Modelling policyholder behaviour

Modelling policyholder behaviour.

Policyholder behaviour is one of the most difficult factors of risk and profit for an insurance company to study. This is due to the lack of baseline insurance data in a context of sudden leaps in rates (which could cause a wave of redemptions) and the fact that it is a risk comprising a human element.  Insurers often begin by assessing the redemption and cancellation risks of a statistical approach based on past observations extrapolated for exceptional situations using a simple approach.  Insurers are increasingly seeking to implement worst-case approaches or financial-based approaches in which policyholders play the insurers to their best advantage as if they have followed in-depth stochastic control courses geared to finance.  Clearly, the first approach tends to under-estimate risk while the second over-estimates it. The truth certainly lies between the two, with a transition towards better informed policyholders which insurers should be wary of.

Using the analogy of the UK mortgage market, attention should also be given to the risk of sheep-like behaviour caused by a fiscal shock, a reputation risk, or by the recommendations of a guru (this can be an insurance comparison website, an online insurance wholesaler or a specialist magazine, etc.). Policyholder decisions will also depend on policies offered in the future and can be positively or negatively influenced by the insurer.  Fraud is also one of the main subjects linked to policyholder behaviour. How can it be better detected and prevented?

It will also be important to look at what data analytics techniques can offer in understanding the policyholder behaviour, but also reciprocally, to understand the effect of using data analytics methods on policyholder behaviour in a new era in which the use of personal data is potentially perceived by policyholders as invasive in relation to their private lives. These questions will be linked to research conducted as part of the Chair’s second main research theme: data analytics in insurance (see below).