Point and interval forecasts of age-specific life expectancies

This paper considers forecasting life table, and proposes a model averaging approach to improve point and interval forecast accuracy. Illustrated by data of eleven countries, we compare point and interval forecasts among ten principal component and two random walk methods. Based on averaged forecast errors, random walk with drift and Lee-Miller methods are two most accurate methods for producing point forecasts. By combining their forecasts, point forecast accuracy is improved. As measured by averaged coverage probability deviance, Hyndman-Ullah methods provide more accurate interval forecasts than Lee-Carter methods. By combining forecasts from two most accurate methods, interval forecast accuracy is improved.