A Review of Demographic Forecasting Models for Mortality

Abstract

The goal of Chapter 1 is to describe and comment on the methods and approaches that have been in use or have emerged in recent years. Section 1.1 introduces the most common classifications of forecasting models for mortality. Section 1.2 is devoted to a brief historical review of parameterisation functions. In this context, attention is paid to prediction based on parameterised age schedules, in particular by using time series models. Section 1.3 focuses on the (statistical association) models of Lee and Carter and Section 1.4 characterises the (log-linear) age-period-cohort models. In Section 1.5 the reader can find a review of the methods used in international statistical practice and in Section 1.6 the importance of uncertainty in forecasting is addressed. Section 1.7 outlines the prospects for modelling and forecasting mortality as seen from the perspective of this chapter. 1.1 | Most Common Classifications of Forecasting Models for Mortality A positive feature of forecasting mortality in developed countries is that adequate historical information is usually available, at least for aggregate measurements. In this case the two salient questions facing a forecaster are finding an accurate description of the past, and secondly, taking on judgemental factors in order to produce plausible forecasts. The first can be viewed as a technical problem essentially concerned with modelling and 1 E. Tabeau et al. (eds.), Forecasting Mortality in Developed Countries, 1-32.