

National Health Insurance Estimates of the Population Download full PDF Document The establishment of an accurate population for South Africa and the nine provinces is not a simple matter and official figures are revised at times. Possible sources of population data and some brief features of each are discussed below and illustrated in Figure 1. Census 2001 from StatsSA: this was the latest full census and provides a single point estimate for 2001. A 10% sample is available for more detailed breakdowns. However this data is now eight years old and there are significant deficiencies in the census which need to be taken into account 1. It is the most detailed provincial, district and municipal data but is not useful for medical scheme membership and income data is poor. The next Census is due in 2011 and future censuses are likely to be at ten year gaps, not five years as in the 1990s. Community Survey 2007 (CS2007) from StatsSA: this “mini-census” was much smaller than Census 2001 with a sample of 2.5% of households but is the largest survey conducted by StatsSA. Information was collected from some 247,000 households on population dynamics, disability and social grants, school attendance and educational attainment, labour force participation and income 2. Some provinces had major changes in total population compared to Census 2001, for example the Western Cape showed a 16.7% increase. Provincial, district and municipal data are available but detailed use of figures at district level should be treated with great caution. StatsSA Mid-Year Population Estimates: these are produced annually and give estimates of the national and provincial populations as at 30 June each year. The latest estimate 3 contained a major revision of all historic figures from 2001 to 2008, using fertility, mortality and migration data derived from CS2007. Three “variants” of the historic total population have now been produced: Low variant: 47.8 million (linked to Census 2001); Medium variant: 48.7 million. “Best estimate” (lower out-migration for Whites, higher in-migration African and Asian, revised mortality and fertility Coloured); and High variant : 49.0 million (CS2007 not adjusted by Census 2001 or 1996). The mid-year estimates have provincial and ethnicity breakdowns by age and gender. There was a substantial adjustment to provincial estimates, informed by the findings of CS2007. These figures have the advantage of being the official population estimates in the country but projections beyond 2008 have not been released as yet. Actuarial Society of SA model (ASSA2003): the Actuarial Society of South Africa (ASSA) has developed a demographic and AIDS model that makes use of data from several sources to project the potential course and demographic impact of the HIV/AIDS epidemic 4. The model is modified as more data and epidemiological information becomes available. The latest version of the model, ASSA2003, was released in November 2005, calibrated with data up to and including the 2003 antenatal prevalence survey and registered deaths up to 2003. A series of nine provincial models was simultaneously released and made available to researchers. The population can be projected for each province from 1985 to a chosen date in the future, for example 2025 as used in this policy brief. General Household Surveys (GHS) and Income and Expenditure Surveys (IES) from StatsSA: these are very small studies (GHS2005 had some 30,000 households) but they have many more variables and are thus useful in social security modelling. It is important to take account of the weighting used for these surveys, for example, the GHS2005 population is weighted using Census 2001 updated to July 2005. The GHS2005 has been used extensively for patterns in social security modelling and these are applied to more robust and more recent measures of the total population like the StatsSA mid-year estimates 2008 or the ASSA population projections. While data from StatsSA is readily available, some of it needs to be used with caution and preferably with adjustments by professional demographers. For example, the Census 2001 data needs substantial adjustment before use 1. This includes dealing with under-enumeration of the 0-4 year olds; too few foreigners identified; age misstatement, particularly age exaggeration (particularly across the pension age for both males and females); too few male in-migrants and/or significant male undercount (relative to the number of females); an excess of teenagers; potential significant undercount of the White population; and slightly greater than expected Coloured population. The graph below compares the StatsSA population estimates for South Africa and the ASSA2003 projection to illustrate the differences in the estimates made at different points in time. Figure 1 : Comparison of Population Estimates and Projections for South Africa The graph shows that there are considerable differences in the projections from StatsSA dated up to 2007 and the results from CS2007. As a result, the historic population was re-calculated back to 2001 and a range of projections given. The ASSA2003 model 5 is by far the most consistent and useful projection available. Despite not yet being adjusted for the findings of CS2007, the projected population is closer to the revised StatsSA 2008 figures than those produced by StatsSA up to 2007. An update of ASSA2003 is under consideration but has not yet begun and it seems unlikely there will be a revised model until late 2009 or perhaps early 2010. The ASSA2003 projection is of course available into the future which is essential for costing the implications of any NHI structure. The version being used allows for separate projections for each province. The ASSA provincial model produces (amongst other results) the expected total population, the population structure by age and gender and expected total births. All of these can be used directly in models of National Health Insurance. The model also provides a projection of the numbers who are HIV+, the numbers of AIDS sick and the numbers expected to need anti-retroviral treatment. Projections of orphans are also directly available for social grant planning by the provinces. It is strongly recommended that all costing work on National Health Insurance be done using the ASSA2003 provincial model and that the costings be updated when a revision to the model is released. The ASSA2003 provincial tables by age and gender from 1985 to 2025, in an age format needed for costing, can be downloaded from the IMSA web-site. << Back to NHI Project List	Contact Details: Innovative Medicines SA Val Beaumont P.O. Box 2008 Houghton, 2041 Tel: +27 11 880-4644 Fax: +27 11 880-5987 Contact Us! IMSA: Published Profile of Member Companies and Position Statements Deloitte Final Report: Feb 2010

National Health Insurance
Estimates of the Population

The establishment of an accurate population for South Africa and the nine provinces is not a simple matter and official figures are revised at times. Possible sources of population data and some brief features of each are discussed below and illustrated in Figure 1.

