Forecasting Active Life Expectancy and Population

Rogers, Andrei

Abstract

This proposal describe a study that would extend multistate demographic methods to a new problem area: the analysis and forecasting of active life expectancy and the future active/inactive populations. By active life expectancy is meant the expected duration of well-being---a status that signifies an individual's independence in the activities of daily living. Current unistate methods for estimating these expectancies are deficient in that they ignore the possibility of a return from dependent to independent status. This is a direct consequence of their reliance on conventional life table models---models that only permit exists from the population and no returns. Multistate life tables, on the other hand, accommodate both decrements and increments and focus simultaneously on several interacting subpopulations. Hence they are uniquely suited for analyses of the evolution of active life processes---processes in which about a quarter of those persons becoming dependent return to independent status. This project seeks to apply multistate models to the study of active life expectancy. It will develop improved life table methods for describing and accounting for observed patterns of active life expectancy, using statistical models that link transition probabilities to explanatory variables. And it will develop improved projection methods for forecasting active life expectancy and the future active/inactive aged populations.