Loss reserves represent an important component of non-life insurance companies' liabilities, making their accurate calculation essential. Generally, loss reserves consist of two components: reserves for already reported claims and reserves for claims that are yet to be reported. The work presents a pragmatic method for calculating reserves for reported claims that utilizes a case-based reasoning paradigm. The proposed method uses historical data to generate stochastic predictions of individual claims developments, including ultimate claim amounts and corresponding developments of both paid and incurred amounts. Unlike most existing methods, the proposed method leverages individual claim information and uses paid as well as incurred data. The proposed method is evaluated using synthetic portfolios. The actual values of synthetic data and the predictions of the chain ladder method—a widely used method for loss reserving—are used as benchmarks. Results indicate that for homogeneous portfolios, both methods' predictions are similar to the actual values of synthetic data. However, because the proposed method uses individual claims information, it shows improved performance over the chain method in the case of heterogeneous portfolios.
|
