Improving the financial sustainability of supplementary health insurances with data analytics

The financial sustainability of the Dutch supplementary health insurance is currently under pressure (DNB, 2017). One of the main reasons is the presence of the adverse selection effect (Winssen, van Kleef and van de Ven, 2017). Adverse selection is the phenomenon that individuals with high health risk are more likely to insure themselves compared to people who have low health risk. The past few years more and more people decided to stop their supplementary insurance which reinforces this effect (Vektis, 2017). Studies conducted so far have only been able to apply price differentiation based on health cost related risk as a mean to counteract adverse selection (Van Winssen, van Kleef and van de Ven, 2017). This thesis tried to provide an alternative by looking whether the risk of churning identified on claim data could be used to differentiate prices among insured. The extent to which future customer churn is predictable based on this claim dataset was assessed by several models. First, it was found that out of the 46 attributes used, nine were significantly related to customer churn. Age and cost of an insured within a collectivity and whether or not the insured moved during the year appeared to be most related to customer churn. However, predictive models based on those attributes were maximally able to correctly predict 24.1 % of the customers who actually churned in 2015 and 9.1% of those who churned in 2016. Therefore, it had to be concluded that this study could not provide a complete alternative for price differentiation based on health cost related risk and is therefore not suited to counteract adverse selection. Nevertheless, this study identified characteristics related to churn that can be used in future studies. Therefore, this study could be seen as a starting point and several recommendations for future research are provided.