Ανθεκτικές μέθοδοι αποθεματοποίησης
The purpose of this thesis is the robust estimation of reserve, in run off triangles with outliers or large jumps. The main reserving method is Chain Ladder, which is sensitive to outliers or large claims. In this thesis, we propose some changes in the way Chain Ladder is applied, in order to give a robust estimation of reserve and not be affected by outliers (Robust Chain Ladder). However, a method which detects outliers in a run off triangle, is very important. Therefore, is present not only a way to detect outliers but also an idea to deal with those outliers or large jumps. Last but not least, it is essential for the standard error of the method to be measured, in order to be sure that the reserve estimated was the real one. Two well-known methods, Mack method and Bootstrap Method, are presented. However, these methods are also sensitive to outliers or large jumps, therefore a robust transform of these methods is also presented. Finally, for each method the theory and an analytical example are given, so that the methods be more understood.