Centre for Actuarial Research
(CARE)
A Research Unit of the University of Cape Town
Population projections for the Western Cape 2001 to 2025
Prepared by: Rob Dorrington
Centre for Actuarial Research
Prepared for: PGWC
Date: November 2005
Centre for Actuarial Research
University of Cape Town
Private Bag
Rondebosch
7701 SOUTH AFRICA
Telephone: +27 (21) 650-2475
Fax: +27 (21) 689-7580
E-mail:
ii
Synopsis
This report describes the results of the first part of the work undertaken to project the population of the Western Cape to the year 2025 taking into account, inter alia, the impact of HIV/AIDS. The report discusses the reconstruction of the 2001 provincial population, the model and underlying demographic and HIV/AIDS parametisation of the model, and the results of projections for the Western Cape as a whole. This is follow by a discussion on the derivation of models for the municipalities and some tables of results.
Population projection at the provincial level is at the best of times an uncertain business, but in the case of this project past events and the fact that the projections have to incorporate the impact of HIV/AIDS conspire to make these projections even less certain. Even if we confine the local level projections to populations supposedly covered by the health districts for which we have some measure of prevalence, there is no certainty that the prevalence is an accurate measure of the prevalence for the health districts. It is quite possible that clinics near the borders of the health districts also service people from neighbouring health districts.
The provincial projections were achieved by adapting the ASSA2003 model developed by the Actuarial Society of South Africa to correspond to the situation in the Western Cape. Projections were carried out under the assumption that MTCTP was introduced in 2001 and HAART was available since 2000 but significantly so since 2003 and that coverage reached 90% and nearly 60% respectively in 2005.
A summary of the main results are as follows:
· The provincial population is expected to reach 5,5 million by 2025. By 2015 about 29% of the population will be black African, less than 20% White, while the Coloured population is expected to remain above 50%. These proportions are expected to change only slowly thereafter..
· The impact of the mother-to-child-transmission prevention programme is having a significant impact on reducing infant and childhood mortality. ART is expected to have an immediate impact on the mortality of adults but the extent of this impact diminishes over time since ART is only expected to increase life expectancy by some five years on average. The impact of the other interventions on mortality is likely to be fairly small in the short-term.
· Prevalence of women attending public antenatal clinics is not expected (with the assumed package of preventative measures) to rise much above 17% while the overall prevalence is not expected to rise much above 6% or 250 000 infected people in total. By 2015 with treatment about 15% of these infected people will be AIDS sick and around 65% will be asymptomatic.
This report is included on a CD which contains the spreadsheets which can be used to make these and other projections. However, the user is warned that there is a substantial margin of uncertainty in the output from the models, a margin which increases with the period of projection being used and as the size of the population being considered reduces, and the results of projections should be used with this in mind.
Table of contents
Synopsis i
Table of contents iii
Introduction 1
Provincial population projection 3
1. Introduction 3
2. Method 4
2.1 Adaptations to the ASSA2003 national model 4
2.2 Orphans 7
3. Results 7
Projections by municipal districts 17
1. Introduction 17
2. Method 17
2.1 Brief description of the method 17
3. Results 18
Conclusions 23
Appendix A: Terms of reference 27
Appendix B: The Base Population as at October 2001 30
2. Deficiencies in the Census 31
3.1 Coloured 33
3.2 African 35
3.3 White 37
3.4 Indian 39
3.5 Total 41
Appendix C: Derivation of net migration by age from the reconciliation of the 1996 census results with the 1991 census results by the 1996 provincial boundaries 42
1. Derive the 1996 mid-year provincial population estimates (by race, sex and age) 42
2. Derive the 1991 mid-year provincial population estimates (by race and sex) 42
3. Derive the 1991-96 net (of mortality and out migration) in-migration (by province, race and sex) 42
4. Derive black net international in-migration (by province, sex and age). 44
5. Derive the number of white net international in-migrants (negative) (by province and sex) 44
6. Derive net provincial provincial in-migration (by province, race, sex and age) 44
7. Determine the trend over time 44
Appendix D: Instructions for running the models 45
1. Instructions for running ASSA2003WC_051124 45
2. Instructions for running OrphansCombined3f.xls 45
Appendix E: Total Provincial Population In The Middle Of The Year 46
1
CARE PGWC Population Projections
Introduction
This report describes the results of work undertaken to project the population of the Western Cape to the year 2025 taking into account, inter alia, the impact of HIV/AIDS. The report is in two parts, the first discussing the reconstruction of the 2001 provincial population, the model and underlying demographic and HIV/AIDS parameterisation of the model, and the results of projections for the Western Cape as a whole. The second part describes extension of these results to provide projections of the population at the municipal level.
