A principal component model for forecasting age- and sex-specific survival probabilities in Western Europe

Abstract

The assessment of future mortality is of high importance in many areas where the allocation of future resources has to be planned in time, especially in social security and private life insurance. This contribution represents an extension of the classic forecasting approaches of Bell–Monsell and Lee–Carter. Based on a forecast of the first two principal components, age- and sex-specific survival probabilities for 18 Western European countries are predicted simultaneously until the year 2070. In addition to the correlations in the mortality trends between the age groups and the genders, international trends in mortality are captured as well. A major improvement in the classic Lee–Carter models is the adequate quantification of the uncertainty associated with the whole system of variables by stochastic simulation of all remaining principal components with simple time series models. The model’s easy applicability to further analyses is illustrated by forecasting the median life span as well as the resulting Gender Gap for Germany, France, and Italy.