Forecasting Mortality in Related Populations Using Lee-Carter Type Models

Some aspects of modern society are planned according to the values of expected future mortality rates. Due to the relevance of this issue, several approaches for treating this problem have been proposed. Among them, one of the most infuential is the Lee-Carter model. The aim of this thesis is to forecast mortality rates of related populations. In order to do this, some models based on Lee-Carter approach are considered; the models are applied to central death rates and to mortality improvement rates. Firstly, the models are discussed in a qualitative way. Secondly, the models are evaluated on a real dataset and their ability in fitting the data and forecasting are compared. The results highlight strengths and weaknesses of the different approaches. A further discussion relates applications of the models on mortality improvement rates. More specifically, the hypothesis of constant variance of the parametric structure is discussed and the impact of changes in this assumption is investigated by means of an application on a real dataset