The terms of reference for this project are given in Appendix A.
Estimating and projecting the South African population has never been an easy task because of the paucity of data and the heterogeneous and migratory nature of the population. Such work is an order of magnitude more difficult at a provincial and sub-provincial level where migration is far more significant and not documented at all, and where there may be poorly understood provincial and regional deviations in fertility and mortality.
However, the task becomes almost impossible when the country is in the process of transformation. Not only do the systems of record keeping change (or in some cases break down completely) but there is also a change in personnel usually leading to a loss of continuity and institutional memory. The impact of such change is magnified when the base population one is trying to project from has to be derived from a census that nationally is estimated to have missed at least one in every six people.
In addition to all this the country is suffering what could turn out to be one of the worst HIV/AIDS epidemics in the world and there is only limited understanding of how this will shape our demographic future. This uncertainty is increased by the possibility of interventions in future and, in the case of the Western Cape, because the province is still in the relatively early stages of the epidemic.
Although, as demographers are careful to point out, the results of such work are projections, in other words the deterministic outcome that results from a given set of assumptions, and not predictions, it is likely that most users will treat them as, in some sense, best estimates. Thus it is important to stress that even if these results are regarded as best estimates there is always a degree of uncertainty surrounding any population estimates, and that this uncertainty can be expected to increase with increasing time horizons of the projections. Obviously the uncertainty is increased by an order of magnitude when trying to anticipate the outcome of a complex and stigmatised epidemic, and again when the national statistics so poorly record demographic events in the country.
A number of files will be included on the CD:
ASSA2003WC_051124.xls (the population projection model for the Western Cape)
ASSA2003CT_051115.xls (the population projection model for the City of Cape Town)
29 files, one for each of the municipalities, with the following naming structure ASSA2003MC_MN.xls where “MC” is the municipal code and “MN” is the name of the municipality
Municipal tables.xls (Excel version of the tables in Part 2 of the report)
OrphansCombined3f.xls (a workbook used to extract output from the projection model which is necessary to project the number of orphans)
Part1graphs.xls (a sheet containing the data used to produce the graphs in Part 1 of the report)
Report.doc (a MSWord version of this report)
ASSA3003UserGuide.doc (manual for running the ASSA2003 models and accessing the output)
OrphansModel.doc (short note on how to produce numbers of orphans using OrphansCombined3f.xls)
Later in the year a training course on use of the models will be offered.
Provincial population projection
1. Introduction
This part of the report describes efforts to reconstruct the 2001 provincial population and the construction and parameterisation of the model for projecting the population of the Western Cape to the year 2025.
In undertaking such an exercise the following problems have to be dealt with:
1. Estimating the underlying projected mortality and fertility had there not been an HIV/AIDS epidemic.
2. Estimating the extent of both international and inter-provincial migration into the province net of migration out of the province.
3. Estimating the expected level and patterns of the HIV/AIDS epidemic in the province, which in turn must take into account:
- Uniqueness of the Western Cape with respect to the demographic mix and usage of public antenatal clinics
- The change in antenatal sampling protocol in 1997
- Efforts of the authorities to intervene, including resultant changes in behaviour
4. Estimating the extent of miscount in the 2001 census.
The full version of the ASSA2003 AIDS and Demographic impact model, developed by the Actuarial Society of South Africa, was adapted to take into account these unique provincial features. The model is included on the accompanying CD (ASSA2003WC_051124.xls) and summary output (WCSummaryOutput.xls) is included as part of this report, which too, is included on the CD (Report.doc).