Census 2001 from StatsSA: this was the latest full census and provides a single point estimate for 2001. A 10% sample is available for more detailed breakdowns. However this data is now eight years old and there are significant deficiencies in the census which need to be taken into account 1. It is the most detailed provincial, district and municipal data but is not useful for medical scheme membership and income data is poor. The next Census is due in 2011 and future censuses are likely to be at ten year gaps, not five years as in the 1990s.

Community Survey 2007 (CS2007) from StatsSA: this “mini-census” was much smaller than Census 2001 with a sample of 2.5% of households but is the largest survey conducted by StatsSA. Information was collected from some 247,000 households on population dynamics, disability and social grants, school attendance and educational attainment, labour force participation and income 2. Some provinces had major changes in total population compared to Census 2001, for example the Western Cape showed a 16.7% increase. Provincial, district and municipal data are available but detailed use of figures at district level should be treated with great caution.

StatsSA Mid-Year Population Estimates: these are produced annually and give estimates of the national and provincial populations as at 30 June each year. The latest estimate 3 contained a major revision of all historic figures from 2001 to 2008, using fertility, mortality and migration data derived from CS2007. Three “variants” of the historic total population have now been produced:

Low variant: 47.8 million (linked to Census 2001);
Medium variant: 48.7 million. “Best estimate” (lower out-migration for Whites, higher in-migration African and Asian, revised mortality and fertility Coloured); and
High variant : 49.0 million (CS2007 not adjusted by Census 2001 or 1996).

The mid-year estimates have provincial and ethnicity breakdowns by age and gender. There was a substantial adjustment to provincial estimates, informed by the findings of CS2007. These figures have the advantage of being the official population estimates in the country but projections beyond 2008 have not been released as yet.

Actuarial Society of SA model (ASSA2003): the Actuarial Society of South Africa (ASSA) has developed a demographic and AIDS model that makes use of data from several sources to project the potential course and demographic impact of the HIV/AIDS epidemic 4. The model is modified as more data and epidemiological information becomes available. The latest version of the model, ASSA2003, was released in November 2005, calibrated with data up to and including the 2003 antenatal prevalence survey and registered deaths up to 2003. A series of nine provincial models was simultaneously released and made available to researchers. The population can be projected for each province from 1985 to a chosen date in the future, for example 2025 as used in this policy brief.

General Household Surveys (GHS) and Income and Expenditure Surveys (IES) from StatsSA: these are very small studies (GHS2005 had some 30,000 households) but they have many more variables and are thus useful in social security modelling. It is important to take account of the weighting used for these surveys, for example, the GHS2005 population is weighted using Census 2001 updated to July 2005. The GHS2005 has been used extensively for patterns in social security modelling and these are applied to more robust and more recent measures of the total population like the StatsSA mid-year estimates 2008 or the ASSA population projections.

While data from StatsSA is readily available, some of it needs to be used with caution and preferably with adjustments by professional demographers. For example, the Census 2001 data needs substantial adjustment before use 1. This includes dealing with under-enumeration of the 0-4 year olds; too few foreigners identified; age misstatement, particularly age exaggeration (particularly across the pension age for both males and females); too few male in-migrants and/or significant male undercount (relative to the number of females); an excess of teenagers; potential significant undercount of the White population; and slightly greater than expected Coloured population.
The graph below compares the StatsSA population estimates for South Africa and the ASSA2003 projection to illustrate the differences in the estimates made at different points in time.

Comparison of populations and estimates for SA (graph)
Figure 1 : Comparison of Population Estimates and Projections for South Africa

The graph shows that there are considerable differences in the projections from StatsSA dated up to 2007 and the results from CS2007. As a result, the historic population was re-calculated back to 2001 and a range of projections given. The ASSA2003 model 5 is by far the most consistent and useful projection available. Despite not yet being adjusted for the findings of CS2007, the projected population is closer to the revised StatsSA 2008 figures than those produced by StatsSA up to 2007. An update of ASSA2003 is under consideration but has not yet begun and it seems unlikely there will be a revised model until late 2009 or perhaps early 2010. The ASSA2003 projection is of course available into the future which is essential for costing the implications of any NHI structure. The version being used allows for separate projections for each province.

The ASSA provincial model produces (amongst other results) the expected total population, the population structure by age and gender and expected total births. All of these can be used directly in models of National Health Insurance. The model also provides a projection of the numbers who are HIV+, the numbers of AIDS sick and the numbers expected to need anti-retroviral treatment. Projections of orphans are also directly available for social grant planning by the provinces.

It is strongly recommended that all costing work on National Health Insurance be done using the ASSA2003 provincial model and that the costings be updated when a revision to the model is released.

The ASSA2003 provincial tables by age and gender from 1985 to 2025, in an age format needed for costing, can be downloaded from the IMSA web-site.

<< Back to NHI Project List

Contact Details:

Innovative Medicines SA
Val Beaumont

P.O. Box 2008
Houghton, 2041

Tel: +27 11 880-4644

Fax: +27 11 880-5987

IMSA: Published Profile of Member Companies and Position Statements

Deloitte Final Report: Feb 2010

National Health Insurance Estimates of the Population

Contact Details:

National Health Insurance
Estimates of the Population