Abstracts Business

A multi-regional adaptation of the ASSA2008 AIDS and Demographic Model for population projections

by Greaves Thabo. Mofokeng

Includes abstract. The traditional models used in population projections rely on the net migration method.The ASSA2008 AIDS and Demographic model is one such model. In this research, the nine provinces and the four population groups are aggregated to give rise to three regions. Using STATA12, the directional migration tables for the years covering the period 1996-2007 between the three regions, by age and sex, based on a 10% sample of the 2001 Census and a 2.5% sample of the 2007 Community Survey, are produced. Using MATLAB 2011a with built-in Levenberg-Marquardt algorithm with nonlinear least squares methods, Rogers-Castro multi-exponential age schedules are fitted to the census/survey migration data in order to obtain parameters used to estimate migration rates in the model for the period 1996-2007. After 2007, migration rates areextrapolated roughly linearly, assuming that migration will trend towards zero over a fixed number of years. The multi-regional adaptation of the ASSA model is tested and found to work, with a minor re-calibration to the HIV data for 2008. The projected regional population age structure and size implied by the model for 1996-2025 are consistent with the same estimates implied by the net migration model, and so are the projected net migration rates per 1,000. The level of the migration rates assumed in the multi-regional model accounts for an average of 89% of the change in the estimates of the population size relative to those generated by the net migration model, and the use of multi-regional modelling itself accounts for 11% of these changes. The proportions of the changes attributable to the level of migration rates assumed in the multi-regional model, and the use of the multi-regional modelling, show that the choice of the method by which population projections are done is important. Finally, the three-region model can be extended to a nine-province model that recognises that each province has unique demographic dynamics, but the construction of such a model requires a significant amount of extra work due to its size and complexity.