Forecasting vital rates from demographic summary measures

Carlo Giovanni G. Camarda, Institut National d'Études Démographiques (INED)
Jose Manuel Aburto, University of Oxford

In population and actuarial sciences, time-trends of summary measures (such as life expectancy or the average number of children per woman) are easy to interpret and predict. Most summary measures are nonlinear functions of the vital rates, the key variable we usually want to estimate and forecast. Furthermore smooth outcomes of future age-specific vital rates are desirable. Therefore, optimization with nonlinear constraints in a smoothing setting is necessary. We propose a methodology that combines Sequential Quadratic Programming and a P-spline approach, allowing to forecast age-specific vital rates when future values of demographic summary measures are provided. We provide an application of the model on Italian mortality and Spanish fertility data.

Keywords: Population projections, forecasts, and estimations, Methodology, Mortality, Fertility and childbirth

See extended abstract.

  Presented in Session 80. Innovations in Methodology