Forecasting Births in Post-Transition Populations: Stochastic Renewal with Serially Correlated Fertility

Demographic forecasting techniques fail with post-transition populations dominated by fluctuating fertility; time series analysis of fertility can improve the forecasts. This article develops the optimal forecast and its variance for births to an age-structured population subject to serially correlated random fertility. The white noise, first-order autoregressive, second-order autoregressive and random walk fertility specifications are analyzed, each leading to different forecasts and very different variances, as shown by illustrative applications to U.S. data, 1917–1972.