Session: 70

Mortality

* * *

Summary

Prediction of the future mortality of life annuity portfolios is one of the major concerns faced by life insurance companies. Prediction of mortality curves is, however, a complicated challenge and a number of prediction models have been developed by scientists, actuaries, demographers and political planners. This paper investigates the simplest and most popular prediction model, namely the prediction model using the currently observed mortality as a prediction for future mortality. We call this predictor the ''at risk'' estimator. This prediction estimator is compared to the actually observed mortality throughout the period 1900-1996 in Denmark. Our investigations show that the maximal error made by the simple prediction model corresponds to an interest rate of 0.7% on life annuities. Another way of formulating this is to say that old age pensioners of age 65 have lived up to two years longer than the ''at risk'' prediction model has indicated.

Session: 76

Mortality

* * *

Summary

We investigate the practical aspects of some recent marker dependent hazard estimators when applied to mortality models. We analyze the development over time of Danish longevity from 1974-1998, whith mortality considered as a two-dimensional function of age and chronological time, respectively. A new method of bootstrapping confidence bands of marker dependent estimator is applied to identify significant changes in the mortality patterns. Functionals of the hazard rate estimator, such as expected remaining lifetime and probability of survival, are also presented. We outline the usefulness of this methodology to analyze the age and time effect on longevity and its implications for life insurance risk management.