Report to the Chief Strategy Officer

Report to the Chief Strategy Officer

A Technical Note on Long-term Population Estimates for eThekwini

Prepared by the Research and Policy Advocacy Department

Draft 1. November 2016.

Contents

Introduction

Short-term estimates

Table 1: SSA Mid-year estimates demographic assumptions.

Table2: SSA Mid-year population estimates: eThekwini

Figure 1: eThekwini Population Pyramid 2016.

Long term estimates.

Table 3: eThekwini / KwaZulu-Natal Population Ratio 2002 to 2020

Figure 2: ETH/KZN Ratios 2002 to 2020

Figure 3: eThekwini Population 2002 to 2060

Figure 4: KZN Population Pyramid 2060

Table 4: eThekwini 2016 and KZN 2060 age bands

Low and High estimates

Figure 5: eThekwini Projections, SSA and Kramer.

Table 5: Kramer Scenario ratios to Best Estimate

Figure 6: eThekwini population Projection Scenarios.

Recommendation

Appendices

Appendix 1: A methodology for Long term population projections: The approach used by Statistics South Africa

Appendix 2: SSA: Short Term Population Forecasts for eThekwini

Appendix 3: KZN and ETH projected population 2021 to 2060

Appendix 4: Kramer population projection scenarios.

Appendix 5: Lower, SSA estimate and Higher Population Projections.

References

Introduction

In collaboration with Statistics South Africa (SSA): Demography Unit the eThekwini Office of Strategy Management has a demographic forecast for eThekwini until 2060. SSA have expressed the view that demographic projections become less dependable the longer the time scale and the smaller the geographic area of projection. For these reasons SSA produces National and Provincial forecasts until 2060 but for Metros and DCs not beyond 5 years from the current year. Statistics South Africa (SSA) utilises fertility, mortality, HIV / AIDS and migration factors to produce official short and long term population projections. A fuller description of SSA’s 2015 methodology for long term population projections is included in Appendix 1.

The purpose of this technical note is to discuss the methodology used to derive eThekwini’s future population from its share of the official KwaZulu-Natal 2060 population projection. The means of calculating the projected eThekwini share will be to plot the ratio of the eThekwini to KZN mid-year estimates from 2002 to 2020 and to use log regression of the ratio to estimate eThekwini’s future share of KZN. The implications of this projection fall outside the scope of this technical note.

Short-term estimates

National and Provincial Mid-year Population estimates are published annually as Statistical Release P0302. As discussed in Annexure 1 SSA uses Spectrum Software and the DemProj module which has been adapted by the SAS Institute for SSA in order to do estimates at national and provincial level. The inputs to DemProj include the following: total fertility rate (TFR) and the life expectancy at birth for all population groups, age-specific fertility rate (ASFR) trend, sex ratios at birth and net international and provincial migration, HIV infections, ratio of new HIV infections.

The demographic assumptions for KZN as published in the mid-year estimates are in Table 1. The key points are that life expectancy is increasing for males and females with 2016 female life expectancy 58.7 years and for males 54 years. Fertility rates are decreasing and are 3.08 in 2016. Net migration is reducing and in 2016 is estimated to be only -684 in 2016.

Table 1: SSA Mid-year estimates demographic assumptions.

In addition to the National and Provincial estimates in P0302 official short term demographic forecasts for Metros and DCs eThekwini are undertaken by Statistics South Africa (SSA) Demographics Unit. The forecast in Table 2 indicates that the population of eThekwini will grow by 175 thousand between 2016 and 2020 when the population total will be 3.85 million. This projection includes population by gender and age cohorts and can be seen in Annexure 2.

Table2: SSA Mid-year population estimates: eThekwini

The eThekwini population pyramid for 2016 according to the SSA estimates can be seen in Figure 1, where 63% are below the age of 35, 28% between 35 and 59 years, and 9% over 60 years.

Figure 1: eThekwini Population Pyramid 2016.

Long term estimates.

SSA estimate the KwaZulu-Natal (KZN) population to be 15.2 million in 2060, which is a growth of 4.1 million from 11.1 million in 2016.

The eThekwini share of the KZN population was calculated using the following steps:

• Calculation of the eThekwini / KZN population ratio based on the short term estimates from 2002 till 2020. See Table 3.
• Calculation of the log regression formula based on the ETH/KZN ratios between 2002 and 2020
• Applying the log regression formula to the KZN long term estimates to derive the eThekwini long term estimate.

The SSA historical short term estimates for ETH and KZN from the periods 2002 to 2020 were collated in Table 3. From these estimates the ETH/KZN ratio was calculated.

Table 3: eThekwini / KwaZulu-Natal Population Ratio 2002 to 2020

The ETH/KZN ratios were plotted and the log regression formula of y = -0.004*LN(x) + 0.3429 was calculated. See Figure 2.

