Un modello di riservazione individuale multistato a tempo discreto

The aim of this dissertation is to propose and to verify the operational use of a Micro-Level approach for the assessment of the outstanding claims reserve IBNR and RBNS based on the empirical application on real data of a Non-Life Italian insurer. The analysis of the results will accompany a comparison with the main Macro-Level models in terms of Expected Value and Prediction Error. The approach is based on a Marked Poisson process piecewise constant for number claims IBNR and regarding the estimate of the amount of claims RBNS we proposed the use of the ratio paid on initial reserve as the dependent variable for the estimate of future payments. The application of the model is completed by the use of bootstrapping technique for variability of model parameters and the use of parallel calculation techniques for MonteCarlo simulation routines. The thesis will also present a possible approach to aggregating variability in the case of using a plurality of models.