Predicting Mortality from Period, Cohort or Cause-specific Trends: A Study of Four European Countries

Abstract

This chapter presents forecasts up to 2020 of overall mortality by period and cohort, and of all-cause mortality obtained by aggregation from cause-specific models (by period) for four developed European countries (France, Italy, the Netherlands and Norway). Dynamic parameterisation models and trend extrapolation are applied to obtain the forecasts. The results show that the three approaches produce different levels of mortality, and consequently of life expectancy in the future. For women, the all-cause approach tends to result in higher life expectancies than the cause-specific approach. For men in France, Italy and the Netherlands, the all-cause approach generally resulted in lower life expectancies by 2020 than cause-specific forecasts, again noted for ages 40, 60 and 80. For Norwegian men, the overall mortality approach produced higher life expectancies than cause-specific forecasts. Forecasts of mortality by cohort are the lowest (in terms of life expectancy) of the three. When comparing the three approaches, criteria such as forecast accuracy, transparency, utility and options for validation should be taken into account. It is suggested, however, that all three approaches are useful and needed. So, choosing among the various approaches is needed only if particular constraints prevent researchers from performing all three types of forecasts.