The contribution of migration to population change tends to vary inversely with the scale of the population being considered. Thus, over a five-year period, net migration into or out of a smallish district could be as high as 20 per cent, while for many countries it will be less than 1 per cent, and for the world it is highly likely to be zero. Because of the difficulty of predicting migration into and out of small areas, the accuracy of forecasts tends to vary inversely with the size of the population being projected.

The sensitivity of forecasts for different age groups to assumptions about each component of growth also varies. For example, in a 20-year forecast:

• the size of the projected population aged less than 20 will depend mainly on assumptions made about fertility
• the size of the projected population aged 20–75 years will depend mainly on forecasts of migration as losses due to mortality at these ages are small. Forecasts for these ages, therefore, tend to be relatively accurate at the national level in most situations
• the size of the projected population age 75 or more will depend to a greater degree on what is assumed about mortality at older ages. Unexpected trends in mortality can introduce large errors in forecasts for the elderly population, especially those aged 85 or more.

The level of migration can change rapidly in response to changes in governmental policies or economic trends.  It can increase or drop by a factor of two in a few years. Occasionally, it changes even more radically than this as a result of changes in the law or political developments. In contrast, fertility typically changes more slowly (perhaps by a factor of two per decade) and mortality yet more slowly again. Thus, migration is the least predictable component of population change in the short-term.

To predict the size of cohorts that have already been born in the base year involves assumptions about only mortality and migration. Thus, inaccuracies in forecasts of fertility have a minor impact on short-term population forecasts of the total size of the population. They matter only for those providing services for pregnant women and young children such as maternity and neonatal care, nursery care, and infant schools.

Nevertheless, the greatest uncertainty in population forecasts usually attaches to cohorts that have not been born in the base year. Thus, estimates of the number of children in the population are usually the most uncertain part of any medium-term forecast. Moreover, the scope for error increases greatly in long-term forecasts as their results come increasingly to depend on the accuracy with which they predict childbearing by women who are themselves not yet born.

Note that, because every one dies eventually, the mortality of cohorts of people who are too old to bear significant numbers of children has no effect on the rate of population growth in the very long term. Mortality in the first 50 years of life does influence long-term growth, but is now quite low in most of the world and relatively easy to predict.

Errors in the estimates of the base population can account for a substantial part of the errors in short-term forecasts. However, the relative importance of this component of total error shrinks as the time horizon of the forecast increases.

Another factor that influences likely errors is the level of fertility, mortality and migration. The lower a population’s vital rates, the smaller will be the absolute errors in the projected population that result from a given proportional error in forecasts of future vital rates.

The trend in mortality tends to be more volatile when mortality is high than when it is low. On the other hand, changes in fertility tend to be more rapid and erratic in contracepting populations than populations where fertility is high.

Thus, in general, medium-term forecasts for developed countries will be more sensitive to errors in the assumptions made about fertility than those for developing countries. In contrast, forecasts for low-income countries will be relatively vulnerable to errors originating in inaccuracies in the forecasts of future mortality.