Description: In addition to gender, age, period and cohort are three main risk factors considered in constructing mortality models. Other than these factors, it is believed marriage status is related to health and longevity. In this study, we use Taiwan’s marital data for the whole population to evaluate if marriage status can be a preferred criteria. In addition to checking if mortality rates vary for different marital statuses, we want to know if this preferred criteria will be valid in the future. We chose Lee-Carter and age-period-cohort models to model the mortality improvements of various marital statuses. Because of linear dependence in the parameters of the age-period-cohort model, we used computer simulation to chose the appropriate estimation method. We found the married have significantly lower mortality rates than the single and, if converting the difference into a life insurance policy, the discount amount is even larger than that for smokers/nonsmokers.

Description: We propose an age-shift model to modify the LC model and deal with the problem of parameters. The proposed method attains smaller estimation errors with respect to mean absolute percentage error or mean square error. We also apply the proposed age-shift model to the mortality rates of these four countries to evaluate the longevity risk in annuity products by measuring life expectancy.