Modelling reserves the Chain Ladder Methods, Bornhuetter Ferguson Method and Cluster Poisson Processes

Specialeforsvar ved Christian Vangsgaard Jørgensen 

Titel: Modelling reserves The Chain Ladder Method, Bornhuetter Ferguson Method and Cluster Poisson Processes

Abstract: A major topic in non-life insurance is loss reserving, the classical model for modelling these reserves is the Chain Ladder model presented by Thomas Mack. This model is easy to apply but yields large relative errors and depends strongly on the present amount of claims, and will therefore easily yield nonsense reserves. Because of the importance of the reserves in a non-life insurance company it is of great interest to improve these relative errors and hence improve the model fit. We investigate the development pattern of a portfolio consisting of claims regarding workers’ compensation. From this we propose a Bornhuetter-Ferguson model, where we instead of using the present amount of claims, use an a priori expected ultimate to predict the development. Furthermore, a Poisson Cluster model is presented, this model introduces dependence between claims and arrivals as opposed to the two other models. Finally we compare the model and give our recommendation as to which model one should use for this specific portfolio 

Vejleder: Olivier Wintenberger
Censor:   Anders Jessen, Gjensidige Forsikring