a co-integration analysis of danish mortality data

Specialeforsvar ved Snorre Jalbjørn

Titel: A co-integration analysis of Danish mortality data

Abstract: The projection of future mortality rates is an indispensable part of many financial decisions and of vital interest to the pensions industry. It is crucial to provide a model treating mortality of several populations jointly, since forecasts obtained through multiple independent single population models are known to be diverging and fail to account for the possible existence of common stochastic trends. In this thesis, we explore the contributions of co-integration methodology in the projection of future mortality rates. We give a detailed review of the popular Lee-Carter model and its multi-population extensions applied to Danish mortality data. We successfully employ a Vector Error Correction Model (VECM) to construct a co-integrated version of the Lee-Carter model in which the mortality of the two genders evolve together. Due to the inbuilt identification issues of the Lee-Carter model, we are unable to test whether a one-to-one relationship exists. We suggest a way to circumvent this issue in a Lee-Carter setting, and also establish the relationship in a CBD-model. Furthermore, we look at various longevity stress scenarios in order to compare the performance of the Lee-Carter model to its co-integrated version

Vejledere: Mogens Steffensen / Søren Fiig Jarner  
Censor:     Niels Haldrup, Aarhus Universitet