Uncertainty in demographic projections and its consequences for the user

Abstract

In this paper, the authors first present the aims of demographic projections and the sources of uncertainty. After that, they discuss the consequences of the uncertainty on the quality of projections, in particular the methods to measure the error due to uncertainty. This paper ends with a presentation of methods to control uncertainty and with some recommendations to encourage users to take uncertainty into account.

New analytical methods, as well as best knowledge of underlying mechanisms modifying the occurrence of fertility, mortality or migrations, may improve to some extend the quality of demographic projections. However, the error avoided by a sophisticated projection method compared to a simple technique is relatively weak. The benefits of sophisticated techniques, with regards to simple techniques (extrapolations, for example), are not proved. Methods like ARIMA may be useful for short-range forecasts, while classical methods based on cohort component methods are to be recommended for long-term projections; structural modelling methods apply to simulation of the policy measures' effects. It may be possible to improve slightly the quality of demographic projections with the help of methods like APC models or with the use of causes of death or of birth order, and thus it may be possible to decrease uncertainty. But the cost to improve the forecast quality is probably higher than the benefit that results from it.

The time-range of the projection has to be adapted to the aim of the projection. Official projections would be revised every five years. Simulations can be hold on a longer range. In all cases, the user should be aware of the increasing uncertainty about the future evolution of the demographic processes. Uncertainty concerning the evolution of fertility, mortality and migratory flows will remain linked to demographic projections as to other forecasts. Users of projections are in the habit of dealing with uncertainty, notably by using stochastic projections. Stochastic projections help the decisionmaking by supposing that one cannot predict what will be the future, but rather give some information on the future demographic evolution.