The heuristic potential of cohort-component projections: Three essays in population dynamics

Most population projections are actually forecasts. Population parameters are meant to estimate the present characteristics of a population and to depict plausible assumptions of future demographic trends. Forecasting, however, should not constitute the exclusive use of population projections. The cohort-component model used in most population projections is first of all a model of population dynamics linking demographic processes and changes in population characteristics. The three essays presented here illustrate how a population projection model can be used to address issues in demographic research besides forecasting. The first essay uses counter-factual simulations, which are pure projections to the extent they are thought experiments, simply developing the implications of assumptions without reference to the actual course of population. Specifically, world population is projected from 1950 to 2000 with fertility and mortality rates constant at their mid-century levels. By difference with actual population, these simulations illuminate the respective contribution of the fertility and mortality declines to today\u27s population. Counterfactual simulations are further used to assess the impact of family planning programs on population size. The second essay attempts a demographic reconstruction of Cambodia in the 1970s. By incorporating the formal relationships between the different components of population change, such reconstruction allows to derive the range of estimates for a given component that is compatible with data available for the other components. This technique is applied to assess the extent of excess mortality during the Khmer Rouge regime. Cohort-component projections also provide an age pattern of excess mortality which allows further analysis of the likely causes of death corresponding to that mortality crisis. The last essay presents simulations of the potential impact of AIDS on population growth in sub-Saharan Africa. It concentrates on the sensitivity analysis of a particular model rather than on the selection of a most likely outcome (prediction). The analysis underscores the main gaps in our current knowledge about the epidemic and priorities for future research, even though accurate predictions are beyond reach due to the paucity of available data