A hierarchical reserving model for reported non-life insurance claims (Q2138622)

Within the wide framework of non-life reserving models, the paper considers the information over the lifetime of a claim, such as covariates describing the policy, claim cause and the whole history collected during a claim's development. The study provides a hierarchical reserving model, which decomposes the joint likelihood of the claim development process over time and registered events, specifying best practices for calibrating this model to insurance data as well its predictive capability in relation to reported, but not yet settled (RBNS) reserves. Then, considering the data registered at the level of individual claims, the aggregate reserving models are derived as special cases of the hierarchical reserving model. A case study models the RBNS reserve for a European home insurance portfolio, when damages to the insured property and its contents resulting from a wide range of causes are reimbursed. Moreover, the research presents two hierarchical reserving models on portfolios simulated along four scenarios generated by the simulation engine proposed throughout the paper, evaluating the predictive performance of the hierarchical reserving model on the portfolios generated by a simulation machine. Technical details concerning the simulation engine are described in Appendix. The R package accompanying the paper allows to use the simulation engine for the application of the hierarchical reserving models.