Figure 2: ETH/KZN Ratios 2002 to 2020

The log regression formula was to calculate the ETH/KZN ratio and the KZN projected population was multiplied by the ratio to derive the ETH population as can be seen in Appendix 3: KZN and ETH projected population 2021 to 2060. These results are graphed in Figure 3 where the population in 2060 is 4.97 million, which is an increase of about 1.3 million compared to the 2016 total.

Figure 3: eThekwini Population 2002 to 2060

The KZN population pyramid for 2060 is in Figure 4. Compared to the eThekwini pyramid for 2016 KZN has a lower percentage population between the ages of 0 to 14 and a higher percentage population over 60 years. See Table 4.

Figure 4: KZN Population Pyramid 2060

Table 4: eThekwini 2016 and KZN 2060 age bands

Low and High estimates

Lower and higher estimates are utilised to cater for uncertainties regarding population projections. Statistics South Africa publish their best estimate only. However, Kramer in 2004, outlines 12 twelve scenarios for eThekwini which can be seen in Appendix 4.

According to Anderson and van Zyl, (2002. p2) “The accuracy of projections can be assessed by comparing the projected population by age and sex for a given date with the actual population at that date”. The relevance of using Kramer’s scenarios to calculate lower and higher scenarios around the SSA projection can be assessed by comparing Kramer’s Best Estimate to the SSA Projection, see Figure 5. Contrasting the two projections over the period 2002 till 2030, Kramer’s total population projection varies from SSA’s projection at most by 3% in a year and the lowest variance was 0.2%.

Figure 5: eThekwini Projections, SSA and Kramer.

Given the closeness of the two projections Kramer’s scenarios will be used to calculate variations from the SSA projections. There are eight scenarios, four above and four below the best estimate for migration and ARV access.

• Higher Migration (15% above best estimate) and All access to ARVs
• Higher Migration (15% above best estimate) and Best estimate access to ARVs
• Higher Migration (15% above best estimate) and No access to ARVs
• Best estimate Migration and All access to ARVs
• Best estimate Migration and Best estimate ARVs
• Best estimate Migration and No access to ARVs
• Lower Migration (15% below best estimate) and All access to ARVs
• Lower Migration (15% below best estimate) and Best estimate access to ARVs
• Lower Migration (15% below best estimate) and No access to ARVs

Kramer’s scenarios included half migration and a zero migration scenarios. These were not considered here but the half migration scenario, which was 50% lower that the best estimate, and the best estimate was used to interpolate what the scenarios at 15% lower than the best estimate would be. In Table 5 the ratios of Kramer’s scenarios compared to the best estimate were calculated. The ratios were then graphed and the log regression trend lines and formulas was displayed and extracted in Table 5..

Table 5: Kramer Scenario ratios to Best Estimate

The results of the scenario calculations are in Annexure 4 and are displayed in Figure 6

Figure 6: eThekwini population Projection Scenarios.

The extent to which the HIV/AIDS pandemic and access to ARV affects these scenarios in most notable in the following:

•Lower Migration (15% below best estimate) and All access to ARVs. Despite having lower migration the full or all access to ARVs results in a population 0.73 million higher than the SSA 2060 projection.

•Higher Migration (15% above best estimate) and No access to ARVs. A combination of higher migration and No access to ARVs results in a population lower than the SSA 2060 projection by 0.9 million.

Recommendation

Given that it is unlikely that ARV rollout will be either full or none it is recommended that the scenarios that may be most probable are:

•Higher Migration (15% above best estimate) and Best estimate access to ARVs

•SSA eThekwini projection derived from the KZN 2060 projection

•Lower Migration (15% below best estimate) and Best estimate access to ARVs

Appendices

Appendix 1: A methodology for Long term population projections: The approach used by Statistics South Africa

In meeting the demand for population projections, Statistics South Africa develops long term population projections annually. It is therefore important to note that population and other demographic data in each release form a new set of time series. Users should therefore compare the time series data in each statistical release and not data between statistical releases.

When developing the National long term population projections, Statistics South Africa uses the SPECTRUM system. In the projection with base-year 1985, fertility, mortality and international migration for the projection period are required. The base from which a population projection is done is very important as it has a big effect on the outcome of a projection. Census information regarding the population structure over time was used as an input in determining the base. The projections are unique for each year due to the assumptions made and the data inputs i.e. fertility, mortality and migration patterns.

We distinguish between two components in spectrum when developing the projections. These components are Demproj, a programme to make population projections using the cohort component method and AIM, a programme to project the consequences of the AIDS epidemic. These components are updated yearly incorporating better information regarding HIV and AIDS.

In Demproj, assumptions are made about the future pattern of fertility in South Africa over the long term, bearing in mind current patterns. The TFR in South Africa has declined over time and expected to continue to decline over the next 3 to 4 decades, from an estimated 2,55 in 2015 to 2,31 in 2030 to 1,77 by 2060 (table1).

