Sensitivity of future life expectancy at birth and at age 65 due to different mortality forecasting methods

Lenny Stoeldraijer, Statistics Netherlands
Coen van Duin, Statistics Netherlands
Leo van Wissen, Netherlands Interdisciplinary Demographic Institute (NIDI) and University of Groningen
Fanny Janssen, University of Groningen

With rapid population aging, mortality forecasting is of increased importance, especially for the insurance and pension industries. However, current approaches to project future mortality differ more and more both by and within countries, leading to different outcomes. We review the current methods behind official mortality forecasts in Europe, compare outcomes and assumptions of different projection methods within one country and compare the outcome of different types of methods for this country using similar explicit assumptions on the historical period and the jump-off rates. We compare direct extrapolation, the Lee-Carter model, the Li-Lee model, a cohort model, the separate projection of smoking- and non-smoking-related mortality and the official Dutch forecast. Statistical offices in Europe mostly predict mortality through extrapolation methods, but also make use of target values, expert opinion and cause-specific mortality. Approaches for the Netherlands include explanatory models, the separate projection of smoking- and non-smoking-related mortality, and projecting the age profile of mortality. Clear differences in explicit assumptions, including the historical period, exist. The life expectancy at birth (e0) in 2050 varies by almost 5.5 years for both males and females. Using Dutch data for 1970-2009, different methods result in a range of 2.1 years for females and 1.8 years for males. For e65, this is 1.4 and 1.9, respectively. Because the choice of the explicit assumptions add more to the differences then the choice of the forecasting approach, the choice for the projection method should not only be based on different approaches, but more importantly on the explicit assumptions.