Nathan Keyfitz

The Limits of Population Forecasting

The user of forecasts needs to know how far published estimates of future population can be relied on. The most compact way of describing accuracy is by showing the average error of past forecasts. From over a thousand comparisons of forecast with realization--nationalregional and global and covering the past 30 years--the conclusion emerges that the odds are two to one that a forecast rate of increase [plus or minus] 0.4 percentage points will straddle the realized rate of increase over future periods. That means in practice useable forecasts for the next 5 to 20 years virtually no information on the population 100 years hence. (summary in FRE SPA)

On the momentum of population growth

AbstractIf age-specific birth rates drop immediately to the level of bare replacement the ultimate stationary number of a population will be given by (9):

What difference would it make if cancer were eradicated? An examination of the taeuber paradox

\left( {{\textstyle{{b\mathop e\limits^ \bullet {}_0} \over {r\mu }}}} \right)\left( {\frac{{R_0 - 1}}{{R_0 }}} \right)

Choice of function for mortality analysis: Effective forecasting depends on a minimum parameter representation

multiplied by the present number, where b is the birth rate, r the rate of increase,

Can knowledge improve forecasts

\mathop e\limits^ \bullet _0

Mortality in a heterogeneous population

the expectation of life, and R0 the Net Reproduction Rate, all before the drop in fertility, and μ the mean age of childbearing afterwards. This expression is derived in the first place for females on the stable assumption; extension to both sexes is provided, and comparison with real populations shows the numerical error to be small where fertility has not yet started to drop. The result (9) tells how the lower limit of the ultimate population depends on parameters of the existing population, and for values typical of underdeveloped countries works out to about 1. 6. If a delay of 15 years occurs before the drop of the birth rate to replacement the population will multiply by over 2. 5 before attaining stationarity. The ultimate population actually reached will be higher insofar as death rates continue to improve. If stability cannot be assumed the ultimate stationary population is provided by the more general expression (7), which is still easier to calculate than a detailed projection.