Forecasting Mortality by Using Statistical Moments

Adam Lenart, Max Planck Odense Center and University of Southern Denmark
Marius Pascariu, Max Planck Odense Center

Forecasting mortality by predicting the moments of the distribution of deaths leads to coherent results that can flexibly incorporate exogenous information such as the mortality experience of neighboring (Li-Lee 2005) or world record countries (Torri-Vaupel, 2012). Forecasting the moments of the distribution of deaths yields the further advantage of reducing the forecast dimension by requiring the projection of a lower number of moments than the number of age classes used by other methods (e.g., Lee-Carter 1992, Renshaw-Haberman 2003). In the present paper, a method allowing to determining the age-schedule of death rates is presented, by forecasting a number of statistical moments and reconstruction of the density function, starting from the obtained results. The method is back tested using US female data.

  See extended abstract

Presented in Poster Session 2: Data and Methods/Applied Demography/ Spatial Demography/ Demography of Crime