A fuzzy approach to grouping by policyholder age in general insurance

Abstract

In general insurance, policyholder age is often treated as a factor with the number of levels requiring that the individual ages of the policyholders be grouped. Although the groups are usually defined by the existing underwriting structure, it should be investigated as part of any premium rating exercise that uses a model to assess past claims experience. It is possible that an incorrect grouping by policyholder age could bias the results of the risk premium estimation. On the other hand, it may not be computationally feasible to use separate ages in the premium model, making some form of grouping necessary. In this paper, we specify a data-based procedure for grouping by age using fuzzy set theory. An example is given that illustrates how the method can be used in practice.