2. Method
2.1 Adaptations to the ASSA2003 national model
The starting point for projecting the population in the Western Cape was to adapt the ASSA2003 full version of the AIDS and Demographic projection model developed by the Actuarial Society of South Africa, to be applicable to the Western Cape. One of the reasons for using this model is that it models the impact by population group separately and thus allows us to incorporate the significantly different racial profile in the Western Cape and particularly its municipal districts.
The model was adapted in the following ways.
1. Demographic assumptions
a. Base population: The 1996 and 2001 census populations for the province were adjusted for deficiencies identified at the national level (Stats Council 1998, 2004) on the assumption that similar deficiencies would be found at the provincial level. Since there were some inconsistencies between the numbers in 1996 and 2001 (e.g. the numbers of teenagers and older people in 2001 do not coincide with the numbers projected from 1996) it was decided to reconstruct the base population in 1985 to be consistent with an estimate of the population in 2001 that takes into account both census estimates. The derivation of this population is described in more detail in Appendix B.
b. Non-HIV/AIDS mortality: For Coloured, Indian and White population groups the mortality rates were assumed to be the same as those for the country as a whole (derived, in the main, so that the level and shape changed linearly from 1985 based mainly on Central Statistics (1987) and rates for 2001 derived by Dorrington, Moultrie and Timaeus (2004)), mainly because the data are too pauce to assert any differences between the provinces. For the African population on the other hand the populations are big enough to distinguish between the provinces, however, the only mortality rates available are those for the country as a whole in 1984-86 (Dorrington, Bradshaw and Wegner 1999) and provincial rates for 2001 (Dorrington, Moultrie and Timaeus (2004), and child mortality rates to up to around 1994 derived from the South African DHS data (Dorrington, Timaeus, Moultrie and Nannan, 2004). Rates for the provinces were derived by deriving a trend in the ratio of provincial child and adult mortality rates to the national child and adult mortality rates over time and applying this to the national mortality curve to produce a smooth set of mortality rates for each year from 1985 to 2001.
c. Fertility: As with the mortality national rates were used for Coloured, Indian and White populations and province-specific rates for the African population. These rates were derived from Moultrie and Dorrington (2004) and Moultrie and Timaeus (2002).
2. Migration
Migration is the demographic variable which is most difficult to predict, but this is even more so when there are no past patterns to use as a starting point. (In this case not only were the questions in the census not particularly successful in identifying migrants, but the current provincial boundaries did not exist many years before the 1996 census.) Thus the first thing that had to be done was to derive estimates for net in-migration (both international and inter-provincial) for the 1985-96 period. The details of the approach adopted, the assumptions made and some summary results appear in Appendix B.
In brief the net international in-migration (positive for the black African population group, negative for White population group and zero for the rest of the population) identified in the reconstruction of the national population for the ASSA2003 model up to 1996, was distributed amongst the black African provincial populations in proportion to the numbers of international in-migrants identified in the census, and amongst the white provincial populations in proportion to numbers of domestic out-migrants identified in the census. The numbers of domestic migrants was determined by comparison of the numbers in the reconstructed 1996 census population with those estimate for 1991 according to the current provincial boundaries, after taking into account the international migration.
For the period between the censuses the migrants were derived from the data on province of previous residence for those who were in different provinces at the two census dates. To the extent that this failed to produce sufficient numbers of Africans in the 20-29 year age group the shortage in the projected numbers was assumed to represent unrecorded immigration. For the White population the numbers of emigrants estimated nationally (in similar vain as those prior to 1996) were distributed by province in proportion to the numbers of emigrants identified by the reconstructions of the numbers of South African born people at the time of the census.