Alternative parameterisations of the Poisson common factor model for modelling mortality jointly for both sexes

In a global environment where mortality continues to decline, it is becoming increasingly important to develop mortality models which are able to account for global trends and relationships while also producing reasonable forecasts. In recent years there has been a growing interest in the co-modelling of multiple populations to address this. One such example is the Poisson common factor model proposed by Li (2013) for modelling mortality jointly for both sexes. This thesis expands on the Poisson common factor model by proposing two alternative parameterisations which relax some of the original assumptions. One variation allows a different number of sex-specific factors for each sex, providing more flexibility in taking into account differing features and trends between females and males. The other variation considers a common age effect shared by both sexes, potentially improving the parsimony of the model's optimal use of parameters. The two extended models are then tested using mortality data from six populations. Model performance is measured using goodness-of-fit and forecasting accuracy. The results indicate that both of the two modifications improve fitting compared to the original model, and slightly improve forecasting accuracy in many cases.