mass to offer good and inexpensive insurance plans.’
Apart from developments in the insurance industry, the public’s improving insight into individual risks could make consumers change their behaviour. People with a low risk profile could group together and create their own, low-cost insurance, or they could decide to do without an insurance altogether. ‘There is a certain link of solidarity between people who have an insurance. If they know individual risks, this may undermine this concept of solidarity.’
Vellekoop believes that for the time being, it will not come to that. ‘Counting on a future when we can make detailed estimates of the risk of loss for every individual is not realistic; there will remain too many uncertainties. That is exactly why it is important to find out how we can get as much reliable information as possible from the various sources of data and better estimate the insured risk of losses and the behaviour of the customers.’
More information? Email: firstname.lastname@example.org.