Uses of projections (cont.)
Typically, a trade off exists for users of population projections between the information they need on each of the three dimensions just discussed: the time horizon of the forecasts and their subject and geographical detail.
For example, in the case of the education sector, short-term forecasts by single years of age are needed at the local level for forecasting enrolments. Primary-school-age children, in particular, usually attend a school within 1 or 2 km of their homes. However, there is little need or point in trying to predict school enrolments more than a few years ahead. In contrast, medium-term forecasts at the national level are needed for informing decisions about teacher training. However, there is no need to estimate future requirements for trained teachers for small geographical areas, as many teachers commute some distance to work and some of them will be prepared to migrate considerable distances in order to obtain a post or a promotion.
To take another example, the body responsible for providing water supplies to a metropolitan area has to consider possible trends in the population of that city in the long-term because it may require many decades to design, obtain funds and approval for, construct, and commission new reservoirs. However, this body will only be concerned on a short-term basis about where residential, commercial and industrial development is occurring within the city and will not be very interested in the ages and other characteristics of the metropolitan population that they serve.
It is difficult to think of any use for population forecasts that requires detailed, long-term forecasts for small areas. Users who think that they need such information probably lack any clear idea of what information they need and how they are going to make use of it.
This variation in the information that users of forecasts require means that the methods used to make projections need to be tailored to the use to which the forecasts will be put. Official forecasts face the challenge of providing for the needs of diverse users. Nevertheless, while some of these users will require long-term forecasts and others forecasts for local areas, it is difficult to identify a need for geographically-detailed long-term forecasts.
Moreover, not only do the outputs required by various users of projections differ, but so will the quantity and accuracy of the input data on the current population and past trends in vital rates that are available to the forecaster. In combination, these differences in the context in which projections are being undertaken implies that there is no single best way of forecasting population – the methods used must be adapted to fit the application.