Dynamic Forecast of U.S. Life Expectancy at Birth within a Bayesian Framework

Christina Bohk, University of Rostock
Roland Rau, University of Rostock

Forecasting US mortality is challenging, particularly for women. From the 1980s to the early 2000s, increases in life expectancy have fallen behind those of other high-income countries. It has been suggested that life style factors, especially smoking, are responsible for these slow increases. Since smoking prevalences are expected to decrease among younger birth cohorts, US mortality is likely to fall more rapidly in the future than during the last 30 years. Consequently, life expectancy will probably approach international trends again. We propose a novel method to forecast mortality, which uses rates of mortality improvement and combines objective and subjective information within a Bayesian framework. Besides a forecast until 2050, we demonstrate in a retrospective application that our approach would have resulted in more accurate estimates for US life expectancy than the original Lee-Carter model and two of its refinements proposed by Renshaw and Haberman.