PAPP103 - S04: Population projections (II): uses and accuracy of forecasts

Ex-ante assessments of the accuracy of forecasts

An ex ante assessment of accuracy is one that is made "in advance", that is at the time that the forecasts are produced. Developing ways of doing this is a difficult issue to address that has attracted a lot of research attention in recent years. The traditional approach to indicating the uncertainty inherent in projections was to produce variant high and low projections as well as the central ones. This strategy often confused users. Most of them used the central projections. It was usually unclear whether the variant projections defined any sort of confidence interval and whether combining extreme assumptions, for example high fertility with low mortality, produced a plausible scenario at all.

Broadly speaking, three other approaches have been proposed as ways of assessing how uncertain forecasts are at the time that they are made.

The first is to base ex-ante estimates of error on ex-post facto ones such as Keyfitz’s assessment of UN forecasts or Field’s analysis of the official British ones that have just been described. The main problems with this approach are:

• obtaining an adequate sample of comparable past forecasts
• most forecasters would like to believe that our forecasting procedures are getting better and that errors in future projections will be smaller than those in past ones.