Prediction intervals for future Pareto record claims (Q6984894)

The study focuses on statistical prediction of extreme claims or losses in insurance industry. The core of the paper is the statistical modeling and prediction of extreme claims in non-life insurance. After introducing the model of record statistics, the use of a Pareto distribution model for upper record values is demonstrated as appropriate for predicting future record claims or losses based on previously observed record values in a sequence of claims across time. The results are applied to real data sets, as the Danish insurance data set, the data set from the research study `sigma-1/2021' by the Swiss Re Institute, and the data from an international non-life reinsurer on its worldwide reinsured catastrophe losses. The examples demonstrate how the proposed models perform and can capture the magnitude of future record claims, even with relatively small numbers of record observations.