A Method for the Forecasting of Mortality

Abstract

In population projections the problem of the estimation of future mortality trends is of central importance. In this paper a new method serving this purpose is applied. After assuming the probabilities of death for large age groups (nqx), a relational technique is applied for the estimation of one-year death probabilities (1qx) of a full life table as proposed by Kostaki (Math Popul Stud 9(1):83–95, 2000). Afterwards, a smoothing procedure of the1qx values is used, based on a 9 parameters relational model originally developed by Heligman and Pollard (J Inst Actuar 107:47–80, 1980) and later on modified by Kostaki (Math Popul Stud 3(4):277–288, 1992) in combination with three subsequent cubic splines. After the age of 84 years the probabilities of death were extrapolated on the basis of the parameters of the last spline used. Results of the analysis indicate that the method applied was on one hand very effective and on the other quite parsimonious in terms of calculations, property which further enhances its applicability.