Measuring in Advance the Accuracy of Population Forecasts

Keyfitz, N. (1989). Measuring in Advance the Accuracy of Population Forecasts. IIASA Working Paper. IIASA, Laxenburg, Austria: WP-89-072

Preview

Text WP-89-072.pdf Download (1MB) | Preview

Abstract

Supposing only that future variability in fertility, mortality improvement, and migration will be the same as past variability, and that it is the variability that creates the uncertainty in population forecasts, permits an ex ante estimate of uncertainty. This is calculated by taking the fertility level of a random past year, the mortality improvement of a random past five-year period, and the net immigration of a random past year. The future population that is shown by 1000 such random choices of each of the three input variables for each projection cycle gives the variability to which individual estimates are subject.

We need to check this ex ante calculation with what actually happens, and are able to do so for past projections, comparing a large number of medium projections made by the United Nations and other bodies with what subsequently occurred. It turned out that this ex post estimate was slightly higher than the ex ante as shown by the simulation.

Thus, contrary to appearances, the method of this paper does not exaggerate the variability hut if anything underestimates it.

Estimating variability is less important for the total than it is for certain ratios. Of these the most sought after is that of old people to the number who are at working age, a ratio that determines the tax rate needed to provide old age pensions. The uncertainty of this and any other function of future population can be found by the same simulation.

Actions (login required)

View Item