Essays in Modelling Mortality Rates

This thesis brings together work carried out over the period 2008-2012 in the area of mortality modelling. We look at various aspects of the modelling of mortality and introduce several new models including, the O’Hare and Li (2012) model in chapter 2, and the O’Hare and French, dynamic factor model in chapter 3. In chapter 4 we investigate and identify a major weakness of extrapolative mortality models to account for structural breaks in the underlying data. In particular we note that this is not limited to the one factor Lee Carter model but is symptomatic of multi-factorial time series models also. We identify the structural breaks in a range of existing models and demonstrate the improvement in forecasting as a result of allowing for such structural breaks. In chapters 5 and 6 we consider the explanatory approach to modelling mortality and shift the focus onto identifying explanatory variables which may impact upon mortality rates. In chapter 5 we carry out an analysis of Northern Ireland mortality rates, linking the variation we see in mortality rates to a range of measures of deprivation that we have for each region. We then consider the question of whether or not additional geographical information has any part to play in explaining mortality variation. In chapter 6 we look at a wider range of countries and carry out a factor analysis for each country identifying the optimal number of factors necessary to explain mortality variation. We then carry out an exercise to link latent factors to identified explanatory variables for each of the countries considered. Finally we forecast these factors to inform the future forecast of mortality rates