Methodes de provision technique et applications aux donnees reelles: Etude de cas

Abstract

Cet article traite des méthodes de provisionnement d’une compagnie d’assurance. Un bref résumé des  différentes méthodes a été donné ainsi que les hypothèses des  modèles. Une étude plus approfondie montre des limites de ces méthodes dans l’approche actuelle dans le choix d’un Tail Factor. En effet, les valeurs trouvées semblent explosées par rapport à celles espérées. Pour y remédier, nous proposons une nouvelle approche en définissant un Initial Factor au lieu d’un Tail Factor et des modèles de type Backwards pour améliorer la convergence des différents modèles.

Mots clés : Inversion du cycle de production, Méthode London Chain, Méthode Chain Ladder, Méthode de Pondération, Run-off Triangle, Méthode Backwards.

English Title: Claims reserving methods and applications to insurance company data: Case study

English Abstract

This paper deals with insurance company claim reserving methods. We have made panorama of these  methods and models. A short summary presentation of Chain Ladder method with its hypothesis.  Furthermore, others cases of Chain Ladder method are given: Weighting method and London Chain  method. In the same way, the claim reserving development method show off. We find some limits of these methods owing that some kinds of data such that IBRN1. A deep data analysis is proceeded in  order to send out the kinds of values and improve the claim paid amount. The results showed that, according to the kind of data, some methods give claim reserving values which overflowed the expected ones and most of the time the appropriate choice of loss reserving method lies on the kind and the data properties. To avoid this situation instead of defining tail factor, new loss reserving models are proposed which help to improve the results. These models are the main contributions
of this paper.

Keywords: Reversal of cycle production, London Chain method, Chain Ladder method, weighting  method, Run-off Triangle, Backwards methods.