The main part of this paper is concerned with the boldest and most comprehensive form of demographic prediction—the forecasting of total populations. That form may be regarded as subsuming many more limited types of prediction. In practice, the method of approach in forecasting total populations has not infrequently been very different from that followed in attempting to predict the behaviour of the specific components of population growth. But conceptually, at least, the prediction of total populations represents an extension of the efforts made since the beginnings of demography to derive order from the data on mortality and fertility. Such order has often been found. But attempts to use it as a basis for prediction have not been very successful, and the broader the front the less successful has been the product. Nor are there grounds for believing that long-range, overall predictions are likely to be much more reliable in the future. Nevertheless, the attempts have not been uninstructive. And it is of interest to consider them in a wider historical context—that of the endeavours of demographers to generalize and to discover regularities relating to human populations. These endeavours, which have naturally varied widely in their explicitness, may be grouped under two heads. First, attempts which, generalizing from a particular situation, assumed that the generalization would have a wide applicability, whether or not the absolute levels of fertility or mortality were different. Secondly, generalizations which explicitly included time and change in their framework. Examples of both types are found at various stages in the development of demography. One example is indeed already present in John Graunt’s
Natural and Political Observations
, the first specifically demographic publication—an attempt to formulate what later came to be referred to as the ̒law of mortality’ (1)*.
Using Graphical Features To Improve Demographic Prediction From Smart Phone Data
