Prospective life tables

Prospective life tables depend on forecasting age-specific mortality. Considerable attention has 2 been paid to methods for forecasting mortality in recent years. Much of this work has grown 3 out of the seminal Lee-Carter method (Lee & Carter 1992). Other extrapolative approaches 4 use Bayesian modelling, generalized linear modelling and state-space approaches. Methods for 5 forecasting mortality have been extensively reviewed by Booth (2006) and Booth & Tickle 6 (2008). This chapter covers various extrapolative methods for forecasting age-specific central 7 death rates. Also covered is the derivation of stochastic life expectancy forecasts based on 8 mortality forecasts. 9 The main packages on CRAN for implementing life tables and mortality modelling are de10 mography (Hyndman 2012) and MortalitySmooth (Camarda 2012) and we will concentrate 11 on the methods implemented in those packages. However, mention is also made of other extrap12 olative approaches, and related R packages where these exist. 13 We will use, as a vehicle of illustration, US mortality data from 1950. This can be extracted 14 from the Human Mortality Database (2013) using the demography package. 15