Mortality patterns in South Africa over time, has been influenced greatly by the impact of HIV and AIDS. This is not to say mortality due to other causes are not of consequence. The life expectancy assumption entered into DemProj should be the life expectancy in the absence of AIDS. AIM calculates the number of AIDS deaths and determines a new life expectancy that incorporates the impact of AIDS. It is necessary to use this two-step process as model life tables (for specifying the age distribution of mortality) do not contain patterns of mortality that reflect the excess deaths caused by AIDS. New versions of the SPECTRUM program incorporate new survey results on HIV/AIDS and the effect on mortality and fertility levels.

To obtain a national epidemic curve, the Estimation and Projection Package (EPP) which now has been incorporated into SPECTRUM is applied using HIV prevalence data for the period 1990 to the most recently avaibable data from DOH, obtained from pregnant women attending antenatal clinics in South Africa. National HIV prevalence surveys such as those conducted by the HSRC are also used as inputs into the projection model. Assumptions are made about the treatment and progression of HIV and AIDS in South Africa over the long term, bearing in mind current patterns found in surveys and health data.

Life expectancy in South Africa, has declined significantly due to the impact of HIV and AIDS, however with the introduction and rollout of ARVs (Anti retroviral therapy ) over time along with improved treatment and coverage, life expectancy has increased since 2006 and is assumed to increase to 61,2 in 2014. The higher HIV prevalence levels, the impact of treatment, increased morbidity, combined with the progression of HIV and AIDS, the Life expectancy in South Africa inclusive of the impact of HIV and AIDS is expected to decline marginally to 58,6 by 2030 and thereafter increase marginally to 60,2 by 2060 (table 1).

Migration can be seen as the most difficult component of growth to forecast accurately as they are subject to much greater volatility than either fertility or mortality rates. The number of international immigrants coming into the country is provided by sex (male and female) for the period 1985 to 2060. Latest census information on immigrants is used to inform assumptions on immigrants. Lack of emigration data in census, requires the use of data sources external to South Africa. To determine the number of international immigrants, assumptions are made taking into account what is being reported from census results in receiving countries such as Australia, USA, UK, Canada, New Zealand etc. An assumption of international immigration is that there are more males than females coming into SA. There is a positive net migration in South Africa in 2015. Between 2015 and 2060 there is a decline in the net migration, though still positive by 2060 (table 1). The census populations on 2001 and 2011 census dates are included inputs into the projections

Table 1: Assumptions for long term projection 2015-2060

Provincial projections

For the provincial projections i.e. TFR, Mortality and migration are done for 5 year periods i.e. 2001-2006, 2006-2011, and 2011-2016, 2016-2021 etc. A cohort component method is used to develop the projection for each 5 year period. There are several principles that must be considered when implementing the cohort component method. To preserve the integrity of the age cohorts as they progress through time, it is helpful to follow basic principles: i.e. the number of years in the projection should be equal to the number of years in the age groups. Also projections by sex are essential in that projection for females determine the projection of births.

When projections for all the regions of a country are desired and the appropriate data are available, a multi-regional approach should be considered, as this is the only way to guarantee that the total migration flows between regions will sum to zero, or to the assumed level of international migration (United Nations, 1992). In South Africa, 2448 (9x8x17x2) migration streams are derived for multi-regional model applied in calculating migration streams by age group (17 in total) and sex for each province.

Using the provincial distribution from census 2001 as well as the projected 2001 national population, the 2001 base provincial populations are developed. Census 2011 provincial distribution is used as an input into the provincial projections as well.

The vital registration data on births and deaths are used as inputs into the TFR and Life expectancies across provinces, and projected forward. Similarly migration streams developed from census 2011 is projected forward assuming a particular progression pattern (table 2).

Table 2: Estimated provincial migration streams

REFERENCES

HSRC, 2009. South African National HIV Prevalence, Incidence, Behaviour and Communication Survey, 2008: A Turning Tide among Teenagers?. HSRC Press, Pretoria.

Stover J. 2003. AIM version 4. A computer program for HIV/AIDS projections and examining the social and economic impacts of AIDS. Spectrum system of Policy Models. The Futures Group International.

United Nations. 2002a. HIV/AIDS and fertility in sub-Saharan Africa: A perspective of the research literature. United Nations, New York.

United Nations. 2002b. Fertility levels and trends in countries with intermediate levels of fertility: A background paper for the Expert Group Meeting on Completing the Fertility Transition. 11-14 March 2002. United Nations, New York.

Willekens F & Rogers A. 1978. Spatial Population Analysis: Methods and Computer Programs. International Institute for Applied System Analysis. Research Report, RR 78-18. Laxenberg, Austria.

World Health Organisation. 2001. Prevention of mother-to-child transmission of HIV: Selection and use of Nevirapine. Technical notes. World Health Organisation, Geneva, Switzerland.

Appendix 2: SSA: Short Term Population Forecasts for eThekwini

Appendix 3: KZN and ETH projected population 2021 to 2060