FIGURE 3-17 Share of population aged 65+ in eight high-income countries, 1950–2050. SOURCE: United Nations (2011).
now to ameliorate the consequences of projected population change far in the future. How certain can we be that the projected changes will actually occur and that action is needed now? To answer this important question, we need an indication of the uncertainty in the projections.
Demographers and statisticians have mainly used four different methods to assess the uncertainty of population projections (see National Research Council, 2000, for a detailed examination of both the accuracy of past projections and the uncertainty of population forecasts). The traditional method, which can be called “scenarios,” is familiar to all: The projections are made in high, medium, and low variants, based on expert opinion about how high or low each of the key inputs—fertility, mortality, and net immigration—might be. This is certainly helpful, but there are difficulties with this approach. It seems to assume that if fertility (for example) is higher than expected in the first year of the projection, then it will also be higher in every subsequent year, and this assumption rules out the kinds of fluctuations that have occurred in the past. Because of this, the scenario method invites us to believe that if we just wait for a few years it will become clear whether the population is evolving according to the high or the low scenario, and uncertainty will be reduced. But this interpretation is mistaken. After a few years, a new set of scenarios would again feature similar high, medium, and low variants.
Construction of the scenarios also requires deciding whether to combine the high fertility assumption with a low mortality assumption or a high mortality assumption, and likewise for migration assumptions. This