On Future Mortality

Population projection is the activity of demographers that is best known to the public. The record shows some successes, in which the projected population came close to the population that subsequently materialized, and some failures, in which the two were far apart. The accurate forecasting of population using nothing hut demographic data is impossible, but two things can be done: marginal improvements in accuracy can be made, and the accuracy likely to be attained can be estimated in advance. Since the future population of any area depends on the three components (future births, deaths, and migration) in a simple accounting identity, its forecasting comes down to forecasting these components. The present paper looks into the mortality component, examining past mortality on the basis of Canadian data for the period 1921 to 1981. The examination shows that which past interval one takes as the indication of the pace of future population improvement is the most important element of the forecast of mortality. One reason that this finding is useful is that it enables the range of uncertainty in future mortality to be estimated from the range within which the future life table falls when we assume the pace of improvement of various past periods. Application of the same principle to fertility and migration will enable a calculation to be made of the uncertainty of population projections. This broader matter will be developed in a paper shortly to appear