Joint Probabilistic Projection of Female and Male Life Expectancy

Adrian Raftery, University of Washington
Nevena Lalic, University of Washington

We propose a methodology for obtaining joint probabilistic projections of female and male life expectancy at birth (e0) for all the countries in the world, building on previous work by Chunn et al. (2010) which produces probabilistic projections for each sex independently. Our approach is to project the gap between female and male e0 based on projections of female e0 from the model of Chunn et al. Our autoregressive model for the gap is a function of female life expectancy and a t-distributed random perturbation, which allows for outliers. Our model takes account of mortality data limitations, is comparable across countries, and accounts for possible shocks such as armed conflicts. We estimate all parameters based on life expectancy estimates for 1950-2010 from the United Nations' World Population Prospects 2008. We evaluate our model using out-of-sample projections for 1995-2010.